diff --git "a/openvino_model.xml" "b/openvino_model.xml" new file mode 100644--- /dev/null +++ "b/openvino_model.xml" @@ -0,0 +1,96594 @@ +<?xml version="1.0"?> +<net name="Model0" version="11"> + <layers> + <layer id="3" name="input_ids" type="Parameter" version="opset1"> + <data shape="?,?" element_type="i64" /> + <output> + <port id="0" precision="I64" names="input_ids"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2" name="attention_mask" type="Parameter" version="opset1"> + <data shape="?,?" element_type="i64" /> + <output> + <port id="0" precision="I64" names="attention_mask"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1" name="position_ids" type="Parameter" version="opset1"> + <data shape="?,?" element_type="i64" /> + <output> + <port id="0" precision="I64" names="position_ids"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="0" name="beam_idx" type="Parameter" version="opset1"> + <data shape="?" element_type="i32" /> + <output> + <port id="0" precision="I32" names="beam_idx"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4" name="Constant_123982_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="5" name="Constant_123982" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="6" name="ShapeOf_123968" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="7" name="Constant_123970" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="8" name="Constant_123972" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="9" name="Gather_123973" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="10" name="Constant_123975" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="11" name="Constant_123977" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="12" name="Constant_123979" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="13" name="Concat_123980" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="14" name="Broadcast_123983" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="15" name="ReadValue_121809" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.9.valuepresent.9.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.9.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="16" name="Constant_119492" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="17" name="Gather_119493" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="18" name="Constant_111226_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="26" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="19" name="Constant_111226" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="20" name="self.model.embed_tokens.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="32000, 2048" offset="4122" size="65536000" /> + <output> + <port id="0" precision="U8"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="21" name="Convert_12039" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="22" name="self.model.embed_tokens.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="32000, 1" offset="65540122" size="32000" /> + <output> + <port id="0" precision="U8"> + <dim>32000</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="23" name="Convert_12042" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>32000</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>32000</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="24" name="self.model.embed_tokens.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>32000</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="25" name="self.model.embed_tokens.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="32000, 1" offset="65572122" size="64000" /> + <output> + <port id="0" precision="FP16"> + <dim>32000</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="26" name="self.model.embed_tokens.weight_compressed/fq_weights_0" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>32000</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="27" name="self.model.embed_tokens.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="28" name="__module.model.embed_tokens/aten::embedding/Convert" type="Convert" version="opset1"> + <data destination_type="i32" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="29" name="__module.model.embed_tokens/aten::embedding/Constant" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="30" name="__module.model.embed_tokens/aten::embedding/Gather" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>32000</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="FP32" names="220,268,279,hidden_states.1,inputs_embeds"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="31" name="Constant_111134_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="65636126" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="32" name="Constant_111134" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="33" name="Constant_111133_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="34" name="Constant_111133" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="35" name="Constant_111131_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="36" name="Constant_111131" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="37" name="__module.model.layers.0.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="269"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="38" name="Constant_289" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="270"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="39" name="__module.model.layers.0.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="271,variance.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="40" name="Constant_111132_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="41" name="Constant_111132" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="42" name="__module.model.layers.0.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="272"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="43" name="__module.model.layers.0.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="44" name="__module.model.layers.0.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="273"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="45" name="__module.model.layers.0.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="274,275,hidden_states.3,hidden_states.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="46" name="__module.model.layers.0.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="276,278,hidden_states.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="47" name="self.model.layers.0.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="65640236" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="48" name="Convert_12069" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="49" name="self.model.layers.0.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="69834540" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="50" name="Convert_12072" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="51" name="self.model.layers.0.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="52" name="self.model.layers.0.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="69836588" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="53" name="self.model.layers.0.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="54" name="self.model.layers.0.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="55" name="__module.model.layers.0.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="288,query_states.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="56" name="__module.model.layers.0.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="57" name="Constant_93174" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="58" name="Constant_93175" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="59" name="Gather_93176" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="60" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_1" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840700" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="61" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_2" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="62" name="__module.model.layers.0.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="293"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="63" name="__module.model.layers.0.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="294"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="64" name="__module.model.layers.0.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="65" name="__module.model.layers.0.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="295,q.1"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="66" name="__module.model.layers.0.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="306"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="67" name="__module.model.layers.0.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="68" name="__module.model.layers.0.self_attn.rotary_emb/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="69" name="Constant_92581" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="70" name="__module.model.layers.0.self_attn.rotary_emb/aten::size/Constant" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="71" name="__module.model.layers.0.self_attn.rotary_emb/aten::size/Gather" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1030,1210,1390,1570,1750,1930,2110,2290,2470,2650,2830,3010,307,3190,3370,3550,3730,3910,4090,490,670,850"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="72" name="__module.model.layers.0.self_attn.rotary_emb/prim::ListConstruct/Reshape_0" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="73" name="__module.model.layers.0.self_attn.rotary_emb/prim::ListConstruct/Reshape_1" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840788" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="74" name="__module.model.layers.0.self_attn.rotary_emb/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1031,1211,1391,1571,1751,1931,2111,2291,2471,2651,2831,3011,308,3191,3371,3551,3731,3911,4091,491,671,851"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="75" name="__module.model.layers.0.self_attn.rotary_emb/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="76" name="__module.model.layers.0.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="309,314,inv_freq_expanded.1,inv_freq_expanded.3"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="77" name="168" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="69840788" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" names="168" /> + </output> + </layer> + <layer id="78" name="__module.model.layers.0.self_attn.rotary_emb/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1034,1035,1214,1215,1394,1395,1574,1575,1754,1755,1934,1935,2114,2115,2294,2295,2474,2475,2654,2655,2834,2835,3014,3015,311,312,3194,3195,3374,3375,3554,3555,3734,3735,3914,3915,4094,4095,494,495,674,675,854,855"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="79" name="__module.model.layers.0.self_attn.rotary_emb/aten::to/Convert_2" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1036,1038,1216,1218,1396,1398,1576,1578,1756,1758,1936,1938,2116,2118,2296,2298,2476,2478,2656,2658,2836,2838,3016,3018,313,315,3196,3198,3376,3378,3556,3558,3736,3738,3916,3918,4096,4098,496,498,676,678,856,858,position_ids_expanded,position_ids_expanded.1,position_ids_expanded.11,position_ids_expanded.13,position_ids_expanded.15,position_ids_expanded.17,position_ids_expanded.19,position_ids_expanded.21,position_ids_expanded.23,position_ids_expanded.25,position_ids_expanded.27,position_ids_expanded.29,position_ids_expanded.3,position_ids_expanded.31,position_ids_expanded.33,position_ids_expanded.35,position_ids_expanded.37,position_ids_expanded.39,position_ids_expanded.41,position_ids_expanded.43,position_ids_expanded.45,position_ids_expanded.47,position_ids_expanded.49,position_ids_expanded.5,position_ids_expanded.51,position_ids_expanded.53,position_ids_expanded.55,position_ids_expanded.57,position_ids_expanded.59,position_ids_expanded.61,position_ids_expanded.63,position_ids_expanded.65,position_ids_expanded.67,position_ids_expanded.69,position_ids_expanded.7,position_ids_expanded.71,position_ids_expanded.73,position_ids_expanded.75,position_ids_expanded.77,position_ids_expanded.79,position_ids_expanded.81,position_ids_expanded.83,position_ids_expanded.85,position_ids_expanded.9"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="80" name="__module.model.layers.0.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="316"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="81" name="__module.model.layers.0.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="82" name="__module.model.layers.0.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="317"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="83" name="__module.model.layers.0.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="319,emb.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="84" name="__module.model.layers.0.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="320,322,325,cos.1,cos.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="85" name="__module.model.layers.0.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="327,cos.5"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="86" name="__module.model.layers.0.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="329"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="87" name="Constant_96056" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="88" name="Constant_96055" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="89" name="__module.model.layers.0.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="90" name="161" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" names="161" /> + </output> + </layer> + <layer id="91" name="__module.model.layers.0.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="92" name="__module.model.layers.0.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="330,335" /> + </output> + </layer> + <layer id="93" name="Constant_37" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" names="155" /> + </output> + </layer> + <layer id="94" name="__module.model.layers.0.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="95" name="__module.model.layers.0.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="332,333,337,338" /> + </output> + </layer> + <layer id="96" name="__module.model.layers.0.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="97" name="__module.model.layers.0.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="98" name="Constant_96054" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="99" name="ScatterUpdate_96057" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="100" name="Constant_96060" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="101" name="Constant_96063" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="102" name="__module.model.layers.0.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="339,x2.1"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="103" name="Constant_111135_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="104" name="Constant_111135" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="105" name="__module.model.layers.0.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="340"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="106" name="Constant_96108" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="107" name="Constant_96113" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="108" name="__module.model.layers.0.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="334"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="109" name="__module.model.layers.0.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="342"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="110" name="__module.model.layers.0.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="321,323,326,sin.1,sin.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="111" name="__module.model.layers.0.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="328,sin.5"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="112" name="__module.model.layers.0.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="343"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="113" name="__module.model.layers.0.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="344"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="114" name="Constant_123995_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="115" name="Constant_123995" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="116" name="Constant_123988" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="117" name="Constant_123990" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="118" name="Constant_123992" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="119" name="Concat_123993" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="120" name="Broadcast_123996" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="121" name="ReadValue_121723" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.0.keypresent.0.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.0.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="122" name="Constant_119435" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="123" name="Gather_119436" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="124" name="self.model.layers.0.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="69840922" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="125" name="Convert_12049" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="126" name="self.model.layers.0.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="70365210" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="127" name="Convert_12052" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="128" name="self.model.layers.0.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="129" name="self.model.layers.0.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="70365466" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="130" name="self.model.layers.0.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="131" name="self.model.layers.0.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="132" name="__module.model.layers.0.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="290,key_states.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="133" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_1_1" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="134" name="__module.model.layers.0.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="296,299"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="135" name="__module.model.layers.0.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="297"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="136" name="__module.model.layers.0.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="137" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="298,k.1"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="138" name="__module.model.layers.0.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="345"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="139" name="Constant_96156" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="140" name="Constant_96155" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="141" name="__module.model.layers.0.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="142" name="__module.model.layers.0.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="143" name="__module.model.layers.0.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="346,351" /> + </output> + </layer> + <layer id="144" name="__module.model.layers.0.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="145" name="__module.model.layers.0.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="348,349,353,354" /> + </output> + </layer> + <layer id="146" name="__module.model.layers.0.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="147" name="__module.model.layers.0.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="148" name="Constant_96154" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="149" name="ScatterUpdate_96157" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="150" name="Constant_96160" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="151" name="Constant_96163" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="152" name="__module.model.layers.0.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="355,x2.3"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="153" name="Constant_111136_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="154" name="Constant_111136" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="155" name="__module.model.layers.0.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="356"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="156" name="Constant_96208" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="157" name="Constant_96213" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="158" name="__module.model.layers.0.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="350"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="159" name="__module.model.layers.0.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="358"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="160" name="__module.model.layers.0.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="359"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="161" name="__module.model.layers.0.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="360,key_states.5"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="162" name="__module.model.layers.0.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.0.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="163" name="170" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="69840848" size="8" /> + <output> + <port id="0" precision="I64" names="170" /> + </output> + </layer> + <layer id="164" name="__module.model.layers.0.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="375,376,377"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="165" name="__module.model.layers.0.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="166" name="Constant_92584" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="167" name="__module.model.layers.0.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="168" name="__module.model.layers.0.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="368"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="169" name="__module.model.layers.0.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="170" name="__module.model.layers.0.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="369" /> + </output> + </layer> + <layer id="171" name="Constant_77293" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="172" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="173" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_1_2" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="70365978" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="174" name="Constant_93184" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="175" name="Constant_93185" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="176" name="Gather_93186" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="177" name="__module.model.layers.0.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="378"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="178" name="__module.model.layers.0.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="179" name="__module.model.layers.0.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="379,hidden_states.11"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="180" name="Constant_34" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="70365978" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" names="152" /> + </output> + </layer> + <layer id="181" name="__module.model.layers.0.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="380,381" /> + </output> + </layer> + <layer id="182" name="Constant_77297" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="183" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="184" name="__module.model.layers.0.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="382"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="185" name="__module.model.layers.0.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="383,key_states.7"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="186" name="Constant_124008_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="187" name="Constant_124008" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="188" name="Constant_124001" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="189" name="Constant_124003" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="190" name="Constant_124005" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="191" name="Concat_124006" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="192" name="Broadcast_124009" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="193" name="ReadValue_121725" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.0.valuepresent.0.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.0.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="194" name="Constant_119438" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="195" name="Gather_119439" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="196" name="self.model.layers.0.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="70366002" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="197" name="Convert_12059" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="198" name="self.model.layers.0.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="70890290" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="199" name="Convert_12062" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="200" name="self.model.layers.0.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="201" name="self.model.layers.0.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="70890546" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="202" name="self.model.layers.0.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="203" name="self.model.layers.0.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="204" name="__module.model.layers.0.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="292,value_states.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="205" name="__module.model.layers.0.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="300"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="206" name="__module.model.layers.0.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="207" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="301"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="208" name="__module.model.layers.0.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.0.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="209" name="__module.model.layers.0.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="391,392,393"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="210" name="__module.model.layers.0.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="211" name="Constant_92593" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="212" name="__module.model.layers.0.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="213" name="__module.model.layers.0.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="384"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="214" name="__module.model.layers.0.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="215" name="__module.model.layers.0.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="385" /> + </output> + </layer> + <layer id="216" name="Constant_77299" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="217" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="218" name="Constant_93194" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="219" name="Constant_93195" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="220" name="Gather_93196" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="221" name="__module.model.layers.0.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="394"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="222" name="__module.model.layers.0.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="223" name="__module.model.layers.0.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="395,hidden_states.15"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="224" name="__module.model.layers.0.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="396,397" /> + </output> + </layer> + <layer id="225" name="Constant_77302" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="226" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="227" name="__module.model.layers.0.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="398"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="228" name="__module.model.layers.0.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="399"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="229" name="__module.model/aten::triu/Constant" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="230" name="Multiply_106246_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="70891058" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="231" name="Multiply_106246" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="232" name="Constant_106251" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70891060" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="233" name="__module.model/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="234" name="__module.model/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="235" name="__module.model/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="226,231" /> + </output> + </layer> + <layer id="236" name="Constant_77303" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="237" name="__module.model/prim::ListConstruct/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="238" name="__module.model/aten::size/ShapeOf_1" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="239" name="__module.model/aten::size/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="240" name="__module.model/aten::size/Gather_1" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="232,250" /> + </output> + </layer> + <layer id="241" name="Constant_77304" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="242" name="__module.model/prim::ListConstruct/Reshape_0" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="243" name="__module.model/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="233"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="244" name="Maximum_106252" type="Maximum" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="245" name="__module.model/aten::mul/Multiply" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="235"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="246" name="__module.model/aten::triu/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i32" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I32"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="247" name="__module.model/aten::triu/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="70891076" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="248" name="__module.model/aten::triu/Gather" type="Gather" version="opset7"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I32"> + <dim>2</dim> + </port> + <port id="1" precision="I32" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I32" /> + </output> + </layer> + <layer id="249" name="__module.model/aten::triu/Constant_2" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="70891080" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="250" name="__module.model/aten::triu/Range" type="Range" version="opset4"> + <data output_type="i32" /> + <input> + <port id="0" precision="I32" /> + <port id="1" precision="I32" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I32"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="251" name="__module.model/aten::triu/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I32"> + <dim>-1</dim> + </port> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I32"> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="252" name="__module.model/aten::triu/Convert" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="70891080" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="253" name="__module.model/aten::triu/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="70891084" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="254" name="__module.model/aten::triu/Gather_1" type="Gather" version="opset7"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I32"> + <dim>2</dim> + </port> + <port id="1" precision="I32" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I32" /> + </output> + </layer> + <layer id="255" name="__module.model/aten::triu/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I32" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I32" /> + </output> + </layer> + <layer id="256" name="__module.model/aten::triu/Range_1" type="Range" version="opset4"> + <data output_type="i32" /> + <input> + <port id="0" precision="I32" /> + <port id="1" precision="I32" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I32"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="257" name="__module.model/aten::triu/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I32"> + <dim>-1</dim> + </port> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I32"> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="258" name="__module.model/aten::triu/GreaterEqual" type="GreaterEqual" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I32"> + <dim>1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="BOOL"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="259" name="__module.model/aten::triu/ConvertLike_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="260" name="__module.model/aten::triu/ConvertLike" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="261" name="__module.model/aten::triu/Select" type="Select" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="BOOL"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="2" precision="FP32" /> + </input> + <output> + <port id="3" precision="FP32" names="236_1"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="262" name="__module.model/aten::arange/Constant" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="263" name="__module.model/aten::arange/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="70891080" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="264" name="__module.model/aten::arange/Range" type="Range" version="opset4"> + <data output_type="f32" /> + <input> + <port id="0" precision="I32" /> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="265" name="__module.model/aten::arange/ConvertLike" type="Convert" version="opset1"> + <data destination_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I64" names="237"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="266" name="__module.model/aten::gt/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="267" name="__module.model/aten::size/ShapeOf_2" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="268" name="__module.model/aten::size/Constant_2" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="269" name="__module.model/aten::size/Gather_2" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="224" /> + </output> + </layer> + <layer id="270" name="__module.model/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="228" /> + </output> + </layer> + <layer id="271" name="__module.model/aten::arange/Constant_2" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="70891080" size="4" /> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="272" name="__module.model/aten::arange/Range_1" type="Range" version="opset4"> + <data output_type="f32" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="FP32" names="230,cache_position"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="273" name="Constant_178" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70891088" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" names="238"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="274" name="__module.model/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="239"> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="275" name="__module.model/aten::gt/Greater" type="Greater" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="BOOL" names="240"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="276" name="__module.model/aten::mul_/ConvertLike" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="BOOL"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="277" name="__module.model/aten::mul_/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="236,causal_mask.1"> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="278" name="163" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" names="163" /> + </output> + </layer> + <layer id="279" name="__module.model/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="242"> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="280" name="__module.model/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="243,244,245"> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="281" name="Constant_92602" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="282" name="__module.model/aten::size/Constant_4" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="283" name="__module.model/aten::size/Gather_4" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="246"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="284" name="__module.model/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="247"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="285" name="__module.model/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="286" name="__module.model/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="248,causal_mask.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="287" name="Constant_77306" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="288" name="SliceAssign_271/Reshape_0" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="289" name="Constant_56731" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="290" name="SliceAssign_271/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="291" name="SliceAssign_271/ReduceProd" type="ReduceProd" version="opset1"> + <data keep_dims="false" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="292" name="Constant_56732" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="69840788" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="293" name="SliceAssign_271/Range" type="Range" version="opset4"> + <data output_type="i64" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="294" name="SliceAssign_271/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="295" name="Constant_96356" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="296" name="Constant_96357" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="297" name="Constant_96353" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="298" name="__module.model/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="299" name="__module.model/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="300" name="Constant_96352" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="301" name="ScatterUpdate_96358" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="302" name="Constant_96361" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="303" name="SliceAssign_271/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="304" name="Constant_56734" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70891088" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="305" name="SliceAssign_271/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="306" name="Constant_96368" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="307" name="Constant_96373" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="308" name="__module.model/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="251,259,261_1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="309" name="Constant_111137_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="0" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="310" name="Constant_111137" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="311" name="__module.model/aten::eq/Equal" type="Equal" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="BOOL" names="252"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="312" name="__module.model/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="254"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="313" name="__module.model/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="255,256"> + <dim>-1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="314" name="__module.model/aten::eq/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <input> + <port id="0" precision="I64"> + <dim>-1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="315" name="Constant_111138_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="0" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="316" name="Constant_111138" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="317" name="__module.model/aten::eq/Equal_1" type="Equal" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="BOOL" names="257"> + <dim>-1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="318" name="__module.model/aten::mul/LogicalAnd" type="LogicalAnd" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="BOOL"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="BOOL"> + <dim>-1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="BOOL" names="258,padding_mask"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="319" name="__module.model/aten::masked_fill/ConvertLike_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="70891058" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="320" name="__module.model/aten::masked_fill/ConvertLike" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="321" name="__module.model/aten::masked_fill/Select" type="Select" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="BOOL"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32" /> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="3" precision="FP32" names="260"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="322" name="__module.model/aten::copy_/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="323" name="__module.model/aten::copy_/Broadcast" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="261"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="324" name="Constant_77307" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="325" name="SliceAssign_271/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="326" name="SliceAssign_271/ScatterNDUpdate" type="ScatterNDUpdate" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + </port> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="327" name="SliceAssign_271/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1121,1122,1123,1301,1302,1303,1481,1482,1483,1661,1662,1663,1841,1842,1843,2021,2022,2023,2201,2202,2203,2381,2382,2383,249,2561,2562,2563,2741,2742,2743,2921,2922,2923,3101,3102,3103,3281,3282,3283,3461,3462,3463,3641,3642,3643,3821,3822,3823,4001,4002,4003,401,402,403,4181,4182,4183,581,582,583,761,762,763,941,942,943"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="328" name="Constant_96418" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="329" name="Constant_96419" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="330" name="Constant_96415" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="331" name="__module.model.layers.0.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="332" name="Constant_92605" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="333" name="__module.model.layers.0.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="334" name="__module.model.layers.0.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="400"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="335" name="Constant_96414" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="336" name="ScatterUpdate_96420" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="337" name="Constant_96423" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="338" name="__module.model.layers.0.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="404"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="339" name="__module.model.layers.0.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="405,attn_output.1"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="340" name="__module.model.layers.0.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="341" name="__module.model.layers.0.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="406"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="342" name="__module.model.layers.0.self_attn/prim::ListConstruct/Reshape_1_7" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="70891104" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="343" name="__module.model.layers.0.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="408"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="344" name="__module.model.layers.0.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="409"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="345" name="self.model.layers.0.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="70891112" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="346" name="Convert_12079" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="347" name="self.model.layers.0.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="75085416" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="348" name="Convert_12082" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="349" name="self.model.layers.0.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="350" name="self.model.layers.0.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="75087464" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="351" name="self.model.layers.0.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="352" name="self.model.layers.0.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="353" name="__module.model.layers.0.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="411,413,hidden_states.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="354" name="__module.model.layers.0/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="416,418,429,hidden_states.19,hidden_states.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="355" name="Constant_111142_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="75091560" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="356" name="Constant_111142" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="357" name="Constant_111141_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="358" name="Constant_111141" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="359" name="Constant_111139_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="360" name="Constant_111139" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="361" name="__module.model.layers.0.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="419"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="362" name="Constant_975" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="420"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="363" name="__module.model.layers.0.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="421,variance.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="364" name="Constant_111140_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="365" name="Constant_111140" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="366" name="__module.model.layers.0.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="422"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="367" name="__module.model.layers.0.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="368" name="__module.model.layers.0.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="423"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="369" name="__module.model.layers.0.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="424,425,hidden_states.23,hidden_states.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="370" name="__module.model.layers.0.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="426,428"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="371" name="self.model.layers.0.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="75095656" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="372" name="Convert_12099" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="373" name="self.model.layers.0.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="86629992" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="374" name="Convert_12102" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="375" name="self.model.layers.0.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="376" name="self.model.layers.0.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="86635624" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="377" name="self.model.layers.0.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="378" name="self.model.layers.0.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="379" name="__module.model.layers.0.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="434,input.1"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="380" name="__module.model.layers.0.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="435"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="381" name="self.model.layers.0.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="86646888" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="382" name="Convert_12089" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="383" name="self.model.layers.0.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="98181224" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="384" name="Convert_12092" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="385" name="self.model.layers.0.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="386" name="self.model.layers.0.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="98186856" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="387" name="self.model.layers.0.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="388" name="self.model.layers.0.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="389" name="__module.model.layers.0.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="437"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="390" name="__module.model.layers.0.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="438"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="391" name="self.model.layers.0.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="98198120" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="392" name="Convert_12109" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="393" name="self.model.layers.0.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="109732456" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="394" name="Convert_12112" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="395" name="self.model.layers.0.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="396" name="self.model.layers.0.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="109734504" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="397" name="self.model.layers.0.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="398" name="self.model.layers.0.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="399" name="__module.model.layers.0.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="440,hidden_states.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="400" name="__module.model.layers.0/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="441,443,451,462,hidden_states.29,hidden_states.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="401" name="Constant_111146_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="109738600" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="402" name="Constant_111146" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="403" name="Constant_111145_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="404" name="Constant_111145" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="405" name="Constant_111143_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="406" name="Constant_111143" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="407" name="__module.model.layers.1.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="452"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="408" name="Constant_1046" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="453"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="409" name="__module.model.layers.1.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="454,variance.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="410" name="Constant_111144_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="411" name="Constant_111144" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="412" name="__module.model.layers.1.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="455"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="413" name="__module.model.layers.1.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="414" name="__module.model.layers.1.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="456"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="415" name="__module.model.layers.1.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="457,458,hidden_states.33,hidden_states.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="416" name="__module.model.layers.1.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="459,461,hidden_states.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="417" name="self.model.layers.1.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="109742696" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="418" name="Convert_12139" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="419" name="self.model.layers.1.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="113937000" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="420" name="Convert_12142" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="421" name="self.model.layers.1.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="422" name="self.model.layers.1.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="113939048" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="423" name="self.model.layers.1.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="424" name="self.model.layers.1.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="425" name="__module.model.layers.1.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="471,query_states.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="426" name="__module.model.layers.1.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="427" name="Constant_93209" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="428" name="Constant_93210" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="429" name="Gather_93211" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="430" name="__module.model.layers.1.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="476"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="431" name="__module.model.layers.1.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="477"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="432" name="__module.model.layers.1.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="433" name="__module.model.layers.1.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="478,q.3"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="434" name="__module.model.layers.1.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="489"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="435" name="__module.model.layers.1.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="436" name="__module.model.layers.1.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="492,497,inv_freq_expanded.5,inv_freq_expanded.7"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="437" name="__module.model.layers.1.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="499"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="438" name="__module.model.layers.1.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="439" name="__module.model.layers.1.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="500"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="440" name="__module.model.layers.1.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="502,emb.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="441" name="__module.model.layers.1.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="503,505,508,cos.7,cos.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="442" name="__module.model.layers.1.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="510,cos.11"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="443" name="__module.model.layers.1.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="512"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="444" name="Constant_96428" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="445" name="Constant_96427" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="446" name="__module.model.layers.1.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="447" name="__module.model.layers.1.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="448" name="__module.model.layers.1.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="513,518" /> + </output> + </layer> + <layer id="449" name="__module.model.layers.1.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="450" name="__module.model.layers.1.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="515,516,520,521" /> + </output> + </layer> + <layer id="451" name="__module.model.layers.1.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="452" name="__module.model.layers.1.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="453" name="Constant_96426" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="454" name="ScatterUpdate_96429" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="455" name="Constant_96432" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="456" name="Constant_96435" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="457" name="__module.model.layers.1.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="522,x2.5"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="458" name="Constant_111147_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="459" name="Constant_111147" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="460" name="__module.model.layers.1.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="523"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="461" name="Constant_96480" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="462" name="Constant_96485" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="463" name="__module.model.layers.1.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="517"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="464" name="__module.model.layers.1.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="525"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="465" name="__module.model.layers.1.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="504,506,509,sin.7,sin.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="466" name="__module.model.layers.1.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="511,sin.11"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="467" name="__module.model.layers.1.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="526"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="468" name="__module.model.layers.1.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="527"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="469" name="Constant_124021_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="470" name="Constant_124021" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="471" name="Constant_124014" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="472" name="Constant_124016" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="473" name="Constant_124018" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="474" name="Concat_124019" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="475" name="Broadcast_124022" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="476" name="ReadValue_121727" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.1.keypresent.1.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.1.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="477" name="Constant_119441" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="478" name="Gather_119442" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="479" name="self.model.layers.1.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="113943144" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="480" name="Convert_12119" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="481" name="self.model.layers.1.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="114467432" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="482" name="Convert_12122" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="483" name="self.model.layers.1.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="484" name="self.model.layers.1.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="114467688" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="485" name="self.model.layers.1.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="486" name="self.model.layers.1.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="487" name="__module.model.layers.1.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="473,key_states.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="488" name="__module.model.layers.1.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="479,482"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="489" name="__module.model.layers.1.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="480"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="490" name="__module.model.layers.1.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="491" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="481,k.3"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="492" name="__module.model.layers.1.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="528"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="493" name="Constant_96528" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="494" name="Constant_96527" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="495" name="__module.model.layers.1.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="496" name="__module.model.layers.1.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="497" name="__module.model.layers.1.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="529,534" /> + </output> + </layer> + <layer id="498" name="__module.model.layers.1.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="499" name="__module.model.layers.1.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="531,532,536,537" /> + </output> + </layer> + <layer id="500" name="__module.model.layers.1.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="501" name="__module.model.layers.1.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="502" name="Constant_96526" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="503" name="ScatterUpdate_96529" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="504" name="Constant_96532" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="505" name="Constant_96535" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="506" name="__module.model.layers.1.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="538,x2.7"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="507" name="Constant_111148_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="508" name="Constant_111148" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="509" name="__module.model.layers.1.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="539"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="510" name="Constant_96580" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="511" name="Constant_96585" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="512" name="__module.model.layers.1.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="533"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="513" name="__module.model.layers.1.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="541"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="514" name="__module.model.layers.1.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="542"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="515" name="__module.model.layers.1.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="543"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="516" name="__module.model.layers.1.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.1.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="517" name="__module.model.layers.1.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="555,556,557"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="518" name="__module.model.layers.1.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="519" name="Constant_92614" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="520" name="__module.model.layers.1.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="521" name="__module.model.layers.1.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="548"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="522" name="__module.model.layers.1.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="523" name="__module.model.layers.1.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="549" /> + </output> + </layer> + <layer id="524" name="Constant_77312" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="525" name="__module.model.layers.1.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="526" name="Constant_93219" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="527" name="Constant_93220" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="528" name="Gather_93221" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="529" name="__module.model.layers.1.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="558"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="530" name="__module.model.layers.1.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="531" name="__module.model.layers.1.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="559,hidden_states.41"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="532" name="__module.model.layers.1.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="560,561" /> + </output> + </layer> + <layer id="533" name="Constant_77315" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="534" name="__module.model.layers.1.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="535" name="__module.model.layers.1.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="562"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="536" name="__module.model.layers.1.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="563,key_states.11"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="537" name="Constant_124034_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="538" name="Constant_124034" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="539" name="Constant_124027" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="540" name="Constant_124029" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="541" name="Constant_124031" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="542" name="Concat_124032" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="543" name="Broadcast_124035" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="544" name="ReadValue_121729" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.1.valuepresent.1.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.1.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="545" name="Constant_119444" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="546" name="Gather_119445" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="547" name="self.model.layers.1.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="114468200" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="548" name="Convert_12129" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="549" name="self.model.layers.1.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="114992488" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="550" name="Convert_12132" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="551" name="self.model.layers.1.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="552" name="self.model.layers.1.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="114992744" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="553" name="self.model.layers.1.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="554" name="self.model.layers.1.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="555" name="__module.model.layers.1.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="475,value_states.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="556" name="__module.model.layers.1.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="483"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="557" name="__module.model.layers.1.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="558" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="484"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="559" name="__module.model.layers.1.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.1.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="560" name="__module.model.layers.1.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="571,572,573"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="561" name="__module.model.layers.1.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="562" name="Constant_92623" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="563" name="__module.model.layers.1.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="564" name="__module.model.layers.1.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="564"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="565" name="__module.model.layers.1.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="566" name="__module.model.layers.1.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="565" /> + </output> + </layer> + <layer id="567" name="Constant_77317" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="568" name="__module.model.layers.1.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="569" name="Constant_93229" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="570" name="Constant_93230" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="571" name="Gather_93231" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="572" name="__module.model.layers.1.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="574"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="573" name="__module.model.layers.1.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="574" name="__module.model.layers.1.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="575,hidden_states.45"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="575" name="__module.model.layers.1.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="576,577" /> + </output> + </layer> + <layer id="576" name="Constant_77320" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="577" name="__module.model.layers.1.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="578" name="__module.model.layers.1.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="578"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="579" name="__module.model.layers.1.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="579"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="580" name="Constant_96704" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="581" name="Constant_96705" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="582" name="Constant_96701" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="583" name="__module.model.layers.1.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="584" name="Constant_92632" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="585" name="__module.model.layers.1.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="586" name="__module.model.layers.1.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="580"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="587" name="Constant_96700" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="588" name="ScatterUpdate_96706" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="589" name="Constant_96709" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="590" name="__module.model.layers.1.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="584"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="591" name="__module.model.layers.1.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="585,attn_output.5"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="592" name="__module.model.layers.1.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="593" name="__module.model.layers.1.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="586"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="594" name="__module.model.layers.1.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="588"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="595" name="__module.model.layers.1.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="589"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="596" name="self.model.layers.1.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="114993256" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="597" name="Convert_12149" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="598" name="self.model.layers.1.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="119187560" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="599" name="Convert_12152" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="600" name="self.model.layers.1.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="601" name="self.model.layers.1.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="119189608" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="602" name="self.model.layers.1.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="603" name="self.model.layers.1.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="604" name="__module.model.layers.1.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="591,593,hidden_states.47"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="605" name="__module.model.layers.1/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="596,598,609,hidden_states.49,hidden_states.51"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="606" name="Constant_111152_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="119193704" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="607" name="Constant_111152" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="608" name="Constant_111151_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="609" name="Constant_111151" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="610" name="Constant_111149_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="611" name="Constant_111149" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="612" name="__module.model.layers.1.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="599"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="613" name="Constant_1726" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="600"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="614" name="__module.model.layers.1.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="601,variance.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="615" name="Constant_111150_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="616" name="Constant_111150" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="617" name="__module.model.layers.1.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="602"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="618" name="__module.model.layers.1.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="619" name="__module.model.layers.1.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="603"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="620" name="__module.model.layers.1.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="604,605,hidden_states.53,hidden_states.55"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="621" name="__module.model.layers.1.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="606,608"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="622" name="self.model.layers.1.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="119197800" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="623" name="Convert_12169" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="624" name="self.model.layers.1.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="130732136" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="625" name="Convert_12172" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="626" name="self.model.layers.1.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="627" name="self.model.layers.1.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="130737768" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="628" name="self.model.layers.1.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="629" name="self.model.layers.1.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="630" name="__module.model.layers.1.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="614,input.3"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="631" name="__module.model.layers.1.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="615"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="632" name="self.model.layers.1.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="130749032" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="633" name="Convert_12159" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="634" name="self.model.layers.1.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="142283368" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="635" name="Convert_12162" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="636" name="self.model.layers.1.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="637" name="self.model.layers.1.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="142289000" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="638" name="self.model.layers.1.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="639" name="self.model.layers.1.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="640" name="__module.model.layers.1.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="617"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="641" name="__module.model.layers.1.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="618"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="642" name="self.model.layers.1.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="142300264" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="643" name="Convert_12179" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="644" name="self.model.layers.1.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="153834600" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="645" name="Convert_12182" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="646" name="self.model.layers.1.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="647" name="self.model.layers.1.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="153836648" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="648" name="self.model.layers.1.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="649" name="self.model.layers.1.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="650" name="__module.model.layers.1.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="620,hidden_states.57"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="651" name="__module.model.layers.1/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="621,623,631,642,hidden_states.59,hidden_states.61"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="652" name="Constant_111156_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="153840744" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="653" name="Constant_111156" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="654" name="Constant_111155_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="655" name="Constant_111155" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="656" name="Constant_111153_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="657" name="Constant_111153" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="658" name="__module.model.layers.2.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="632"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="659" name="Constant_1797" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="633"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="660" name="__module.model.layers.2.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="634,variance.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="661" name="Constant_111154_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="662" name="Constant_111154" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="663" name="__module.model.layers.2.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="635"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="664" name="__module.model.layers.2.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="665" name="__module.model.layers.2.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="636"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="666" name="__module.model.layers.2.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="637,638,hidden_states.63,hidden_states.65"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="667" name="__module.model.layers.2.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="639,641,hidden_states.67"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="668" name="self.model.layers.2.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="153844840" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="669" name="Convert_12209" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="670" name="self.model.layers.2.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="158039144" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="671" name="Convert_12212" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="672" name="self.model.layers.2.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="673" name="self.model.layers.2.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="158041192" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="674" name="self.model.layers.2.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="675" name="self.model.layers.2.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="676" name="__module.model.layers.2.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="651,query_states.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="677" name="__module.model.layers.2.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="678" name="Constant_93244" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="679" name="Constant_93245" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="680" name="Gather_93246" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="681" name="__module.model.layers.2.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="656"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="682" name="__module.model.layers.2.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="657"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="683" name="__module.model.layers.2.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="684" name="__module.model.layers.2.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="658,q.5"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="685" name="__module.model.layers.2.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="669"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="686" name="__module.model.layers.2.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="687" name="__module.model.layers.2.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="672,677,inv_freq_expanded.11,inv_freq_expanded.9"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="688" name="__module.model.layers.2.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="679"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="689" name="__module.model.layers.2.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="690" name="__module.model.layers.2.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="680"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="691" name="__module.model.layers.2.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="682,emb.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="692" name="__module.model.layers.2.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="683,685,688,cos.13,cos.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="693" name="__module.model.layers.2.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="690,cos.17"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="694" name="__module.model.layers.2.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="692"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="695" name="Constant_96714" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="696" name="Constant_96713" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="697" name="__module.model.layers.2.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="698" name="__module.model.layers.2.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="699" name="__module.model.layers.2.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="693,698" /> + </output> + </layer> + <layer id="700" name="__module.model.layers.2.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="701" name="__module.model.layers.2.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="695,696,700,701" /> + </output> + </layer> + <layer id="702" name="__module.model.layers.2.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="703" name="__module.model.layers.2.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="704" name="Constant_96712" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="705" name="ScatterUpdate_96715" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="706" name="Constant_96718" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="707" name="Constant_96721" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="708" name="__module.model.layers.2.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="702,x2.9"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="709" name="Constant_111157_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="710" name="Constant_111157" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="711" name="__module.model.layers.2.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="703"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="712" name="Constant_96766" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="713" name="Constant_96771" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="714" name="__module.model.layers.2.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="697"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="715" name="__module.model.layers.2.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="705"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="716" name="__module.model.layers.2.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="684,686,689,sin.13,sin.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="717" name="__module.model.layers.2.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="691,sin.17"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="718" name="__module.model.layers.2.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="706"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="719" name="__module.model.layers.2.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="707"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="720" name="Constant_124047_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="721" name="Constant_124047" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="722" name="Constant_124040" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="723" name="Constant_124042" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="724" name="Constant_124044" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="725" name="Concat_124045" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="726" name="Broadcast_124048" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="727" name="ReadValue_121771" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.2.keypresent.2.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.2.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="728" name="Constant_119447" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="729" name="Gather_119448" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="730" name="self.model.layers.2.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="158045288" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="731" name="Convert_12189" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="732" name="self.model.layers.2.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="158569576" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="733" name="Convert_12192" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="734" name="self.model.layers.2.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="735" name="self.model.layers.2.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="158569832" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="736" name="self.model.layers.2.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="737" name="self.model.layers.2.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="738" name="__module.model.layers.2.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="653,key_states.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="739" name="__module.model.layers.2.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="659,662"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="740" name="__module.model.layers.2.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="660"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="741" name="__module.model.layers.2.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="742" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="661,k.5"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="743" name="__module.model.layers.2.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="708"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="744" name="Constant_96814" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="745" name="Constant_96813" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="746" name="__module.model.layers.2.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="747" name="__module.model.layers.2.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="748" name="__module.model.layers.2.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="709,714" /> + </output> + </layer> + <layer id="749" name="__module.model.layers.2.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="750" name="__module.model.layers.2.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="711,712,716,717" /> + </output> + </layer> + <layer id="751" name="__module.model.layers.2.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="752" name="__module.model.layers.2.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="753" name="Constant_96812" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="754" name="ScatterUpdate_96815" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="755" name="Constant_96818" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="756" name="Constant_96821" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="757" name="__module.model.layers.2.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="718,x2.11"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="758" name="Constant_111158_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="759" name="Constant_111158" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="760" name="__module.model.layers.2.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="719"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="761" name="Constant_96866" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="762" name="Constant_96871" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="763" name="__module.model.layers.2.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="713"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="764" name="__module.model.layers.2.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="721"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="765" name="__module.model.layers.2.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="722"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="766" name="__module.model.layers.2.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="723"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="767" name="__module.model.layers.2.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.2.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="768" name="__module.model.layers.2.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="735,736,737"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="769" name="__module.model.layers.2.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="770" name="Constant_92641" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="771" name="__module.model.layers.2.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="772" name="__module.model.layers.2.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="728"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="773" name="__module.model.layers.2.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="774" name="__module.model.layers.2.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="729" /> + </output> + </layer> + <layer id="775" name="Constant_77324" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="776" name="__module.model.layers.2.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="777" name="Constant_93254" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="778" name="Constant_93255" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="779" name="Gather_93256" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="780" name="__module.model.layers.2.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="738"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="781" name="__module.model.layers.2.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="782" name="__module.model.layers.2.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="739,hidden_states.71"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="783" name="__module.model.layers.2.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="740,741" /> + </output> + </layer> + <layer id="784" name="Constant_77327" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="785" name="__module.model.layers.2.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="786" name="__module.model.layers.2.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="742"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="787" name="__module.model.layers.2.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="743,key_states.15"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="788" name="Constant_124060_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="789" name="Constant_124060" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="790" name="Constant_124053" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="791" name="Constant_124055" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="792" name="Constant_124057" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="793" name="Concat_124058" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="794" name="Broadcast_124061" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="795" name="ReadValue_121773" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.2.valuepresent.2.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.2.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="796" name="Constant_119450" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="797" name="Gather_119451" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="798" name="self.model.layers.2.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="158570344" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="799" name="Convert_12199" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="800" name="self.model.layers.2.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="159094632" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="801" name="Convert_12202" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="802" name="self.model.layers.2.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="803" name="self.model.layers.2.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="159094888" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="804" name="self.model.layers.2.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="805" name="self.model.layers.2.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="806" name="__module.model.layers.2.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="655,value_states.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="807" name="__module.model.layers.2.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="663"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="808" name="__module.model.layers.2.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="809" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="664"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="810" name="__module.model.layers.2.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.2.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="811" name="__module.model.layers.2.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="751,752,753"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="812" name="__module.model.layers.2.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="813" name="Constant_92650" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="814" name="__module.model.layers.2.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="815" name="__module.model.layers.2.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="744"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="816" name="__module.model.layers.2.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="817" name="__module.model.layers.2.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="745" /> + </output> + </layer> + <layer id="818" name="Constant_77329" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="819" name="__module.model.layers.2.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="820" name="Constant_93264" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="821" name="Constant_93265" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="822" name="Gather_93266" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="823" name="__module.model.layers.2.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="754"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="824" name="__module.model.layers.2.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="825" name="__module.model.layers.2.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="755,hidden_states.75"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="826" name="__module.model.layers.2.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="756,757" /> + </output> + </layer> + <layer id="827" name="Constant_77332" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="828" name="__module.model.layers.2.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="829" name="__module.model.layers.2.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="758"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="830" name="__module.model.layers.2.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="759"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="831" name="Constant_96990" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="832" name="Constant_96991" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="833" name="Constant_96987" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="834" name="__module.model.layers.2.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="835" name="Constant_92659" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="836" name="__module.model.layers.2.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="837" name="__module.model.layers.2.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="760"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="838" name="Constant_96986" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="839" name="ScatterUpdate_96992" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="840" name="Constant_96995" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="841" name="__module.model.layers.2.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="764"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="842" name="__module.model.layers.2.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="765,attn_output.9"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="843" name="__module.model.layers.2.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="844" name="__module.model.layers.2.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="766"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="845" name="__module.model.layers.2.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="768"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="846" name="__module.model.layers.2.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="769"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="847" name="self.model.layers.2.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="159095400" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="848" name="Convert_12219" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="849" name="self.model.layers.2.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="163289704" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="850" name="Convert_12222" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="851" name="self.model.layers.2.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="852" name="self.model.layers.2.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="163291752" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="853" name="self.model.layers.2.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="854" name="self.model.layers.2.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="855" name="__module.model.layers.2.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="771,773,hidden_states.77"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="856" name="__module.model.layers.2/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="776,778,789,hidden_states.79,hidden_states.81"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="857" name="Constant_111162_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="163295848" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="858" name="Constant_111162" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="859" name="Constant_111161_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="860" name="Constant_111161" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="861" name="Constant_111159_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="862" name="Constant_111159" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="863" name="__module.model.layers.2.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="779"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="864" name="Constant_2477" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="780"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="865" name="__module.model.layers.2.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="781,variance.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="866" name="Constant_111160_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="867" name="Constant_111160" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="868" name="__module.model.layers.2.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="782"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="869" name="__module.model.layers.2.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="870" name="__module.model.layers.2.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="783"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="871" name="__module.model.layers.2.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="784,785,hidden_states.83,hidden_states.85"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="872" name="__module.model.layers.2.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="786,788"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="873" name="self.model.layers.2.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="163299944" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="874" name="Convert_12239" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="875" name="self.model.layers.2.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="174834280" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="876" name="Convert_12242" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="877" name="self.model.layers.2.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="878" name="self.model.layers.2.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="174839912" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="879" name="self.model.layers.2.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="880" name="self.model.layers.2.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="881" name="__module.model.layers.2.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="794,input.5"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="882" name="__module.model.layers.2.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="795"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="883" name="self.model.layers.2.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="174851176" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="884" name="Convert_12229" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="885" name="self.model.layers.2.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="186385512" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="886" name="Convert_12232" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="887" name="self.model.layers.2.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="888" name="self.model.layers.2.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="186391144" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="889" name="self.model.layers.2.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="890" name="self.model.layers.2.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="891" name="__module.model.layers.2.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="797"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="892" name="__module.model.layers.2.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="798"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="893" name="self.model.layers.2.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="186402408" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="894" name="Convert_12249" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="895" name="self.model.layers.2.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="197936744" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="896" name="Convert_12252" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="897" name="self.model.layers.2.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="898" name="self.model.layers.2.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="197938792" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="899" name="self.model.layers.2.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="900" name="self.model.layers.2.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="901" name="__module.model.layers.2.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="800,hidden_states.87"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="902" name="__module.model.layers.2/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="801,803,811,822,hidden_states.89,hidden_states.91"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="903" name="Constant_111166_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="197942888" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="904" name="Constant_111166" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="905" name="Constant_111165_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="906" name="Constant_111165" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="907" name="Constant_111163_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="908" name="Constant_111163" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="909" name="__module.model.layers.3.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="812"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="910" name="Constant_2548" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="813"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="911" name="__module.model.layers.3.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="814,variance.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="912" name="Constant_111164_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="913" name="Constant_111164" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="914" name="__module.model.layers.3.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="815"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="915" name="__module.model.layers.3.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="916" name="__module.model.layers.3.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="816"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="917" name="__module.model.layers.3.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="817,818,hidden_states.93,hidden_states.95"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="918" name="__module.model.layers.3.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="819,821,hidden_states.97"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="919" name="self.model.layers.3.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="197946984" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="920" name="Convert_12279" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="921" name="self.model.layers.3.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="202141288" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="922" name="Convert_12282" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="923" name="self.model.layers.3.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="924" name="self.model.layers.3.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="202143336" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="925" name="self.model.layers.3.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="926" name="self.model.layers.3.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="927" name="__module.model.layers.3.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="831,query_states.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="928" name="__module.model.layers.3.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="929" name="Constant_93279" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="930" name="Constant_93280" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="931" name="Gather_93281" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="932" name="__module.model.layers.3.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="836"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="933" name="__module.model.layers.3.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="837"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="934" name="__module.model.layers.3.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="935" name="__module.model.layers.3.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="838,q.7"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="936" name="__module.model.layers.3.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="849"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="937" name="__module.model.layers.3.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="938" name="__module.model.layers.3.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="852,857,inv_freq_expanded.13,inv_freq_expanded.15"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="939" name="__module.model.layers.3.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="859"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="940" name="__module.model.layers.3.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="941" name="__module.model.layers.3.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="860"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="942" name="__module.model.layers.3.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="862,emb.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="943" name="__module.model.layers.3.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="863,865,868,cos.19,cos.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="944" name="__module.model.layers.3.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="870,cos.23"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="945" name="__module.model.layers.3.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="872"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="946" name="Constant_97000" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="947" name="Constant_96999" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="948" name="__module.model.layers.3.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="949" name="__module.model.layers.3.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="950" name="__module.model.layers.3.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="873,878" /> + </output> + </layer> + <layer id="951" name="__module.model.layers.3.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="952" name="__module.model.layers.3.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="875,876,880,881" /> + </output> + </layer> + <layer id="953" name="__module.model.layers.3.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="954" name="__module.model.layers.3.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="955" name="Constant_96998" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="956" name="ScatterUpdate_97001" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="957" name="Constant_97004" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="958" name="Constant_97007" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="959" name="__module.model.layers.3.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="882,x2.13"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="960" name="Constant_111167_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="961" name="Constant_111167" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="962" name="__module.model.layers.3.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="883"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="963" name="Constant_97052" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="964" name="Constant_97057" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="965" name="__module.model.layers.3.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="877"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="966" name="__module.model.layers.3.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="885"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="967" name="__module.model.layers.3.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="864,866,869,sin.19,sin.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="968" name="__module.model.layers.3.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="871,sin.23"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="969" name="__module.model.layers.3.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="886"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="970" name="__module.model.layers.3.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="887"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="971" name="Constant_124073_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="972" name="Constant_124073" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="973" name="Constant_124066" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="974" name="Constant_124068" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="975" name="Constant_124070" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="976" name="Concat_124071" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="977" name="Broadcast_124074" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="978" name="ReadValue_121783" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.3.keypresent.3.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.3.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="979" name="Constant_119453" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="980" name="Gather_119454" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="981" name="self.model.layers.3.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="202147432" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="982" name="Convert_12259" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="983" name="self.model.layers.3.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="202671720" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="984" name="Convert_12262" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="985" name="self.model.layers.3.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="986" name="self.model.layers.3.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="202671976" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="987" name="self.model.layers.3.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="988" name="self.model.layers.3.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="989" name="__module.model.layers.3.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="833,key_states.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="990" name="__module.model.layers.3.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="839,842"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="991" name="__module.model.layers.3.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="840"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="992" name="__module.model.layers.3.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="993" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="841,k.7"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="994" name="__module.model.layers.3.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="888"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="995" name="Constant_97100" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="996" name="Constant_97099" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="997" name="__module.model.layers.3.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="998" name="__module.model.layers.3.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="999" name="__module.model.layers.3.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="889,894" /> + </output> + </layer> + <layer id="1000" name="__module.model.layers.3.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="1001" name="__module.model.layers.3.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="891,892,896,897" /> + </output> + </layer> + <layer id="1002" name="__module.model.layers.3.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1003" name="__module.model.layers.3.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1004" name="Constant_97098" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1005" name="ScatterUpdate_97101" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1006" name="Constant_97104" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1007" name="Constant_97107" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1008" name="__module.model.layers.3.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="898,x2.15"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1009" name="Constant_111168_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1010" name="Constant_111168" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1011" name="__module.model.layers.3.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="899"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1012" name="Constant_97152" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1013" name="Constant_97157" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1014" name="__module.model.layers.3.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="893"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1015" name="__module.model.layers.3.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="901"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1016" name="__module.model.layers.3.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="902"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1017" name="__module.model.layers.3.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="903"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1018" name="__module.model.layers.3.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.3.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1019" name="__module.model.layers.3.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="915,916,917"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1020" name="__module.model.layers.3.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1021" name="Constant_92668" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1022" name="__module.model.layers.3.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1023" name="__module.model.layers.3.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="908"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1024" name="__module.model.layers.3.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1025" name="__module.model.layers.3.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="909" /> + </output> + </layer> + <layer id="1026" name="Constant_77336" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1027" name="__module.model.layers.3.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1028" name="Constant_93289" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1029" name="Constant_93290" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1030" name="Gather_93291" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1031" name="__module.model.layers.3.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="918"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1032" name="__module.model.layers.3.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1033" name="__module.model.layers.3.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="919,hidden_states.101"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1034" name="__module.model.layers.3.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="920,921" /> + </output> + </layer> + <layer id="1035" name="Constant_77339" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1036" name="__module.model.layers.3.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1037" name="__module.model.layers.3.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="922"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1038" name="__module.model.layers.3.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="923,key_states.19"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1039" name="Constant_124086_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="1040" name="Constant_124086" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="1041" name="Constant_124079" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1042" name="Constant_124081" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1043" name="Constant_124083" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1044" name="Concat_124084" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1045" name="Broadcast_124087" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1046" name="ReadValue_121785" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.3.valuepresent.3.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.3.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1047" name="Constant_119456" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1048" name="Gather_119457" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1049" name="self.model.layers.3.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="202672488" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1050" name="Convert_12269" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1051" name="self.model.layers.3.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="203196776" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1052" name="Convert_12272" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1053" name="self.model.layers.3.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1054" name="self.model.layers.3.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="203197032" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1055" name="self.model.layers.3.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1056" name="self.model.layers.3.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1057" name="__module.model.layers.3.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="835,value_states.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="1058" name="__module.model.layers.3.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="843"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1059" name="__module.model.layers.3.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1060" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="844"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1061" name="__module.model.layers.3.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.3.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1062" name="__module.model.layers.3.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="931,932,933"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1063" name="__module.model.layers.3.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1064" name="Constant_92677" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1065" name="__module.model.layers.3.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1066" name="__module.model.layers.3.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="924"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1067" name="__module.model.layers.3.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1068" name="__module.model.layers.3.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="925" /> + </output> + </layer> + <layer id="1069" name="Constant_77341" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1070" name="__module.model.layers.3.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1071" name="Constant_93299" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1072" name="Constant_93300" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1073" name="Gather_93301" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1074" name="__module.model.layers.3.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="934"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1075" name="__module.model.layers.3.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1076" name="__module.model.layers.3.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="935,hidden_states.105"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1077" name="__module.model.layers.3.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="936,937" /> + </output> + </layer> + <layer id="1078" name="Constant_77344" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1079" name="__module.model.layers.3.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1080" name="__module.model.layers.3.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="938"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1081" name="__module.model.layers.3.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="939"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1082" name="Constant_97276" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1083" name="Constant_97277" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1084" name="Constant_97273" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1085" name="__module.model.layers.3.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1086" name="Constant_92686" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1087" name="__module.model.layers.3.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1088" name="__module.model.layers.3.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="940"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1089" name="Constant_97272" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1090" name="ScatterUpdate_97278" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1091" name="Constant_97281" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1092" name="__module.model.layers.3.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="944"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1093" name="__module.model.layers.3.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="945,attn_output.13"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1094" name="__module.model.layers.3.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1095" name="__module.model.layers.3.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="946"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1096" name="__module.model.layers.3.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="948"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1097" name="__module.model.layers.3.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="949"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1098" name="self.model.layers.3.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="203197544" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1099" name="Convert_12289" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1100" name="self.model.layers.3.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="207391848" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1101" name="Convert_12292" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1102" name="self.model.layers.3.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1103" name="self.model.layers.3.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="207393896" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1104" name="self.model.layers.3.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1105" name="self.model.layers.3.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1106" name="__module.model.layers.3.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="951,953,hidden_states.107"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1107" name="__module.model.layers.3/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="956,958,969,hidden_states.109,hidden_states.111"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1108" name="Constant_111172_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="207397992" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1109" name="Constant_111172" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1110" name="Constant_111171_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1111" name="Constant_111171" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1112" name="Constant_111169_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1113" name="Constant_111169" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1114" name="__module.model.layers.3.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="959"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1115" name="Constant_3228" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="960"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1116" name="__module.model.layers.3.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="961,variance.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1117" name="Constant_111170_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1118" name="Constant_111170" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1119" name="__module.model.layers.3.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="962"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1120" name="__module.model.layers.3.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1121" name="__module.model.layers.3.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="963"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1122" name="__module.model.layers.3.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="964,965,hidden_states.113,hidden_states.115"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1123" name="__module.model.layers.3.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="966,968"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1124" name="self.model.layers.3.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="207402088" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1125" name="Convert_12309" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1126" name="self.model.layers.3.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="218936424" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1127" name="Convert_12312" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1128" name="self.model.layers.3.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1129" name="self.model.layers.3.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="218942056" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1130" name="self.model.layers.3.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1131" name="self.model.layers.3.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1132" name="__module.model.layers.3.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="974,input.7"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1133" name="__module.model.layers.3.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="975"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1134" name="self.model.layers.3.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="218953320" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1135" name="Convert_12299" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1136" name="self.model.layers.3.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="230487656" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1137" name="Convert_12302" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1138" name="self.model.layers.3.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1139" name="self.model.layers.3.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="230493288" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1140" name="self.model.layers.3.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1141" name="self.model.layers.3.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1142" name="__module.model.layers.3.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="977"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1143" name="__module.model.layers.3.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="978"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1144" name="self.model.layers.3.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="230504552" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1145" name="Convert_12319" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1146" name="self.model.layers.3.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="242038888" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1147" name="Convert_12322" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1148" name="self.model.layers.3.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1149" name="self.model.layers.3.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="242040936" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1150" name="self.model.layers.3.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1151" name="self.model.layers.3.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1152" name="__module.model.layers.3.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="980,hidden_states.117"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1153" name="__module.model.layers.3/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1002,981,983,991,hidden_states.119,hidden_states.121"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1154" name="Constant_111176_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="242045032" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1155" name="Constant_111176" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1156" name="Constant_111175_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1157" name="Constant_111175" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1158" name="Constant_111173_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1159" name="Constant_111173" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1160" name="__module.model.layers.4.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="992"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1161" name="Constant_3299" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="993"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1162" name="__module.model.layers.4.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="994,variance.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1163" name="Constant_111174_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1164" name="Constant_111174" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1165" name="__module.model.layers.4.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="995"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1166" name="__module.model.layers.4.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1167" name="__module.model.layers.4.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="996"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1168" name="__module.model.layers.4.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="997,998,hidden_states.123,hidden_states.125"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1169" name="__module.model.layers.4.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1001,999,hidden_states.127"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1170" name="self.model.layers.4.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="242049128" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1171" name="Convert_12349" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1172" name="self.model.layers.4.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="246243432" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1173" name="Convert_12352" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1174" name="self.model.layers.4.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1175" name="self.model.layers.4.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="246245480" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1176" name="self.model.layers.4.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1177" name="self.model.layers.4.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1178" name="__module.model.layers.4.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1011,query_states.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1179" name="__module.model.layers.4.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1180" name="Constant_93314" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1181" name="Constant_93315" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1182" name="Gather_93316" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1183" name="__module.model.layers.4.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1016"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1184" name="__module.model.layers.4.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1017"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1185" name="__module.model.layers.4.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1186" name="__module.model.layers.4.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1018,q.9"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1187" name="__module.model.layers.4.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="1029"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1188" name="__module.model.layers.4.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1189" name="__module.model.layers.4.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1032,1037,inv_freq_expanded.17,inv_freq_expanded.19"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1190" name="__module.model.layers.4.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1039"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1191" name="__module.model.layers.4.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1192" name="__module.model.layers.4.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1040"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1193" name="__module.model.layers.4.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1042,emb.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1194" name="__module.model.layers.4.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1043,1045,1048,cos.25,cos.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1195" name="__module.model.layers.4.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1050,cos.29"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1196" name="__module.model.layers.4.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1052"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1197" name="Constant_97286" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1198" name="Constant_97285" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1199" name="__module.model.layers.4.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1200" name="__module.model.layers.4.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1201" name="__module.model.layers.4.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1053,1058" /> + </output> + </layer> + <layer id="1202" name="__module.model.layers.4.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="1203" name="__module.model.layers.4.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1055,1056,1060,1061" /> + </output> + </layer> + <layer id="1204" name="__module.model.layers.4.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1205" name="__module.model.layers.4.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1206" name="Constant_97284" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1207" name="ScatterUpdate_97287" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1208" name="Constant_97290" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1209" name="Constant_97293" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1210" name="__module.model.layers.4.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1062,x2.17"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1211" name="Constant_111177_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1212" name="Constant_111177" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1213" name="__module.model.layers.4.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1063"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1214" name="Constant_97338" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1215" name="Constant_97343" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1216" name="__module.model.layers.4.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1057"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1217" name="__module.model.layers.4.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1065"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1218" name="__module.model.layers.4.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1044,1046,1049,sin.25,sin.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1219" name="__module.model.layers.4.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1051,sin.29"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1220" name="__module.model.layers.4.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1066"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1221" name="__module.model.layers.4.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1067"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1222" name="Constant_124099_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="1223" name="Constant_124099" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="1224" name="Constant_124092" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1225" name="Constant_124094" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1226" name="Constant_124096" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1227" name="Concat_124097" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1228" name="Broadcast_124100" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1229" name="ReadValue_121787" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.4.keypresent.4.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.4.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1230" name="Constant_119459" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1231" name="Gather_119460" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1232" name="self.model.layers.4.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="246249576" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1233" name="Convert_12329" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1234" name="self.model.layers.4.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="246773864" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1235" name="Convert_12332" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1236" name="self.model.layers.4.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1237" name="self.model.layers.4.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="246774120" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1238" name="self.model.layers.4.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1239" name="self.model.layers.4.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1240" name="__module.model.layers.4.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1013,key_states.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="1241" name="__module.model.layers.4.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1019,1022"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1242" name="__module.model.layers.4.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1020"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1243" name="__module.model.layers.4.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1244" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1021,k.9"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1245" name="__module.model.layers.4.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1068"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1246" name="Constant_97386" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1247" name="Constant_97385" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1248" name="__module.model.layers.4.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1249" name="__module.model.layers.4.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1250" name="__module.model.layers.4.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1069,1074" /> + </output> + </layer> + <layer id="1251" name="__module.model.layers.4.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="1252" name="__module.model.layers.4.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1071,1072,1076,1077" /> + </output> + </layer> + <layer id="1253" name="__module.model.layers.4.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1254" name="__module.model.layers.4.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1255" name="Constant_97384" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1256" name="ScatterUpdate_97387" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1257" name="Constant_97390" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1258" name="Constant_97393" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1259" name="__module.model.layers.4.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1078,x2.19"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1260" name="Constant_111178_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1261" name="Constant_111178" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1262" name="__module.model.layers.4.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1079"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1263" name="Constant_97438" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1264" name="Constant_97443" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1265" name="__module.model.layers.4.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1073"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1266" name="__module.model.layers.4.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1081"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1267" name="__module.model.layers.4.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1082"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1268" name="__module.model.layers.4.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1083"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1269" name="__module.model.layers.4.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.4.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1270" name="__module.model.layers.4.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1095,1096,1097"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1271" name="__module.model.layers.4.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1272" name="Constant_92695" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1273" name="__module.model.layers.4.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1274" name="__module.model.layers.4.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1088"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1275" name="__module.model.layers.4.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1276" name="__module.model.layers.4.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1089" /> + </output> + </layer> + <layer id="1277" name="Constant_77348" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1278" name="__module.model.layers.4.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1279" name="Constant_93324" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1280" name="Constant_93325" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1281" name="Gather_93326" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1282" name="__module.model.layers.4.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1098"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1283" name="__module.model.layers.4.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1284" name="__module.model.layers.4.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1099,hidden_states.131"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1285" name="__module.model.layers.4.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1100,1101" /> + </output> + </layer> + <layer id="1286" name="Constant_77351" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1287" name="__module.model.layers.4.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1288" name="__module.model.layers.4.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1102"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1289" name="__module.model.layers.4.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1103,key_states.23"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1290" name="Constant_124112_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="1291" name="Constant_124112" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="1292" name="Constant_124105" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1293" name="Constant_124107" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1294" name="Constant_124109" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1295" name="Concat_124110" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1296" name="Broadcast_124113" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1297" name="ReadValue_121789" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.4.valuepresent.4.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.4.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1298" name="Constant_119462" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1299" name="Gather_119463" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1300" name="self.model.layers.4.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="246774632" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1301" name="Convert_12339" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1302" name="self.model.layers.4.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="247298920" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1303" name="Convert_12342" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1304" name="self.model.layers.4.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1305" name="self.model.layers.4.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="247299176" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1306" name="self.model.layers.4.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1307" name="self.model.layers.4.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1308" name="__module.model.layers.4.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1015,value_states.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="1309" name="__module.model.layers.4.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1023"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1310" name="__module.model.layers.4.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1311" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1024"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1312" name="__module.model.layers.4.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.4.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1313" name="__module.model.layers.4.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1111,1112,1113"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1314" name="__module.model.layers.4.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1315" name="Constant_92704" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1316" name="__module.model.layers.4.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1317" name="__module.model.layers.4.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1104"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1318" name="__module.model.layers.4.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1319" name="__module.model.layers.4.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1105" /> + </output> + </layer> + <layer id="1320" name="Constant_77353" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1321" name="__module.model.layers.4.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1322" name="Constant_93334" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1323" name="Constant_93335" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1324" name="Gather_93336" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1325" name="__module.model.layers.4.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1114"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1326" name="__module.model.layers.4.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1327" name="__module.model.layers.4.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1115,hidden_states.135"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1328" name="__module.model.layers.4.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1116,1117" /> + </output> + </layer> + <layer id="1329" name="Constant_77356" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1330" name="__module.model.layers.4.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1331" name="__module.model.layers.4.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1118"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1332" name="__module.model.layers.4.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1119"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1333" name="Constant_97562" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1334" name="Constant_97563" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1335" name="Constant_97559" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1336" name="__module.model.layers.4.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1337" name="Constant_92713" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1338" name="__module.model.layers.4.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1339" name="__module.model.layers.4.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1120"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1340" name="Constant_97558" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1341" name="ScatterUpdate_97564" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1342" name="Constant_97567" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1343" name="__module.model.layers.4.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1124"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1344" name="__module.model.layers.4.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1125,attn_output.17"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1345" name="__module.model.layers.4.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1346" name="__module.model.layers.4.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1126"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1347" name="__module.model.layers.4.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="1128"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1348" name="__module.model.layers.4.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1129"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1349" name="self.model.layers.4.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="247299688" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1350" name="Convert_12359" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1351" name="self.model.layers.4.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="251493992" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1352" name="Convert_12362" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1353" name="self.model.layers.4.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1354" name="self.model.layers.4.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="251496040" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1355" name="self.model.layers.4.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1356" name="self.model.layers.4.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1357" name="__module.model.layers.4.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1131,1133,hidden_states.137"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1358" name="__module.model.layers.4/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1136,1138,1149,hidden_states.139,hidden_states.141"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1359" name="Constant_111182_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="251500136" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1360" name="Constant_111182" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1361" name="Constant_111181_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1362" name="Constant_111181" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1363" name="Constant_111179_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1364" name="Constant_111179" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1365" name="__module.model.layers.4.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1139"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1366" name="Constant_3979" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1140"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1367" name="__module.model.layers.4.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1141,variance.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1368" name="Constant_111180_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1369" name="Constant_111180" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1370" name="__module.model.layers.4.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1142"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1371" name="__module.model.layers.4.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1372" name="__module.model.layers.4.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1143"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1373" name="__module.model.layers.4.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1144,1145,hidden_states.143,hidden_states.145"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1374" name="__module.model.layers.4.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1146,1148"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1375" name="self.model.layers.4.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="251504232" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1376" name="Convert_12379" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1377" name="self.model.layers.4.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="263038568" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1378" name="Convert_12382" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1379" name="self.model.layers.4.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1380" name="self.model.layers.4.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="263044200" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1381" name="self.model.layers.4.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1382" name="self.model.layers.4.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1383" name="__module.model.layers.4.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1154,input.9"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1384" name="__module.model.layers.4.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1155"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1385" name="self.model.layers.4.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="263055464" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1386" name="Convert_12369" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1387" name="self.model.layers.4.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="274589800" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1388" name="Convert_12372" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1389" name="self.model.layers.4.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1390" name="self.model.layers.4.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="274595432" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1391" name="self.model.layers.4.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1392" name="self.model.layers.4.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1393" name="__module.model.layers.4.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1157"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1394" name="__module.model.layers.4.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1158"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1395" name="self.model.layers.4.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="274606696" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1396" name="Convert_12389" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1397" name="self.model.layers.4.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="286141032" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1398" name="Convert_12392" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1399" name="self.model.layers.4.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1400" name="self.model.layers.4.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="286143080" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1401" name="self.model.layers.4.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1402" name="self.model.layers.4.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1403" name="__module.model.layers.4.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1160,hidden_states.147"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1404" name="__module.model.layers.4/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1161,1163,1171,1182,hidden_states.149,hidden_states.151"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1405" name="Constant_111186_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="286147176" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1406" name="Constant_111186" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1407" name="Constant_111185_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1408" name="Constant_111185" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1409" name="Constant_111183_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1410" name="Constant_111183" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1411" name="__module.model.layers.5.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1172"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1412" name="Constant_4050" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1173"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1413" name="__module.model.layers.5.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1174,variance.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1414" name="Constant_111184_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1415" name="Constant_111184" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1416" name="__module.model.layers.5.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1175"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1417" name="__module.model.layers.5.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1418" name="__module.model.layers.5.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1176"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1419" name="__module.model.layers.5.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1177,1178,hidden_states.153,hidden_states.155"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1420" name="__module.model.layers.5.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1179,1181,hidden_states.157"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1421" name="self.model.layers.5.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="286151272" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1422" name="Convert_12419" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1423" name="self.model.layers.5.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="290345576" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1424" name="Convert_12422" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1425" name="self.model.layers.5.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1426" name="self.model.layers.5.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="290347624" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1427" name="self.model.layers.5.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1428" name="self.model.layers.5.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1429" name="__module.model.layers.5.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1191,query_states.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1430" name="__module.model.layers.5.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1431" name="Constant_93349" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1432" name="Constant_93350" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1433" name="Gather_93351" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1434" name="__module.model.layers.5.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1196"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1435" name="__module.model.layers.5.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1197"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1436" name="__module.model.layers.5.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1437" name="__module.model.layers.5.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1198,q.11"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1438" name="__module.model.layers.5.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="1209"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1439" name="__module.model.layers.5.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1440" name="__module.model.layers.5.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1212,1217,inv_freq_expanded.21,inv_freq_expanded.23"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1441" name="__module.model.layers.5.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1219"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1442" name="__module.model.layers.5.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1443" name="__module.model.layers.5.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1220"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1444" name="__module.model.layers.5.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1222,emb.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1445" name="__module.model.layers.5.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1223,1225,1228,cos.31,cos.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1446" name="__module.model.layers.5.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1230,cos.35"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1447" name="__module.model.layers.5.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1232"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1448" name="Constant_97572" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1449" name="Constant_97571" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1450" name="__module.model.layers.5.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1451" name="__module.model.layers.5.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1452" name="__module.model.layers.5.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1233,1238" /> + </output> + </layer> + <layer id="1453" name="__module.model.layers.5.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="1454" name="__module.model.layers.5.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1235,1236,1240,1241" /> + </output> + </layer> + <layer id="1455" name="__module.model.layers.5.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1456" name="__module.model.layers.5.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1457" name="Constant_97570" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1458" name="ScatterUpdate_97573" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1459" name="Constant_97576" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1460" name="Constant_97579" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1461" name="__module.model.layers.5.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1242,x2.21"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1462" name="Constant_111187_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1463" name="Constant_111187" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1464" name="__module.model.layers.5.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1243"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1465" name="Constant_97624" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1466" name="Constant_97629" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1467" name="__module.model.layers.5.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1237"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1468" name="__module.model.layers.5.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1245"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1469" name="__module.model.layers.5.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1224,1226,1229,sin.31,sin.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1470" name="__module.model.layers.5.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1231,sin.35"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1471" name="__module.model.layers.5.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1246"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1472" name="__module.model.layers.5.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1247"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1473" name="Constant_124125_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="1474" name="Constant_124125" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="1475" name="Constant_124118" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1476" name="Constant_124120" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1477" name="Constant_124122" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1478" name="Concat_124123" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1479" name="Broadcast_124126" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1480" name="ReadValue_121791" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.5.keypresent.5.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.5.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1481" name="Constant_119465" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1482" name="Gather_119466" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1483" name="self.model.layers.5.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="290351720" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1484" name="Convert_12399" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1485" name="self.model.layers.5.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="290876008" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1486" name="Convert_12402" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1487" name="self.model.layers.5.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1488" name="self.model.layers.5.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="290876264" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1489" name="self.model.layers.5.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1490" name="self.model.layers.5.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1491" name="__module.model.layers.5.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1193,key_states.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="1492" name="__module.model.layers.5.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1199,1202"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1493" name="__module.model.layers.5.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1200"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1494" name="__module.model.layers.5.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1495" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1201,k.11"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1496" name="__module.model.layers.5.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1248"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1497" name="Constant_97672" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1498" name="Constant_97671" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1499" name="__module.model.layers.5.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1500" name="__module.model.layers.5.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1501" name="__module.model.layers.5.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1249,1254" /> + </output> + </layer> + <layer id="1502" name="__module.model.layers.5.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="1503" name="__module.model.layers.5.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1251,1252,1256,1257" /> + </output> + </layer> + <layer id="1504" name="__module.model.layers.5.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1505" name="__module.model.layers.5.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1506" name="Constant_97670" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1507" name="ScatterUpdate_97673" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1508" name="Constant_97676" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1509" name="Constant_97679" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1510" name="__module.model.layers.5.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1258,x2.23"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1511" name="Constant_111188_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1512" name="Constant_111188" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1513" name="__module.model.layers.5.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1259"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1514" name="Constant_97724" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1515" name="Constant_97729" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1516" name="__module.model.layers.5.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1253"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1517" name="__module.model.layers.5.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1261"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1518" name="__module.model.layers.5.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1262"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1519" name="__module.model.layers.5.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1263"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1520" name="__module.model.layers.5.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.5.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1521" name="__module.model.layers.5.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1275,1276,1277"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1522" name="__module.model.layers.5.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1523" name="Constant_92722" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1524" name="__module.model.layers.5.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1525" name="__module.model.layers.5.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1268"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1526" name="__module.model.layers.5.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1527" name="__module.model.layers.5.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1269" /> + </output> + </layer> + <layer id="1528" name="Constant_77360" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1529" name="__module.model.layers.5.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1530" name="Constant_93359" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1531" name="Constant_93360" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1532" name="Gather_93361" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1533" name="__module.model.layers.5.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1278"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1534" name="__module.model.layers.5.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1535" name="__module.model.layers.5.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1279,hidden_states.161"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1536" name="__module.model.layers.5.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1280,1281" /> + </output> + </layer> + <layer id="1537" name="Constant_77363" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1538" name="__module.model.layers.5.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1539" name="__module.model.layers.5.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1282"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1540" name="__module.model.layers.5.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1283,key_states.27"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1541" name="Constant_124138_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="1542" name="Constant_124138" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="1543" name="Constant_124131" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1544" name="Constant_124133" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1545" name="Constant_124135" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1546" name="Concat_124136" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1547" name="Broadcast_124139" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1548" name="ReadValue_121793" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.5.valuepresent.5.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.5.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1549" name="Constant_119468" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1550" name="Gather_119469" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1551" name="self.model.layers.5.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="290876776" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1552" name="Convert_12409" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1553" name="self.model.layers.5.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="291401064" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1554" name="Convert_12412" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1555" name="self.model.layers.5.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1556" name="self.model.layers.5.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="291401320" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1557" name="self.model.layers.5.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1558" name="self.model.layers.5.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1559" name="__module.model.layers.5.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1195,value_states.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="1560" name="__module.model.layers.5.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1203"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1561" name="__module.model.layers.5.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1562" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1204"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1563" name="__module.model.layers.5.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.5.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1564" name="__module.model.layers.5.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1291,1292,1293"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1565" name="__module.model.layers.5.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1566" name="Constant_92731" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1567" name="__module.model.layers.5.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1568" name="__module.model.layers.5.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1284"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1569" name="__module.model.layers.5.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1570" name="__module.model.layers.5.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1285" /> + </output> + </layer> + <layer id="1571" name="Constant_77365" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1572" name="__module.model.layers.5.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1573" name="Constant_93369" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1574" name="Constant_93370" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1575" name="Gather_93371" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1576" name="__module.model.layers.5.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1294"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1577" name="__module.model.layers.5.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1578" name="__module.model.layers.5.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1295,hidden_states.165"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1579" name="__module.model.layers.5.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1296,1297" /> + </output> + </layer> + <layer id="1580" name="Constant_77368" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1581" name="__module.model.layers.5.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1582" name="__module.model.layers.5.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1298"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1583" name="__module.model.layers.5.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1299"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1584" name="Constant_97848" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1585" name="Constant_97849" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1586" name="Constant_97845" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1587" name="__module.model.layers.5.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1588" name="Constant_92740" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1589" name="__module.model.layers.5.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1590" name="__module.model.layers.5.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1300"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1591" name="Constant_97844" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1592" name="ScatterUpdate_97850" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1593" name="Constant_97853" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1594" name="__module.model.layers.5.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1304"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1595" name="__module.model.layers.5.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1305,attn_output.21"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1596" name="__module.model.layers.5.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1597" name="__module.model.layers.5.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1306"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1598" name="__module.model.layers.5.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="1308"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1599" name="__module.model.layers.5.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1309"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1600" name="self.model.layers.5.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="291401832" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1601" name="Convert_12429" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1602" name="self.model.layers.5.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="295596136" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1603" name="Convert_12432" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1604" name="self.model.layers.5.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1605" name="self.model.layers.5.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="295598184" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1606" name="self.model.layers.5.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1607" name="self.model.layers.5.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1608" name="__module.model.layers.5.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1311,1313,hidden_states.167"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1609" name="__module.model.layers.5/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1316,1318,1329,hidden_states.169,hidden_states.171"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1610" name="Constant_111192_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="295602280" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1611" name="Constant_111192" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1612" name="Constant_111191_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1613" name="Constant_111191" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1614" name="Constant_111189_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1615" name="Constant_111189" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1616" name="__module.model.layers.5.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1319"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1617" name="Constant_4730" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1320"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1618" name="__module.model.layers.5.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1321,variance.23"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1619" name="Constant_111190_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1620" name="Constant_111190" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1621" name="__module.model.layers.5.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1322"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1622" name="__module.model.layers.5.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1623" name="__module.model.layers.5.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1323"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1624" name="__module.model.layers.5.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1324,1325,hidden_states.173,hidden_states.175"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1625" name="__module.model.layers.5.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1326,1328"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1626" name="self.model.layers.5.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="295606376" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1627" name="Convert_12449" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1628" name="self.model.layers.5.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="307140712" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1629" name="Convert_12452" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1630" name="self.model.layers.5.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1631" name="self.model.layers.5.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="307146344" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1632" name="self.model.layers.5.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1633" name="self.model.layers.5.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1634" name="__module.model.layers.5.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1334,input.11"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1635" name="__module.model.layers.5.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1335"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1636" name="self.model.layers.5.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="307157608" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1637" name="Convert_12439" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1638" name="self.model.layers.5.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="318691944" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1639" name="Convert_12442" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1640" name="self.model.layers.5.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1641" name="self.model.layers.5.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="318697576" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1642" name="self.model.layers.5.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1643" name="self.model.layers.5.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1644" name="__module.model.layers.5.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1337"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1645" name="__module.model.layers.5.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1338"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1646" name="self.model.layers.5.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="318708840" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1647" name="Convert_12459" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1648" name="self.model.layers.5.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="330243176" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1649" name="Convert_12462" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1650" name="self.model.layers.5.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1651" name="self.model.layers.5.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="330245224" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1652" name="self.model.layers.5.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1653" name="self.model.layers.5.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1654" name="__module.model.layers.5.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1340,hidden_states.177"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1655" name="__module.model.layers.5/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1341,1343,1351,1362,hidden_states.179,hidden_states.181"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1656" name="Constant_111196_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="330249320" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1657" name="Constant_111196" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1658" name="Constant_111195_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1659" name="Constant_111195" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1660" name="Constant_111193_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1661" name="Constant_111193" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1662" name="__module.model.layers.6.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1352"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1663" name="Constant_4801" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1353"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1664" name="__module.model.layers.6.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1354,variance.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1665" name="Constant_111194_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1666" name="Constant_111194" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1667" name="__module.model.layers.6.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1355"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1668" name="__module.model.layers.6.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1669" name="__module.model.layers.6.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1356"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1670" name="__module.model.layers.6.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1357,1358,hidden_states.183,hidden_states.185"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1671" name="__module.model.layers.6.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1359,1361,hidden_states.187"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1672" name="self.model.layers.6.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="330253416" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1673" name="Convert_12489" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1674" name="self.model.layers.6.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="334447720" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1675" name="Convert_12492" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1676" name="self.model.layers.6.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1677" name="self.model.layers.6.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="334449768" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1678" name="self.model.layers.6.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1679" name="self.model.layers.6.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1680" name="__module.model.layers.6.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1371,query_states.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1681" name="__module.model.layers.6.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1682" name="Constant_93384" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1683" name="Constant_93385" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1684" name="Gather_93386" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1685" name="__module.model.layers.6.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1376"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1686" name="__module.model.layers.6.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1377"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1687" name="__module.model.layers.6.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1688" name="__module.model.layers.6.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1378,q.13"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1689" name="__module.model.layers.6.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="1389"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1690" name="__module.model.layers.6.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1691" name="__module.model.layers.6.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1392,1397,inv_freq_expanded.25,inv_freq_expanded.27"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1692" name="__module.model.layers.6.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1399"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1693" name="__module.model.layers.6.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1694" name="__module.model.layers.6.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1400"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1695" name="__module.model.layers.6.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1402,emb.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1696" name="__module.model.layers.6.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1403,1405,1408,cos.37,cos.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1697" name="__module.model.layers.6.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1410,cos.41"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1698" name="__module.model.layers.6.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1412"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1699" name="Constant_97858" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1700" name="Constant_97857" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1701" name="__module.model.layers.6.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1702" name="__module.model.layers.6.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1703" name="__module.model.layers.6.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1413,1418" /> + </output> + </layer> + <layer id="1704" name="__module.model.layers.6.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="1705" name="__module.model.layers.6.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1415,1416,1420,1421" /> + </output> + </layer> + <layer id="1706" name="__module.model.layers.6.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1707" name="__module.model.layers.6.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1708" name="Constant_97856" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1709" name="ScatterUpdate_97859" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1710" name="Constant_97862" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1711" name="Constant_97865" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1712" name="__module.model.layers.6.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1422,x2.25"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1713" name="Constant_111197_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1714" name="Constant_111197" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1715" name="__module.model.layers.6.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1423"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1716" name="Constant_97910" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1717" name="Constant_97915" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1718" name="__module.model.layers.6.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1417"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1719" name="__module.model.layers.6.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1425"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1720" name="__module.model.layers.6.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1404,1406,1409,sin.37,sin.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1721" name="__module.model.layers.6.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1411,sin.41"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1722" name="__module.model.layers.6.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1426"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1723" name="__module.model.layers.6.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1427"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1724" name="Constant_124151_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="1725" name="Constant_124151" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="1726" name="Constant_124144" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1727" name="Constant_124146" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1728" name="Constant_124148" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1729" name="Concat_124149" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1730" name="Broadcast_124152" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1731" name="ReadValue_121795" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.6.keypresent.6.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.6.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1732" name="Constant_119471" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1733" name="Gather_119472" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1734" name="self.model.layers.6.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="334453864" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1735" name="Convert_12469" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1736" name="self.model.layers.6.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="334978152" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1737" name="Convert_12472" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1738" name="self.model.layers.6.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1739" name="self.model.layers.6.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="334978408" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1740" name="self.model.layers.6.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1741" name="self.model.layers.6.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1742" name="__module.model.layers.6.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1373,key_states.29"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="1743" name="__module.model.layers.6.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1379,1382"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1744" name="__module.model.layers.6.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1380"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1745" name="__module.model.layers.6.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1746" name="__module.model.layers.6.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1381,k.13"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1747" name="__module.model.layers.6.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1428"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1748" name="Constant_97958" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1749" name="Constant_97957" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1750" name="__module.model.layers.6.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1751" name="__module.model.layers.6.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1752" name="__module.model.layers.6.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1429,1434" /> + </output> + </layer> + <layer id="1753" name="__module.model.layers.6.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="1754" name="__module.model.layers.6.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1431,1432,1436,1437" /> + </output> + </layer> + <layer id="1755" name="__module.model.layers.6.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1756" name="__module.model.layers.6.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1757" name="Constant_97956" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1758" name="ScatterUpdate_97959" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1759" name="Constant_97962" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1760" name="Constant_97965" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1761" name="__module.model.layers.6.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1438,x2.27"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1762" name="Constant_111198_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1763" name="Constant_111198" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1764" name="__module.model.layers.6.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1439"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1765" name="Constant_98010" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1766" name="Constant_98015" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1767" name="__module.model.layers.6.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1433"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1768" name="__module.model.layers.6.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1441"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1769" name="__module.model.layers.6.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1442"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1770" name="__module.model.layers.6.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1443"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1771" name="__module.model.layers.6.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.6.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1772" name="__module.model.layers.6.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1455,1456,1457"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1773" name="__module.model.layers.6.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1774" name="Constant_92749" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1775" name="__module.model.layers.6.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1776" name="__module.model.layers.6.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1448"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1777" name="__module.model.layers.6.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1778" name="__module.model.layers.6.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1449" /> + </output> + </layer> + <layer id="1779" name="Constant_77372" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1780" name="__module.model.layers.6.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1781" name="Constant_93394" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1782" name="Constant_93395" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1783" name="Gather_93396" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1784" name="__module.model.layers.6.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1458"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1785" name="__module.model.layers.6.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1786" name="__module.model.layers.6.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1459,hidden_states.191"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1787" name="__module.model.layers.6.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1460,1461" /> + </output> + </layer> + <layer id="1788" name="Constant_77375" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1789" name="__module.model.layers.6.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1790" name="__module.model.layers.6.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1462"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1791" name="__module.model.layers.6.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1463,key_states.31"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1792" name="Constant_124164_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="1793" name="Constant_124164" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="1794" name="Constant_124157" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1795" name="Constant_124159" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1796" name="Constant_124161" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1797" name="Concat_124162" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1798" name="Broadcast_124165" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1799" name="ReadValue_121797" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.6.valuepresent.6.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.6.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1800" name="Constant_119474" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1801" name="Gather_119475" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1802" name="self.model.layers.6.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="334978920" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1803" name="Convert_12479" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1804" name="self.model.layers.6.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="335503208" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1805" name="Convert_12482" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1806" name="self.model.layers.6.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1807" name="self.model.layers.6.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="335503464" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1808" name="self.model.layers.6.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1809" name="self.model.layers.6.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1810" name="__module.model.layers.6.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1375,value_states.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="1811" name="__module.model.layers.6.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1383"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1812" name="__module.model.layers.6.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1813" name="__module.model.layers.6.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1384"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1814" name="__module.model.layers.6.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.6.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1815" name="__module.model.layers.6.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1471,1472,1473"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1816" name="__module.model.layers.6.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1817" name="Constant_92758" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1818" name="__module.model.layers.6.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1819" name="__module.model.layers.6.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1464"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1820" name="__module.model.layers.6.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1821" name="__module.model.layers.6.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1465" /> + </output> + </layer> + <layer id="1822" name="Constant_77377" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1823" name="__module.model.layers.6.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1824" name="Constant_93404" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1825" name="Constant_93405" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1826" name="Gather_93406" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1827" name="__module.model.layers.6.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1474"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1828" name="__module.model.layers.6.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="1829" name="__module.model.layers.6.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1475,hidden_states.195"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1830" name="__module.model.layers.6.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1476,1477" /> + </output> + </layer> + <layer id="1831" name="Constant_77380" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1832" name="__module.model.layers.6.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1833" name="__module.model.layers.6.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1478"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1834" name="__module.model.layers.6.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1479"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1835" name="Constant_98134" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1836" name="Constant_98135" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1837" name="Constant_98131" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1838" name="__module.model.layers.6.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1839" name="Constant_92767" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1840" name="__module.model.layers.6.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1841" name="__module.model.layers.6.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1480"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1842" name="Constant_98130" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1843" name="ScatterUpdate_98136" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1844" name="Constant_98139" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1845" name="__module.model.layers.6.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1484"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1846" name="__module.model.layers.6.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1485,attn_output.25"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1847" name="__module.model.layers.6.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1848" name="__module.model.layers.6.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1486"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1849" name="__module.model.layers.6.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="1488"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1850" name="__module.model.layers.6.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1489"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1851" name="self.model.layers.6.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="335503976" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1852" name="Convert_12499" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1853" name="self.model.layers.6.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="339698280" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1854" name="Convert_12502" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1855" name="self.model.layers.6.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1856" name="self.model.layers.6.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="339700328" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1857" name="self.model.layers.6.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1858" name="self.model.layers.6.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1859" name="__module.model.layers.6.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1491,1493,hidden_states.197"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1860" name="__module.model.layers.6/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1496,1498,1509,hidden_states.199,hidden_states.201"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1861" name="Constant_111202_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="339704424" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1862" name="Constant_111202" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1863" name="Constant_111201_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1864" name="Constant_111201" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1865" name="Constant_111199_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1866" name="Constant_111199" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1867" name="__module.model.layers.6.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1499"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1868" name="Constant_5481" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1500"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1869" name="__module.model.layers.6.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1501,variance.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1870" name="Constant_111200_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1871" name="Constant_111200" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1872" name="__module.model.layers.6.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1502"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1873" name="__module.model.layers.6.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1874" name="__module.model.layers.6.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1503"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1875" name="__module.model.layers.6.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1504,1505,hidden_states.203,hidden_states.205"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1876" name="__module.model.layers.6.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1506,1508"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1877" name="self.model.layers.6.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="339708520" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1878" name="Convert_12519" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1879" name="self.model.layers.6.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="351242856" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1880" name="Convert_12522" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1881" name="self.model.layers.6.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1882" name="self.model.layers.6.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="351248488" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1883" name="self.model.layers.6.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1884" name="self.model.layers.6.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1885" name="__module.model.layers.6.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1514,input.13"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1886" name="__module.model.layers.6.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1515"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1887" name="self.model.layers.6.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="351259752" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1888" name="Convert_12509" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1889" name="self.model.layers.6.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="362794088" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1890" name="Convert_12512" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1891" name="self.model.layers.6.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1892" name="self.model.layers.6.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="362799720" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1893" name="self.model.layers.6.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1894" name="self.model.layers.6.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1895" name="__module.model.layers.6.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1517"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1896" name="__module.model.layers.6.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1518"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1897" name="self.model.layers.6.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="362810984" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1898" name="Convert_12529" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1899" name="self.model.layers.6.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="374345320" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1900" name="Convert_12532" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1901" name="self.model.layers.6.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1902" name="self.model.layers.6.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="374347368" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1903" name="self.model.layers.6.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1904" name="self.model.layers.6.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="1905" name="__module.model.layers.6.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1520,hidden_states.207"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1906" name="__module.model.layers.6/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1521,1523,1531,1542,hidden_states.209,hidden_states.211"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1907" name="Constant_111206_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="374351464" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1908" name="Constant_111206" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1909" name="Constant_111205_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1910" name="Constant_111205" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1911" name="Constant_111203_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1912" name="Constant_111203" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1913" name="__module.model.layers.7.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1532"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1914" name="Constant_5552" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1533"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1915" name="__module.model.layers.7.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1534,variance.29"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1916" name="Constant_111204_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1917" name="Constant_111204" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1918" name="__module.model.layers.7.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1535"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1919" name="__module.model.layers.7.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1920" name="__module.model.layers.7.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1536"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1921" name="__module.model.layers.7.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1537,1538,hidden_states.213,hidden_states.215"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1922" name="__module.model.layers.7.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1539,1541,hidden_states.217"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1923" name="self.model.layers.7.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="374355560" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1924" name="Convert_12559" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1925" name="self.model.layers.7.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="378549864" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1926" name="Convert_12562" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1927" name="self.model.layers.7.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1928" name="self.model.layers.7.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="378551912" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1929" name="self.model.layers.7.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1930" name="self.model.layers.7.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1931" name="__module.model.layers.7.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1551,query_states.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1932" name="__module.model.layers.7.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1933" name="Constant_93419" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1934" name="Constant_93420" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1935" name="Gather_93421" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="1936" name="__module.model.layers.7.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1556"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1937" name="__module.model.layers.7.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1557"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1938" name="__module.model.layers.7.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1939" name="__module.model.layers.7.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1558,q.15"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1940" name="__module.model.layers.7.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="1569"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1941" name="__module.model.layers.7.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1942" name="__module.model.layers.7.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1572,1577,inv_freq_expanded.29,inv_freq_expanded.31"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1943" name="__module.model.layers.7.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1579"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="1944" name="__module.model.layers.7.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="1945" name="__module.model.layers.7.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1580"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1946" name="__module.model.layers.7.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1582,emb.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1947" name="__module.model.layers.7.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1583,1585,1588,cos.43,cos.45"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1948" name="__module.model.layers.7.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1590,cos.47"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1949" name="__module.model.layers.7.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1592"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1950" name="Constant_98144" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1951" name="Constant_98143" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1952" name="__module.model.layers.7.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1953" name="__module.model.layers.7.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1954" name="__module.model.layers.7.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1593,1598" /> + </output> + </layer> + <layer id="1955" name="__module.model.layers.7.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="1956" name="__module.model.layers.7.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1595,1596,1600,1601" /> + </output> + </layer> + <layer id="1957" name="__module.model.layers.7.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="1958" name="__module.model.layers.7.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1959" name="Constant_98142" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1960" name="ScatterUpdate_98145" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1961" name="Constant_98148" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1962" name="Constant_98151" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1963" name="__module.model.layers.7.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1602,x2.29"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1964" name="Constant_111207_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1965" name="Constant_111207" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1966" name="__module.model.layers.7.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1603"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1967" name="Constant_98196" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1968" name="Constant_98201" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1969" name="__module.model.layers.7.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1597"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="1970" name="__module.model.layers.7.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1605"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1971" name="__module.model.layers.7.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1584,1586,1589,sin.43,sin.45"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1972" name="__module.model.layers.7.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1591,sin.47"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1973" name="__module.model.layers.7.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1606"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1974" name="__module.model.layers.7.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1607"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1975" name="Constant_124177_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="1976" name="Constant_124177" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="1977" name="Constant_124170" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1978" name="Constant_124172" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1979" name="Constant_124174" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1980" name="Concat_124175" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1981" name="Broadcast_124178" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1982" name="ReadValue_121799" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.7.keypresent.7.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.7.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1983" name="Constant_119477" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="1984" name="Gather_119478" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1985" name="self.model.layers.7.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="378556008" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1986" name="Convert_12539" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1987" name="self.model.layers.7.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="379080296" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1988" name="Convert_12542" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1989" name="self.model.layers.7.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1990" name="self.model.layers.7.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="379080552" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="1991" name="self.model.layers.7.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1992" name="self.model.layers.7.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="1993" name="__module.model.layers.7.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1553,key_states.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="1994" name="__module.model.layers.7.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1559,1562"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1995" name="__module.model.layers.7.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1560"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1996" name="__module.model.layers.7.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="1997" name="__module.model.layers.7.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1561,k.15"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1998" name="__module.model.layers.7.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1608"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="1999" name="Constant_98244" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2000" name="Constant_98243" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2001" name="__module.model.layers.7.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2002" name="__module.model.layers.7.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2003" name="__module.model.layers.7.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1609,1614" /> + </output> + </layer> + <layer id="2004" name="__module.model.layers.7.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="2005" name="__module.model.layers.7.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1611,1612,1616,1617" /> + </output> + </layer> + <layer id="2006" name="__module.model.layers.7.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2007" name="__module.model.layers.7.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2008" name="Constant_98242" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2009" name="ScatterUpdate_98245" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2010" name="Constant_98248" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2011" name="Constant_98251" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2012" name="__module.model.layers.7.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1618,x2.31"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2013" name="Constant_111208_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2014" name="Constant_111208" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2015" name="__module.model.layers.7.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1619"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2016" name="Constant_98296" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2017" name="Constant_98301" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2018" name="__module.model.layers.7.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1613"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2019" name="__module.model.layers.7.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1621"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2020" name="__module.model.layers.7.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1622"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2021" name="__module.model.layers.7.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1623"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2022" name="__module.model.layers.7.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.7.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2023" name="__module.model.layers.7.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1635,1636,1637"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2024" name="__module.model.layers.7.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2025" name="Constant_92776" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2026" name="__module.model.layers.7.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2027" name="__module.model.layers.7.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1628"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2028" name="__module.model.layers.7.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2029" name="__module.model.layers.7.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1629" /> + </output> + </layer> + <layer id="2030" name="Constant_77384" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2031" name="__module.model.layers.7.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2032" name="Constant_93429" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2033" name="Constant_93430" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2034" name="Gather_93431" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2035" name="__module.model.layers.7.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1638"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2036" name="__module.model.layers.7.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2037" name="__module.model.layers.7.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1639,hidden_states.221"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2038" name="__module.model.layers.7.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1640,1641" /> + </output> + </layer> + <layer id="2039" name="Constant_77387" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2040" name="__module.model.layers.7.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2041" name="__module.model.layers.7.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1642"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2042" name="__module.model.layers.7.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1643,key_states.35"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2043" name="Constant_124190_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="2044" name="Constant_124190" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="2045" name="Constant_124183" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2046" name="Constant_124185" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2047" name="Constant_124187" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2048" name="Concat_124188" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2049" name="Broadcast_124191" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2050" name="ReadValue_121801" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.7.valuepresent.7.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.7.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2051" name="Constant_119480" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2052" name="Gather_119481" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2053" name="self.model.layers.7.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="379081064" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2054" name="Convert_12549" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2055" name="self.model.layers.7.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="379605352" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2056" name="Convert_12552" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2057" name="self.model.layers.7.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2058" name="self.model.layers.7.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="379605608" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2059" name="self.model.layers.7.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2060" name="self.model.layers.7.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2061" name="__module.model.layers.7.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1555,value_states.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="2062" name="__module.model.layers.7.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1563"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2063" name="__module.model.layers.7.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2064" name="__module.model.layers.7.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1564"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2065" name="__module.model.layers.7.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.7.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2066" name="__module.model.layers.7.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1651,1652,1653"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2067" name="__module.model.layers.7.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2068" name="Constant_92785" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2069" name="__module.model.layers.7.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2070" name="__module.model.layers.7.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1644"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2071" name="__module.model.layers.7.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2072" name="__module.model.layers.7.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1645" /> + </output> + </layer> + <layer id="2073" name="Constant_77389" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2074" name="__module.model.layers.7.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2075" name="Constant_93439" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2076" name="Constant_93440" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2077" name="Gather_93441" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2078" name="__module.model.layers.7.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1654"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2079" name="__module.model.layers.7.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2080" name="__module.model.layers.7.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1655,hidden_states.225"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2081" name="__module.model.layers.7.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1656,1657" /> + </output> + </layer> + <layer id="2082" name="Constant_77392" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2083" name="__module.model.layers.7.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2084" name="__module.model.layers.7.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1658"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2085" name="__module.model.layers.7.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1659"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2086" name="Constant_98420" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2087" name="Constant_98421" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2088" name="Constant_98417" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2089" name="__module.model.layers.7.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2090" name="Constant_92794" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2091" name="__module.model.layers.7.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2092" name="__module.model.layers.7.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1660"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2093" name="Constant_98416" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2094" name="ScatterUpdate_98422" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2095" name="Constant_98425" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2096" name="__module.model.layers.7.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1664"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2097" name="__module.model.layers.7.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1665,attn_output.29"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2098" name="__module.model.layers.7.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2099" name="__module.model.layers.7.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1666"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2100" name="__module.model.layers.7.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="1668"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2101" name="__module.model.layers.7.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1669"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2102" name="self.model.layers.7.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="379606120" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2103" name="Convert_12569" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2104" name="self.model.layers.7.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="383800424" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2105" name="Convert_12572" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2106" name="self.model.layers.7.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2107" name="self.model.layers.7.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="383802472" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2108" name="self.model.layers.7.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2109" name="self.model.layers.7.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2110" name="__module.model.layers.7.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1671,1673,hidden_states.227"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2111" name="__module.model.layers.7/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1676,1678,1689,hidden_states.229,hidden_states.231"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2112" name="Constant_111212_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="383806568" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2113" name="Constant_111212" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2114" name="Constant_111211_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2115" name="Constant_111211" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2116" name="Constant_111209_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2117" name="Constant_111209" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2118" name="__module.model.layers.7.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1679"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2119" name="Constant_6232" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1680"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2120" name="__module.model.layers.7.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1681,variance.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2121" name="Constant_111210_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2122" name="Constant_111210" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2123" name="__module.model.layers.7.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1682"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2124" name="__module.model.layers.7.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2125" name="__module.model.layers.7.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1683"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2126" name="__module.model.layers.7.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1684,1685,hidden_states.233,hidden_states.235"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2127" name="__module.model.layers.7.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1686,1688"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2128" name="self.model.layers.7.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="383810664" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2129" name="Convert_12589" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2130" name="self.model.layers.7.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="395345000" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2131" name="Convert_12592" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2132" name="self.model.layers.7.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2133" name="self.model.layers.7.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="395350632" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2134" name="self.model.layers.7.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2135" name="self.model.layers.7.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2136" name="__module.model.layers.7.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1694,input.15"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2137" name="__module.model.layers.7.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1695"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2138" name="self.model.layers.7.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="395361896" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2139" name="Convert_12579" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2140" name="self.model.layers.7.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="406896232" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2141" name="Convert_12582" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2142" name="self.model.layers.7.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2143" name="self.model.layers.7.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="406901864" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2144" name="self.model.layers.7.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2145" name="self.model.layers.7.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2146" name="__module.model.layers.7.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1697"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2147" name="__module.model.layers.7.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1698"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2148" name="self.model.layers.7.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="406913128" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2149" name="Convert_12599" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2150" name="self.model.layers.7.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="418447464" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2151" name="Convert_12602" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2152" name="self.model.layers.7.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2153" name="self.model.layers.7.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="418449512" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2154" name="self.model.layers.7.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2155" name="self.model.layers.7.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2156" name="__module.model.layers.7.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1700,hidden_states.237"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2157" name="__module.model.layers.7/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1701,1703,1711,1722,hidden_states.239,hidden_states.241"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2158" name="Constant_111216_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="418453608" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2159" name="Constant_111216" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2160" name="Constant_111215_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2161" name="Constant_111215" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2162" name="Constant_111213_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2163" name="Constant_111213" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2164" name="__module.model.layers.8.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1712"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2165" name="Constant_6303" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1713"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2166" name="__module.model.layers.8.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1714,variance.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2167" name="Constant_111214_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2168" name="Constant_111214" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2169" name="__module.model.layers.8.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1715"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2170" name="__module.model.layers.8.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2171" name="__module.model.layers.8.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1716"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2172" name="__module.model.layers.8.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1717,1718,hidden_states.243,hidden_states.245"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2173" name="__module.model.layers.8.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1719,1721,hidden_states.247"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2174" name="self.model.layers.8.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="418457704" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2175" name="Convert_12629" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2176" name="self.model.layers.8.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="422652008" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2177" name="Convert_12632" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2178" name="self.model.layers.8.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2179" name="self.model.layers.8.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="422654056" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2180" name="self.model.layers.8.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2181" name="self.model.layers.8.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2182" name="__module.model.layers.8.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1731,query_states.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2183" name="__module.model.layers.8.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2184" name="Constant_93454" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2185" name="Constant_93455" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2186" name="Gather_93456" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2187" name="__module.model.layers.8.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1736"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2188" name="__module.model.layers.8.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1737"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2189" name="__module.model.layers.8.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2190" name="__module.model.layers.8.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1738,q.17"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2191" name="__module.model.layers.8.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="1749"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2192" name="__module.model.layers.8.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2193" name="__module.model.layers.8.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1752,1757,inv_freq_expanded.33,inv_freq_expanded.35"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2194" name="__module.model.layers.8.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1759"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2195" name="__module.model.layers.8.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2196" name="__module.model.layers.8.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1760"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2197" name="__module.model.layers.8.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1762,emb.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2198" name="__module.model.layers.8.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1763,1765,1768,cos.49,cos.51"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2199" name="__module.model.layers.8.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1770,cos.53"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2200" name="__module.model.layers.8.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1772"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2201" name="Constant_98430" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2202" name="Constant_98429" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2203" name="__module.model.layers.8.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2204" name="__module.model.layers.8.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2205" name="__module.model.layers.8.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1773,1778" /> + </output> + </layer> + <layer id="2206" name="__module.model.layers.8.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="2207" name="__module.model.layers.8.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1775,1776,1780,1781" /> + </output> + </layer> + <layer id="2208" name="__module.model.layers.8.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2209" name="__module.model.layers.8.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2210" name="Constant_98428" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2211" name="ScatterUpdate_98431" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2212" name="Constant_98434" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2213" name="Constant_98437" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2214" name="__module.model.layers.8.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1782,x2.33"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2215" name="Constant_111217_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2216" name="Constant_111217" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2217" name="__module.model.layers.8.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1783"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2218" name="Constant_98482" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2219" name="Constant_98487" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2220" name="__module.model.layers.8.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1777"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2221" name="__module.model.layers.8.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1785"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2222" name="__module.model.layers.8.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1764,1766,1769,sin.49,sin.51"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2223" name="__module.model.layers.8.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1771,sin.53"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2224" name="__module.model.layers.8.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1786"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2225" name="__module.model.layers.8.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1787"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2226" name="Constant_124203_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="2227" name="Constant_124203" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="2228" name="Constant_124196" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2229" name="Constant_124198" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2230" name="Constant_124200" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2231" name="Concat_124201" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2232" name="Broadcast_124204" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2233" name="ReadValue_121803" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.8.keypresent.8.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.8.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2234" name="Constant_119483" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2235" name="Gather_119484" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2236" name="self.model.layers.8.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="422658152" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2237" name="Convert_12609" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2238" name="self.model.layers.8.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="423182440" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2239" name="Convert_12612" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2240" name="self.model.layers.8.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2241" name="self.model.layers.8.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="423182696" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2242" name="self.model.layers.8.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2243" name="self.model.layers.8.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2244" name="__module.model.layers.8.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1733,key_states.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="2245" name="__module.model.layers.8.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1739,1742"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2246" name="__module.model.layers.8.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1740"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2247" name="__module.model.layers.8.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2248" name="__module.model.layers.8.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1741,k.17"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2249" name="__module.model.layers.8.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1788"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2250" name="Constant_98530" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2251" name="Constant_98529" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2252" name="__module.model.layers.8.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2253" name="__module.model.layers.8.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2254" name="__module.model.layers.8.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1789,1794" /> + </output> + </layer> + <layer id="2255" name="__module.model.layers.8.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="2256" name="__module.model.layers.8.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1791,1792,1796,1797" /> + </output> + </layer> + <layer id="2257" name="__module.model.layers.8.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2258" name="__module.model.layers.8.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2259" name="Constant_98528" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2260" name="ScatterUpdate_98531" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2261" name="Constant_98534" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2262" name="Constant_98537" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2263" name="__module.model.layers.8.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1798,x2.35"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2264" name="Constant_111218_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2265" name="Constant_111218" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2266" name="__module.model.layers.8.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1799"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2267" name="Constant_98582" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2268" name="Constant_98587" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2269" name="__module.model.layers.8.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1793"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2270" name="__module.model.layers.8.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1801"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2271" name="__module.model.layers.8.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1802"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2272" name="__module.model.layers.8.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1803"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2273" name="__module.model.layers.8.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.8.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2274" name="__module.model.layers.8.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1815,1816,1817"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2275" name="__module.model.layers.8.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2276" name="Constant_92803" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2277" name="__module.model.layers.8.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2278" name="__module.model.layers.8.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1808"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2279" name="__module.model.layers.8.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2280" name="__module.model.layers.8.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1809" /> + </output> + </layer> + <layer id="2281" name="Constant_77396" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2282" name="__module.model.layers.8.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2283" name="Constant_93464" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2284" name="Constant_93465" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2285" name="Gather_93466" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2286" name="__module.model.layers.8.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1818"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2287" name="__module.model.layers.8.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2288" name="__module.model.layers.8.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1819,hidden_states.251"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2289" name="__module.model.layers.8.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1820,1821" /> + </output> + </layer> + <layer id="2290" name="Constant_77399" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2291" name="__module.model.layers.8.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2292" name="__module.model.layers.8.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1822"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2293" name="__module.model.layers.8.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1823,key_states.39"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2294" name="Constant_124216_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="2295" name="Constant_124216" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="2296" name="Constant_124209" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2297" name="Constant_124211" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2298" name="Constant_124213" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2299" name="Concat_124214" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2300" name="Broadcast_124217" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2301" name="ReadValue_121805" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.8.valuepresent.8.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.8.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2302" name="Constant_119486" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2303" name="Gather_119487" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2304" name="self.model.layers.8.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="423183208" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2305" name="Convert_12619" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2306" name="self.model.layers.8.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="423707496" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2307" name="Convert_12622" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2308" name="self.model.layers.8.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2309" name="self.model.layers.8.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="423707752" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2310" name="self.model.layers.8.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2311" name="self.model.layers.8.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2312" name="__module.model.layers.8.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1735,value_states.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="2313" name="__module.model.layers.8.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1743"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2314" name="__module.model.layers.8.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2315" name="__module.model.layers.8.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1744"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2316" name="__module.model.layers.8.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.8.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2317" name="__module.model.layers.8.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1831,1832,1833"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2318" name="__module.model.layers.8.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2319" name="Constant_92812" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2320" name="__module.model.layers.8.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2321" name="__module.model.layers.8.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1824"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2322" name="__module.model.layers.8.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2323" name="__module.model.layers.8.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1825" /> + </output> + </layer> + <layer id="2324" name="Constant_77401" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2325" name="__module.model.layers.8.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2326" name="Constant_93474" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2327" name="Constant_93475" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2328" name="Gather_93476" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2329" name="__module.model.layers.8.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1834"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2330" name="__module.model.layers.8.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2331" name="__module.model.layers.8.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1835,hidden_states.255"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2332" name="__module.model.layers.8.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="1836,1837" /> + </output> + </layer> + <layer id="2333" name="Constant_77404" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2334" name="__module.model.layers.8.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2335" name="__module.model.layers.8.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1838"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2336" name="__module.model.layers.8.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1839"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2337" name="Constant_98706" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2338" name="Constant_98707" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2339" name="Constant_98703" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2340" name="__module.model.layers.8.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2341" name="Constant_92821" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2342" name="__module.model.layers.8.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2343" name="__module.model.layers.8.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1840"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2344" name="Constant_98702" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2345" name="ScatterUpdate_98708" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2346" name="Constant_98711" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2347" name="__module.model.layers.8.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1844"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2348" name="__module.model.layers.8.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1845,attn_output.33"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2349" name="__module.model.layers.8.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2350" name="__module.model.layers.8.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1846"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2351" name="__module.model.layers.8.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="1848"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2352" name="__module.model.layers.8.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1849"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2353" name="self.model.layers.8.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="423708264" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2354" name="Convert_12639" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2355" name="self.model.layers.8.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="427902568" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2356" name="Convert_12642" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2357" name="self.model.layers.8.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2358" name="self.model.layers.8.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="427904616" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2359" name="self.model.layers.8.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2360" name="self.model.layers.8.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2361" name="__module.model.layers.8.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1851,1853,hidden_states.257"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2362" name="__module.model.layers.8/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1856,1858,1869,hidden_states.259,hidden_states.261"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2363" name="Constant_111222_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="427908712" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2364" name="Constant_111222" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2365" name="Constant_111221_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2366" name="Constant_111221" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2367" name="Constant_111219_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2368" name="Constant_111219" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2369" name="__module.model.layers.8.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1859"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2370" name="Constant_6983" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1860"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2371" name="__module.model.layers.8.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1861,variance.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2372" name="Constant_111220_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2373" name="Constant_111220" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2374" name="__module.model.layers.8.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1862"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2375" name="__module.model.layers.8.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2376" name="__module.model.layers.8.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1863"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2377" name="__module.model.layers.8.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1864,1865,hidden_states.263,hidden_states.265"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2378" name="__module.model.layers.8.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1866,1868"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2379" name="self.model.layers.8.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="427912808" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2380" name="Convert_12659" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2381" name="self.model.layers.8.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="439447144" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2382" name="Convert_12662" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2383" name="self.model.layers.8.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2384" name="self.model.layers.8.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="439452776" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2385" name="self.model.layers.8.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2386" name="self.model.layers.8.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2387" name="__module.model.layers.8.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1874,input.17"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2388" name="__module.model.layers.8.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1875"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2389" name="self.model.layers.8.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="439464040" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2390" name="Convert_12649" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2391" name="self.model.layers.8.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="450998376" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2392" name="Convert_12652" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2393" name="self.model.layers.8.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2394" name="self.model.layers.8.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="451004008" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2395" name="self.model.layers.8.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2396" name="self.model.layers.8.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2397" name="__module.model.layers.8.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1877"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2398" name="__module.model.layers.8.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1878"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2399" name="self.model.layers.8.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="451015272" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2400" name="Convert_12669" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2401" name="self.model.layers.8.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="462549608" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2402" name="Convert_12672" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2403" name="self.model.layers.8.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2404" name="self.model.layers.8.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="462551656" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2405" name="self.model.layers.8.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2406" name="self.model.layers.8.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2407" name="__module.model.layers.8.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1880,hidden_states.267"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2408" name="__module.model.layers.8/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1881,1883,1891,1902,hidden_states.269,hidden_states.271"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2409" name="Constant_111225_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2410" name="Constant_111225" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2411" name="Constant_111223_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2412" name="Constant_111223" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2413" name="__module.model.layers.9.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1892"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2414" name="Constant_7054" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="1893"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2415" name="__module.model.layers.9.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1894,variance.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2416" name="Constant_111224_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2417" name="Constant_111224" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2418" name="__module.model.layers.9.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1895"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2419" name="__module.model.layers.9.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2420" name="__module.model.layers.9.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1896"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2421" name="__module.model.layers.9.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1897,1898,hidden_states.273,hidden_states.275"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2422" name="__module.model.layers.9.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1899,1901,hidden_states.277"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2423" name="self.model.layers.9.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="462555752" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2424" name="Convert_12679" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2425" name="self.model.layers.9.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="463080040" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2426" name="Convert_12682" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2427" name="self.model.layers.9.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2428" name="self.model.layers.9.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="463080296" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2429" name="self.model.layers.9.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2430" name="self.model.layers.9.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2431" name="__module.model.layers.9.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1915,value_states.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="2432" name="__module.model.layers.9.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2433" name="Constant_93489" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2434" name="Constant_93490" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2435" name="Gather_93491" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2436" name="__module.model.layers.9.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1919,1922"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2437" name="__module.model.layers.9.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1923"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2438" name="__module.model.layers.9.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2439" name="__module.model.layers.9.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1924"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2440" name="__module.model.layers.9.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.9.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2442" name="Constant_124229_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="2443" name="Constant_124229" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="2444" name="Constant_124222" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2445" name="Constant_124224" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2446" name="Constant_124226" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2447" name="Concat_124227" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2448" name="Broadcast_124230" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2449" name="ReadValue_121807" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.9.keypresent.9.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.9.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2450" name="Constant_119489" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2451" name="Gather_119490" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2452" name="self.model.layers.9.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="463080808" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2453" name="Convert_12689" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2454" name="self.model.layers.9.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="463605096" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2455" name="Convert_12692" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2456" name="self.model.layers.9.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2457" name="self.model.layers.9.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="463605352" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2458" name="self.model.layers.9.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2459" name="self.model.layers.9.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2460" name="__module.model.layers.9.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1913,key_states.41"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="2461" name="__module.model.layers.9.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1920"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2462" name="__module.model.layers.9.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2463" name="__module.model.layers.9.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1921,k.19"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2464" name="__module.model.layers.9.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="1929"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2465" name="__module.model.layers.9.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2466" name="__module.model.layers.9.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1932,1937,inv_freq_expanded.37,inv_freq_expanded.39"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2467" name="__module.model.layers.9.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1939"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2468" name="__module.model.layers.9.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2469" name="__module.model.layers.9.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1940"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2470" name="__module.model.layers.9.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1942,emb.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2471" name="__module.model.layers.9.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1943,1945,1948,cos.55,cos.57"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2472" name="__module.model.layers.9.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1950,cos.59"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2473" name="__module.model.layers.9.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1968"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2474" name="Constant_98816" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2475" name="Constant_98815" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2476" name="__module.model.layers.9.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2477" name="__module.model.layers.9.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2478" name="__module.model.layers.9.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1969,1974" /> + </output> + </layer> + <layer id="2479" name="__module.model.layers.9.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="2480" name="__module.model.layers.9.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1971,1972,1976,1977" /> + </output> + </layer> + <layer id="2481" name="__module.model.layers.9.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2482" name="__module.model.layers.9.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2483" name="Constant_98814" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2484" name="ScatterUpdate_98817" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2485" name="Constant_98820" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2486" name="Constant_98823" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2487" name="__module.model.layers.9.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1978,x2.39"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2488" name="Constant_111228_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2489" name="Constant_111228" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2490" name="__module.model.layers.9.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1979"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2491" name="Constant_98868" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2492" name="Constant_98873" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2493" name="__module.model.layers.9.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1973"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2494" name="__module.model.layers.9.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1981"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2495" name="__module.model.layers.9.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="1944,1946,1949,sin.55,sin.57"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2496" name="__module.model.layers.9.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1951,sin.59"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2497" name="__module.model.layers.9.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1982"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2498" name="__module.model.layers.9.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1983"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2499" name="__module.model.layers.9.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.9.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2513" name="Constant_124242_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="2514" name="Constant_124242" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="2515" name="Constant_124235" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2516" name="Constant_124237" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2517" name="Constant_124239" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2518" name="Concat_124240" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2519" name="Broadcast_124243" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2520" name="ReadValue_121781" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.21.valuepresent.21.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.21.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2521" name="Constant_119564" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2522" name="Gather_119565" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2523" name="Constant_111346_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="463605864" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2524" name="Constant_111346" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2525" name="self.model.layers.9.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="463609960" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2526" name="Convert_12699" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2527" name="self.model.layers.9.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="467804264" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2528" name="Convert_12702" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2529" name="self.model.layers.9.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2530" name="self.model.layers.9.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="467806312" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2531" name="self.model.layers.9.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2532" name="self.model.layers.9.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2533" name="__module.model.layers.9.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1911,query_states.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2534" name="__module.model.layers.9.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="1916"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2535" name="__module.model.layers.9.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1917"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2536" name="__module.model.layers.9.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2537" name="__module.model.layers.9.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1918,q.19"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2538" name="__module.model.layers.9.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1952"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2539" name="Constant_98716" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2540" name="Constant_98715" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2541" name="__module.model.layers.9.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2542" name="__module.model.layers.9.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2543" name="__module.model.layers.9.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1953,1958" /> + </output> + </layer> + <layer id="2544" name="__module.model.layers.9.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="2545" name="__module.model.layers.9.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="1955,1956,1960,1961" /> + </output> + </layer> + <layer id="2546" name="__module.model.layers.9.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2547" name="__module.model.layers.9.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2548" name="Constant_98714" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2549" name="ScatterUpdate_98717" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2550" name="Constant_98720" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2551" name="Constant_98723" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2552" name="__module.model.layers.9.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1962,x2.37"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2553" name="Constant_111227_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2554" name="Constant_111227" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2555" name="__module.model.layers.9.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1963"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2556" name="Constant_98768" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2557" name="Constant_98773" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2558" name="__module.model.layers.9.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="1957"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2559" name="__module.model.layers.9.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1965"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2560" name="__module.model.layers.9.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1966"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2561" name="__module.model.layers.9.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1967"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2562" name="__module.model.layers.9.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="1995,1996,1997"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2563" name="__module.model.layers.9.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2564" name="Constant_92830" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2565" name="__module.model.layers.9.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2566" name="__module.model.layers.9.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1988"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2567" name="__module.model.layers.9.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2568" name="__module.model.layers.9.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="1989" /> + </output> + </layer> + <layer id="2569" name="Constant_77408" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2570" name="__module.model.layers.9.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2571" name="Constant_93499" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2572" name="Constant_93500" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2573" name="Gather_93501" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2574" name="__module.model.layers.9.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="1998"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2575" name="__module.model.layers.9.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2576" name="__module.model.layers.9.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="1999,hidden_states.281"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2577" name="__module.model.layers.9.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2000,2001" /> + </output> + </layer> + <layer id="2578" name="Constant_77411" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2579" name="__module.model.layers.9.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2580" name="__module.model.layers.9.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2002"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2581" name="__module.model.layers.9.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2003,key_states.43"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2582" name="__module.model.layers.9.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2011,2012,2013"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2583" name="__module.model.layers.9.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2584" name="Constant_92839" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2585" name="__module.model.layers.9.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2586" name="__module.model.layers.9.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2004"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2587" name="__module.model.layers.9.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2588" name="__module.model.layers.9.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2005" /> + </output> + </layer> + <layer id="2589" name="Constant_77413" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2590" name="__module.model.layers.9.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2591" name="Constant_93509" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2592" name="Constant_93510" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2593" name="Gather_93511" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2594" name="__module.model.layers.9.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2014"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2595" name="__module.model.layers.9.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2596" name="__module.model.layers.9.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2015,hidden_states.285"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2597" name="__module.model.layers.9.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2016,2017" /> + </output> + </layer> + <layer id="2598" name="Constant_77416" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2599" name="__module.model.layers.9.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2600" name="__module.model.layers.9.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2018"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2601" name="__module.model.layers.9.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2019"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2602" name="Constant_98992" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2603" name="Constant_98993" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2604" name="Constant_98989" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2605" name="__module.model.layers.9.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2606" name="Constant_92848" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2607" name="__module.model.layers.9.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2608" name="__module.model.layers.9.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2020"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2609" name="Constant_98988" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2610" name="ScatterUpdate_98994" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2611" name="Constant_98997" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2612" name="__module.model.layers.9.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2024"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2613" name="__module.model.layers.9.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2025,attn_output.37"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2614" name="__module.model.layers.9.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2615" name="__module.model.layers.9.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2026"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2616" name="__module.model.layers.9.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="2028"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2617" name="__module.model.layers.9.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2029"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2618" name="self.model.layers.9.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="467810408" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2619" name="Convert_12709" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2620" name="self.model.layers.9.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="472004712" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2621" name="Convert_12712" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2622" name="self.model.layers.9.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2623" name="self.model.layers.9.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="472006760" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2624" name="self.model.layers.9.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2625" name="self.model.layers.9.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2626" name="__module.model.layers.9.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2031,2033,hidden_states.287"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2627" name="__module.model.layers.9/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2036,2038,2049,hidden_states.289,hidden_states.291"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2628" name="Constant_111232_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="472010856" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2629" name="Constant_111232" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2630" name="Constant_111231_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2631" name="Constant_111231" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2632" name="Constant_111229_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2633" name="Constant_111229" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2634" name="__module.model.layers.9.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2039"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2635" name="Constant_7734" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2040"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2636" name="__module.model.layers.9.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2041,variance.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2637" name="Constant_111230_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2638" name="Constant_111230" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2639" name="__module.model.layers.9.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2042"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2640" name="__module.model.layers.9.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2641" name="__module.model.layers.9.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2043"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2642" name="__module.model.layers.9.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2044,2045,hidden_states.293,hidden_states.295"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2643" name="__module.model.layers.9.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2046,2048"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2644" name="self.model.layers.9.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="472014952" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2645" name="Convert_12729" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2646" name="self.model.layers.9.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="483549288" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2647" name="Convert_12732" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2648" name="self.model.layers.9.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2649" name="self.model.layers.9.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="483554920" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2650" name="self.model.layers.9.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2651" name="self.model.layers.9.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2652" name="__module.model.layers.9.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2054,input.19"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2653" name="__module.model.layers.9.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2055"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2654" name="self.model.layers.9.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="483566184" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2655" name="Convert_12719" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2656" name="self.model.layers.9.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="495100520" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2657" name="Convert_12722" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2658" name="self.model.layers.9.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2659" name="self.model.layers.9.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="495106152" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2660" name="self.model.layers.9.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2661" name="self.model.layers.9.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2662" name="__module.model.layers.9.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2057"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2663" name="__module.model.layers.9.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2058"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2664" name="self.model.layers.9.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="495117416" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2665" name="Convert_12739" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2666" name="self.model.layers.9.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="506651752" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2667" name="Convert_12742" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2668" name="self.model.layers.9.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2669" name="self.model.layers.9.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="506653800" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2670" name="self.model.layers.9.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2671" name="self.model.layers.9.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2672" name="__module.model.layers.9.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2060,hidden_states.297"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2673" name="__module.model.layers.9/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2061,2063,2071,2082,hidden_states.299,hidden_states.301"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2674" name="Constant_111236_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="506657896" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2675" name="Constant_111236" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2676" name="Constant_111235_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2677" name="Constant_111235" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2678" name="Constant_111233_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2679" name="Constant_111233" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2680" name="__module.model.layers.10.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2072"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2681" name="Constant_7805" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2073"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2682" name="__module.model.layers.10.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2074,variance.41"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2683" name="Constant_111234_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2684" name="Constant_111234" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2685" name="__module.model.layers.10.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2075"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2686" name="__module.model.layers.10.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2687" name="__module.model.layers.10.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2076"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2688" name="__module.model.layers.10.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2077,2078,hidden_states.303,hidden_states.305"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2689" name="__module.model.layers.10.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2079,2081,hidden_states.307"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2690" name="self.model.layers.10.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="506661992" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2691" name="Convert_12769" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2692" name="self.model.layers.10.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="510856296" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2693" name="Convert_12772" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2694" name="self.model.layers.10.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2695" name="self.model.layers.10.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="510858344" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2696" name="self.model.layers.10.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2697" name="self.model.layers.10.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2698" name="__module.model.layers.10.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2091,query_states.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2699" name="__module.model.layers.10.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2700" name="Constant_93524" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2701" name="Constant_93525" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2702" name="Gather_93526" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2703" name="__module.model.layers.10.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2096"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2704" name="__module.model.layers.10.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2097"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2705" name="__module.model.layers.10.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2706" name="__module.model.layers.10.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2098,q.21"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2707" name="__module.model.layers.10.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="2109"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2708" name="__module.model.layers.10.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2709" name="__module.model.layers.10.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2112,2117,inv_freq_expanded.41,inv_freq_expanded.43"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2710" name="__module.model.layers.10.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2119"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2711" name="__module.model.layers.10.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2712" name="__module.model.layers.10.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2120"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2713" name="__module.model.layers.10.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2122,emb.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2714" name="__module.model.layers.10.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2123,2125,2128,cos.61,cos.63"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2715" name="__module.model.layers.10.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2130,cos.65"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2716" name="__module.model.layers.10.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2132"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2717" name="Constant_99002" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2718" name="Constant_99001" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2719" name="__module.model.layers.10.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2720" name="__module.model.layers.10.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2721" name="__module.model.layers.10.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2133,2138" /> + </output> + </layer> + <layer id="2722" name="__module.model.layers.10.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="2723" name="__module.model.layers.10.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2135,2136,2140,2141" /> + </output> + </layer> + <layer id="2724" name="__module.model.layers.10.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2725" name="__module.model.layers.10.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2726" name="Constant_99000" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2727" name="ScatterUpdate_99003" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2728" name="Constant_99006" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2729" name="Constant_99009" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2730" name="__module.model.layers.10.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2142,x2.41"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2731" name="Constant_111237_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2732" name="Constant_111237" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2733" name="__module.model.layers.10.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2143"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2734" name="Constant_99054" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2735" name="Constant_99059" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2736" name="__module.model.layers.10.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2137"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2737" name="__module.model.layers.10.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2145"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2738" name="__module.model.layers.10.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2124,2126,2129,sin.61,sin.63"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2739" name="__module.model.layers.10.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2131,sin.65"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2740" name="__module.model.layers.10.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2146"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2741" name="__module.model.layers.10.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2147"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2742" name="Constant_124255_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="2743" name="Constant_124255" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="2744" name="Constant_124248" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2745" name="Constant_124250" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2746" name="Constant_124252" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2747" name="Concat_124253" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2748" name="Broadcast_124256" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2749" name="ReadValue_121731" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.10.keypresent.10.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.10.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2750" name="Constant_119495" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2751" name="Gather_119496" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2752" name="self.model.layers.10.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="510862440" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2753" name="Convert_12749" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2754" name="self.model.layers.10.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="511386728" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2755" name="Convert_12752" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2756" name="self.model.layers.10.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2757" name="self.model.layers.10.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="511386984" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2758" name="self.model.layers.10.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2759" name="self.model.layers.10.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2760" name="__module.model.layers.10.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2093,key_states.45"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="2761" name="__module.model.layers.10.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2099,2102"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2762" name="__module.model.layers.10.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2100"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2763" name="__module.model.layers.10.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2764" name="__module.model.layers.10.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2101,k.21"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2765" name="__module.model.layers.10.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2148"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2766" name="Constant_99102" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2767" name="Constant_99101" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2768" name="__module.model.layers.10.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2769" name="__module.model.layers.10.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2770" name="__module.model.layers.10.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2149,2154" /> + </output> + </layer> + <layer id="2771" name="__module.model.layers.10.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="2772" name="__module.model.layers.10.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2151,2152,2156,2157" /> + </output> + </layer> + <layer id="2773" name="__module.model.layers.10.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2774" name="__module.model.layers.10.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2775" name="Constant_99100" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2776" name="ScatterUpdate_99103" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2777" name="Constant_99106" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2778" name="Constant_99109" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2779" name="__module.model.layers.10.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2158,x2.43"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2780" name="Constant_111238_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2781" name="Constant_111238" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2782" name="__module.model.layers.10.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2159"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2783" name="Constant_99154" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2784" name="Constant_99159" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2785" name="__module.model.layers.10.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2153"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2786" name="__module.model.layers.10.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2161"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2787" name="__module.model.layers.10.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2162"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2788" name="__module.model.layers.10.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2163"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2789" name="__module.model.layers.10.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.10.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2790" name="__module.model.layers.10.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2175,2176,2177"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2791" name="__module.model.layers.10.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2792" name="Constant_92857" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2793" name="__module.model.layers.10.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2794" name="__module.model.layers.10.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2168"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2795" name="__module.model.layers.10.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2796" name="__module.model.layers.10.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2169" /> + </output> + </layer> + <layer id="2797" name="Constant_77420" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2798" name="__module.model.layers.10.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2799" name="Constant_93534" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2800" name="Constant_93535" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2801" name="Gather_93536" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2802" name="__module.model.layers.10.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2178"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2803" name="__module.model.layers.10.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2804" name="__module.model.layers.10.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2179,hidden_states.311"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2805" name="__module.model.layers.10.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2180,2181" /> + </output> + </layer> + <layer id="2806" name="Constant_77423" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2807" name="__module.model.layers.10.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2808" name="__module.model.layers.10.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2182"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2809" name="__module.model.layers.10.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2183,key_states.47"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2810" name="Constant_124268_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="2811" name="Constant_124268" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="2812" name="Constant_124261" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2813" name="Constant_124263" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2814" name="Constant_124265" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2815" name="Concat_124266" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2816" name="Broadcast_124269" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2817" name="ReadValue_121733" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.10.valuepresent.10.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.10.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2818" name="Constant_119498" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2819" name="Gather_119499" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2820" name="self.model.layers.10.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="511387496" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2821" name="Convert_12759" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2822" name="self.model.layers.10.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="511911784" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2823" name="Convert_12762" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2824" name="self.model.layers.10.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2825" name="self.model.layers.10.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="511912040" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2826" name="self.model.layers.10.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2827" name="self.model.layers.10.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2828" name="__module.model.layers.10.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2095,value_states.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="2829" name="__module.model.layers.10.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2103"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2830" name="__module.model.layers.10.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2831" name="__module.model.layers.10.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2104"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2832" name="__module.model.layers.10.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.10.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2833" name="__module.model.layers.10.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2191,2192,2193"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2834" name="__module.model.layers.10.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2835" name="Constant_92866" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2836" name="__module.model.layers.10.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2837" name="__module.model.layers.10.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2184"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2838" name="__module.model.layers.10.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2839" name="__module.model.layers.10.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2185" /> + </output> + </layer> + <layer id="2840" name="Constant_77425" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2841" name="__module.model.layers.10.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2842" name="Constant_93544" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2843" name="Constant_93545" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2844" name="Gather_93546" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2845" name="__module.model.layers.10.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2194"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2846" name="__module.model.layers.10.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="2847" name="__module.model.layers.10.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2195,hidden_states.315"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2848" name="__module.model.layers.10.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2196,2197" /> + </output> + </layer> + <layer id="2849" name="Constant_77428" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2850" name="__module.model.layers.10.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2851" name="__module.model.layers.10.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2198"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2852" name="__module.model.layers.10.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2199"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2853" name="Constant_99278" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2854" name="Constant_99279" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2855" name="Constant_99275" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2856" name="__module.model.layers.10.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2857" name="Constant_92875" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2858" name="__module.model.layers.10.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2859" name="__module.model.layers.10.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2200"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2860" name="Constant_99274" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2861" name="ScatterUpdate_99280" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2862" name="Constant_99283" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2863" name="__module.model.layers.10.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2204"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2864" name="__module.model.layers.10.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2205,attn_output.41"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2865" name="__module.model.layers.10.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2866" name="__module.model.layers.10.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2206"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2867" name="__module.model.layers.10.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="2208"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2868" name="__module.model.layers.10.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2209"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2869" name="self.model.layers.10.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="511912552" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2870" name="Convert_12779" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2871" name="self.model.layers.10.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="516106856" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2872" name="Convert_12782" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2873" name="self.model.layers.10.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2874" name="self.model.layers.10.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="516108904" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2875" name="self.model.layers.10.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2876" name="self.model.layers.10.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2877" name="__module.model.layers.10.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2211,2213,hidden_states.317"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2878" name="__module.model.layers.10/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2216,2218,2229,hidden_states.319,hidden_states.321"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2879" name="Constant_111242_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="516113000" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2880" name="Constant_111242" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2881" name="Constant_111241_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2882" name="Constant_111241" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2883" name="Constant_111239_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2884" name="Constant_111239" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2885" name="__module.model.layers.10.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2219"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2886" name="Constant_8485" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2220"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2887" name="__module.model.layers.10.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2221,variance.43"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2888" name="Constant_111240_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2889" name="Constant_111240" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2890" name="__module.model.layers.10.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2222"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2891" name="__module.model.layers.10.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2892" name="__module.model.layers.10.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2223"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2893" name="__module.model.layers.10.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2224,2225,hidden_states.323,hidden_states.325"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2894" name="__module.model.layers.10.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2226,2228"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2895" name="self.model.layers.10.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="516117096" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2896" name="Convert_12799" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2897" name="self.model.layers.10.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="527651432" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2898" name="Convert_12802" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2899" name="self.model.layers.10.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2900" name="self.model.layers.10.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="527657064" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2901" name="self.model.layers.10.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2902" name="self.model.layers.10.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2903" name="__module.model.layers.10.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2234,input.21"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2904" name="__module.model.layers.10.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2235"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2905" name="self.model.layers.10.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="527668328" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2906" name="Convert_12789" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2907" name="self.model.layers.10.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="539202664" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2908" name="Convert_12792" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2909" name="self.model.layers.10.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2910" name="self.model.layers.10.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="539208296" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2911" name="self.model.layers.10.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2912" name="self.model.layers.10.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2913" name="__module.model.layers.10.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2237"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2914" name="__module.model.layers.10.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2238"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2915" name="self.model.layers.10.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="539219560" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2916" name="Convert_12809" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2917" name="self.model.layers.10.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="550753896" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2918" name="Convert_12812" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2919" name="self.model.layers.10.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2920" name="self.model.layers.10.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="550755944" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2921" name="self.model.layers.10.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2922" name="self.model.layers.10.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="2923" name="__module.model.layers.10.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2240,hidden_states.327"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2924" name="__module.model.layers.10/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2241,2243,2251,2262,hidden_states.329,hidden_states.331"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2925" name="Constant_111246_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="550760040" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2926" name="Constant_111246" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2927" name="Constant_111245_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2928" name="Constant_111245" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2929" name="Constant_111243_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2930" name="Constant_111243" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2931" name="__module.model.layers.11.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2252"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2932" name="Constant_8556" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2253"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2933" name="__module.model.layers.11.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2254,variance.45"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2934" name="Constant_111244_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2935" name="Constant_111244" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2936" name="__module.model.layers.11.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2255"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2937" name="__module.model.layers.11.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2938" name="__module.model.layers.11.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2256"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2939" name="__module.model.layers.11.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2257,2258,hidden_states.333,hidden_states.335"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2940" name="__module.model.layers.11.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2259,2261,hidden_states.337"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2941" name="self.model.layers.11.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="550764136" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2942" name="Convert_12839" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2943" name="self.model.layers.11.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="554958440" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2944" name="Convert_12842" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2945" name="self.model.layers.11.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2946" name="self.model.layers.11.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="554960488" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2947" name="self.model.layers.11.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2948" name="self.model.layers.11.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2949" name="__module.model.layers.11.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2271,query_states.23"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="2950" name="__module.model.layers.11.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2951" name="Constant_93559" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2952" name="Constant_93560" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="2953" name="Gather_93561" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="2954" name="__module.model.layers.11.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2276"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2955" name="__module.model.layers.11.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2277"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2956" name="__module.model.layers.11.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2957" name="__module.model.layers.11.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2278,q.23"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2958" name="__module.model.layers.11.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="2289"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2959" name="__module.model.layers.11.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2960" name="__module.model.layers.11.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2292,2297,inv_freq_expanded.45,inv_freq_expanded.47"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2961" name="__module.model.layers.11.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2299"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="2962" name="__module.model.layers.11.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="2963" name="__module.model.layers.11.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2300"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2964" name="__module.model.layers.11.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2302,emb.23"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2965" name="__module.model.layers.11.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2303,2305,2308,cos.67,cos.69"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2966" name="__module.model.layers.11.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2310,cos.71"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2967" name="__module.model.layers.11.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2312"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2968" name="Constant_99288" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2969" name="Constant_99287" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2970" name="__module.model.layers.11.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2971" name="__module.model.layers.11.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2972" name="__module.model.layers.11.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2313,2318" /> + </output> + </layer> + <layer id="2973" name="__module.model.layers.11.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="2974" name="__module.model.layers.11.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2315,2316,2320,2321" /> + </output> + </layer> + <layer id="2975" name="__module.model.layers.11.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="2976" name="__module.model.layers.11.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2977" name="Constant_99286" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2978" name="ScatterUpdate_99289" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2979" name="Constant_99292" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2980" name="Constant_99295" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2981" name="__module.model.layers.11.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2322,x2.45"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2982" name="Constant_111247_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2983" name="Constant_111247" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2984" name="__module.model.layers.11.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2323"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2985" name="Constant_99340" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2986" name="Constant_99345" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2987" name="__module.model.layers.11.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2317"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="2988" name="__module.model.layers.11.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2325"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2989" name="__module.model.layers.11.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2304,2306,2309,sin.67,sin.69"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2990" name="__module.model.layers.11.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2311,sin.71"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2991" name="__module.model.layers.11.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2326"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2992" name="__module.model.layers.11.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2327"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2993" name="Constant_124281_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="2994" name="Constant_124281" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="2995" name="Constant_124274" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2996" name="Constant_124276" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2997" name="Constant_124278" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="2998" name="Concat_124279" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="2999" name="Broadcast_124282" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3000" name="ReadValue_121735" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.11.keypresent.11.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.11.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3001" name="Constant_119501" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3002" name="Gather_119502" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3003" name="self.model.layers.11.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="554964584" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3004" name="Convert_12819" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3005" name="self.model.layers.11.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="555488872" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3006" name="Convert_12822" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3007" name="self.model.layers.11.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3008" name="self.model.layers.11.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="555489128" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3009" name="self.model.layers.11.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3010" name="self.model.layers.11.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3011" name="__module.model.layers.11.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2273,key_states.49"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="3012" name="__module.model.layers.11.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2279,2282"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3013" name="__module.model.layers.11.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2280"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3014" name="__module.model.layers.11.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3015" name="__module.model.layers.11.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2281,k.23"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3016" name="__module.model.layers.11.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2328"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3017" name="Constant_99388" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3018" name="Constant_99387" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3019" name="__module.model.layers.11.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3020" name="__module.model.layers.11.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3021" name="__module.model.layers.11.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2329,2334" /> + </output> + </layer> + <layer id="3022" name="__module.model.layers.11.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="3023" name="__module.model.layers.11.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2331,2332,2336,2337" /> + </output> + </layer> + <layer id="3024" name="__module.model.layers.11.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3025" name="__module.model.layers.11.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3026" name="Constant_99386" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3027" name="ScatterUpdate_99389" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3028" name="Constant_99392" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3029" name="Constant_99395" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3030" name="__module.model.layers.11.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2338,x2.47"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3031" name="Constant_111248_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3032" name="Constant_111248" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3033" name="__module.model.layers.11.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2339"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3034" name="Constant_99440" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3035" name="Constant_99445" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3036" name="__module.model.layers.11.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2333"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3037" name="__module.model.layers.11.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2341"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3038" name="__module.model.layers.11.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2342"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3039" name="__module.model.layers.11.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2343"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3040" name="__module.model.layers.11.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.11.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3041" name="__module.model.layers.11.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2355,2356,2357"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3042" name="__module.model.layers.11.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3043" name="Constant_92884" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3044" name="__module.model.layers.11.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3045" name="__module.model.layers.11.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2348"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3046" name="__module.model.layers.11.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3047" name="__module.model.layers.11.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2349" /> + </output> + </layer> + <layer id="3048" name="Constant_77432" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3049" name="__module.model.layers.11.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3050" name="Constant_93569" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3051" name="Constant_93570" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3052" name="Gather_93571" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3053" name="__module.model.layers.11.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2358"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3054" name="__module.model.layers.11.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3055" name="__module.model.layers.11.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2359,hidden_states.341"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3056" name="__module.model.layers.11.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2360,2361" /> + </output> + </layer> + <layer id="3057" name="Constant_77435" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3058" name="__module.model.layers.11.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3059" name="__module.model.layers.11.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2362"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3060" name="__module.model.layers.11.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2363,key_states.51"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3061" name="Constant_124294_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="3062" name="Constant_124294" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="3063" name="Constant_124287" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3064" name="Constant_124289" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3065" name="Constant_124291" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3066" name="Concat_124292" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3067" name="Broadcast_124295" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3068" name="ReadValue_121737" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.11.valuepresent.11.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.11.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3069" name="Constant_119504" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3070" name="Gather_119505" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3071" name="self.model.layers.11.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="555489640" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3072" name="Convert_12829" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3073" name="self.model.layers.11.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="556013928" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3074" name="Convert_12832" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3075" name="self.model.layers.11.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3076" name="self.model.layers.11.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="556014184" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3077" name="self.model.layers.11.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3078" name="self.model.layers.11.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3079" name="__module.model.layers.11.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2275,value_states.23"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="3080" name="__module.model.layers.11.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2283"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3081" name="__module.model.layers.11.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3082" name="__module.model.layers.11.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2284"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3083" name="__module.model.layers.11.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.11.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3084" name="__module.model.layers.11.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2371,2372,2373"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3085" name="__module.model.layers.11.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3086" name="Constant_92893" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3087" name="__module.model.layers.11.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3088" name="__module.model.layers.11.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2364"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3089" name="__module.model.layers.11.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3090" name="__module.model.layers.11.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2365" /> + </output> + </layer> + <layer id="3091" name="Constant_77437" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3092" name="__module.model.layers.11.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3093" name="Constant_93579" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3094" name="Constant_93580" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3095" name="Gather_93581" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3096" name="__module.model.layers.11.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2374"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3097" name="__module.model.layers.11.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3098" name="__module.model.layers.11.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2375,hidden_states.345"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3099" name="__module.model.layers.11.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2376,2377" /> + </output> + </layer> + <layer id="3100" name="Constant_77440" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3101" name="__module.model.layers.11.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3102" name="__module.model.layers.11.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2378"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3103" name="__module.model.layers.11.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2379"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3104" name="Constant_99564" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3105" name="Constant_99565" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3106" name="Constant_99561" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3107" name="__module.model.layers.11.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3108" name="Constant_92902" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3109" name="__module.model.layers.11.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3110" name="__module.model.layers.11.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2380"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3111" name="Constant_99560" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3112" name="ScatterUpdate_99566" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3113" name="Constant_99569" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3114" name="__module.model.layers.11.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2384"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3115" name="__module.model.layers.11.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2385,attn_output.45"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3116" name="__module.model.layers.11.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3117" name="__module.model.layers.11.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2386"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3118" name="__module.model.layers.11.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="2388"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3119" name="__module.model.layers.11.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2389"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3120" name="self.model.layers.11.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="556014696" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3121" name="Convert_12849" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3122" name="self.model.layers.11.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="560209000" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3123" name="Convert_12852" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3124" name="self.model.layers.11.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3125" name="self.model.layers.11.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="560211048" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3126" name="self.model.layers.11.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3127" name="self.model.layers.11.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3128" name="__module.model.layers.11.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2391,2393,hidden_states.347"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3129" name="__module.model.layers.11/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2396,2398,2409,hidden_states.349,hidden_states.351"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3130" name="Constant_111252_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="560215144" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3131" name="Constant_111252" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3132" name="Constant_111251_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3133" name="Constant_111251" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3134" name="Constant_111249_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3135" name="Constant_111249" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3136" name="__module.model.layers.11.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2399"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3137" name="Constant_9236" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2400"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3138" name="__module.model.layers.11.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2401,variance.47"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3139" name="Constant_111250_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3140" name="Constant_111250" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3141" name="__module.model.layers.11.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2402"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3142" name="__module.model.layers.11.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3143" name="__module.model.layers.11.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2403"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3144" name="__module.model.layers.11.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2404,2405,hidden_states.353,hidden_states.355"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3145" name="__module.model.layers.11.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2406,2408"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3146" name="self.model.layers.11.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="560219240" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3147" name="Convert_12869" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3148" name="self.model.layers.11.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="571753576" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3149" name="Convert_12872" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3150" name="self.model.layers.11.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3151" name="self.model.layers.11.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="571759208" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3152" name="self.model.layers.11.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3153" name="self.model.layers.11.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3154" name="__module.model.layers.11.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2414,input.23"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3155" name="__module.model.layers.11.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2415"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3156" name="self.model.layers.11.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="571770472" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3157" name="Convert_12859" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3158" name="self.model.layers.11.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="583304808" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3159" name="Convert_12862" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3160" name="self.model.layers.11.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3161" name="self.model.layers.11.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="583310440" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3162" name="self.model.layers.11.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3163" name="self.model.layers.11.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3164" name="__module.model.layers.11.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2417"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3165" name="__module.model.layers.11.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2418"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3166" name="self.model.layers.11.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="583321704" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3167" name="Convert_12879" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3168" name="self.model.layers.11.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="594856040" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3169" name="Convert_12882" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3170" name="self.model.layers.11.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3171" name="self.model.layers.11.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="594858088" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3172" name="self.model.layers.11.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3173" name="self.model.layers.11.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3174" name="__module.model.layers.11.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2420,hidden_states.357"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3175" name="__module.model.layers.11/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2421,2423,2431,2442,hidden_states.359,hidden_states.361"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3176" name="Constant_111256_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="594862184" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3177" name="Constant_111256" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3178" name="Constant_111255_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3179" name="Constant_111255" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3180" name="Constant_111253_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3181" name="Constant_111253" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3182" name="__module.model.layers.12.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2432"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3183" name="Constant_9307" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2433"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3184" name="__module.model.layers.12.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2434,variance.49"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3185" name="Constant_111254_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3186" name="Constant_111254" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3187" name="__module.model.layers.12.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2435"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3188" name="__module.model.layers.12.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3189" name="__module.model.layers.12.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2436"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3190" name="__module.model.layers.12.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2437,2438,hidden_states.363,hidden_states.365"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3191" name="__module.model.layers.12.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2439,2441,hidden_states.367"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3192" name="self.model.layers.12.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="594866280" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3193" name="Convert_12909" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3194" name="self.model.layers.12.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="599060584" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3195" name="Convert_12912" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3196" name="self.model.layers.12.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3197" name="self.model.layers.12.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="599062632" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3198" name="self.model.layers.12.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3199" name="self.model.layers.12.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3200" name="__module.model.layers.12.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2451,query_states.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3201" name="__module.model.layers.12.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3202" name="Constant_93594" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3203" name="Constant_93595" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3204" name="Gather_93596" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3205" name="__module.model.layers.12.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2456"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3206" name="__module.model.layers.12.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2457"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3207" name="__module.model.layers.12.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3208" name="__module.model.layers.12.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2458,q.25"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3209" name="__module.model.layers.12.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="2469"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3210" name="__module.model.layers.12.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3211" name="__module.model.layers.12.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2472,2477,inv_freq_expanded.49,inv_freq_expanded.51"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3212" name="__module.model.layers.12.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2479"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3213" name="__module.model.layers.12.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3214" name="__module.model.layers.12.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2480"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3215" name="__module.model.layers.12.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2482,emb.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3216" name="__module.model.layers.12.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2483,2485,2488,cos.73,cos.75"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3217" name="__module.model.layers.12.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2490,cos.77"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3218" name="__module.model.layers.12.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2492"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3219" name="Constant_99574" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3220" name="Constant_99573" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3221" name="__module.model.layers.12.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3222" name="__module.model.layers.12.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3223" name="__module.model.layers.12.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2493,2498" /> + </output> + </layer> + <layer id="3224" name="__module.model.layers.12.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="3225" name="__module.model.layers.12.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2495,2496,2500,2501" /> + </output> + </layer> + <layer id="3226" name="__module.model.layers.12.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3227" name="__module.model.layers.12.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3228" name="Constant_99572" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3229" name="ScatterUpdate_99575" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3230" name="Constant_99578" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3231" name="Constant_99581" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3232" name="__module.model.layers.12.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2502,x2.49"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3233" name="Constant_111257_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3234" name="Constant_111257" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3235" name="__module.model.layers.12.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2503"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3236" name="Constant_99626" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3237" name="Constant_99631" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3238" name="__module.model.layers.12.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2497"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3239" name="__module.model.layers.12.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2505"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3240" name="__module.model.layers.12.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2484,2486,2489,sin.73,sin.75"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3241" name="__module.model.layers.12.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2491,sin.77"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3242" name="__module.model.layers.12.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2506"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3243" name="__module.model.layers.12.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2507"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3244" name="Constant_124307_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="3245" name="Constant_124307" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="3246" name="Constant_124300" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3247" name="Constant_124302" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3248" name="Constant_124304" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3249" name="Concat_124305" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3250" name="Broadcast_124308" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3251" name="ReadValue_121739" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.12.keypresent.12.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.12.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3252" name="Constant_119507" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3253" name="Gather_119508" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3254" name="self.model.layers.12.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="599066728" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3255" name="Convert_12889" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3256" name="self.model.layers.12.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="599591016" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3257" name="Convert_12892" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3258" name="self.model.layers.12.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3259" name="self.model.layers.12.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="599591272" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3260" name="self.model.layers.12.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3261" name="self.model.layers.12.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3262" name="__module.model.layers.12.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2453,key_states.53"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="3263" name="__module.model.layers.12.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2459,2462"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3264" name="__module.model.layers.12.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2460"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3265" name="__module.model.layers.12.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3266" name="__module.model.layers.12.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2461,k.25"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3267" name="__module.model.layers.12.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2508"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3268" name="Constant_99674" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3269" name="Constant_99673" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3270" name="__module.model.layers.12.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3271" name="__module.model.layers.12.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3272" name="__module.model.layers.12.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2509,2514" /> + </output> + </layer> + <layer id="3273" name="__module.model.layers.12.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="3274" name="__module.model.layers.12.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2511,2512,2516,2517" /> + </output> + </layer> + <layer id="3275" name="__module.model.layers.12.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3276" name="__module.model.layers.12.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3277" name="Constant_99672" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3278" name="ScatterUpdate_99675" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3279" name="Constant_99678" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3280" name="Constant_99681" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3281" name="__module.model.layers.12.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2518,x2.51"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3282" name="Constant_111258_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3283" name="Constant_111258" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3284" name="__module.model.layers.12.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2519"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3285" name="Constant_99726" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3286" name="Constant_99731" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3287" name="__module.model.layers.12.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2513"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3288" name="__module.model.layers.12.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2521"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3289" name="__module.model.layers.12.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2522"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3290" name="__module.model.layers.12.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2523"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3291" name="__module.model.layers.12.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.12.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3292" name="__module.model.layers.12.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2535,2536,2537"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3293" name="__module.model.layers.12.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3294" name="Constant_92911" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3295" name="__module.model.layers.12.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3296" name="__module.model.layers.12.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2528"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3297" name="__module.model.layers.12.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3298" name="__module.model.layers.12.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2529" /> + </output> + </layer> + <layer id="3299" name="Constant_77444" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3300" name="__module.model.layers.12.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3301" name="Constant_93604" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3302" name="Constant_93605" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3303" name="Gather_93606" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3304" name="__module.model.layers.12.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2538"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3305" name="__module.model.layers.12.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3306" name="__module.model.layers.12.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2539,hidden_states.371"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3307" name="__module.model.layers.12.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2540,2541" /> + </output> + </layer> + <layer id="3308" name="Constant_77447" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3309" name="__module.model.layers.12.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3310" name="__module.model.layers.12.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2542"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3311" name="__module.model.layers.12.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2543,key_states.55"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3312" name="Constant_124320_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="3313" name="Constant_124320" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="3314" name="Constant_124313" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3315" name="Constant_124315" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3316" name="Constant_124317" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3317" name="Concat_124318" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3318" name="Broadcast_124321" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3319" name="ReadValue_121741" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.12.valuepresent.12.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.12.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3320" name="Constant_119510" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3321" name="Gather_119511" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3322" name="self.model.layers.12.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="599591784" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3323" name="Convert_12899" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3324" name="self.model.layers.12.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="600116072" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3325" name="Convert_12902" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3326" name="self.model.layers.12.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3327" name="self.model.layers.12.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="600116328" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3328" name="self.model.layers.12.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3329" name="self.model.layers.12.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3330" name="__module.model.layers.12.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2455,value_states.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="3331" name="__module.model.layers.12.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2463"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3332" name="__module.model.layers.12.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3333" name="__module.model.layers.12.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2464"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3334" name="__module.model.layers.12.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.12.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3335" name="__module.model.layers.12.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2551,2552,2553"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3336" name="__module.model.layers.12.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3337" name="Constant_92920" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3338" name="__module.model.layers.12.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3339" name="__module.model.layers.12.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2544"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3340" name="__module.model.layers.12.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3341" name="__module.model.layers.12.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2545" /> + </output> + </layer> + <layer id="3342" name="Constant_77449" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3343" name="__module.model.layers.12.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3344" name="Constant_93614" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3345" name="Constant_93615" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3346" name="Gather_93616" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3347" name="__module.model.layers.12.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2554"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3348" name="__module.model.layers.12.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3349" name="__module.model.layers.12.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2555,hidden_states.375"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3350" name="__module.model.layers.12.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2556,2557" /> + </output> + </layer> + <layer id="3351" name="Constant_77452" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3352" name="__module.model.layers.12.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3353" name="__module.model.layers.12.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2558"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3354" name="__module.model.layers.12.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2559"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3355" name="Constant_99850" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3356" name="Constant_99851" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3357" name="Constant_99847" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3358" name="__module.model.layers.12.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3359" name="Constant_92929" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3360" name="__module.model.layers.12.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3361" name="__module.model.layers.12.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2560"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3362" name="Constant_99846" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3363" name="ScatterUpdate_99852" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3364" name="Constant_99855" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3365" name="__module.model.layers.12.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2564"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3366" name="__module.model.layers.12.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2565,attn_output.49"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3367" name="__module.model.layers.12.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3368" name="__module.model.layers.12.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2566"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3369" name="__module.model.layers.12.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="2568"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3370" name="__module.model.layers.12.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2569"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3371" name="self.model.layers.12.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="600116840" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3372" name="Convert_12919" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3373" name="self.model.layers.12.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="604311144" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3374" name="Convert_12922" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3375" name="self.model.layers.12.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3376" name="self.model.layers.12.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="604313192" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3377" name="self.model.layers.12.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3378" name="self.model.layers.12.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3379" name="__module.model.layers.12.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2571,2573,hidden_states.377"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3380" name="__module.model.layers.12/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2576,2578,2589,hidden_states.379,hidden_states.381"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3381" name="Constant_111262_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="604317288" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3382" name="Constant_111262" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3383" name="Constant_111261_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3384" name="Constant_111261" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3385" name="Constant_111259_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3386" name="Constant_111259" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3387" name="__module.model.layers.12.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2579"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3388" name="Constant_9987" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2580"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3389" name="__module.model.layers.12.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2581,variance.51"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3390" name="Constant_111260_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3391" name="Constant_111260" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3392" name="__module.model.layers.12.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2582"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3393" name="__module.model.layers.12.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3394" name="__module.model.layers.12.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2583"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3395" name="__module.model.layers.12.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2584,2585,hidden_states.383,hidden_states.385"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3396" name="__module.model.layers.12.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2586,2588"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3397" name="self.model.layers.12.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="604321384" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3398" name="Convert_12939" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3399" name="self.model.layers.12.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="615855720" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3400" name="Convert_12942" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3401" name="self.model.layers.12.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3402" name="self.model.layers.12.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="615861352" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3403" name="self.model.layers.12.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3404" name="self.model.layers.12.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3405" name="__module.model.layers.12.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2594,input.25"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3406" name="__module.model.layers.12.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2595"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3407" name="self.model.layers.12.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="615872616" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3408" name="Convert_12929" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3409" name="self.model.layers.12.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="627406952" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3410" name="Convert_12932" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3411" name="self.model.layers.12.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3412" name="self.model.layers.12.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="627412584" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3413" name="self.model.layers.12.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3414" name="self.model.layers.12.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3415" name="__module.model.layers.12.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2597"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3416" name="__module.model.layers.12.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2598"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3417" name="self.model.layers.12.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="627423848" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3418" name="Convert_12949" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3419" name="self.model.layers.12.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="638958184" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3420" name="Convert_12952" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3421" name="self.model.layers.12.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3422" name="self.model.layers.12.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="638960232" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3423" name="self.model.layers.12.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3424" name="self.model.layers.12.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3425" name="__module.model.layers.12.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2600,hidden_states.387"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3426" name="__module.model.layers.12/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2601,2603,2611,2622,hidden_states.389,hidden_states.391"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3427" name="Constant_111266_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="638964328" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3428" name="Constant_111266" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3429" name="Constant_111265_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3430" name="Constant_111265" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3431" name="Constant_111263_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3432" name="Constant_111263" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3433" name="__module.model.layers.13.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2612"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3434" name="Constant_10058" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2613"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3435" name="__module.model.layers.13.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2614,variance.53"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3436" name="Constant_111264_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3437" name="Constant_111264" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3438" name="__module.model.layers.13.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2615"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3439" name="__module.model.layers.13.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3440" name="__module.model.layers.13.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2616"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3441" name="__module.model.layers.13.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2617,2618,hidden_states.393,hidden_states.395"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3442" name="__module.model.layers.13.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2619,2621,hidden_states.397"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3443" name="self.model.layers.13.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="638968424" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3444" name="Convert_12979" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3445" name="self.model.layers.13.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="643162728" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3446" name="Convert_12982" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3447" name="self.model.layers.13.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3448" name="self.model.layers.13.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="643164776" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3449" name="self.model.layers.13.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3450" name="self.model.layers.13.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3451" name="__module.model.layers.13.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2631,query_states.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3452" name="__module.model.layers.13.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3453" name="Constant_93629" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3454" name="Constant_93630" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3455" name="Gather_93631" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3456" name="__module.model.layers.13.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2636"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3457" name="__module.model.layers.13.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2637"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3458" name="__module.model.layers.13.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3459" name="__module.model.layers.13.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2638,q.27"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3460" name="__module.model.layers.13.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="2649"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3461" name="__module.model.layers.13.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3462" name="__module.model.layers.13.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2652,2657,inv_freq_expanded.53,inv_freq_expanded.55"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3463" name="__module.model.layers.13.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2659"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3464" name="__module.model.layers.13.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3465" name="__module.model.layers.13.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2660"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3466" name="__module.model.layers.13.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2662,emb.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3467" name="__module.model.layers.13.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2663,2665,2668,cos.79,cos.81"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3468" name="__module.model.layers.13.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2670,cos.83"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3469" name="__module.model.layers.13.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2672"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3470" name="Constant_99860" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3471" name="Constant_99859" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3472" name="__module.model.layers.13.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3473" name="__module.model.layers.13.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3474" name="__module.model.layers.13.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2673,2678" /> + </output> + </layer> + <layer id="3475" name="__module.model.layers.13.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="3476" name="__module.model.layers.13.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2675,2676,2680,2681" /> + </output> + </layer> + <layer id="3477" name="__module.model.layers.13.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3478" name="__module.model.layers.13.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3479" name="Constant_99858" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3480" name="ScatterUpdate_99861" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3481" name="Constant_99864" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3482" name="Constant_99867" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3483" name="__module.model.layers.13.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2682,x2.53"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3484" name="Constant_111267_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3485" name="Constant_111267" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3486" name="__module.model.layers.13.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2683"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3487" name="Constant_99912" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3488" name="Constant_99917" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3489" name="__module.model.layers.13.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2677"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3490" name="__module.model.layers.13.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2685"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3491" name="__module.model.layers.13.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2664,2666,2669,sin.79,sin.81"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3492" name="__module.model.layers.13.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2671,sin.83"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3493" name="__module.model.layers.13.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2686"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3494" name="__module.model.layers.13.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2687"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3495" name="Constant_124333_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="3496" name="Constant_124333" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="3497" name="Constant_124326" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3498" name="Constant_124328" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3499" name="Constant_124330" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3500" name="Concat_124331" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3501" name="Broadcast_124334" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3502" name="ReadValue_121743" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.13.keypresent.13.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.13.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3503" name="Constant_119513" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3504" name="Gather_119514" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3505" name="self.model.layers.13.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="643168872" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3506" name="Convert_12959" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3507" name="self.model.layers.13.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="643693160" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3508" name="Convert_12962" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3509" name="self.model.layers.13.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3510" name="self.model.layers.13.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="643693416" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3511" name="self.model.layers.13.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3512" name="self.model.layers.13.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3513" name="__module.model.layers.13.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2633,key_states.57"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="3514" name="__module.model.layers.13.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2639,2642"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3515" name="__module.model.layers.13.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2640"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3516" name="__module.model.layers.13.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3517" name="__module.model.layers.13.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2641,k.27"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3518" name="__module.model.layers.13.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2688"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3519" name="Constant_99960" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3520" name="Constant_99959" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3521" name="__module.model.layers.13.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3522" name="__module.model.layers.13.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3523" name="__module.model.layers.13.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2689,2694" /> + </output> + </layer> + <layer id="3524" name="__module.model.layers.13.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="3525" name="__module.model.layers.13.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2691,2692,2696,2697" /> + </output> + </layer> + <layer id="3526" name="__module.model.layers.13.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3527" name="__module.model.layers.13.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3528" name="Constant_99958" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3529" name="ScatterUpdate_99961" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3530" name="Constant_99964" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3531" name="Constant_99967" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3532" name="__module.model.layers.13.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2698,x2.55"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3533" name="Constant_111268_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3534" name="Constant_111268" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3535" name="__module.model.layers.13.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2699"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3536" name="Constant_100012" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3537" name="Constant_100017" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3538" name="__module.model.layers.13.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2693"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3539" name="__module.model.layers.13.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2701"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3540" name="__module.model.layers.13.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2702"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3541" name="__module.model.layers.13.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2703"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3542" name="__module.model.layers.13.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.13.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3543" name="__module.model.layers.13.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2715,2716,2717"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3544" name="__module.model.layers.13.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3545" name="Constant_92938" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3546" name="__module.model.layers.13.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3547" name="__module.model.layers.13.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2708"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3548" name="__module.model.layers.13.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3549" name="__module.model.layers.13.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2709" /> + </output> + </layer> + <layer id="3550" name="Constant_77456" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3551" name="__module.model.layers.13.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3552" name="Constant_93639" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3553" name="Constant_93640" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3554" name="Gather_93641" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3555" name="__module.model.layers.13.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2718"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3556" name="__module.model.layers.13.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3557" name="__module.model.layers.13.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2719,hidden_states.401"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3558" name="__module.model.layers.13.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2720,2721" /> + </output> + </layer> + <layer id="3559" name="Constant_77459" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3560" name="__module.model.layers.13.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3561" name="__module.model.layers.13.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2722"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3562" name="__module.model.layers.13.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2723,key_states.59"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3563" name="Constant_124346_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="3564" name="Constant_124346" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="3565" name="Constant_124339" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3566" name="Constant_124341" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3567" name="Constant_124343" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3568" name="Concat_124344" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3569" name="Broadcast_124347" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3570" name="ReadValue_121745" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.13.valuepresent.13.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.13.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3571" name="Constant_119516" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3572" name="Gather_119517" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3573" name="self.model.layers.13.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="643693928" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3574" name="Convert_12969" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3575" name="self.model.layers.13.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="644218216" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3576" name="Convert_12972" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3577" name="self.model.layers.13.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3578" name="self.model.layers.13.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="644218472" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3579" name="self.model.layers.13.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3580" name="self.model.layers.13.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3581" name="__module.model.layers.13.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2635,value_states.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="3582" name="__module.model.layers.13.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2643"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3583" name="__module.model.layers.13.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3584" name="__module.model.layers.13.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2644"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3585" name="__module.model.layers.13.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.13.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3586" name="__module.model.layers.13.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2731,2732,2733"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3587" name="__module.model.layers.13.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3588" name="Constant_92947" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3589" name="__module.model.layers.13.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3590" name="__module.model.layers.13.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2724"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3591" name="__module.model.layers.13.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3592" name="__module.model.layers.13.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2725" /> + </output> + </layer> + <layer id="3593" name="Constant_77461" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3594" name="__module.model.layers.13.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3595" name="Constant_93649" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3596" name="Constant_93650" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3597" name="Gather_93651" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3598" name="__module.model.layers.13.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2734"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3599" name="__module.model.layers.13.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3600" name="__module.model.layers.13.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2735,hidden_states.405"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3601" name="__module.model.layers.13.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2736,2737" /> + </output> + </layer> + <layer id="3602" name="Constant_77464" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3603" name="__module.model.layers.13.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3604" name="__module.model.layers.13.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2738"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3605" name="__module.model.layers.13.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2739"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3606" name="Constant_100136" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3607" name="Constant_100137" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3608" name="Constant_100133" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3609" name="__module.model.layers.13.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3610" name="Constant_92956" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3611" name="__module.model.layers.13.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3612" name="__module.model.layers.13.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2740"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3613" name="Constant_100132" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3614" name="ScatterUpdate_100138" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3615" name="Constant_100141" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3616" name="__module.model.layers.13.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2744"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3617" name="__module.model.layers.13.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2745,attn_output.53"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3618" name="__module.model.layers.13.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3619" name="__module.model.layers.13.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2746"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3620" name="__module.model.layers.13.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="2748"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3621" name="__module.model.layers.13.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2749"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3622" name="self.model.layers.13.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="644218984" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3623" name="Convert_12989" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3624" name="self.model.layers.13.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="648413288" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3625" name="Convert_12992" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3626" name="self.model.layers.13.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3627" name="self.model.layers.13.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="648415336" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3628" name="self.model.layers.13.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3629" name="self.model.layers.13.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3630" name="__module.model.layers.13.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2751,2753,hidden_states.407"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3631" name="__module.model.layers.13/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2756,2758,2769,hidden_states.409,hidden_states.411"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3632" name="Constant_111272_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="648419432" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3633" name="Constant_111272" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3634" name="Constant_111271_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3635" name="Constant_111271" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3636" name="Constant_111269_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3637" name="Constant_111269" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3638" name="__module.model.layers.13.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2759"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3639" name="Constant_10738" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2760"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3640" name="__module.model.layers.13.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2761,variance.55"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3641" name="Constant_111270_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3642" name="Constant_111270" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3643" name="__module.model.layers.13.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2762"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3644" name="__module.model.layers.13.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3645" name="__module.model.layers.13.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2763"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3646" name="__module.model.layers.13.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2764,2765,hidden_states.413,hidden_states.415"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3647" name="__module.model.layers.13.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2766,2768"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3648" name="self.model.layers.13.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="648423528" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3649" name="Convert_13009" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3650" name="self.model.layers.13.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="659957864" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3651" name="Convert_13012" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3652" name="self.model.layers.13.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3653" name="self.model.layers.13.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="659963496" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3654" name="self.model.layers.13.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3655" name="self.model.layers.13.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3656" name="__module.model.layers.13.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2774,input.27"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3657" name="__module.model.layers.13.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2775"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3658" name="self.model.layers.13.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="659974760" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3659" name="Convert_12999" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3660" name="self.model.layers.13.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="671509096" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3661" name="Convert_13002" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3662" name="self.model.layers.13.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3663" name="self.model.layers.13.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="671514728" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3664" name="self.model.layers.13.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3665" name="self.model.layers.13.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3666" name="__module.model.layers.13.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2777"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3667" name="__module.model.layers.13.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2778"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3668" name="self.model.layers.13.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="671525992" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3669" name="Convert_13019" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3670" name="self.model.layers.13.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="683060328" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3671" name="Convert_13022" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3672" name="self.model.layers.13.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3673" name="self.model.layers.13.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="683062376" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3674" name="self.model.layers.13.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3675" name="self.model.layers.13.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3676" name="__module.model.layers.13.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2780,hidden_states.417"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3677" name="__module.model.layers.13/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2781,2783,2791,2802,hidden_states.419,hidden_states.421"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3678" name="Constant_111276_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="683066472" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3679" name="Constant_111276" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3680" name="Constant_111275_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3681" name="Constant_111275" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3682" name="Constant_111273_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3683" name="Constant_111273" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3684" name="__module.model.layers.14.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2792"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3685" name="Constant_10809" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2793"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3686" name="__module.model.layers.14.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2794,variance.57"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3687" name="Constant_111274_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3688" name="Constant_111274" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3689" name="__module.model.layers.14.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2795"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3690" name="__module.model.layers.14.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3691" name="__module.model.layers.14.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2796"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3692" name="__module.model.layers.14.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2797,2798,hidden_states.423,hidden_states.425"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3693" name="__module.model.layers.14.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2799,2801,hidden_states.427"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3694" name="self.model.layers.14.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="683070568" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3695" name="Convert_13049" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3696" name="self.model.layers.14.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="687264872" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3697" name="Convert_13052" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3698" name="self.model.layers.14.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3699" name="self.model.layers.14.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="687266920" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3700" name="self.model.layers.14.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3701" name="self.model.layers.14.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3702" name="__module.model.layers.14.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2811,query_states.29"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3703" name="__module.model.layers.14.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3704" name="Constant_93664" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3705" name="Constant_93665" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3706" name="Gather_93666" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3707" name="__module.model.layers.14.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2816"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3708" name="__module.model.layers.14.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2817"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3709" name="__module.model.layers.14.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3710" name="__module.model.layers.14.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2818,q.29"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3711" name="__module.model.layers.14.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="2829"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3712" name="__module.model.layers.14.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3713" name="__module.model.layers.14.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2832,2837,inv_freq_expanded.57,inv_freq_expanded.59"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3714" name="__module.model.layers.14.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2839"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3715" name="__module.model.layers.14.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3716" name="__module.model.layers.14.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2840"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3717" name="__module.model.layers.14.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2842,emb.29"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3718" name="__module.model.layers.14.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2843,2845,2848,cos.85,cos.87"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3719" name="__module.model.layers.14.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2850,cos.89"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3720" name="__module.model.layers.14.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2852"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3721" name="Constant_100146" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3722" name="Constant_100145" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3723" name="__module.model.layers.14.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3724" name="__module.model.layers.14.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3725" name="__module.model.layers.14.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2853,2858" /> + </output> + </layer> + <layer id="3726" name="__module.model.layers.14.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="3727" name="__module.model.layers.14.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2855,2856,2860,2861" /> + </output> + </layer> + <layer id="3728" name="__module.model.layers.14.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3729" name="__module.model.layers.14.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3730" name="Constant_100144" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3731" name="ScatterUpdate_100147" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3732" name="Constant_100150" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3733" name="Constant_100153" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3734" name="__module.model.layers.14.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2862,x2.57"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3735" name="Constant_111277_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3736" name="Constant_111277" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3737" name="__module.model.layers.14.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2863"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3738" name="Constant_100198" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3739" name="Constant_100203" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3740" name="__module.model.layers.14.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2857"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3741" name="__module.model.layers.14.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2865"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3742" name="__module.model.layers.14.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2844,2846,2849,sin.85,sin.87"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3743" name="__module.model.layers.14.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2851,sin.89"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3744" name="__module.model.layers.14.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2866"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3745" name="__module.model.layers.14.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2867"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3746" name="Constant_124359_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="3747" name="Constant_124359" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="3748" name="Constant_124352" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3749" name="Constant_124354" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3750" name="Constant_124356" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3751" name="Concat_124357" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3752" name="Broadcast_124360" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3753" name="ReadValue_121747" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.14.keypresent.14.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.14.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3754" name="Constant_119519" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3755" name="Gather_119520" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3756" name="self.model.layers.14.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="687271016" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3757" name="Convert_13029" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3758" name="self.model.layers.14.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="687795304" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3759" name="Convert_13032" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3760" name="self.model.layers.14.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3761" name="self.model.layers.14.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="687795560" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3762" name="self.model.layers.14.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3763" name="self.model.layers.14.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3764" name="__module.model.layers.14.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2813,key_states.61"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="3765" name="__module.model.layers.14.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2819,2822"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3766" name="__module.model.layers.14.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2820"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3767" name="__module.model.layers.14.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3768" name="__module.model.layers.14.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2821,k.29"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3769" name="__module.model.layers.14.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2868"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3770" name="Constant_100246" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3771" name="Constant_100245" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3772" name="__module.model.layers.14.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3773" name="__module.model.layers.14.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3774" name="__module.model.layers.14.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2869,2874" /> + </output> + </layer> + <layer id="3775" name="__module.model.layers.14.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="3776" name="__module.model.layers.14.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="2871,2872,2876,2877" /> + </output> + </layer> + <layer id="3777" name="__module.model.layers.14.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3778" name="__module.model.layers.14.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3779" name="Constant_100244" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3780" name="ScatterUpdate_100247" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3781" name="Constant_100250" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3782" name="Constant_100253" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3783" name="__module.model.layers.14.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2878,x2.59"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3784" name="Constant_111278_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3785" name="Constant_111278" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3786" name="__module.model.layers.14.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2879"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3787" name="Constant_100298" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3788" name="Constant_100303" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3789" name="__module.model.layers.14.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2873"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3790" name="__module.model.layers.14.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2881"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3791" name="__module.model.layers.14.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2882"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3792" name="__module.model.layers.14.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2883"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3793" name="__module.model.layers.14.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.14.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3794" name="__module.model.layers.14.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2895,2896,2897"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3795" name="__module.model.layers.14.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3796" name="Constant_92965" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3797" name="__module.model.layers.14.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3798" name="__module.model.layers.14.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2888"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3799" name="__module.model.layers.14.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3800" name="__module.model.layers.14.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2889" /> + </output> + </layer> + <layer id="3801" name="Constant_77468" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3802" name="__module.model.layers.14.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3803" name="Constant_93674" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3804" name="Constant_93675" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3805" name="Gather_93676" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3806" name="__module.model.layers.14.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2898"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3807" name="__module.model.layers.14.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3808" name="__module.model.layers.14.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2899,hidden_states.431"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3809" name="__module.model.layers.14.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2900,2901" /> + </output> + </layer> + <layer id="3810" name="Constant_77471" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3811" name="__module.model.layers.14.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3812" name="__module.model.layers.14.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2902"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3813" name="__module.model.layers.14.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2903,key_states.63"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3814" name="Constant_124372_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="3815" name="Constant_124372" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="3816" name="Constant_124365" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3817" name="Constant_124367" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3818" name="Constant_124369" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3819" name="Concat_124370" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3820" name="Broadcast_124373" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3821" name="ReadValue_121749" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.14.valuepresent.14.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.14.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3822" name="Constant_119522" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3823" name="Gather_119523" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3824" name="self.model.layers.14.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="687796072" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3825" name="Convert_13039" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3826" name="self.model.layers.14.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="688320360" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3827" name="Convert_13042" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3828" name="self.model.layers.14.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3829" name="self.model.layers.14.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="688320616" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3830" name="self.model.layers.14.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3831" name="self.model.layers.14.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3832" name="__module.model.layers.14.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2815,value_states.29"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="3833" name="__module.model.layers.14.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2823"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3834" name="__module.model.layers.14.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3835" name="__module.model.layers.14.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2824"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3836" name="__module.model.layers.14.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.14.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3837" name="__module.model.layers.14.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="2911,2912,2913"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3838" name="__module.model.layers.14.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3839" name="Constant_92974" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3840" name="__module.model.layers.14.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3841" name="__module.model.layers.14.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2904"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3842" name="__module.model.layers.14.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3843" name="__module.model.layers.14.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2905" /> + </output> + </layer> + <layer id="3844" name="Constant_77473" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3845" name="__module.model.layers.14.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3846" name="Constant_93684" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3847" name="Constant_93685" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3848" name="Gather_93686" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3849" name="__module.model.layers.14.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="2914"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3850" name="__module.model.layers.14.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="3851" name="__module.model.layers.14.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2915,hidden_states.435"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3852" name="__module.model.layers.14.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="2916,2917" /> + </output> + </layer> + <layer id="3853" name="Constant_77476" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3854" name="__module.model.layers.14.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3855" name="__module.model.layers.14.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2918"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3856" name="__module.model.layers.14.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2919"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3857" name="Constant_100422" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3858" name="Constant_100423" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3859" name="Constant_100419" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3860" name="__module.model.layers.14.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3861" name="Constant_92983" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3862" name="__module.model.layers.14.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3863" name="__module.model.layers.14.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="2920"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3864" name="Constant_100418" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3865" name="ScatterUpdate_100424" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3866" name="Constant_100427" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3867" name="__module.model.layers.14.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2924"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3868" name="__module.model.layers.14.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="2925,attn_output.57"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3869" name="__module.model.layers.14.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3870" name="__module.model.layers.14.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2926"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3871" name="__module.model.layers.14.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="2928"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3872" name="__module.model.layers.14.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2929"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3873" name="self.model.layers.14.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="688321128" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3874" name="Convert_13059" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3875" name="self.model.layers.14.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="692515432" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3876" name="Convert_13062" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3877" name="self.model.layers.14.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3878" name="self.model.layers.14.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="692517480" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3879" name="self.model.layers.14.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3880" name="self.model.layers.14.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3881" name="__module.model.layers.14.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2931,2933,hidden_states.437"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3882" name="__module.model.layers.14/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2936,2938,2949,hidden_states.439,hidden_states.441"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3883" name="Constant_111282_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="692521576" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3884" name="Constant_111282" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3885" name="Constant_111281_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3886" name="Constant_111281" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3887" name="Constant_111279_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3888" name="Constant_111279" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3889" name="__module.model.layers.14.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2939"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3890" name="Constant_11489" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2940"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3891" name="__module.model.layers.14.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2941,variance.59"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3892" name="Constant_111280_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3893" name="Constant_111280" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3894" name="__module.model.layers.14.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2942"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3895" name="__module.model.layers.14.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3896" name="__module.model.layers.14.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2943"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3897" name="__module.model.layers.14.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2944,2945,hidden_states.443,hidden_states.445"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3898" name="__module.model.layers.14.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2946,2948"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3899" name="self.model.layers.14.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="692525672" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3900" name="Convert_13079" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3901" name="self.model.layers.14.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="704060008" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3902" name="Convert_13082" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3903" name="self.model.layers.14.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3904" name="self.model.layers.14.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="704065640" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3905" name="self.model.layers.14.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3906" name="self.model.layers.14.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3907" name="__module.model.layers.14.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2954,input.29"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3908" name="__module.model.layers.14.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="2955"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3909" name="self.model.layers.14.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="704076904" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3910" name="Convert_13069" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3911" name="self.model.layers.14.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="715611240" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3912" name="Convert_13072" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3913" name="self.model.layers.14.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3914" name="self.model.layers.14.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="715616872" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3915" name="self.model.layers.14.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3916" name="self.model.layers.14.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3917" name="__module.model.layers.14.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2957"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3918" name="__module.model.layers.14.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2958"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3919" name="self.model.layers.14.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="715628136" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3920" name="Convert_13089" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3921" name="self.model.layers.14.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="727162472" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3922" name="Convert_13092" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3923" name="self.model.layers.14.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3924" name="self.model.layers.14.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="727164520" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3925" name="self.model.layers.14.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3926" name="self.model.layers.14.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="3927" name="__module.model.layers.14.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2960,hidden_states.447"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3928" name="__module.model.layers.14/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2961,2963,2971,2982,hidden_states.449,hidden_states.451"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3929" name="Constant_111286_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="727168616" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3930" name="Constant_111286" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3931" name="Constant_111285_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3932" name="Constant_111285" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3933" name="Constant_111283_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3934" name="Constant_111283" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3935" name="__module.model.layers.15.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2972"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3936" name="Constant_11560" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="2973"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3937" name="__module.model.layers.15.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2974,variance.61"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3938" name="Constant_111284_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3939" name="Constant_111284" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3940" name="__module.model.layers.15.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2975"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3941" name="__module.model.layers.15.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3942" name="__module.model.layers.15.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2976"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3943" name="__module.model.layers.15.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2977,2978,hidden_states.453,hidden_states.455"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3944" name="__module.model.layers.15.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2979,2981,hidden_states.457"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3945" name="self.model.layers.15.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="727172712" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3946" name="Convert_13119" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3947" name="self.model.layers.15.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="731367016" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3948" name="Convert_13122" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3949" name="self.model.layers.15.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3950" name="self.model.layers.15.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="731369064" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3951" name="self.model.layers.15.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3952" name="self.model.layers.15.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3953" name="__module.model.layers.15.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2991,query_states.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="3954" name="__module.model.layers.15.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3955" name="Constant_93699" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3956" name="Constant_93700" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="3957" name="Gather_93701" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="3958" name="__module.model.layers.15.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2996"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3959" name="__module.model.layers.15.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2997"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3960" name="__module.model.layers.15.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3961" name="__module.model.layers.15.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2998,q.31"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3962" name="__module.model.layers.15.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="3009"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3963" name="__module.model.layers.15.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3964" name="__module.model.layers.15.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3012,3017,inv_freq_expanded.61,inv_freq_expanded.63"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3965" name="__module.model.layers.15.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3019"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="3966" name="__module.model.layers.15.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="3967" name="__module.model.layers.15.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3020"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3968" name="__module.model.layers.15.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3022,emb.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3969" name="__module.model.layers.15.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3023,3025,3028,cos.91,cos.93"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3970" name="__module.model.layers.15.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3030,cos.95"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3971" name="__module.model.layers.15.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3032"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3972" name="Constant_100432" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3973" name="Constant_100431" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3974" name="__module.model.layers.15.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3975" name="__module.model.layers.15.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3976" name="__module.model.layers.15.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3033,3038" /> + </output> + </layer> + <layer id="3977" name="__module.model.layers.15.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="3978" name="__module.model.layers.15.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3035,3036,3040,3041" /> + </output> + </layer> + <layer id="3979" name="__module.model.layers.15.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="3980" name="__module.model.layers.15.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3981" name="Constant_100430" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3982" name="ScatterUpdate_100433" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3983" name="Constant_100436" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3984" name="Constant_100439" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3985" name="__module.model.layers.15.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3042,x2.61"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3986" name="Constant_111287_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3987" name="Constant_111287" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="3988" name="__module.model.layers.15.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3043"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3989" name="Constant_100484" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3990" name="Constant_100489" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="3991" name="__module.model.layers.15.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3037"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="3992" name="__module.model.layers.15.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3045"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3993" name="__module.model.layers.15.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3024,3026,3029,sin.91,sin.93"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3994" name="__module.model.layers.15.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3031,sin.95"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3995" name="__module.model.layers.15.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3046"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3996" name="__module.model.layers.15.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3047"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="3997" name="Constant_124385_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="3998" name="Constant_124385" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="3999" name="Constant_124378" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4000" name="Constant_124380" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4001" name="Constant_124382" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4002" name="Concat_124383" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4003" name="Broadcast_124386" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4004" name="ReadValue_121751" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.15.keypresent.15.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.15.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4005" name="Constant_119525" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4006" name="Gather_119526" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4007" name="self.model.layers.15.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="731373160" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4008" name="Convert_13099" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4009" name="self.model.layers.15.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="731897448" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4010" name="Convert_13102" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4011" name="self.model.layers.15.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4012" name="self.model.layers.15.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="731897704" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4013" name="self.model.layers.15.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4014" name="self.model.layers.15.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4015" name="__module.model.layers.15.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2993,key_states.65"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="4016" name="__module.model.layers.15.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="2999,3002"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4017" name="__module.model.layers.15.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3000"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4018" name="__module.model.layers.15.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4019" name="__module.model.layers.15.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3001,k.31"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4020" name="__module.model.layers.15.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3048"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4021" name="Constant_100532" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4022" name="Constant_100531" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4023" name="__module.model.layers.15.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4024" name="__module.model.layers.15.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4025" name="__module.model.layers.15.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3049,3054" /> + </output> + </layer> + <layer id="4026" name="__module.model.layers.15.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="4027" name="__module.model.layers.15.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3051,3052,3056,3057" /> + </output> + </layer> + <layer id="4028" name="__module.model.layers.15.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4029" name="__module.model.layers.15.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4030" name="Constant_100530" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4031" name="ScatterUpdate_100533" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4032" name="Constant_100536" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4033" name="Constant_100539" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4034" name="__module.model.layers.15.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3058,x2.63"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4035" name="Constant_111288_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4036" name="Constant_111288" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4037" name="__module.model.layers.15.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3059"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4038" name="Constant_100584" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4039" name="Constant_100589" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4040" name="__module.model.layers.15.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3053"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4041" name="__module.model.layers.15.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3061"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4042" name="__module.model.layers.15.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3062"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4043" name="__module.model.layers.15.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3063"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4044" name="__module.model.layers.15.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.15.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4045" name="__module.model.layers.15.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3075,3076,3077"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4046" name="__module.model.layers.15.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4047" name="Constant_92992" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4048" name="__module.model.layers.15.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4049" name="__module.model.layers.15.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3068"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4050" name="__module.model.layers.15.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4051" name="__module.model.layers.15.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3069" /> + </output> + </layer> + <layer id="4052" name="Constant_77480" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4053" name="__module.model.layers.15.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4054" name="Constant_93709" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4055" name="Constant_93710" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4056" name="Gather_93711" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4057" name="__module.model.layers.15.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3078"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4058" name="__module.model.layers.15.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4059" name="__module.model.layers.15.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3079,hidden_states.461"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4060" name="__module.model.layers.15.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3080,3081" /> + </output> + </layer> + <layer id="4061" name="Constant_77483" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4062" name="__module.model.layers.15.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4063" name="__module.model.layers.15.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3082"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4064" name="__module.model.layers.15.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3083,key_states.67"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4065" name="Constant_124398_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="4066" name="Constant_124398" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="4067" name="Constant_124391" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4068" name="Constant_124393" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4069" name="Constant_124395" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4070" name="Concat_124396" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4071" name="Broadcast_124399" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4072" name="ReadValue_121753" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.15.valuepresent.15.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.15.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4073" name="Constant_119528" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4074" name="Gather_119529" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4075" name="self.model.layers.15.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="731898216" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4076" name="Convert_13109" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4077" name="self.model.layers.15.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="732422504" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4078" name="Convert_13112" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4079" name="self.model.layers.15.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4080" name="self.model.layers.15.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="732422760" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4081" name="self.model.layers.15.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4082" name="self.model.layers.15.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4083" name="__module.model.layers.15.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="2995,value_states.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="4084" name="__module.model.layers.15.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3003"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4085" name="__module.model.layers.15.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4086" name="__module.model.layers.15.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3004"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4087" name="__module.model.layers.15.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.15.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4088" name="__module.model.layers.15.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3091,3092,3093"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4089" name="__module.model.layers.15.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4090" name="Constant_93001" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4091" name="__module.model.layers.15.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4092" name="__module.model.layers.15.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3084"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4093" name="__module.model.layers.15.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4094" name="__module.model.layers.15.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3085" /> + </output> + </layer> + <layer id="4095" name="Constant_77485" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4096" name="__module.model.layers.15.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4097" name="Constant_93719" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4098" name="Constant_93720" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4099" name="Gather_93721" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4100" name="__module.model.layers.15.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3094"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4101" name="__module.model.layers.15.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4102" name="__module.model.layers.15.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3095,hidden_states.465"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4103" name="__module.model.layers.15.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3096,3097" /> + </output> + </layer> + <layer id="4104" name="Constant_77488" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4105" name="__module.model.layers.15.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4106" name="__module.model.layers.15.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3098"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4107" name="__module.model.layers.15.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3099"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4108" name="Constant_100708" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4109" name="Constant_100709" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4110" name="Constant_100705" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4111" name="__module.model.layers.15.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4112" name="Constant_93010" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4113" name="__module.model.layers.15.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4114" name="__module.model.layers.15.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3100"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4115" name="Constant_100704" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4116" name="ScatterUpdate_100710" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4117" name="Constant_100713" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4118" name="__module.model.layers.15.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3104"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4119" name="__module.model.layers.15.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3105,attn_output.61"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4120" name="__module.model.layers.15.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4121" name="__module.model.layers.15.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3106"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4122" name="__module.model.layers.15.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="3108"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4123" name="__module.model.layers.15.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3109"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4124" name="self.model.layers.15.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="732423272" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4125" name="Convert_13129" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4126" name="self.model.layers.15.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="736617576" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4127" name="Convert_13132" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4128" name="self.model.layers.15.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4129" name="self.model.layers.15.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="736619624" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4130" name="self.model.layers.15.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4131" name="self.model.layers.15.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4132" name="__module.model.layers.15.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3111,3113,hidden_states.467"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4133" name="__module.model.layers.15/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3116,3118,3129,hidden_states.469,hidden_states.471"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4134" name="Constant_111292_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="736623720" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4135" name="Constant_111292" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4136" name="Constant_111291_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4137" name="Constant_111291" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4138" name="Constant_111289_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4139" name="Constant_111289" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4140" name="__module.model.layers.15.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3119"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4141" name="Constant_12240" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3120"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4142" name="__module.model.layers.15.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3121,variance.63"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4143" name="Constant_111290_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4144" name="Constant_111290" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4145" name="__module.model.layers.15.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3122"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4146" name="__module.model.layers.15.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4147" name="__module.model.layers.15.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3123"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4148" name="__module.model.layers.15.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3124,3125,hidden_states.473,hidden_states.475"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4149" name="__module.model.layers.15.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3126,3128"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4150" name="self.model.layers.15.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="736627816" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4151" name="Convert_13149" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4152" name="self.model.layers.15.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="748162152" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4153" name="Convert_13152" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4154" name="self.model.layers.15.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4155" name="self.model.layers.15.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="748167784" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4156" name="self.model.layers.15.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4157" name="self.model.layers.15.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4158" name="__module.model.layers.15.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3134,input.31"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4159" name="__module.model.layers.15.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3135"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4160" name="self.model.layers.15.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="748179048" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4161" name="Convert_13139" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4162" name="self.model.layers.15.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="759713384" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4163" name="Convert_13142" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4164" name="self.model.layers.15.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4165" name="self.model.layers.15.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="759719016" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4166" name="self.model.layers.15.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4167" name="self.model.layers.15.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4168" name="__module.model.layers.15.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3137"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4169" name="__module.model.layers.15.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3138"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4170" name="self.model.layers.15.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="759730280" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4171" name="Convert_13159" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4172" name="self.model.layers.15.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="771264616" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4173" name="Convert_13162" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4174" name="self.model.layers.15.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4175" name="self.model.layers.15.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="771266664" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4176" name="self.model.layers.15.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4177" name="self.model.layers.15.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4178" name="__module.model.layers.15.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3140,hidden_states.477"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4179" name="__module.model.layers.15/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3141,3143,3151,3162,hidden_states.479,hidden_states.481"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4180" name="Constant_111296_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="771270760" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4181" name="Constant_111296" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4182" name="Constant_111295_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4183" name="Constant_111295" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4184" name="Constant_111293_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4185" name="Constant_111293" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4186" name="__module.model.layers.16.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3152"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4187" name="Constant_12311" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3153"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4188" name="__module.model.layers.16.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3154,variance.65"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4189" name="Constant_111294_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4190" name="Constant_111294" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4191" name="__module.model.layers.16.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3155"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4192" name="__module.model.layers.16.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4193" name="__module.model.layers.16.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3156"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4194" name="__module.model.layers.16.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3157,3158,hidden_states.483,hidden_states.485"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4195" name="__module.model.layers.16.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3159,3161,hidden_states.487"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4196" name="self.model.layers.16.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="771274856" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4197" name="Convert_13189" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4198" name="self.model.layers.16.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="775469160" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4199" name="Convert_13192" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4200" name="self.model.layers.16.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4201" name="self.model.layers.16.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="775471208" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4202" name="self.model.layers.16.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4203" name="self.model.layers.16.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4204" name="__module.model.layers.16.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3171,query_states.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4205" name="__module.model.layers.16.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4206" name="Constant_93734" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4207" name="Constant_93735" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4208" name="Gather_93736" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4209" name="__module.model.layers.16.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3176"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4210" name="__module.model.layers.16.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3177"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4211" name="__module.model.layers.16.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4212" name="__module.model.layers.16.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3178,q.33"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4213" name="__module.model.layers.16.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="3189"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4214" name="__module.model.layers.16.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4215" name="__module.model.layers.16.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3192,3197,inv_freq_expanded.65,inv_freq_expanded.67"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4216" name="__module.model.layers.16.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3199"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4217" name="__module.model.layers.16.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4218" name="__module.model.layers.16.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3200"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4219" name="__module.model.layers.16.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3202,emb.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4220" name="__module.model.layers.16.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3203,3205,3208,cos.97,cos.99"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4221" name="__module.model.layers.16.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3210,cos.101"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4222" name="__module.model.layers.16.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3212"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4223" name="Constant_100718" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4224" name="Constant_100717" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4225" name="__module.model.layers.16.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4226" name="__module.model.layers.16.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4227" name="__module.model.layers.16.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3213,3218" /> + </output> + </layer> + <layer id="4228" name="__module.model.layers.16.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="4229" name="__module.model.layers.16.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3215,3216,3220,3221" /> + </output> + </layer> + <layer id="4230" name="__module.model.layers.16.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4231" name="__module.model.layers.16.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4232" name="Constant_100716" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4233" name="ScatterUpdate_100719" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4234" name="Constant_100722" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4235" name="Constant_100725" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4236" name="__module.model.layers.16.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3222,x2.65"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4237" name="Constant_111297_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4238" name="Constant_111297" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4239" name="__module.model.layers.16.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3223"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4240" name="Constant_100770" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4241" name="Constant_100775" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4242" name="__module.model.layers.16.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3217"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4243" name="__module.model.layers.16.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3225"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4244" name="__module.model.layers.16.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3204,3206,3209,sin.97,sin.99"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4245" name="__module.model.layers.16.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3211,sin.101"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4246" name="__module.model.layers.16.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3226"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4247" name="__module.model.layers.16.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3227"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4248" name="Constant_124411_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="4249" name="Constant_124411" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="4250" name="Constant_124404" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4251" name="Constant_124406" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4252" name="Constant_124408" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4253" name="Concat_124409" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4254" name="Broadcast_124412" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4255" name="ReadValue_121755" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.16.keypresent.16.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.16.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4256" name="Constant_119531" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4257" name="Gather_119532" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4258" name="self.model.layers.16.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="775475304" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4259" name="Convert_13169" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4260" name="self.model.layers.16.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="775999592" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4261" name="Convert_13172" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4262" name="self.model.layers.16.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4263" name="self.model.layers.16.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="775999848" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4264" name="self.model.layers.16.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4265" name="self.model.layers.16.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4266" name="__module.model.layers.16.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3173,key_states.69"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="4267" name="__module.model.layers.16.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3179,3182"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4268" name="__module.model.layers.16.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3180"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4269" name="__module.model.layers.16.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4270" name="__module.model.layers.16.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3181,k.33"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4271" name="__module.model.layers.16.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3228"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4272" name="Constant_100818" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4273" name="Constant_100817" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4274" name="__module.model.layers.16.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4275" name="__module.model.layers.16.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4276" name="__module.model.layers.16.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3229,3234" /> + </output> + </layer> + <layer id="4277" name="__module.model.layers.16.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="4278" name="__module.model.layers.16.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3231,3232,3236,3237" /> + </output> + </layer> + <layer id="4279" name="__module.model.layers.16.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4280" name="__module.model.layers.16.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4281" name="Constant_100816" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4282" name="ScatterUpdate_100819" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4283" name="Constant_100822" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4284" name="Constant_100825" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4285" name="__module.model.layers.16.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3238,x2.67"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4286" name="Constant_111298_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4287" name="Constant_111298" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4288" name="__module.model.layers.16.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3239"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4289" name="Constant_100870" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4290" name="Constant_100875" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4291" name="__module.model.layers.16.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3233"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4292" name="__module.model.layers.16.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3241"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4293" name="__module.model.layers.16.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3242"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4294" name="__module.model.layers.16.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3243"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4295" name="__module.model.layers.16.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.16.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4296" name="__module.model.layers.16.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3255,3256,3257"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4297" name="__module.model.layers.16.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4298" name="Constant_93019" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4299" name="__module.model.layers.16.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4300" name="__module.model.layers.16.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3248"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4301" name="__module.model.layers.16.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4302" name="__module.model.layers.16.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3249" /> + </output> + </layer> + <layer id="4303" name="Constant_77492" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4304" name="__module.model.layers.16.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4305" name="Constant_93744" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4306" name="Constant_93745" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4307" name="Gather_93746" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4308" name="__module.model.layers.16.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3258"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4309" name="__module.model.layers.16.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4310" name="__module.model.layers.16.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3259,hidden_states.491"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4311" name="__module.model.layers.16.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3260,3261" /> + </output> + </layer> + <layer id="4312" name="Constant_77495" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4313" name="__module.model.layers.16.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4314" name="__module.model.layers.16.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3262"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4315" name="__module.model.layers.16.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3263,key_states.71"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4316" name="Constant_124424_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="4317" name="Constant_124424" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="4318" name="Constant_124417" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4319" name="Constant_124419" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4320" name="Constant_124421" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4321" name="Concat_124422" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4322" name="Broadcast_124425" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4323" name="ReadValue_121757" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.16.valuepresent.16.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.16.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4324" name="Constant_119534" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4325" name="Gather_119535" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4326" name="self.model.layers.16.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="776000360" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4327" name="Convert_13179" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4328" name="self.model.layers.16.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="776524648" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4329" name="Convert_13182" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4330" name="self.model.layers.16.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4331" name="self.model.layers.16.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="776524904" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4332" name="self.model.layers.16.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4333" name="self.model.layers.16.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4334" name="__module.model.layers.16.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3175,value_states.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="4335" name="__module.model.layers.16.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3183"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4336" name="__module.model.layers.16.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4337" name="__module.model.layers.16.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3184"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4338" name="__module.model.layers.16.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.16.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4339" name="__module.model.layers.16.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3271,3272,3273"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4340" name="__module.model.layers.16.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4341" name="Constant_93028" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4342" name="__module.model.layers.16.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4343" name="__module.model.layers.16.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3264"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4344" name="__module.model.layers.16.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4345" name="__module.model.layers.16.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3265" /> + </output> + </layer> + <layer id="4346" name="Constant_77497" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4347" name="__module.model.layers.16.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4348" name="Constant_93754" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4349" name="Constant_93755" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4350" name="Gather_93756" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4351" name="__module.model.layers.16.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3274"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4352" name="__module.model.layers.16.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4353" name="__module.model.layers.16.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3275,hidden_states.495"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4354" name="__module.model.layers.16.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3276,3277" /> + </output> + </layer> + <layer id="4355" name="Constant_77500" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4356" name="__module.model.layers.16.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4357" name="__module.model.layers.16.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3278"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4358" name="__module.model.layers.16.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3279"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4359" name="Constant_100994" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4360" name="Constant_100995" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4361" name="Constant_100991" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4362" name="__module.model.layers.16.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4363" name="Constant_93037" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4364" name="__module.model.layers.16.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4365" name="__module.model.layers.16.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3280"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4366" name="Constant_100990" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4367" name="ScatterUpdate_100996" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4368" name="Constant_100999" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4369" name="__module.model.layers.16.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3284"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4370" name="__module.model.layers.16.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3285,attn_output.65"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4371" name="__module.model.layers.16.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4372" name="__module.model.layers.16.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3286"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4373" name="__module.model.layers.16.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="3288"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4374" name="__module.model.layers.16.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3289"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4375" name="self.model.layers.16.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="776525416" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4376" name="Convert_13199" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4377" name="self.model.layers.16.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="780719720" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4378" name="Convert_13202" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4379" name="self.model.layers.16.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4380" name="self.model.layers.16.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="780721768" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4381" name="self.model.layers.16.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4382" name="self.model.layers.16.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4383" name="__module.model.layers.16.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3291,3293,hidden_states.497"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4384" name="__module.model.layers.16/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3296,3298,3309,hidden_states.499,hidden_states.501"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4385" name="Constant_111302_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="780725864" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4386" name="Constant_111302" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4387" name="Constant_111301_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4388" name="Constant_111301" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4389" name="Constant_111299_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4390" name="Constant_111299" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4391" name="__module.model.layers.16.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3299"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4392" name="Constant_12991" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3300"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4393" name="__module.model.layers.16.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3301,variance.67"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4394" name="Constant_111300_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4395" name="Constant_111300" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4396" name="__module.model.layers.16.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3302"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4397" name="__module.model.layers.16.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4398" name="__module.model.layers.16.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3303"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4399" name="__module.model.layers.16.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3304,3305,hidden_states.503,hidden_states.505"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4400" name="__module.model.layers.16.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3306,3308"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4401" name="self.model.layers.16.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="780729960" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4402" name="Convert_13219" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4403" name="self.model.layers.16.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="792264296" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4404" name="Convert_13222" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4405" name="self.model.layers.16.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4406" name="self.model.layers.16.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="792269928" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4407" name="self.model.layers.16.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4408" name="self.model.layers.16.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4409" name="__module.model.layers.16.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3314,input.33"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4410" name="__module.model.layers.16.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3315"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4411" name="self.model.layers.16.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="792281192" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4412" name="Convert_13209" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4413" name="self.model.layers.16.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="803815528" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4414" name="Convert_13212" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4415" name="self.model.layers.16.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4416" name="self.model.layers.16.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="803821160" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4417" name="self.model.layers.16.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4418" name="self.model.layers.16.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4419" name="__module.model.layers.16.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3317"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4420" name="__module.model.layers.16.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3318"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4421" name="self.model.layers.16.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="803832424" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4422" name="Convert_13229" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4423" name="self.model.layers.16.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="815366760" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4424" name="Convert_13232" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4425" name="self.model.layers.16.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4426" name="self.model.layers.16.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="815368808" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4427" name="self.model.layers.16.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4428" name="self.model.layers.16.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4429" name="__module.model.layers.16.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3320,hidden_states.507"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4430" name="__module.model.layers.16/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3321,3323,3331,3342,hidden_states.509,hidden_states.511"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4431" name="Constant_111306_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="815372904" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4432" name="Constant_111306" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4433" name="Constant_111305_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4434" name="Constant_111305" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4435" name="Constant_111303_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4436" name="Constant_111303" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4437" name="__module.model.layers.17.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3332"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4438" name="Constant_13062" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3333"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4439" name="__module.model.layers.17.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3334,variance.69"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4440" name="Constant_111304_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4441" name="Constant_111304" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4442" name="__module.model.layers.17.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3335"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4443" name="__module.model.layers.17.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4444" name="__module.model.layers.17.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3336"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4445" name="__module.model.layers.17.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3337,3338,hidden_states.513,hidden_states.515"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4446" name="__module.model.layers.17.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3339,3341,hidden_states.517"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4447" name="self.model.layers.17.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="815377000" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4448" name="Convert_13259" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4449" name="self.model.layers.17.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="819571304" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4450" name="Convert_13262" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4451" name="self.model.layers.17.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4452" name="self.model.layers.17.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="819573352" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4453" name="self.model.layers.17.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4454" name="self.model.layers.17.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4455" name="__module.model.layers.17.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3351,query_states.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4456" name="__module.model.layers.17.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4457" name="Constant_93769" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4458" name="Constant_93770" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4459" name="Gather_93771" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4460" name="__module.model.layers.17.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3356"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4461" name="__module.model.layers.17.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3357"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4462" name="__module.model.layers.17.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4463" name="__module.model.layers.17.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3358,q.35"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4464" name="__module.model.layers.17.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="3369"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4465" name="__module.model.layers.17.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4466" name="__module.model.layers.17.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3372,3377,inv_freq_expanded.69,inv_freq_expanded.71"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4467" name="__module.model.layers.17.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3379"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4468" name="__module.model.layers.17.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4469" name="__module.model.layers.17.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3380"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4470" name="__module.model.layers.17.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3382,emb.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4471" name="__module.model.layers.17.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3383,3385,3388,cos.103,cos.105"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4472" name="__module.model.layers.17.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3390,cos.107"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4473" name="__module.model.layers.17.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3392"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4474" name="Constant_101004" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4475" name="Constant_101003" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4476" name="__module.model.layers.17.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4477" name="__module.model.layers.17.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4478" name="__module.model.layers.17.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3393,3398" /> + </output> + </layer> + <layer id="4479" name="__module.model.layers.17.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="4480" name="__module.model.layers.17.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3395,3396,3400,3401" /> + </output> + </layer> + <layer id="4481" name="__module.model.layers.17.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4482" name="__module.model.layers.17.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4483" name="Constant_101002" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4484" name="ScatterUpdate_101005" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4485" name="Constant_101008" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4486" name="Constant_101011" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4487" name="__module.model.layers.17.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3402,x2.69"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4488" name="Constant_111307_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4489" name="Constant_111307" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4490" name="__module.model.layers.17.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3403"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4491" name="Constant_101056" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4492" name="Constant_101061" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4493" name="__module.model.layers.17.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3397"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4494" name="__module.model.layers.17.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3405"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4495" name="__module.model.layers.17.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3384,3386,3389,sin.103,sin.105"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4496" name="__module.model.layers.17.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3391,sin.107"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4497" name="__module.model.layers.17.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3406"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4498" name="__module.model.layers.17.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3407"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4499" name="Constant_124437_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="4500" name="Constant_124437" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="4501" name="Constant_124430" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4502" name="Constant_124432" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4503" name="Constant_124434" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4504" name="Concat_124435" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4505" name="Broadcast_124438" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4506" name="ReadValue_121759" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.17.keypresent.17.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.17.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4507" name="Constant_119537" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4508" name="Gather_119538" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4509" name="self.model.layers.17.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="819577448" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4510" name="Convert_13239" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4511" name="self.model.layers.17.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="820101736" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4512" name="Convert_13242" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4513" name="self.model.layers.17.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4514" name="self.model.layers.17.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="820101992" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4515" name="self.model.layers.17.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4516" name="self.model.layers.17.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4517" name="__module.model.layers.17.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3353,key_states.73"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="4518" name="__module.model.layers.17.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3359,3362"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4519" name="__module.model.layers.17.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3360"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4520" name="__module.model.layers.17.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4521" name="__module.model.layers.17.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3361,k.35"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4522" name="__module.model.layers.17.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3408"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4523" name="Constant_101104" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4524" name="Constant_101103" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4525" name="__module.model.layers.17.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4526" name="__module.model.layers.17.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4527" name="__module.model.layers.17.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3409,3414" /> + </output> + </layer> + <layer id="4528" name="__module.model.layers.17.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="4529" name="__module.model.layers.17.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3411,3412,3416,3417" /> + </output> + </layer> + <layer id="4530" name="__module.model.layers.17.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4531" name="__module.model.layers.17.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4532" name="Constant_101102" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4533" name="ScatterUpdate_101105" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4534" name="Constant_101108" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4535" name="Constant_101111" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4536" name="__module.model.layers.17.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3418,x2.71"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4537" name="Constant_111308_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4538" name="Constant_111308" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4539" name="__module.model.layers.17.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3419"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4540" name="Constant_101156" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4541" name="Constant_101161" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4542" name="__module.model.layers.17.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3413"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4543" name="__module.model.layers.17.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3421"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4544" name="__module.model.layers.17.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3422"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4545" name="__module.model.layers.17.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3423"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4546" name="__module.model.layers.17.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.17.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4547" name="__module.model.layers.17.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3435,3436,3437"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4548" name="__module.model.layers.17.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4549" name="Constant_93046" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4550" name="__module.model.layers.17.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4551" name="__module.model.layers.17.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3428"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4552" name="__module.model.layers.17.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4553" name="__module.model.layers.17.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3429" /> + </output> + </layer> + <layer id="4554" name="Constant_77504" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4555" name="__module.model.layers.17.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4556" name="Constant_93779" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4557" name="Constant_93780" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4558" name="Gather_93781" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4559" name="__module.model.layers.17.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3438"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4560" name="__module.model.layers.17.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4561" name="__module.model.layers.17.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3439,hidden_states.521"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4562" name="__module.model.layers.17.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3440,3441" /> + </output> + </layer> + <layer id="4563" name="Constant_77507" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4564" name="__module.model.layers.17.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4565" name="__module.model.layers.17.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3442"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4566" name="__module.model.layers.17.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3443,key_states.75"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4567" name="Constant_124450_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="4568" name="Constant_124450" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="4569" name="Constant_124443" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4570" name="Constant_124445" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4571" name="Constant_124447" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4572" name="Concat_124448" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4573" name="Broadcast_124451" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4574" name="ReadValue_121761" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.17.valuepresent.17.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.17.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4575" name="Constant_119540" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4576" name="Gather_119541" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4577" name="self.model.layers.17.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="820102504" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4578" name="Convert_13249" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4579" name="self.model.layers.17.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="820626792" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4580" name="Convert_13252" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4581" name="self.model.layers.17.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4582" name="self.model.layers.17.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="820627048" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4583" name="self.model.layers.17.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4584" name="self.model.layers.17.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4585" name="__module.model.layers.17.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3355,value_states.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="4586" name="__module.model.layers.17.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3363"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4587" name="__module.model.layers.17.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4588" name="__module.model.layers.17.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3364"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4589" name="__module.model.layers.17.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.17.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4590" name="__module.model.layers.17.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3451,3452,3453"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4591" name="__module.model.layers.17.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4592" name="Constant_93055" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4593" name="__module.model.layers.17.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4594" name="__module.model.layers.17.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3444"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4595" name="__module.model.layers.17.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4596" name="__module.model.layers.17.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3445" /> + </output> + </layer> + <layer id="4597" name="Constant_77509" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4598" name="__module.model.layers.17.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4599" name="Constant_93789" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4600" name="Constant_93790" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4601" name="Gather_93791" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4602" name="__module.model.layers.17.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3454"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4603" name="__module.model.layers.17.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4604" name="__module.model.layers.17.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3455,hidden_states.525"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4605" name="__module.model.layers.17.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3456,3457" /> + </output> + </layer> + <layer id="4606" name="Constant_77512" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4607" name="__module.model.layers.17.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4608" name="__module.model.layers.17.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3458"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4609" name="__module.model.layers.17.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3459"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4610" name="Constant_101280" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4611" name="Constant_101281" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4612" name="Constant_101277" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4613" name="__module.model.layers.17.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4614" name="Constant_93064" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4615" name="__module.model.layers.17.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4616" name="__module.model.layers.17.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3460"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4617" name="Constant_101276" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4618" name="ScatterUpdate_101282" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4619" name="Constant_101285" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4620" name="__module.model.layers.17.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3464"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4621" name="__module.model.layers.17.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3465,attn_output.69"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4622" name="__module.model.layers.17.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4623" name="__module.model.layers.17.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3466"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4624" name="__module.model.layers.17.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="3468"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4625" name="__module.model.layers.17.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3469"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4626" name="self.model.layers.17.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="820627560" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4627" name="Convert_13269" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4628" name="self.model.layers.17.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="824821864" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4629" name="Convert_13272" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4630" name="self.model.layers.17.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4631" name="self.model.layers.17.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="824823912" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4632" name="self.model.layers.17.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4633" name="self.model.layers.17.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4634" name="__module.model.layers.17.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3471,3473,hidden_states.527"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4635" name="__module.model.layers.17/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3476,3478,3489,hidden_states.529,hidden_states.531"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4636" name="Constant_111312_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="824828008" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4637" name="Constant_111312" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4638" name="Constant_111311_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4639" name="Constant_111311" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4640" name="Constant_111309_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4641" name="Constant_111309" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4642" name="__module.model.layers.17.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3479"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4643" name="Constant_13742" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3480"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4644" name="__module.model.layers.17.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3481,variance.71"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4645" name="Constant_111310_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4646" name="Constant_111310" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4647" name="__module.model.layers.17.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3482"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4648" name="__module.model.layers.17.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4649" name="__module.model.layers.17.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3483"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4650" name="__module.model.layers.17.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3484,3485,hidden_states.533,hidden_states.535"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4651" name="__module.model.layers.17.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3486,3488"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4652" name="self.model.layers.17.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="824832104" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4653" name="Convert_13289" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4654" name="self.model.layers.17.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="836366440" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4655" name="Convert_13292" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4656" name="self.model.layers.17.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4657" name="self.model.layers.17.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="836372072" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4658" name="self.model.layers.17.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4659" name="self.model.layers.17.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4660" name="__module.model.layers.17.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3494,input.35"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4661" name="__module.model.layers.17.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3495"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4662" name="self.model.layers.17.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="836383336" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4663" name="Convert_13279" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4664" name="self.model.layers.17.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="847917672" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4665" name="Convert_13282" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4666" name="self.model.layers.17.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4667" name="self.model.layers.17.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="847923304" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4668" name="self.model.layers.17.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4669" name="self.model.layers.17.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4670" name="__module.model.layers.17.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3497"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4671" name="__module.model.layers.17.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3498"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4672" name="self.model.layers.17.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="847934568" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4673" name="Convert_13299" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4674" name="self.model.layers.17.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="859468904" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4675" name="Convert_13302" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4676" name="self.model.layers.17.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4677" name="self.model.layers.17.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="859470952" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4678" name="self.model.layers.17.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4679" name="self.model.layers.17.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4680" name="__module.model.layers.17.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3500,hidden_states.537"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4681" name="__module.model.layers.17/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3501,3503,3511,3522,hidden_states.539,hidden_states.541"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4682" name="Constant_111316_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="859475048" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4683" name="Constant_111316" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4684" name="Constant_111315_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4685" name="Constant_111315" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4686" name="Constant_111313_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4687" name="Constant_111313" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4688" name="__module.model.layers.18.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3512"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4689" name="Constant_13813" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3513"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4690" name="__module.model.layers.18.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3514,variance.73"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4691" name="Constant_111314_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4692" name="Constant_111314" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4693" name="__module.model.layers.18.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3515"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4694" name="__module.model.layers.18.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4695" name="__module.model.layers.18.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3516"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4696" name="__module.model.layers.18.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3517,3518,hidden_states.543,hidden_states.545"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4697" name="__module.model.layers.18.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3519,3521,hidden_states.547"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4698" name="self.model.layers.18.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="859479144" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4699" name="Convert_13329" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4700" name="self.model.layers.18.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="863673448" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4701" name="Convert_13332" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4702" name="self.model.layers.18.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4703" name="self.model.layers.18.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="863675496" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4704" name="self.model.layers.18.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4705" name="self.model.layers.18.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4706" name="__module.model.layers.18.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3531,query_states.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4707" name="__module.model.layers.18.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4708" name="Constant_93804" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4709" name="Constant_93805" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4710" name="Gather_93806" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4711" name="__module.model.layers.18.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3536"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4712" name="__module.model.layers.18.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3537"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4713" name="__module.model.layers.18.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4714" name="__module.model.layers.18.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3538,q.37"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4715" name="__module.model.layers.18.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="3549"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4716" name="__module.model.layers.18.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4717" name="__module.model.layers.18.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3552,3557,inv_freq_expanded.73,inv_freq_expanded.75"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4718" name="__module.model.layers.18.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3559"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4719" name="__module.model.layers.18.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4720" name="__module.model.layers.18.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3560"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4721" name="__module.model.layers.18.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3562,emb.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4722" name="__module.model.layers.18.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3563,3565,3568,cos.109,cos.111"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4723" name="__module.model.layers.18.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3570,cos.113"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4724" name="__module.model.layers.18.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3572"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4725" name="Constant_101290" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4726" name="Constant_101289" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4727" name="__module.model.layers.18.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4728" name="__module.model.layers.18.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4729" name="__module.model.layers.18.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3573,3578" /> + </output> + </layer> + <layer id="4730" name="__module.model.layers.18.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="4731" name="__module.model.layers.18.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3575,3576,3580,3581" /> + </output> + </layer> + <layer id="4732" name="__module.model.layers.18.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4733" name="__module.model.layers.18.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4734" name="Constant_101288" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4735" name="ScatterUpdate_101291" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4736" name="Constant_101294" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4737" name="Constant_101297" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4738" name="__module.model.layers.18.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3582,x2.73"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4739" name="Constant_111317_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4740" name="Constant_111317" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4741" name="__module.model.layers.18.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3583"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4742" name="Constant_101342" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4743" name="Constant_101347" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4744" name="__module.model.layers.18.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3577"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4745" name="__module.model.layers.18.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3585"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4746" name="__module.model.layers.18.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3564,3566,3569,sin.109,sin.111"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4747" name="__module.model.layers.18.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3571,sin.113"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4748" name="__module.model.layers.18.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3586"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4749" name="__module.model.layers.18.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3587"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4750" name="Constant_124463_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="4751" name="Constant_124463" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="4752" name="Constant_124456" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4753" name="Constant_124458" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4754" name="Constant_124460" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4755" name="Concat_124461" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4756" name="Broadcast_124464" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4757" name="ReadValue_121763" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.18.keypresent.18.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.18.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4758" name="Constant_119543" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4759" name="Gather_119544" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4760" name="self.model.layers.18.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="863679592" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4761" name="Convert_13309" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4762" name="self.model.layers.18.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="864203880" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4763" name="Convert_13312" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4764" name="self.model.layers.18.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4765" name="self.model.layers.18.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="864204136" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4766" name="self.model.layers.18.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4767" name="self.model.layers.18.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4768" name="__module.model.layers.18.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3533,key_states.77"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="4769" name="__module.model.layers.18.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3539,3542"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4770" name="__module.model.layers.18.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3540"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4771" name="__module.model.layers.18.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4772" name="__module.model.layers.18.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3541,k.37"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4773" name="__module.model.layers.18.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3588"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4774" name="Constant_101390" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4775" name="Constant_101389" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4776" name="__module.model.layers.18.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4777" name="__module.model.layers.18.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4778" name="__module.model.layers.18.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3589,3594" /> + </output> + </layer> + <layer id="4779" name="__module.model.layers.18.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="4780" name="__module.model.layers.18.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3591,3592,3596,3597" /> + </output> + </layer> + <layer id="4781" name="__module.model.layers.18.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4782" name="__module.model.layers.18.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4783" name="Constant_101388" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4784" name="ScatterUpdate_101391" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4785" name="Constant_101394" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4786" name="Constant_101397" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4787" name="__module.model.layers.18.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3598,x2.75"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4788" name="Constant_111318_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4789" name="Constant_111318" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4790" name="__module.model.layers.18.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3599"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4791" name="Constant_101442" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4792" name="Constant_101447" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4793" name="__module.model.layers.18.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3593"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4794" name="__module.model.layers.18.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3601"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4795" name="__module.model.layers.18.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3602"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4796" name="__module.model.layers.18.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3603"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4797" name="__module.model.layers.18.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.18.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4798" name="__module.model.layers.18.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3615,3616,3617"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4799" name="__module.model.layers.18.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4800" name="Constant_93073" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4801" name="__module.model.layers.18.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4802" name="__module.model.layers.18.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3608"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4803" name="__module.model.layers.18.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4804" name="__module.model.layers.18.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3609" /> + </output> + </layer> + <layer id="4805" name="Constant_77516" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4806" name="__module.model.layers.18.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4807" name="Constant_93814" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4808" name="Constant_93815" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4809" name="Gather_93816" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4810" name="__module.model.layers.18.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3618"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4811" name="__module.model.layers.18.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4812" name="__module.model.layers.18.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3619,hidden_states.551"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4813" name="__module.model.layers.18.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3620,3621" /> + </output> + </layer> + <layer id="4814" name="Constant_77519" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4815" name="__module.model.layers.18.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4816" name="__module.model.layers.18.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3622"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4817" name="__module.model.layers.18.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3623,key_states.79"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4818" name="Constant_124476_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="4819" name="Constant_124476" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="4820" name="Constant_124469" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4821" name="Constant_124471" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4822" name="Constant_124473" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4823" name="Concat_124474" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4824" name="Broadcast_124477" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4825" name="ReadValue_121765" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.18.valuepresent.18.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.18.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4826" name="Constant_119546" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4827" name="Gather_119547" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4828" name="self.model.layers.18.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="864204648" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4829" name="Convert_13319" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4830" name="self.model.layers.18.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="864728936" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4831" name="Convert_13322" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4832" name="self.model.layers.18.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4833" name="self.model.layers.18.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="864729192" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4834" name="self.model.layers.18.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4835" name="self.model.layers.18.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4836" name="__module.model.layers.18.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3535,value_states.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="4837" name="__module.model.layers.18.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3543"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4838" name="__module.model.layers.18.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4839" name="__module.model.layers.18.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3544"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4840" name="__module.model.layers.18.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.18.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4841" name="__module.model.layers.18.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3631,3632,3633"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4842" name="__module.model.layers.18.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4843" name="Constant_93082" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4844" name="__module.model.layers.18.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4845" name="__module.model.layers.18.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3624"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4846" name="__module.model.layers.18.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4847" name="__module.model.layers.18.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3625" /> + </output> + </layer> + <layer id="4848" name="Constant_77521" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4849" name="__module.model.layers.18.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4850" name="Constant_93824" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4851" name="Constant_93825" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4852" name="Gather_93826" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4853" name="__module.model.layers.18.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3634"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4854" name="__module.model.layers.18.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="4855" name="__module.model.layers.18.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3635,hidden_states.555"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4856" name="__module.model.layers.18.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3636,3637" /> + </output> + </layer> + <layer id="4857" name="Constant_77524" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4858" name="__module.model.layers.18.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4859" name="__module.model.layers.18.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3638"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4860" name="__module.model.layers.18.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3639"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4861" name="Constant_101566" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4862" name="Constant_101567" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4863" name="Constant_101563" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4864" name="__module.model.layers.18.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4865" name="Constant_93091" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4866" name="__module.model.layers.18.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4867" name="__module.model.layers.18.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3640"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4868" name="Constant_101562" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4869" name="ScatterUpdate_101568" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4870" name="Constant_101571" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4871" name="__module.model.layers.18.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3644"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4872" name="__module.model.layers.18.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3645,attn_output.73"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4873" name="__module.model.layers.18.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4874" name="__module.model.layers.18.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3646"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4875" name="__module.model.layers.18.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="3648"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4876" name="__module.model.layers.18.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3649"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4877" name="self.model.layers.18.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="864729704" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4878" name="Convert_13339" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4879" name="self.model.layers.18.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="868924008" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4880" name="Convert_13342" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4881" name="self.model.layers.18.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4882" name="self.model.layers.18.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="868926056" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4883" name="self.model.layers.18.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4884" name="self.model.layers.18.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4885" name="__module.model.layers.18.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3651,3653,hidden_states.557"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4886" name="__module.model.layers.18/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3656,3658,3669,hidden_states.559,hidden_states.561"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4887" name="Constant_111322_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="868930152" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4888" name="Constant_111322" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4889" name="Constant_111321_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4890" name="Constant_111321" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4891" name="Constant_111319_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4892" name="Constant_111319" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4893" name="__module.model.layers.18.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3659"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4894" name="Constant_14493" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3660"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4895" name="__module.model.layers.18.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3661,variance.75"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4896" name="Constant_111320_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4897" name="Constant_111320" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4898" name="__module.model.layers.18.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3662"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4899" name="__module.model.layers.18.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4900" name="__module.model.layers.18.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3663"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4901" name="__module.model.layers.18.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3664,3665,hidden_states.563,hidden_states.565"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4902" name="__module.model.layers.18.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3666,3668"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4903" name="self.model.layers.18.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="868934248" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4904" name="Convert_13359" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4905" name="self.model.layers.18.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="880468584" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4906" name="Convert_13362" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4907" name="self.model.layers.18.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4908" name="self.model.layers.18.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="880474216" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4909" name="self.model.layers.18.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4910" name="self.model.layers.18.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4911" name="__module.model.layers.18.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3674,input.37"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4912" name="__module.model.layers.18.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3675"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4913" name="self.model.layers.18.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="880485480" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4914" name="Convert_13349" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4915" name="self.model.layers.18.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="892019816" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4916" name="Convert_13352" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4917" name="self.model.layers.18.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4918" name="self.model.layers.18.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="892025448" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4919" name="self.model.layers.18.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4920" name="self.model.layers.18.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4921" name="__module.model.layers.18.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3677"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4922" name="__module.model.layers.18.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3678"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4923" name="self.model.layers.18.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="892036712" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4924" name="Convert_13369" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4925" name="self.model.layers.18.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="903571048" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4926" name="Convert_13372" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4927" name="self.model.layers.18.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4928" name="self.model.layers.18.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="903573096" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4929" name="self.model.layers.18.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4930" name="self.model.layers.18.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="4931" name="__module.model.layers.18.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3680,hidden_states.567"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4932" name="__module.model.layers.18/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3681,3683,3691,3702,hidden_states.569,hidden_states.571"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4933" name="Constant_111326_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="903577192" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4934" name="Constant_111326" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4935" name="Constant_111325_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4936" name="Constant_111325" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4937" name="Constant_111323_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4938" name="Constant_111323" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4939" name="__module.model.layers.19.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3692"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4940" name="Constant_14564" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3693"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4941" name="__module.model.layers.19.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3694,variance.77"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4942" name="Constant_111324_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4943" name="Constant_111324" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4944" name="__module.model.layers.19.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3695"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4945" name="__module.model.layers.19.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4946" name="__module.model.layers.19.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3696"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4947" name="__module.model.layers.19.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3697,3698,hidden_states.573,hidden_states.575"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4948" name="__module.model.layers.19.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3699,3701,hidden_states.577"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4949" name="self.model.layers.19.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="903581288" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4950" name="Convert_13399" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4951" name="self.model.layers.19.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="907775592" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4952" name="Convert_13402" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4953" name="self.model.layers.19.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4954" name="self.model.layers.19.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="907777640" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4955" name="self.model.layers.19.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4956" name="self.model.layers.19.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4957" name="__module.model.layers.19.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3711,query_states.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="4958" name="__module.model.layers.19.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4959" name="Constant_93839" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4960" name="Constant_93840" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="4961" name="Gather_93841" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="4962" name="__module.model.layers.19.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3716"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4963" name="__module.model.layers.19.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3717"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4964" name="__module.model.layers.19.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4965" name="__module.model.layers.19.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3718,q.39"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4966" name="__module.model.layers.19.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="3729"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4967" name="__module.model.layers.19.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4968" name="__module.model.layers.19.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3732,3737,inv_freq_expanded.77,inv_freq_expanded.79"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4969" name="__module.model.layers.19.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3739"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="4970" name="__module.model.layers.19.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="4971" name="__module.model.layers.19.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3740"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4972" name="__module.model.layers.19.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3742,emb.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4973" name="__module.model.layers.19.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3743,3745,3748,cos.115,cos.117"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4974" name="__module.model.layers.19.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3750,cos.119"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4975" name="__module.model.layers.19.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3752"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4976" name="Constant_101576" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4977" name="Constant_101575" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4978" name="__module.model.layers.19.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4979" name="__module.model.layers.19.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4980" name="__module.model.layers.19.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3753,3758" /> + </output> + </layer> + <layer id="4981" name="__module.model.layers.19.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="4982" name="__module.model.layers.19.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3755,3756,3760,3761" /> + </output> + </layer> + <layer id="4983" name="__module.model.layers.19.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="4984" name="__module.model.layers.19.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4985" name="Constant_101574" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4986" name="ScatterUpdate_101577" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4987" name="Constant_101580" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4988" name="Constant_101583" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4989" name="__module.model.layers.19.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3762,x2.77"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4990" name="Constant_111327_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4991" name="Constant_111327" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="4992" name="__module.model.layers.19.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3763"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4993" name="Constant_101628" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4994" name="Constant_101633" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="4995" name="__module.model.layers.19.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3757"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="4996" name="__module.model.layers.19.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3765"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4997" name="__module.model.layers.19.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3744,3746,3749,sin.115,sin.117"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4998" name="__module.model.layers.19.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3751,sin.119"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="4999" name="__module.model.layers.19.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3766"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5000" name="__module.model.layers.19.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3767"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5001" name="Constant_124489_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="5002" name="Constant_124489" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="5003" name="Constant_124482" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5004" name="Constant_124484" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5005" name="Constant_124486" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5006" name="Concat_124487" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5007" name="Broadcast_124490" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5008" name="ReadValue_121767" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.19.keypresent.19.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.19.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5009" name="Constant_119549" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5010" name="Gather_119550" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5011" name="self.model.layers.19.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="907781736" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5012" name="Convert_13379" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5013" name="self.model.layers.19.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="908306024" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5014" name="Convert_13382" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5015" name="self.model.layers.19.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5016" name="self.model.layers.19.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="908306280" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5017" name="self.model.layers.19.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5018" name="self.model.layers.19.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5019" name="__module.model.layers.19.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3713,key_states.81"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="5020" name="__module.model.layers.19.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3719,3722"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5021" name="__module.model.layers.19.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3720"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5022" name="__module.model.layers.19.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5023" name="__module.model.layers.19.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3721,k.39"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5024" name="__module.model.layers.19.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3768"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5025" name="Constant_101676" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5026" name="Constant_101675" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5027" name="__module.model.layers.19.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5028" name="__module.model.layers.19.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5029" name="__module.model.layers.19.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3769,3774" /> + </output> + </layer> + <layer id="5030" name="__module.model.layers.19.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="5031" name="__module.model.layers.19.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3771,3772,3776,3777" /> + </output> + </layer> + <layer id="5032" name="__module.model.layers.19.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5033" name="__module.model.layers.19.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5034" name="Constant_101674" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5035" name="ScatterUpdate_101677" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5036" name="Constant_101680" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5037" name="Constant_101683" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5038" name="__module.model.layers.19.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3778,x2.79"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5039" name="Constant_111328_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5040" name="Constant_111328" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5041" name="__module.model.layers.19.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3779"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5042" name="Constant_101728" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5043" name="Constant_101733" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5044" name="__module.model.layers.19.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3773"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5045" name="__module.model.layers.19.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3781"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5046" name="__module.model.layers.19.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3782"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5047" name="__module.model.layers.19.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3783"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5048" name="__module.model.layers.19.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.19.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5049" name="__module.model.layers.19.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3795,3796,3797"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5050" name="__module.model.layers.19.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5051" name="Constant_93100" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5052" name="__module.model.layers.19.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5053" name="__module.model.layers.19.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3788"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5054" name="__module.model.layers.19.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5055" name="__module.model.layers.19.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3789" /> + </output> + </layer> + <layer id="5056" name="Constant_77528" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5057" name="__module.model.layers.19.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5058" name="Constant_93849" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5059" name="Constant_93850" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5060" name="Gather_93851" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5061" name="__module.model.layers.19.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3798"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5062" name="__module.model.layers.19.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5063" name="__module.model.layers.19.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3799,hidden_states.581"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5064" name="__module.model.layers.19.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3800,3801" /> + </output> + </layer> + <layer id="5065" name="Constant_77531" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5066" name="__module.model.layers.19.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5067" name="__module.model.layers.19.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3802"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5068" name="__module.model.layers.19.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3803,key_states.83"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5069" name="Constant_124502_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="5070" name="Constant_124502" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="5071" name="Constant_124495" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5072" name="Constant_124497" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5073" name="Constant_124499" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5074" name="Concat_124500" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5075" name="Broadcast_124503" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5076" name="ReadValue_121769" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.19.valuepresent.19.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.19.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5077" name="Constant_119552" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5078" name="Gather_119553" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5079" name="self.model.layers.19.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="908306792" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5080" name="Convert_13389" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5081" name="self.model.layers.19.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="908831080" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5082" name="Convert_13392" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5083" name="self.model.layers.19.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5084" name="self.model.layers.19.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="908831336" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5085" name="self.model.layers.19.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5086" name="self.model.layers.19.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5087" name="__module.model.layers.19.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3715,value_states.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="5088" name="__module.model.layers.19.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3723"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5089" name="__module.model.layers.19.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5090" name="__module.model.layers.19.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3724"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5091" name="__module.model.layers.19.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.19.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5092" name="__module.model.layers.19.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3811,3812,3813"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5093" name="__module.model.layers.19.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5094" name="Constant_93109" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5095" name="__module.model.layers.19.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5096" name="__module.model.layers.19.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3804"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5097" name="__module.model.layers.19.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5098" name="__module.model.layers.19.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3805" /> + </output> + </layer> + <layer id="5099" name="Constant_77533" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5100" name="__module.model.layers.19.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5101" name="Constant_93859" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5102" name="Constant_93860" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5103" name="Gather_93861" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5104" name="__module.model.layers.19.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3814"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5105" name="__module.model.layers.19.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5106" name="__module.model.layers.19.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3815,hidden_states.585"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5107" name="__module.model.layers.19.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3816,3817" /> + </output> + </layer> + <layer id="5108" name="Constant_77536" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5109" name="__module.model.layers.19.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5110" name="__module.model.layers.19.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3818"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5111" name="__module.model.layers.19.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3819"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5112" name="Constant_101852" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5113" name="Constant_101853" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5114" name="Constant_101849" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5115" name="__module.model.layers.19.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5116" name="Constant_93118" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5117" name="__module.model.layers.19.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5118" name="__module.model.layers.19.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3820"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5119" name="Constant_101848" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5120" name="ScatterUpdate_101854" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5121" name="Constant_101857" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5122" name="__module.model.layers.19.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3824"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="5123" name="__module.model.layers.19.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3825,attn_output.77"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5124" name="__module.model.layers.19.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5125" name="__module.model.layers.19.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3826"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5126" name="__module.model.layers.19.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="3828"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="5127" name="__module.model.layers.19.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3829"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5128" name="self.model.layers.19.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="908831848" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5129" name="Convert_13409" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5130" name="self.model.layers.19.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="913026152" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5131" name="Convert_13412" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5132" name="self.model.layers.19.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5133" name="self.model.layers.19.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="913028200" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5134" name="self.model.layers.19.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5135" name="self.model.layers.19.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5136" name="__module.model.layers.19.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3831,3833,hidden_states.587"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5137" name="__module.model.layers.19/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3836,3838,3849,hidden_states.589,hidden_states.591"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5138" name="Constant_111332_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="913032296" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5139" name="Constant_111332" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5140" name="Constant_111331_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5141" name="Constant_111331" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5142" name="Constant_111329_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5143" name="Constant_111329" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5144" name="__module.model.layers.19.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3839"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5145" name="Constant_15244" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3840"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5146" name="__module.model.layers.19.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3841,variance.79"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5147" name="Constant_111330_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5148" name="Constant_111330" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5149" name="__module.model.layers.19.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3842"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5150" name="__module.model.layers.19.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5151" name="__module.model.layers.19.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3843"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5152" name="__module.model.layers.19.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3844,3845,hidden_states.593,hidden_states.595"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5153" name="__module.model.layers.19.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3846,3848"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5154" name="self.model.layers.19.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="913036392" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5155" name="Convert_13429" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5156" name="self.model.layers.19.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="924570728" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5157" name="Convert_13432" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5158" name="self.model.layers.19.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5159" name="self.model.layers.19.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="924576360" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5160" name="self.model.layers.19.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5161" name="self.model.layers.19.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5162" name="__module.model.layers.19.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3854,input.39"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5163" name="__module.model.layers.19.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3855"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5164" name="self.model.layers.19.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="924587624" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5165" name="Convert_13419" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5166" name="self.model.layers.19.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="936121960" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5167" name="Convert_13422" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5168" name="self.model.layers.19.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5169" name="self.model.layers.19.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="936127592" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5170" name="self.model.layers.19.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5171" name="self.model.layers.19.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5172" name="__module.model.layers.19.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3857"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5173" name="__module.model.layers.19.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3858"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5174" name="self.model.layers.19.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="936138856" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5175" name="Convert_13439" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5176" name="self.model.layers.19.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="947673192" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5177" name="Convert_13442" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5178" name="self.model.layers.19.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5179" name="self.model.layers.19.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="947675240" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5180" name="self.model.layers.19.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5181" name="self.model.layers.19.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5182" name="__module.model.layers.19.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3860,hidden_states.597"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5183" name="__module.model.layers.19/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3861,3863,3871,3882,hidden_states.599,hidden_states.601"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5184" name="Constant_111336_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="947679336" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5185" name="Constant_111336" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5186" name="Constant_111335_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5187" name="Constant_111335" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5188" name="Constant_111333_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5189" name="Constant_111333" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5190" name="__module.model.layers.20.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3872"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5191" name="Constant_15315" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="3873"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5192" name="__module.model.layers.20.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3874,variance.81"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5193" name="Constant_111334_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5194" name="Constant_111334" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5195" name="__module.model.layers.20.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3875"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5196" name="__module.model.layers.20.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5197" name="__module.model.layers.20.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3876"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5198" name="__module.model.layers.20.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3877,3878,hidden_states.603,hidden_states.605"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5199" name="__module.model.layers.20.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3879,3881,hidden_states.607"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5200" name="self.model.layers.20.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="947683432" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5201" name="Convert_13469" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5202" name="self.model.layers.20.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="951877736" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5203" name="Convert_13472" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5204" name="self.model.layers.20.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5205" name="self.model.layers.20.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="951879784" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5206" name="self.model.layers.20.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5207" name="self.model.layers.20.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5208" name="__module.model.layers.20.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3891,query_states.41"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5209" name="__module.model.layers.20.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="5210" name="Constant_93874" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5211" name="Constant_93875" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5212" name="Gather_93876" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5213" name="__module.model.layers.20.self_attn/prim::ListConstruct/Concat" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3896"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5214" name="__module.model.layers.20.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3897"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5215" name="__module.model.layers.20.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5216" name="__module.model.layers.20.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3898,q.41"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5217" name="__module.model.layers.20.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="3909"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5218" name="__module.model.layers.20.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5219" name="__module.model.layers.20.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3912,3917,inv_freq_expanded.81,inv_freq_expanded.83"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="5220" name="__module.model.layers.20.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3919"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="5221" name="__module.model.layers.20.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="5222" name="__module.model.layers.20.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3920"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5223" name="__module.model.layers.20.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3922,emb.41"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5224" name="__module.model.layers.20.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3923,3925,3928,cos.121,cos.123"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5225" name="__module.model.layers.20.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3930,cos.125"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5226" name="__module.model.layers.20.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3932"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5227" name="Constant_101862" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5228" name="Constant_101861" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5229" name="__module.model.layers.20.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5230" name="__module.model.layers.20.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5231" name="__module.model.layers.20.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3933,3938" /> + </output> + </layer> + <layer id="5232" name="__module.model.layers.20.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="5233" name="__module.model.layers.20.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3935,3936,3940,3941" /> + </output> + </layer> + <layer id="5234" name="__module.model.layers.20.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5235" name="__module.model.layers.20.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5236" name="Constant_101860" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5237" name="ScatterUpdate_101863" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5238" name="Constant_101866" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5239" name="Constant_101869" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5240" name="__module.model.layers.20.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3942,x2.81"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5241" name="Constant_111337_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5242" name="Constant_111337" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5243" name="__module.model.layers.20.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3943"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5244" name="Constant_101914" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5245" name="Constant_101919" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5246" name="__module.model.layers.20.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3937"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5247" name="__module.model.layers.20.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3945"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5248" name="__module.model.layers.20.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="3924,3926,3929,sin.121,sin.123"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5249" name="__module.model.layers.20.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3931,sin.125"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5250" name="__module.model.layers.20.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3946"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5251" name="__module.model.layers.20.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3947"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5252" name="Constant_124515_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="5253" name="Constant_124515" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="5254" name="Constant_124508" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5255" name="Constant_124510" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5256" name="Constant_124512" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5257" name="Concat_124513" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5258" name="Broadcast_124516" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5259" name="ReadValue_121775" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.20.keypresent.20.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.20.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5260" name="Constant_119555" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5261" name="Gather_119556" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5262" name="self.model.layers.20.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="951883880" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5263" name="Convert_13449" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5264" name="self.model.layers.20.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="952408168" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5265" name="Convert_13452" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5266" name="self.model.layers.20.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5267" name="self.model.layers.20.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="952408424" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5268" name="self.model.layers.20.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5269" name="self.model.layers.20.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5270" name="__module.model.layers.20.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3893,key_states.85"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="5271" name="__module.model.layers.20.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3899,3902"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5272" name="__module.model.layers.20.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3900"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5273" name="__module.model.layers.20.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5274" name="__module.model.layers.20.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3901,k.41"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5275" name="__module.model.layers.20.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3948"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5276" name="Constant_101962" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5277" name="Constant_101961" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5278" name="__module.model.layers.20.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5279" name="__module.model.layers.20.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5280" name="__module.model.layers.20.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3949,3954" /> + </output> + </layer> + <layer id="5281" name="__module.model.layers.20.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="5282" name="__module.model.layers.20.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="3951,3952,3956,3957" /> + </output> + </layer> + <layer id="5283" name="__module.model.layers.20.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5284" name="__module.model.layers.20.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5285" name="Constant_101960" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5286" name="ScatterUpdate_101963" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5287" name="Constant_101966" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5288" name="Constant_101969" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5289" name="__module.model.layers.20.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3958,x2.83"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5290" name="Constant_111338_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5291" name="Constant_111338" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5292" name="__module.model.layers.20.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3959"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5293" name="Constant_102014" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5294" name="Constant_102019" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5295" name="__module.model.layers.20.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="3953"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5296" name="__module.model.layers.20.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3961"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5297" name="__module.model.layers.20.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3962"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5298" name="__module.model.layers.20.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3963"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5299" name="__module.model.layers.20.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.20.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5300" name="__module.model.layers.20.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3975,3976,3977"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5301" name="__module.model.layers.20.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5302" name="Constant_93127" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5303" name="__module.model.layers.20.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5304" name="__module.model.layers.20.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3968"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5305" name="__module.model.layers.20.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5306" name="__module.model.layers.20.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3969" /> + </output> + </layer> + <layer id="5307" name="Constant_77540" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5308" name="__module.model.layers.20.self_attn/prim::ListConstruct/Reshape_0_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5309" name="Constant_93884" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5310" name="Constant_93885" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5311" name="Gather_93886" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5312" name="__module.model.layers.20.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3978"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5313" name="__module.model.layers.20.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5314" name="__module.model.layers.20.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3979,hidden_states.611"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5315" name="__module.model.layers.20.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3980,3981" /> + </output> + </layer> + <layer id="5316" name="Constant_77543" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5317" name="__module.model.layers.20.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5318" name="__module.model.layers.20.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3982"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5319" name="__module.model.layers.20.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3983,key_states.87"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5320" name="Constant_124528_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="5321" name="Constant_124528" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="5322" name="Constant_124521" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5323" name="Constant_124523" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5324" name="Constant_124525" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5325" name="Concat_124526" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5326" name="Broadcast_124529" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5327" name="ReadValue_121777" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.20.valuepresent.20.value" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.20.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5328" name="Constant_119558" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5329" name="Gather_119559" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5330" name="self.model.layers.20.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="952408936" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5331" name="Convert_13459" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5332" name="self.model.layers.20.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="952933224" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5333" name="Convert_13462" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5334" name="self.model.layers.20.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5335" name="self.model.layers.20.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="952933480" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5336" name="self.model.layers.20.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5337" name="self.model.layers.20.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5338" name="__module.model.layers.20.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3895,value_states.41"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="5339" name="__module.model.layers.20.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3903"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5340" name="__module.model.layers.20.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5341" name="__module.model.layers.20.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3904"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5342" name="__module.model.layers.20.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.20.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5343" name="__module.model.layers.20.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="3991,3992,3993"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5344" name="__module.model.layers.20.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5345" name="Constant_93136" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5346" name="__module.model.layers.20.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5347" name="__module.model.layers.20.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3984"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5348" name="__module.model.layers.20.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5349" name="__module.model.layers.20.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="3985" /> + </output> + </layer> + <layer id="5350" name="Constant_77545" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5351" name="__module.model.layers.20.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5352" name="Constant_93894" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5353" name="Constant_93895" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5354" name="Gather_93896" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5355" name="__module.model.layers.20.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="3994"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5356" name="__module.model.layers.20.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5357" name="__module.model.layers.20.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3995,hidden_states.615"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5358" name="__module.model.layers.20.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="3996,3997" /> + </output> + </layer> + <layer id="5359" name="Constant_77548" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5360" name="__module.model.layers.20.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5361" name="__module.model.layers.20.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="3998"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5362" name="__module.model.layers.20.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="3999"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5363" name="Constant_102138" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5364" name="Constant_102139" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5365" name="Constant_102135" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5366" name="__module.model.layers.20.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5367" name="Constant_93145" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5368" name="__module.model.layers.20.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5369" name="__module.model.layers.20.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="4000"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5370" name="Constant_102134" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5371" name="ScatterUpdate_102140" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5372" name="Constant_102143" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5373" name="__module.model.layers.20.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="4004"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="5374" name="__module.model.layers.20.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="4005,attn_output.81"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5375" name="__module.model.layers.20.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5376" name="__module.model.layers.20.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4006"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5377" name="__module.model.layers.20.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="4008"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="5378" name="__module.model.layers.20.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4009"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5379" name="self.model.layers.20.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="952933992" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5380" name="Convert_13479" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5381" name="self.model.layers.20.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="957128296" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5382" name="Convert_13482" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5383" name="self.model.layers.20.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5384" name="self.model.layers.20.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="957130344" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5385" name="self.model.layers.20.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5386" name="self.model.layers.20.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5387" name="__module.model.layers.20.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4011,4013,hidden_states.617"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5388" name="__module.model.layers.20/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4016,4018,4029,hidden_states.619,hidden_states.621"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5389" name="Constant_111342_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="957134440" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5390" name="Constant_111342" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5391" name="Constant_111341_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5392" name="Constant_111341" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5393" name="Constant_111339_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5394" name="Constant_111339" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5395" name="__module.model.layers.20.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4019"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5396" name="Constant_15995" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="4020"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5397" name="__module.model.layers.20.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4021,variance.83"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5398" name="Constant_111340_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5399" name="Constant_111340" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5400" name="__module.model.layers.20.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4022"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5401" name="__module.model.layers.20.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5402" name="__module.model.layers.20.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4023"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5403" name="__module.model.layers.20.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4024,4025,hidden_states.623,hidden_states.625"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5404" name="__module.model.layers.20.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4026,4028"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5405" name="self.model.layers.20.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="957138536" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5406" name="Convert_13499" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5407" name="self.model.layers.20.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="968672872" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5408" name="Convert_13502" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5409" name="self.model.layers.20.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5410" name="self.model.layers.20.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="968678504" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5411" name="self.model.layers.20.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5412" name="self.model.layers.20.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5413" name="__module.model.layers.20.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4034,input.41"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5414" name="__module.model.layers.20.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="4035"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5415" name="self.model.layers.20.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="968689768" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5416" name="Convert_13489" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5417" name="self.model.layers.20.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="980224104" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5418" name="Convert_13492" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5419" name="self.model.layers.20.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5420" name="self.model.layers.20.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="980229736" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5421" name="self.model.layers.20.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5422" name="self.model.layers.20.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5423" name="__module.model.layers.20.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4037"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5424" name="__module.model.layers.20.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4038"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5425" name="self.model.layers.20.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="980241000" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5426" name="Convert_13509" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5427" name="self.model.layers.20.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="991775336" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5428" name="Convert_13512" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5429" name="self.model.layers.20.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5430" name="self.model.layers.20.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="991777384" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5431" name="self.model.layers.20.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5432" name="self.model.layers.20.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5433" name="__module.model.layers.20.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4040,hidden_states.627"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5434" name="__module.model.layers.20/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4041,4043,4051,4062,hidden_states.629,hidden_states.631"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5435" name="Constant_111345_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5436" name="Constant_111345" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5437" name="Constant_111343_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5438" name="Constant_111343" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5439" name="__module.model.layers.21.input_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4052"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5440" name="Constant_16066" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="4053"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5441" name="__module.model.layers.21.input_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4054,variance.85"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5442" name="Constant_111344_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5443" name="Constant_111344" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5444" name="__module.model.layers.21.input_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4055"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5445" name="__module.model.layers.21.input_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5446" name="__module.model.layers.21.input_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4056"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5447" name="__module.model.layers.21.input_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4057,4058,hidden_states.633,hidden_states.635"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5448" name="__module.model.layers.21.input_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4059,4061,hidden_states.637"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5449" name="self.model.layers.21.self_attn.v_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="991781480" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5450" name="Convert_13519" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5451" name="self.model.layers.21.self_attn.v_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="992305768" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5452" name="Convert_13522" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5453" name="self.model.layers.21.self_attn.v_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5454" name="self.model.layers.21.self_attn.v_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="992306024" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5455" name="self.model.layers.21.self_attn.v_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5456" name="self.model.layers.21.self_attn.v_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5457" name="__module.model.layers.21.self_attn.v_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4075,value_states"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="5458" name="__module.model.layers.21.self_attn/aten::size/ShapeOf" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="5459" name="Constant_93909" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="69840684" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5460" name="Constant_93910" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5461" name="Gather_93911" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>3</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5462" name="__module.model.layers.21.self_attn/prim::ListConstruct/Concat_1" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="4079,4082"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5463" name="__module.model.layers.21.self_attn/aten::view/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4083"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5464" name="__module.model.layers.21.self_attn/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5465" name="__module.model.layers.21.self_attn/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4084"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5466" name="__module.model.layers.21.self_attn/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.21.value"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5468" name="Constant_124541_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="" offset="0" size="2" /> + <output> + <port id="0" precision="FP16" /> + </output> + </layer> + <layer id="5469" name="Constant_124541" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16" /> + </input> + <output> + <port id="1" precision="FP32" /> + </output> + </layer> + <layer id="5470" name="Constant_124534" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="10" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5471" name="Constant_124536" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5472" name="Constant_124538" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="18" size="8" /> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5473" name="Concat_124539" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5474" name="Broadcast_124542" type="Broadcast" version="opset3"> + <data mode="numpy" /> + <input> + <port id="0" precision="FP32" /> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5475" name="ReadValue_121779" type="ReadValue" version="opset6"> + <data variable_id="past_key_values.21.keypresent.21.key" variable_type="f32" variable_shape="?,4,?,64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>0</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="past_key_values.21.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5476" name="Constant_119561" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5477" name="Gather_119562" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>-1</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5478" name="self.model.layers.21.self_attn.k_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="256, 2048" offset="992306536" size="524288" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5479" name="Convert_13529" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5480" name="self.model.layers.21.self_attn.k_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="256, 1" offset="992830824" size="256" /> + <output> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5481" name="Convert_13532" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5482" name="self.model.layers.21.self_attn.k_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5483" name="self.model.layers.21.self_attn.k_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="256, 1" offset="992831080" size="512" /> + <output> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5484" name="self.model.layers.21.self_attn.k_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>256</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5485" name="self.model.layers.21.self_attn.k_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5486" name="__module.model.layers.21.self_attn.k_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>256</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4073,key_states.89"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + </output> + </layer> + <layer id="5487" name="__module.model.layers.21.self_attn/aten::view/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>256</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4080"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5488" name="__module.model.layers.21.self_attn/aten::transpose/ScatterElementsUpdate_1" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5489" name="__module.model.layers.21.self_attn/aten::transpose/Transpose_1" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>4</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4081,k"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5490" name="__module.model.layers.21.self_attn.rotary_emb/aten::to/Convert_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 32, 1" offset="69840724" size="64" /> + <output> + <port id="0" precision="FP16" names="4089"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5491" name="__module.model.layers.21.self_attn.rotary_emb/aten::to/Convert" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5492" name="__module.model.layers.21.self_attn.rotary_emb/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>32</dim> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4092,4097,inv_freq_expanded,inv_freq_expanded.85"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="5493" name="__module.model.layers.21.self_attn.rotary_emb/aten::matmul/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4099"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="5494" name="__module.model.layers.21.self_attn.rotary_emb/aten::transpose/ScatterElementsUpdate" type="Const" version="opset1"> + <data element_type="i32" shape="3" offset="69840796" size="12" /> + <output> + <port id="0" precision="I32"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="5495" name="__module.model.layers.21.self_attn.rotary_emb/aten::transpose/Transpose" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I32"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4100"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5496" name="__module.model.layers.21.self_attn.rotary_emb/aten::cat/Concat" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4102,emb"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5497" name="__module.model.layers.21.self_attn.rotary_emb/aten::cos/Cos" type="Cos" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="4103,4105,4108,cos.127,cos.129"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5498" name="__module.model.layers.21.self_attn/aten::unsqueeze/Unsqueeze" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="4110,cos"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5499" name="__module.model.layers.21.self_attn/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4128"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5500" name="Constant_102148" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5501" name="Constant_102147" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5502" name="__module.model.layers.21.self_attn/aten::size/ShapeOf_3" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5503" name="__module.model.layers.21.self_attn/aten::size/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5504" name="__module.model.layers.21.self_attn/aten::size/Gather_3" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="4129,4134" /> + </output> + </layer> + <layer id="5505" name="__module.model.layers.21.self_attn/aten::floor_divide/Divide_1" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="5506" name="__module.model.layers.21.self_attn/aten::floor_divide/Floor_1" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="4131,4132,4136,4137" /> + </output> + </layer> + <layer id="5507" name="__module.model.layers.21.self_attn/aten::slice/Constant_1" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5508" name="__module.model.layers.21.self_attn/aten::slice/Unsqueeze_5" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5509" name="Constant_102146" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5510" name="ScatterUpdate_102149" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5511" name="Constant_102152" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5512" name="Constant_102155" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5513" name="__module.model.layers.21.self_attn/aten::slice/Slice" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="4138,x2"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5514" name="Constant_111347_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5515" name="Constant_111347" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5516" name="__module.model.layers.21.self_attn/aten::neg/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4139"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5517" name="Constant_102200" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5518" name="Constant_102205" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5519" name="__module.model.layers.21.self_attn/aten::slice/Slice_1" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="4133"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5520" name="__module.model.layers.21.self_attn/aten::cat/Concat_1" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4141"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5521" name="__module.model.layers.21.self_attn.rotary_emb/aten::sin/Sin" type="Sin" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="4104,4106,4109,sin.127,sin.129"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5522" name="__module.model.layers.21.self_attn/aten::unsqueeze/Unsqueeze_1" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="4111,sin"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5523" name="__module.model.layers.21.self_attn/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4142"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5524" name="__module.model.layers.21.self_attn/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4143"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5525" name="__module.model.layers.21.self_attn/aten::cat/Concat_2" type="Concat" version="opset1"> + <data axis="-2" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="present.21.key"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5555" name="Constant_111356_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="992831592" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5556" name="Constant_111356" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5557" name="self.model.layers.21.self_attn.q_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="992835688" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5558" name="Convert_13539" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5559" name="self.model.layers.21.self_attn.q_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="997029992" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5560" name="Convert_13542" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5561" name="self.model.layers.21.self_attn.q_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5562" name="self.model.layers.21.self_attn.q_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="997032040" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5563" name="self.model.layers.21.self_attn.q_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5564" name="self.model.layers.21.self_attn.q_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5565" name="__module.model.layers.21.self_attn.q_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4071,query_states"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5566" name="__module.model.layers.21.self_attn/prim::ListConstruct/Concat_2" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="4076"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5567" name="__module.model.layers.21.self_attn/aten::view/Reshape_2" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4077"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5568" name="__module.model.layers.21.self_attn/aten::transpose/ScatterElementsUpdate_2" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5569" name="__module.model.layers.21.self_attn/aten::transpose/Transpose_2" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4078,q"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5570" name="__module.model.layers.21.self_attn/aten::mul/Multiply_2" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4112"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5571" name="Constant_102248" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5572" name="Constant_102247" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5573" name="__module.model.layers.21.self_attn/aten::size/ShapeOf_5" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5574" name="__module.model.layers.21.self_attn/aten::size/Constant_5" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5575" name="__module.model.layers.21.self_attn/aten::size/Gather_5" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="4113,4118" /> + </output> + </layer> + <layer id="5576" name="__module.model.layers.21.self_attn/aten::floor_divide/Divide_3" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" /> + </output> + </layer> + <layer id="5577" name="__module.model.layers.21.self_attn/aten::floor_divide/Floor_3" type="Floor" version="opset1"> + <input> + <port id="0" precision="I64" /> + </input> + <output> + <port id="1" precision="I64" names="4115,4116,4120,4121" /> + </output> + </layer> + <layer id="5578" name="__module.model.layers.21.self_attn/aten::slice/Constant_3" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5579" name="__module.model.layers.21.self_attn/aten::slice/Unsqueeze_13" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I32" /> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5580" name="Constant_102246" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5581" name="ScatterUpdate_102249" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5582" name="Constant_102252" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840856" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5583" name="Constant_102255" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5584" name="__module.model.layers.21.self_attn/aten::slice/Slice_2" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="4122,x2.85"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5585" name="Constant_111348_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1, 1" offset="69840920" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5586" name="Constant_111348" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5587" name="__module.model.layers.21.self_attn/aten::neg/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4123"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5588" name="Constant_102300" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5589" name="Constant_102305" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5590" name="__module.model.layers.21.self_attn/aten::slice/Slice_3" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="4117"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </output> + </layer> + <layer id="5591" name="__module.model.layers.21.self_attn/aten::cat/Concat_3" type="Concat" version="opset1"> + <data axis="-1" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>32</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4125"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5592" name="__module.model.layers.21.self_attn/aten::mul/Multiply_3" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4126"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5593" name="__module.model.layers.21.self_attn/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4127"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5594" name="__module.model.layers.21.self_attn/aten::unsqueeze/Unsqueeze_2" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="4155,4156,4157"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5595" name="__module.model.layers.21.self_attn/aten::size/ShapeOf_6" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5596" name="Constant_93154" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5597" name="__module.model.layers.21.self_attn/aten::size/Constant_6" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5598" name="__module.model.layers.21.self_attn/aten::size/Gather_6" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="4148"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5599" name="__module.model.layers.21.self_attn/aten::size/Constant_7" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5600" name="__module.model.layers.21.self_attn/aten::size/Gather_7" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="4149" /> + </output> + </layer> + <layer id="5601" name="Constant_77552" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5602" name="__module.model.layers.21.self_attn/prim::ListConstruct/Reshape_0_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5603" name="Constant_93919" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5604" name="Constant_93920" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5605" name="Gather_93921" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5606" name="__module.model.layers.21.self_attn/prim::ListConstruct/Concat_3" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="4158"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5607" name="__module.model.layers.21.self_attn/aten::expand/Abs" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5608" name="__module.model.layers.21.self_attn/aten::expand/Broadcast" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4159,hidden_states.641"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5609" name="__module.model.layers.21.self_attn/aten::mul/Multiply_4" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="4160,4161" /> + </output> + </layer> + <layer id="5610" name="Constant_77555" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5611" name="__module.model.layers.21.self_attn/prim::ListConstruct/Reshape_0_4" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5612" name="__module.model.layers.21.self_attn/prim::ListConstruct/Concat_4" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="4162"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5613" name="__module.model.layers.21.self_attn/aten::reshape/Reshape" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4163,key_states"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5614" name="__module.model.layers.21.self_attn/aten::unsqueeze/Unsqueeze_3" type="Unsqueeze" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="FP32" names="4171,4172,4173"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5615" name="__module.model.layers.21.self_attn/aten::size/ShapeOf_10" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5616" name="Constant_93163" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5617" name="__module.model.layers.21.self_attn/aten::size/Constant_10" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5618" name="__module.model.layers.21.self_attn/aten::size/Gather_10" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="4164"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5619" name="__module.model.layers.21.self_attn/aten::size/Constant_11" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5620" name="__module.model.layers.21.self_attn/aten::size/Gather_11" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64" /> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="4165" /> + </output> + </layer> + <layer id="5621" name="Constant_77557" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5622" name="__module.model.layers.21.self_attn/prim::ListConstruct/Reshape_0_5" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5623" name="Constant_93929" type="Const" version="opset1"> + <data element_type="i64" shape="2" offset="70365986" size="16" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5624" name="Constant_93930" type="Const" version="opset1"> + <data element_type="i64" shape="" offset="2" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64" /> + </output> + </layer> + <layer id="5625" name="Gather_93931" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>2</dim> + </port> + <port id="2" precision="I64" /> + </input> + <output> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </output> + </layer> + <layer id="5626" name="__module.model.layers.21.self_attn/prim::ListConstruct/Concat_5" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="4" precision="I64" names="4174"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5627" name="__module.model.layers.21.self_attn/aten::expand/Abs_1" type="Abs" version="opset1"> + <input> + <port id="0" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </output> + </layer> + <layer id="5628" name="__module.model.layers.21.self_attn/aten::expand/Broadcast_1" type="Broadcast" version="opset3"> + <data mode="bidirectional" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>5</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4175,hidden_states.645"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5629" name="__module.model.layers.21.self_attn/aten::mul/Multiply_5" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64" /> + </input> + <output> + <port id="2" precision="I64" names="4176,4177" /> + </output> + </layer> + <layer id="5630" name="Constant_77560" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5631" name="__module.model.layers.21.self_attn/prim::ListConstruct/Reshape_0_6" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="I64" /> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5632" name="__module.model.layers.21.self_attn/prim::ListConstruct/Concat_6" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>2</dim> + </port> + </input> + <output> + <port id="3" precision="I64" names="4178"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5633" name="__module.model.layers.21.self_attn/aten::reshape/Reshape_1" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4179"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5634" name="Constant_102424" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5635" name="Constant_102425" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840808" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5636" name="Constant_102421" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840840" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5637" name="__module.model.layers.21.self_attn/aten::size/ShapeOf_14" type="ShapeOf" version="opset3"> + <data output_type="i64" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5638" name="Constant_93172" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="69840848" size="8" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5639" name="__module.model.layers.21.self_attn/aten::size/Constant_14" type="Const" version="opset1"> + <data element_type="i32" shape="" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32" /> + </output> + </layer> + <layer id="5640" name="__module.model.layers.21.self_attn/aten::size/Gather_14" type="Gather" version="opset8"> + <data batch_dims="0" /> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I32" /> + </input> + <output> + <port id="3" precision="I64" names="4180"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5641" name="Constant_102420" type="Const" version="opset1"> + <data element_type="i32" shape="1" offset="65636122" size="4" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I32"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5642" name="ScatterUpdate_102426" type="ScatterUpdate" version="opset3"> + <input> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + <port id="2" precision="I64"> + <dim>1</dim> + </port> + <port id="3" precision="I32"> + <dim>1</dim> + </port> + </input> + <output> + <port id="4" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5643" name="Constant_102429" type="Const" version="opset1"> + <data element_type="i64" shape="4" offset="69840888" size="32" /> + <rt_info> + <attribute name="precise" version="0" /> + </rt_info> + <output> + <port id="0" precision="I64"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5644" name="__module.model.layers.21.self_attn/aten::slice/Slice_15" type="StridedSlice" version="opset1"> + <data begin_mask="1, 1, 1, 0" end_mask="1, 1, 1, 0" new_axis_mask="" shrink_axis_mask="" ellipsis_mask="" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + <port id="1" precision="I64"> + <dim>4</dim> + </port> + <port id="2" precision="I64"> + <dim>4</dim> + </port> + <port id="3" precision="I64"> + <dim>4</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="4184"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </output> + </layer> + <layer id="5645" name="__module.model.layers.21.self_attn/aten::scaled_dot_product_attention/ScaledDotProductAttention" type="ScaledDotProductAttention" version="opset13"> + <data causal="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="2" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="3" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + <dim>-1</dim> + </port> + </input> + <output> + <port id="4" precision="FP32" names="4185,attn_output.85"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5646" name="__module.model.layers.21.self_attn/aten::transpose/ScatterElementsUpdate_3" type="Const" version="opset1"> + <data element_type="i32" shape="4" offset="69840708" size="16" /> + <output> + <port id="0" precision="I32"> + <dim>4</dim> + </port> + </output> + </layer> + <layer id="5647" name="__module.model.layers.21.self_attn/aten::transpose/Transpose_3" type="Transpose" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>32</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + <port id="1" precision="I32"> + <dim>4</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4186"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5648" name="__module.model.layers.21.self_attn/prim::ListConstruct/Concat_7" type="Concat" version="opset1"> + <data axis="0" /> + <input> + <port id="0" precision="I64"> + <dim>2</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="I64" names="4188"> + <dim>3</dim> + </port> + </output> + </layer> + <layer id="5649" name="__module.model.layers.21.self_attn/aten::view/Reshape_3" type="Reshape" version="opset1"> + <data special_zero="false" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32</dim> + <dim>64</dim> + </port> + <port id="1" precision="I64"> + <dim>3</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4189"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5650" name="self.model.layers.21.self_attn.o_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 2048" offset="997036136" size="4194304" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5651" name="Convert_13549" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5652" name="self.model.layers.21.self_attn.o_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="1001230440" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5653" name="Convert_13552" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5654" name="self.model.layers.21.self_attn.o_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5655" name="self.model.layers.21.self_attn.o_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="1001232488" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5656" name="self.model.layers.21.self_attn.o_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5657" name="self.model.layers.21.self_attn.o_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5658" name="__module.model.layers.21.self_attn.o_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4191,4193,hidden_states.647"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5659" name="__module.model.layers.21/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4196,4198,4209,hidden_states.649,hidden_states.651"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5660" name="Constant_111352_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 2048" offset="1001236584" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5661" name="Constant_111352" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5662" name="Constant_111351_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5663" name="Constant_111351" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5664" name="Constant_111349_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5665" name="Constant_111349" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5666" name="__module.model.layers.21.post_attention_layernorm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4199"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5667" name="Constant_16746" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="4200"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5668" name="__module.model.layers.21.post_attention_layernorm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4201,variance.87"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5669" name="Constant_111350_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5670" name="Constant_111350" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5671" name="__module.model.layers.21.post_attention_layernorm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4202"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5672" name="__module.model.layers.21.post_attention_layernorm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5673" name="__module.model.layers.21.post_attention_layernorm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4203"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5674" name="__module.model.layers.21.post_attention_layernorm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4204,4205,hidden_states.653,hidden_states.655"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5675" name="__module.model.layers.21.post_attention_layernorm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4206,4208"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5676" name="self.model.layers.21.mlp.gate_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="1001240680" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5677" name="Convert_13569" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5678" name="self.model.layers.21.mlp.gate_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="1012775016" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5679" name="Convert_13572" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5680" name="self.model.layers.21.mlp.gate_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5681" name="self.model.layers.21.mlp.gate_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="1012780648" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5682" name="self.model.layers.21.mlp.gate_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5683" name="self.model.layers.21.mlp.gate_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5684" name="__module.model.layers.21.mlp.gate_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4214,input"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5685" name="__module.model.layers.21.mlp.act_fn/aten::silu/Swish" type="Swish" version="opset4"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32" names="4215"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5686" name="self.model.layers.21.mlp.up_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 2048" offset="1012791912" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5687" name="Convert_13559" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5688" name="self.model.layers.21.mlp.up_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="5632, 1" offset="1024326248" size="5632" /> + <output> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5689" name="Convert_13562" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5690" name="self.model.layers.21.mlp.up_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5691" name="self.model.layers.21.mlp.up_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="5632, 1" offset="1024331880" size="11264" /> + <output> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5692" name="self.model.layers.21.mlp.up_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>5632</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5693" name="self.model.layers.21.mlp.up_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5694" name="__module.model.layers.21.mlp.up_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>5632</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4217"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5695" name="__module.model.layers.21.mlp/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4218"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5696" name="self.model.layers.21.mlp.down_proj.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 5632" offset="1024343144" size="11534336" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5697" name="Convert_13579" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5698" name="self.model.layers.21.mlp.down_proj.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="2048, 1" offset="1035877480" size="2048" /> + <output> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5699" name="Convert_13582" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5700" name="self.model.layers.21.mlp.down_proj.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5701" name="self.model.layers.21.mlp.down_proj.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="2048, 1" offset="1035879528" size="4096" /> + <output> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5702" name="self.model.layers.21.mlp.down_proj.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP16"> + <dim>2048</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5703" name="self.model.layers.21.mlp.down_proj.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </output> + </layer> + <layer id="5704" name="__module.model.layers.21.mlp.down_proj/aten::linear/MatMul" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>5632</dim> + </port> + <port id="1" precision="FP32"> + <dim>2048</dim> + <dim>5632</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4220,hidden_states.657"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5705" name="__module.model.layers.21/aten::add/Add_1" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4221,4223,4227,hidden_states.659,hidden_states.661"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5706" name="Constant_111355_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640222" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5707" name="Constant_111355" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5708" name="Constant_111353_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640224" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5709" name="Constant_111353" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5710" name="__module.model.norm/aten::pow/Power" type="Power" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4228"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5711" name="Constant_16813" type="Const" version="opset1"> + <data element_type="i64" shape="1" offset="65640226" size="8" /> + <output> + <port id="0" precision="I64" names="4229"> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5712" name="__module.model.norm/aten::mean/ReduceMean" type="ReduceMean" version="opset1"> + <data keep_dims="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="I64"> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4230,variance"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5713" name="Constant_111354_compressed" type="Const" version="opset1"> + <data element_type="f16" shape="1, 1, 1" offset="65640234" size="2" /> + <output> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5714" name="Constant_111354" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5715" name="__module.model.norm/aten::add/Add" type="Add" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4231"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5716" name="__module.model.norm/aten::rsqrt/Sqrt" type="Sqrt" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5717" name="__module.model.norm/aten::rsqrt/Divide" type="Divide" version="opset1"> + <data auto_broadcast="numpy" m_pythondiv="true" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>1</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4232"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5718" name="__module.model.norm/aten::mul/Multiply" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4233,4234,hidden_states,hidden_states.663"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5719" name="__module.model.norm/aten::mul/Multiply_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP32"> + <dim>1</dim> + <dim>1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="4235,74"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5720" name="self.lm_head.weight_compressed" type="Const" version="opset1"> + <data element_type="u8" shape="32000, 2048" offset="1035883624" size="65536000" /> + <output> + <port id="0" precision="U8"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5721" name="Convert_13589" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5722" name="self.lm_head.weight_compressed/zero_point" type="Const" version="opset1"> + <data element_type="u8" shape="32000, 1" offset="1101419624" size="32000" /> + <output> + <port id="0" precision="U8"> + <dim>32000</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5723" name="Convert_13592" type="Convert" version="opset1"> + <data destination_type="f16" /> + <input> + <port id="0" precision="U8"> + <dim>32000</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="1" precision="FP16"> + <dim>32000</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5724" name="self.lm_head.weight_compressed/zero_point/subtract" type="Subtract" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>32000</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5725" name="self.lm_head.weight_compressed/scale" type="Const" version="opset1"> + <data element_type="f16" shape="32000, 1" offset="1101451624" size="64000" /> + <output> + <port id="0" precision="FP16"> + <dim>32000</dim> + <dim>1</dim> + </port> + </output> + </layer> + <layer id="5726" name="self.lm_head.weight_compressed/fq_weights_1" type="Multiply" version="opset1"> + <data auto_broadcast="numpy" /> + <input> + <port id="0" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP16"> + <dim>32000</dim> + <dim>1</dim> + </port> + </input> + <output> + <port id="2" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5727" name="self.lm_head.weight" type="Convert" version="opset1"> + <data destination_type="f32" /> + <rt_info> + <attribute name="decompression" version="0" /> + </rt_info> + <input> + <port id="0" precision="FP16"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </output> + </layer> + <layer id="5728" name="aten::to/Convert" type="MatMul" version="opset1"> + <data transpose_a="false" transpose_b="true" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>2048</dim> + </port> + <port id="1" precision="FP32"> + <dim>32000</dim> + <dim>2048</dim> + </port> + </input> + <output> + <port id="2" precision="FP32" names="logits"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32000</dim> + </port> + </output> + </layer> + <layer id="5554" name="Assign_121724" type="Assign" version="opset6"> + <data variable_id="past_key_values.0.keypresent.0.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5553" name="Assign_121726" type="Assign" version="opset6"> + <data variable_id="past_key_values.0.valuepresent.0.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5552" name="Assign_121728" type="Assign" version="opset6"> + <data variable_id="past_key_values.1.keypresent.1.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5551" name="Assign_121730" type="Assign" version="opset6"> + <data variable_id="past_key_values.1.valuepresent.1.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5550" name="Assign_121732" type="Assign" version="opset6"> + <data variable_id="past_key_values.10.keypresent.10.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5549" name="Assign_121734" type="Assign" version="opset6"> + <data variable_id="past_key_values.10.valuepresent.10.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5548" name="Assign_121736" type="Assign" version="opset6"> + <data variable_id="past_key_values.11.keypresent.11.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5547" name="Assign_121738" type="Assign" version="opset6"> + <data variable_id="past_key_values.11.valuepresent.11.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5546" name="Assign_121740" type="Assign" version="opset6"> + <data variable_id="past_key_values.12.keypresent.12.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5545" name="Assign_121742" type="Assign" version="opset6"> + <data variable_id="past_key_values.12.valuepresent.12.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5544" name="Assign_121744" type="Assign" version="opset6"> + <data variable_id="past_key_values.13.keypresent.13.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5543" name="Assign_121746" type="Assign" version="opset6"> + <data variable_id="past_key_values.13.valuepresent.13.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5542" name="Assign_121748" type="Assign" version="opset6"> + <data variable_id="past_key_values.14.keypresent.14.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5541" name="Assign_121750" type="Assign" version="opset6"> + <data variable_id="past_key_values.14.valuepresent.14.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5540" name="Assign_121752" type="Assign" version="opset6"> + <data variable_id="past_key_values.15.keypresent.15.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5539" name="Assign_121754" type="Assign" version="opset6"> + <data variable_id="past_key_values.15.valuepresent.15.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5538" name="Assign_121756" type="Assign" version="opset6"> + <data variable_id="past_key_values.16.keypresent.16.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5537" name="Assign_121758" type="Assign" version="opset6"> + <data variable_id="past_key_values.16.valuepresent.16.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5536" name="Assign_121760" type="Assign" version="opset6"> + <data variable_id="past_key_values.17.keypresent.17.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5535" name="Assign_121762" type="Assign" version="opset6"> + <data variable_id="past_key_values.17.valuepresent.17.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5534" name="Assign_121764" type="Assign" version="opset6"> + <data variable_id="past_key_values.18.keypresent.18.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5533" name="Assign_121766" type="Assign" version="opset6"> + <data variable_id="past_key_values.18.valuepresent.18.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5532" name="Assign_121768" type="Assign" version="opset6"> + <data variable_id="past_key_values.19.keypresent.19.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5531" name="Assign_121770" type="Assign" version="opset6"> + <data variable_id="past_key_values.19.valuepresent.19.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5530" name="Assign_121772" type="Assign" version="opset6"> + <data variable_id="past_key_values.2.keypresent.2.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5529" name="Assign_121774" type="Assign" version="opset6"> + <data variable_id="past_key_values.2.valuepresent.2.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5528" name="Assign_121776" type="Assign" version="opset6"> + <data variable_id="past_key_values.20.keypresent.20.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5527" name="Assign_121778" type="Assign" version="opset6"> + <data variable_id="past_key_values.20.valuepresent.20.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5526" name="Assign_121780" type="Assign" version="opset6"> + <data variable_id="past_key_values.21.keypresent.21.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5467" name="Assign_121782" type="Assign" version="opset6"> + <data variable_id="past_key_values.21.valuepresent.21.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2512" name="Assign_121784" type="Assign" version="opset6"> + <data variable_id="past_key_values.3.keypresent.3.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2511" name="Assign_121786" type="Assign" version="opset6"> + <data variable_id="past_key_values.3.valuepresent.3.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2510" name="Assign_121788" type="Assign" version="opset6"> + <data variable_id="past_key_values.4.keypresent.4.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2509" name="Assign_121790" type="Assign" version="opset6"> + <data variable_id="past_key_values.4.valuepresent.4.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2508" name="Assign_121792" type="Assign" version="opset6"> + <data variable_id="past_key_values.5.keypresent.5.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2507" name="Assign_121794" type="Assign" version="opset6"> + <data variable_id="past_key_values.5.valuepresent.5.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2506" name="Assign_121796" type="Assign" version="opset6"> + <data variable_id="past_key_values.6.keypresent.6.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2505" name="Assign_121798" type="Assign" version="opset6"> + <data variable_id="past_key_values.6.valuepresent.6.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2504" name="Assign_121800" type="Assign" version="opset6"> + <data variable_id="past_key_values.7.keypresent.7.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2503" name="Assign_121802" type="Assign" version="opset6"> + <data variable_id="past_key_values.7.valuepresent.7.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2502" name="Assign_121804" type="Assign" version="opset6"> + <data variable_id="past_key_values.8.keypresent.8.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2501" name="Assign_121806" type="Assign" version="opset6"> + <data variable_id="past_key_values.8.valuepresent.8.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2500" name="Assign_121808" type="Assign" version="opset6"> + <data variable_id="past_key_values.9.keypresent.9.key" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="2441" name="Assign_121810" type="Assign" version="opset6"> + <data variable_id="past_key_values.9.valuepresent.9.value" /> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </input> + <output> + <port id="1" precision="FP32"> + <dim>-1</dim> + <dim>4</dim> + <dim>-1</dim> + <dim>64</dim> + </port> + </output> + </layer> + <layer id="5729" name="Result_51270" type="Result" version="opset1"> + <input> + <port id="0" precision="FP32"> + <dim>-1</dim> + <dim>-1</dim> + <dim>32000</dim> + </port> + </input> + </layer> + </layers> + <edges> + <edge from-layer="0" from-port="0" to-layer="1482" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="5477" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="2751" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="2522" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="2451" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="2303" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="2235" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="2052" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="1984" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="1801" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="1733" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="1550" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="3002" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="1299" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="1231" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="1048" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="980" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="797" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="729" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="546" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="478" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="195" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="123" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="5261" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="3070" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="3253" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="3321" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="3504" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="3572" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="3755" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="3823" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="4006" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="4074" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="4257" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="4325" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="17" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="4508" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="4576" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="4759" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="4827" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="2819" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="5010" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="5078" to-port="1" /> + <edge from-layer="0" from-port="0" to-layer="5329" to-port="1" /> + <edge from-layer="1" from-port="0" to-layer="78" to-port="0" /> + <edge from-layer="1" from-port="0" to-layer="68" to-port="0" /> + <edge from-layer="2" from-port="0" to-layer="312" to-port="0" /> + <edge from-layer="2" from-port="0" to-layer="238" to-port="0" /> + <edge from-layer="3" from-port="0" to-layer="6" to-port="0" /> + <edge from-layer="3" from-port="0" to-layer="28" to-port="0" /> + <edge from-layer="4" from-port="0" to-layer="5" to-port="0" /> + <edge from-layer="5" from-port="1" to-layer="14" to-port="0" /> + <edge from-layer="6" from-port="1" to-layer="9" to-port="0" /> + <edge from-layer="7" from-port="0" to-layer="9" to-port="1" /> + <edge from-layer="8" from-port="0" to-layer="9" to-port="2" /> + <edge from-layer="9" from-port="3" to-layer="3751" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="3568" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="1797" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="2747" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="119" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="5473" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="3819" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="5325" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="5257" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="4002" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="5074" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="4070" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="1729" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="5006" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="2815" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="4253" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="4321" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="4823" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="4504" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="4755" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="4572" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="2998" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="191" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="474" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="542" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="725" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="793" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="976" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="1044" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="1227" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="1980" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="1295" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="1478" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="2048" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="3066" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="2231" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="1546" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="2299" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="3249" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="3317" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="2447" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="2518" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="3500" to-port="0" /> + <edge from-layer="9" from-port="3" to-layer="13" to-port="0" /> + <edge from-layer="10" from-port="0" to-layer="13" to-port="1" /> + <edge from-layer="11" from-port="0" to-layer="13" to-port="2" /> + <edge from-layer="12" from-port="0" to-layer="13" to-port="3" /> + <edge from-layer="13" from-port="4" to-layer="14" to-port="1" /> + <edge from-layer="14" from-port="2" to-layer="15" to-port="0" /> + <edge from-layer="15" from-port="1" to-layer="17" to-port="0" /> + <edge from-layer="16" from-port="0" to-layer="17" to-port="2" /> + <edge from-layer="17" from-port="3" to-layer="2440" to-port="0" /> + <edge from-layer="18" from-port="0" to-layer="19" to-port="0" /> + <edge from-layer="19" from-port="1" to-layer="2422" to-port="0" /> + <edge from-layer="20" from-port="0" to-layer="21" to-port="0" /> + <edge from-layer="21" from-port="1" to-layer="24" to-port="0" /> + <edge from-layer="22" from-port="0" to-layer="23" to-port="0" /> + <edge from-layer="23" from-port="1" to-layer="24" to-port="1" /> + <edge from-layer="24" from-port="2" to-layer="26" to-port="0" /> + <edge from-layer="25" from-port="0" to-layer="26" to-port="1" /> + <edge from-layer="26" from-port="2" to-layer="27" to-port="0" /> + <edge from-layer="27" from-port="1" to-layer="30" to-port="0" /> + <edge from-layer="28" from-port="1" to-layer="30" to-port="1" /> + <edge from-layer="29" from-port="0" to-layer="30" to-port="2" /> + <edge from-layer="30" from-port="3" to-layer="233" to-port="0" /> + <edge from-layer="30" from-port="3" to-layer="354" to-port="0" /> + <edge from-layer="30" from-port="3" to-layer="37" to-port="0" /> + <edge from-layer="30" from-port="3" to-layer="45" to-port="0" /> + <edge from-layer="31" from-port="0" to-layer="32" to-port="0" /> + <edge from-layer="32" from-port="1" to-layer="46" to-port="0" /> + <edge from-layer="33" from-port="0" to-layer="34" to-port="0" /> + <edge from-layer="34" from-port="1" to-layer="44" to-port="0" /> + <edge from-layer="35" from-port="0" to-layer="36" to-port="0" /> + <edge from-layer="36" from-port="1" to-layer="37" to-port="1" /> + <edge from-layer="37" from-port="2" to-layer="39" to-port="0" /> + <edge from-layer="38" from-port="0" to-layer="39" to-port="1" /> + <edge from-layer="39" from-port="2" to-layer="42" to-port="0" /> + <edge from-layer="40" from-port="0" to-layer="41" to-port="0" /> + <edge from-layer="41" from-port="1" to-layer="42" to-port="1" /> + <edge from-layer="42" from-port="2" to-layer="43" to-port="0" /> + <edge from-layer="43" from-port="1" to-layer="44" to-port="1" /> + <edge from-layer="44" from-port="2" to-layer="45" to-port="1" /> + <edge from-layer="45" from-port="2" to-layer="46" to-port="1" /> + <edge from-layer="46" from-port="2" to-layer="56" to-port="0" /> + <edge from-layer="46" from-port="2" to-layer="55" to-port="0" /> + <edge from-layer="46" from-port="2" to-layer="132" to-port="0" /> + <edge from-layer="46" from-port="2" to-layer="204" to-port="0" /> + <edge from-layer="47" from-port="0" to-layer="48" to-port="0" /> + <edge from-layer="48" from-port="1" to-layer="51" to-port="0" /> + <edge from-layer="49" from-port="0" to-layer="50" to-port="0" /> + <edge from-layer="50" from-port="1" to-layer="51" to-port="1" /> + <edge from-layer="51" from-port="2" to-layer="53" to-port="0" /> + <edge from-layer="52" from-port="0" to-layer="53" to-port="1" /> + <edge from-layer="53" from-port="2" to-layer="54" to-port="0" /> + <edge from-layer="54" from-port="1" to-layer="55" to-port="1" /> + <edge from-layer="55" from-port="2" to-layer="63" to-port="0" /> + <edge from-layer="56" from-port="1" to-layer="59" to-port="0" /> + <edge from-layer="57" from-port="0" to-layer="59" to-port="1" /> + <edge from-layer="58" from-port="0" to-layer="59" to-port="2" /> + <edge from-layer="59" from-port="3" to-layer="134" to-port="0" /> + <edge from-layer="59" from-port="3" to-layer="62" to-port="0" /> + <edge from-layer="59" from-port="3" to-layer="343" to-port="0" /> + <edge from-layer="60" from-port="0" to-layer="1434" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="4711" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="4460" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="1936" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="4962" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="4209" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="5213" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="430" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="3958" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="1685" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="5566" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="3707" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="681" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="2703" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="2534" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="2954" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="932" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="3456" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="3205" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="1183" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="2187" to-port="1" /> + <edge from-layer="60" from-port="0" to-layer="62" to-port="1" /> + <edge from-layer="61" from-port="0" to-layer="2187" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="430" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="134" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="4016" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="2761" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="3205" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="3707" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="62" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="1994" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="4267" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="4962" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="5566" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="3263" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="2703" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="4711" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="4769" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="2245" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="4209" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="3514" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="2534" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="3456" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="2436" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="2954" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="1936" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="1434" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="488" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="1743" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="4518" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="3958" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="681" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="5213" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="1183" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="4460" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="739" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="1241" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="1492" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="5462" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="3765" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="932" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="5020" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="990" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="5271" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="1685" to-port="2" /> + <edge from-layer="61" from-port="0" to-layer="3012" to-port="2" /> + <edge from-layer="62" from-port="3" to-layer="63" to-port="1" /> + <edge from-layer="63" from-port="2" to-layer="65" to-port="0" /> + <edge from-layer="64" from-port="0" to-layer="65" to-port="1" /> + <edge from-layer="65" from-port="2" to-layer="86" to-port="0" /> + <edge from-layer="65" from-port="2" to-layer="102" to-port="0" /> + <edge from-layer="65" from-port="2" to-layer="108" to-port="0" /> + <edge from-layer="65" from-port="2" to-layer="89" to-port="0" /> + <edge from-layer="66" from-port="0" to-layer="67" to-port="0" /> + <edge from-layer="67" from-port="1" to-layer="76" to-port="0" /> + <edge from-layer="68" from-port="1" to-layer="71" to-port="0" /> + <edge from-layer="69" from-port="0" to-layer="71" to-port="1" /> + <edge from-layer="70" from-port="0" to-layer="71" to-port="2" /> + <edge from-layer="71" from-port="3" to-layer="74" to-port="0" /> + <edge from-layer="72" from-port="0" to-layer="74" to-port="1" /> + <edge from-layer="72" from-port="0" to-layer="284" to-port="2" /> + <edge from-layer="72" from-port="0" to-layer="284" to-port="3" /> + <edge from-layer="73" from-port="0" to-layer="74" to-port="2" /> + <edge from-layer="73" from-port="0" to-layer="284" to-port="1" /> + <edge from-layer="74" from-port="3" to-layer="75" to-port="0" /> + <edge from-layer="75" from-port="1" to-layer="2709" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="1189" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="5219" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="1691" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="3211" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="3713" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="4968" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="938" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="5492" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="1942" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="436" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="2193" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="4717" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="687" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="4466" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="2466" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="3462" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="4215" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="76" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="2960" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="1440" to-port="1" /> + <edge from-layer="75" from-port="1" to-layer="3964" to-port="1" /> + <edge from-layer="76" from-port="2" to-layer="80" to-port="0" /> + <edge from-layer="77" from-port="0" to-layer="280" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4553" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5055" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4596" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4496" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4998" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1721" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4974" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="312" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1195" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4847" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2839" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4723" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2588" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1219" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4747" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1697" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="968" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2323" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2280" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2223" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2199" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2072" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2029" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1025" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1972" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1948" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1068" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="944" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="442" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="817" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="774" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="717" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="693" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="566" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="523" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="466" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5600" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2796" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5225" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5249" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5306" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="85" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5349" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5498" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5522" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1778" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5098" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="5620" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2739" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2715" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4804" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2568" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="78" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1821" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2496" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2472" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2990" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4245" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="240" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1319" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3743" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4221" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3468" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3970" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1570" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1527" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4094" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3047" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1276" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="111" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4051" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3341" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3090" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3217" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="2966" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3800" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3241" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3994" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3298" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="235" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3843" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1446" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="215" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3592" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3549" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="170" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3719" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="1470" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4345" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4302" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="3492" to-port="1" /> + <edge from-layer="77" from-port="0" to-layer="4472" to-port="1" /> + <edge from-layer="78" from-port="2" to-layer="79" to-port="0" /> + <edge from-layer="79" from-port="1" to-layer="1943" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="1441" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="1692" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="1190" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="3212" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="4467" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="2961" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="4969" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="2710" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="939" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="688" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="4216" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="3463" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="80" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="437" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="3965" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="2194" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="4718" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="5493" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="3714" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="2467" to-port="1" /> + <edge from-layer="79" from-port="1" to-layer="5220" to-port="1" /> + <edge from-layer="80" from-port="2" to-layer="82" to-port="0" /> + <edge from-layer="81" from-port="0" to-layer="82" to-port="1" /> + <edge from-layer="82" from-port="2" to-layer="83" to-port="1" /> + <edge from-layer="82" from-port="2" to-layer="83" to-port="0" /> + <edge from-layer="83" from-port="2" to-layer="84" to-port="0" /> + <edge from-layer="83" from-port="2" to-layer="110" to-port="0" /> + <edge from-layer="84" from-port="1" to-layer="85" to-port="0" /> + <edge from-layer="85" from-port="2" to-layer="86" to-port="1" /> + <edge from-layer="85" from-port="2" to-layer="138" to-port="1" /> + <edge from-layer="86" from-port="2" to-layer="113" to-port="0" /> + <edge from-layer="87" from-port="0" to-layer="99" to-port="0" /> + <edge from-layer="88" from-port="0" to-layer="99" to-port="1" /> + <edge from-layer="89" from-port="1" to-layer="92" to-port="0" /> + <edge from-layer="90" from-port="0" to-layer="2205" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="1752" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="4478" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="2972" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="2254" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="5280" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="3774" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="2770" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="5504" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="3474" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="2478" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="3725" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="5575" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="2721" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="2543" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="3523" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="1501" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="1201" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="999" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="4778" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="3021" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="1703" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="4227" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="4729" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="950" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="143" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="5231" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="4276" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="448" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="748" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="699" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="1452" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="4527" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="1250" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="497" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="4025" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="1954" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="4980" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="3272" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="5029" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="3223" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="2003" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="3976" to-port="1" /> + <edge from-layer="90" from-port="0" to-layer="92" to-port="1" /> + <edge from-layer="91" from-port="0" to-layer="92" to-port="2" /> + <edge from-layer="92" from-port="3" to-layer="94" to-port="0" /> + <edge from-layer="93" from-port="0" to-layer="2206" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="498" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="1453" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="2973" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="3273" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="700" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="2544" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="3475" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="1251" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="449" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="749" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="144" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="2479" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="1502" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="2255" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="3224" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="2004" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="951" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="3022" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="1955" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="1000" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="5505" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="5281" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="4228" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="3775" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="4277" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="4730" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="1753" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="5232" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="3977" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="5030" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="2771" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="1704" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="4779" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="3726" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="1202" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="4528" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="5576" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="94" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="2722" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="4981" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="4026" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="4479" to-port="1" /> + <edge from-layer="93" from-port="0" to-layer="3524" to-port="1" /> + <edge from-layer="94" from-port="2" to-layer="95" to-port="0" /> + <edge from-layer="95" from-port="1" to-layer="97" to-port="0" /> + <edge from-layer="96" from-port="0" to-layer="97" to-port="1" /> + <edge from-layer="97" from-port="2" to-layer="99" to-port="2" /> + <edge from-layer="98" from-port="0" to-layer="99" to-port="3" /> + <edge from-layer="99" from-port="4" to-layer="102" to-port="1" /> + <edge from-layer="99" from-port="4" to-layer="108" to-port="2" /> + <edge from-layer="100" from-port="0" to-layer="102" to-port="2" /> + <edge from-layer="101" from-port="0" to-layer="102" to-port="3" /> + <edge from-layer="102" from-port="4" to-layer="105" to-port="0" /> + <edge from-layer="103" from-port="0" to-layer="104" to-port="0" /> + <edge from-layer="104" from-port="1" to-layer="105" to-port="1" /> + <edge from-layer="105" from-port="2" to-layer="109" to-port="0" /> + <edge from-layer="106" from-port="0" to-layer="108" to-port="1" /> + <edge from-layer="107" from-port="0" to-layer="108" to-port="3" /> + <edge from-layer="108" from-port="4" to-layer="109" to-port="1" /> + <edge from-layer="109" from-port="2" to-layer="112" to-port="0" /> + <edge from-layer="110" from-port="1" to-layer="111" to-port="0" /> + <edge from-layer="111" from-port="2" to-layer="160" to-port="1" /> + <edge from-layer="111" from-port="2" to-layer="112" to-port="1" /> + <edge from-layer="112" from-port="2" to-layer="113" to-port="1" /> + <edge from-layer="113" from-port="2" to-layer="339" to-port="0" /> + <edge from-layer="114" from-port="0" to-layer="115" to-port="0" /> + <edge from-layer="115" from-port="1" to-layer="120" to-port="0" /> + <edge from-layer="116" from-port="0" to-layer="119" to-port="1" /> + <edge from-layer="117" from-port="0" to-layer="119" to-port="2" /> + <edge from-layer="118" from-port="0" to-layer="119" to-port="3" /> + <edge from-layer="119" from-port="4" to-layer="120" to-port="1" /> + <edge from-layer="120" from-port="2" to-layer="121" to-port="0" /> + <edge from-layer="121" from-port="1" to-layer="123" to-port="0" /> + <edge from-layer="122" from-port="0" to-layer="123" to-port="2" /> + <edge from-layer="123" from-port="3" to-layer="162" to-port="0" /> + <edge from-layer="123" from-port="3" to-layer="267" to-port="0" /> + <edge from-layer="124" from-port="0" to-layer="125" to-port="0" /> + <edge from-layer="125" from-port="1" to-layer="128" to-port="0" /> + <edge from-layer="126" from-port="0" to-layer="127" to-port="0" /> + <edge from-layer="127" from-port="1" to-layer="128" to-port="1" /> + <edge from-layer="128" from-port="2" to-layer="130" to-port="0" /> + <edge from-layer="129" from-port="0" to-layer="130" to-port="1" /> + <edge from-layer="130" from-port="2" to-layer="131" to-port="0" /> + <edge from-layer="131" from-port="1" to-layer="132" to-port="1" /> + <edge from-layer="132" from-port="2" to-layer="135" to-port="0" /> + <edge from-layer="133" from-port="0" to-layer="3514" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="488" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="2761" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="1241" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="2436" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="739" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="3765" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="5462" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="4518" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="1492" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="4267" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="134" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="1994" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="5020" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="3263" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="4769" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="1743" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="990" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="5271" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="2245" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="3012" to-port="1" /> + <edge from-layer="133" from-port="0" to-layer="4016" to-port="1" /> + <edge from-layer="134" from-port="3" to-layer="205" to-port="1" /> + <edge from-layer="134" from-port="3" to-layer="135" to-port="1" /> + <edge from-layer="135" from-port="2" to-layer="137" to-port="0" /> + <edge from-layer="136" from-port="0" to-layer="137" to-port="1" /> + <edge from-layer="137" from-port="2" to-layer="152" to-port="0" /> + <edge from-layer="137" from-port="2" to-layer="141" to-port="0" /> + <edge from-layer="137" from-port="2" to-layer="158" to-port="0" /> + <edge from-layer="137" from-port="2" to-layer="138" to-port="0" /> + <edge from-layer="138" from-port="2" to-layer="161" to-port="0" /> + <edge from-layer="139" from-port="0" to-layer="149" to-port="0" /> + <edge from-layer="140" from-port="0" to-layer="149" to-port="1" /> + <edge from-layer="141" from-port="1" to-layer="143" to-port="0" /> + <edge from-layer="142" from-port="0" to-layer="143" to-port="2" /> + <edge from-layer="143" from-port="3" to-layer="144" to-port="0" /> + <edge from-layer="144" from-port="2" to-layer="145" to-port="0" /> + <edge from-layer="145" from-port="1" to-layer="147" to-port="0" /> + <edge from-layer="146" from-port="0" to-layer="147" to-port="1" /> + <edge from-layer="147" from-port="2" to-layer="149" to-port="2" /> + <edge from-layer="148" from-port="0" to-layer="149" to-port="3" /> + <edge from-layer="149" from-port="4" to-layer="152" to-port="1" /> + <edge from-layer="149" from-port="4" to-layer="158" to-port="2" /> + <edge from-layer="150" from-port="0" to-layer="152" to-port="2" /> + <edge from-layer="151" from-port="0" to-layer="152" to-port="3" /> + <edge from-layer="152" from-port="4" to-layer="155" to-port="0" /> + <edge from-layer="153" from-port="0" to-layer="154" to-port="0" /> + <edge from-layer="154" from-port="1" to-layer="155" to-port="1" /> + <edge from-layer="155" from-port="2" to-layer="159" to-port="0" /> + <edge from-layer="156" from-port="0" to-layer="158" to-port="1" /> + <edge from-layer="157" from-port="0" to-layer="158" to-port="3" /> + <edge from-layer="158" from-port="4" to-layer="159" to-port="1" /> + <edge from-layer="159" from-port="2" to-layer="160" to-port="0" /> + <edge from-layer="160" from-port="2" to-layer="161" to-port="1" /> + <edge from-layer="161" from-port="2" to-layer="162" to-port="1" /> + <edge from-layer="162" from-port="2" to-layer="165" to-port="0" /> + <edge from-layer="162" from-port="2" to-layer="164" to-port="0" /> + <edge from-layer="162" from-port="2" to-layer="5554" to-port="0" /> + <edge from-layer="163" from-port="0" to-layer="2582" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="1564" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="560" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="1062" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="3335" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="4339" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="517" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="1772" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="768" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="4296" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="1270" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="313" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="5343" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="811" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="2562" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="5594" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="4841" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="3292" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="1521" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="2274" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="2317" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="269" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="4547" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="4798" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="209" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="5614" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="3586" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="2833" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="4590" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="164" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="3543" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="2790" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="3041" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="1019" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="5049" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="4045" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="3084" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="2023" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="3837" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="4088" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="2066" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="3794" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="1815" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="1313" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="5092" to-port="1" /> + <edge from-layer="163" from-port="0" to-layer="5300" to-port="1" /> + <edge from-layer="164" from-port="2" to-layer="179" to-port="0" /> + <edge from-layer="165" from-port="1" to-layer="168" to-port="0" /> + <edge from-layer="165" from-port="1" to-layer="170" to-port="0" /> + <edge from-layer="165" from-port="1" to-layer="176" to-port="0" /> + <edge from-layer="166" from-port="0" to-layer="168" to-port="1" /> + <edge from-layer="167" from-port="0" to-layer="168" to-port="2" /> + <edge from-layer="168" from-port="3" to-layer="177" to-port="0" /> + <edge from-layer="168" from-port="3" to-layer="184" to-port="0" /> + <edge from-layer="169" from-port="0" to-layer="170" to-port="2" /> + <edge from-layer="170" from-port="3" to-layer="172" to-port="0" /> + <edge from-layer="170" from-port="3" to-layer="181" to-port="0" /> + <edge from-layer="171" from-port="0" to-layer="172" to-port="1" /> + <edge from-layer="172" from-port="2" to-layer="177" to-port="1" /> + <edge from-layer="173" from-port="0" to-layer="1074" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="3555" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="1827" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="2845" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="4602" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="4559" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="823" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="2594" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="3849" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="3598" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="572" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="1325" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="2329" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="4100" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="177" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="4057" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="3806" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="221" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="4351" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="4308" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="2574" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="1282" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="5104" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="5355" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="2035" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="3096" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="3304" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="780" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="1784" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="3347" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="1031" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="5312" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="5061" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="4853" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="2078" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="2802" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="5606" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="529" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="1533" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="4810" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="5626" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="2286" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="3053" to-port="2" /> + <edge from-layer="173" from-port="0" to-layer="1576" to-port="2" /> + <edge from-layer="174" from-port="0" to-layer="176" to-port="1" /> + <edge from-layer="175" from-port="0" to-layer="176" to-port="2" /> + <edge from-layer="176" from-port="3" to-layer="184" to-port="2" /> + <edge from-layer="176" from-port="3" to-layer="177" to-port="3" /> + <edge from-layer="177" from-port="4" to-layer="178" to-port="0" /> + <edge from-layer="178" from-port="1" to-layer="179" to-port="1" /> + <edge from-layer="179" from-port="2" to-layer="185" to-port="0" /> + <edge from-layer="180" from-port="0" to-layer="5064" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="4060" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="1285" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="2805" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="1077" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="4311" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="4103" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="5107" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="5315" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="5358" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="4605" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="4856" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="4354" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="2577" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="783" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="1787" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="4813" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="5609" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="2848" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="4562" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="5629" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="2597" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="1579" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="3601" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="2081" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="2289" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="1034" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="3350" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="224" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="2332" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="3307" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="532" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="1830" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="3809" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="575" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="1328" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="3852" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="2038" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="3558" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="3056" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="181" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="3099" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="826" to-port="1" /> + <edge from-layer="180" from-port="0" to-layer="1536" to-port="1" /> + <edge from-layer="181" from-port="2" to-layer="183" to-port="0" /> + <edge from-layer="182" from-port="0" to-layer="183" to-port="1" /> + <edge from-layer="183" from-port="2" to-layer="184" to-port="1" /> + <edge from-layer="184" from-port="3" to-layer="185" to-port="1" /> + <edge from-layer="185" from-port="2" to-layer="331" to-port="0" /> + <edge from-layer="185" from-port="2" to-layer="339" to-port="1" /> + <edge from-layer="186" from-port="0" to-layer="187" to-port="0" /> + <edge from-layer="187" from-port="1" to-layer="192" to-port="0" /> + <edge from-layer="188" from-port="0" to-layer="191" to-port="1" /> + <edge from-layer="189" from-port="0" to-layer="191" to-port="2" /> + <edge from-layer="190" from-port="0" to-layer="191" to-port="3" /> + <edge from-layer="191" from-port="4" to-layer="192" to-port="1" /> + <edge from-layer="192" from-port="2" to-layer="193" to-port="0" /> + <edge from-layer="193" from-port="1" to-layer="195" to-port="0" /> + <edge from-layer="194" from-port="0" to-layer="195" to-port="2" /> + <edge from-layer="195" from-port="3" to-layer="208" to-port="0" /> + <edge from-layer="196" from-port="0" to-layer="197" to-port="0" /> + <edge from-layer="197" from-port="1" to-layer="200" to-port="0" /> + <edge from-layer="198" from-port="0" to-layer="199" to-port="0" /> + <edge from-layer="199" from-port="1" to-layer="200" to-port="1" /> + <edge from-layer="200" from-port="2" to-layer="202" to-port="0" /> + <edge from-layer="201" from-port="0" to-layer="202" to-port="1" /> + <edge from-layer="202" from-port="2" to-layer="203" to-port="0" /> + <edge from-layer="203" from-port="1" to-layer="204" to-port="1" /> + <edge from-layer="204" from-port="2" to-layer="205" to-port="0" /> + <edge from-layer="205" from-port="2" to-layer="207" to-port="0" /> + <edge from-layer="206" from-port="0" to-layer="207" to-port="1" /> + <edge from-layer="207" from-port="2" to-layer="208" to-port="1" /> + <edge from-layer="208" from-port="2" to-layer="209" to-port="0" /> + <edge from-layer="208" from-port="2" to-layer="5553" to-port="0" /> + <edge from-layer="208" from-port="2" to-layer="210" to-port="0" /> + <edge from-layer="209" from-port="2" to-layer="223" to-port="0" /> + <edge from-layer="210" from-port="1" to-layer="220" to-port="0" /> + <edge from-layer="210" from-port="1" to-layer="215" to-port="0" /> + <edge from-layer="210" from-port="1" to-layer="213" to-port="0" /> + <edge from-layer="211" from-port="0" to-layer="213" to-port="1" /> + <edge from-layer="212" from-port="0" to-layer="213" to-port="2" /> + <edge from-layer="213" from-port="3" to-layer="227" to-port="0" /> + <edge from-layer="213" from-port="3" to-layer="221" to-port="0" /> + <edge from-layer="214" from-port="0" to-layer="215" to-port="2" /> + <edge from-layer="215" from-port="3" to-layer="224" to-port="0" /> + <edge from-layer="215" from-port="3" to-layer="217" to-port="0" /> + <edge from-layer="216" from-port="0" to-layer="217" to-port="1" /> + <edge from-layer="217" from-port="2" to-layer="221" to-port="1" /> + <edge from-layer="218" from-port="0" to-layer="220" to-port="1" /> + <edge from-layer="219" from-port="0" to-layer="220" to-port="2" /> + <edge from-layer="220" from-port="3" to-layer="227" to-port="2" /> + <edge from-layer="220" from-port="3" to-layer="221" to-port="3" /> + <edge from-layer="221" from-port="4" to-layer="222" to-port="0" /> + <edge from-layer="222" from-port="1" to-layer="223" to-port="1" /> + <edge from-layer="223" from-port="2" to-layer="228" to-port="0" /> + <edge from-layer="224" from-port="2" to-layer="226" to-port="0" /> + <edge from-layer="225" from-port="0" to-layer="226" to-port="1" /> + <edge from-layer="226" from-port="2" to-layer="227" to-port="1" /> + <edge from-layer="227" from-port="3" to-layer="228" to-port="1" /> + <edge from-layer="228" from-port="2" to-layer="339" to-port="2" /> + <edge from-layer="229" from-port="0" to-layer="254" to-port="2" /> + <edge from-layer="229" from-port="0" to-layer="251" to-port="1" /> + <edge from-layer="229" from-port="0" to-layer="248" to-port="2" /> + <edge from-layer="229" from-port="0" to-layer="250" to-port="0" /> + <edge from-layer="230" from-port="0" to-layer="231" to-port="0" /> + <edge from-layer="231" from-port="1" to-layer="245" to-port="0" /> + <edge from-layer="232" from-port="0" to-layer="244" to-port="0" /> + <edge from-layer="233" from-port="1" to-layer="235" to-port="0" /> + <edge from-layer="233" from-port="1" to-layer="283" to-port="0" /> + <edge from-layer="234" from-port="0" to-layer="235" to-port="2" /> + <edge from-layer="235" from-port="3" to-layer="237" to-port="0" /> + <edge from-layer="235" from-port="3" to-layer="270" to-port="1" /> + <edge from-layer="236" from-port="0" to-layer="237" to-port="1" /> + <edge from-layer="237" from-port="2" to-layer="243" to-port="0" /> + <edge from-layer="238" from-port="1" to-layer="240" to-port="0" /> + <edge from-layer="239" from-port="0" to-layer="240" to-port="2" /> + <edge from-layer="240" from-port="3" to-layer="299" to-port="0" /> + <edge from-layer="240" from-port="3" to-layer="264" to-port="1" /> + <edge from-layer="240" from-port="3" to-layer="242" to-port="0" /> + <edge from-layer="241" from-port="0" to-layer="242" to-port="1" /> + <edge from-layer="242" from-port="2" to-layer="243" to-port="1" /> + <edge from-layer="243" from-port="2" to-layer="244" to-port="1" /> + <edge from-layer="244" from-port="2" to-layer="245" to-port="1" /> + <edge from-layer="245" from-port="2" to-layer="261" to-port="1" /> + <edge from-layer="245" from-port="2" to-layer="246" to-port="0" /> + <edge from-layer="246" from-port="1" to-layer="254" to-port="0" /> + <edge from-layer="246" from-port="1" to-layer="248" to-port="0" /> + <edge from-layer="247" from-port="0" to-layer="248" to-port="1" /> + <edge from-layer="248" from-port="3" to-layer="250" to-port="1" /> + <edge from-layer="249" from-port="0" to-layer="250" to-port="2" /> + <edge from-layer="249" from-port="0" to-layer="257" to-port="1" /> + <edge from-layer="249" from-port="0" to-layer="256" to-port="2" /> + <edge from-layer="250" from-port="3" to-layer="251" to-port="0" /> + <edge from-layer="251" from-port="2" to-layer="258" to-port="0" /> + <edge from-layer="252" from-port="0" to-layer="255" to-port="1" /> + <edge from-layer="252" from-port="0" to-layer="256" to-port="0" /> + <edge from-layer="253" from-port="0" to-layer="254" to-port="1" /> + <edge from-layer="254" from-port="3" to-layer="255" to-port="0" /> + <edge from-layer="255" from-port="2" to-layer="256" to-port="1" /> + <edge from-layer="256" from-port="3" to-layer="257" to-port="0" /> + <edge from-layer="257" from-port="2" to-layer="258" to-port="1" /> + <edge from-layer="258" from-port="2" to-layer="261" to-port="0" /> + <edge from-layer="259" from-port="0" to-layer="260" to-port="0" /> + <edge from-layer="260" from-port="1" to-layer="261" to-port="2" /> + <edge from-layer="261" from-port="3" to-layer="277" to-port="0" /> + <edge from-layer="262" from-port="0" to-layer="264" to-port="0" /> + <edge from-layer="263" from-port="0" to-layer="264" to-port="2" /> + <edge from-layer="264" from-port="3" to-layer="265" to-port="0" /> + <edge from-layer="265" from-port="1" to-layer="266" to-port="0" /> + <edge from-layer="266" from-port="1" to-layer="275" to-port="0" /> + <edge from-layer="267" from-port="1" to-layer="269" to-port="0" /> + <edge from-layer="268" from-port="0" to-layer="269" to-port="2" /> + <edge from-layer="269" from-port="3" to-layer="272" to-port="0" /> + <edge from-layer="269" from-port="3" to-layer="270" to-port="0" /> + <edge from-layer="270" from-port="2" to-layer="272" to-port="1" /> + <edge from-layer="271" from-port="0" to-layer="272" to-port="2" /> + <edge from-layer="272" from-port="3" to-layer="274" to-port="0" /> + <edge from-layer="273" from-port="0" to-layer="274" to-port="1" /> + <edge from-layer="274" from-port="2" to-layer="275" to-port="1" /> + <edge from-layer="275" from-port="2" to-layer="276" to-port="0" /> + <edge from-layer="276" from-port="1" to-layer="277" to-port="1" /> + <edge from-layer="277" from-port="2" to-layer="279" to-port="0" /> + <edge from-layer="278" from-port="0" to-layer="279" to-port="1" /> + <edge from-layer="279" from-port="2" to-layer="280" to-port="0" /> + <edge from-layer="280" from-port="2" to-layer="286" to-port="0" /> + <edge from-layer="281" from-port="0" to-layer="283" to-port="1" /> + <edge from-layer="282" from-port="0" to-layer="283" to-port="2" /> + <edge from-layer="283" from-port="3" to-layer="284" to-port="0" /> + <edge from-layer="284" from-port="4" to-layer="285" to-port="0" /> + <edge from-layer="285" from-port="1" to-layer="286" to-port="1" /> + <edge from-layer="286" from-port="2" to-layer="308" to-port="0" /> + <edge from-layer="286" from-port="2" to-layer="288" to-port="0" /> + <edge from-layer="286" from-port="2" to-layer="290" to-port="0" /> + <edge from-layer="287" from-port="0" to-layer="288" to-port="1" /> + <edge from-layer="288" from-port="2" to-layer="326" to-port="0" /> + <edge from-layer="289" from-port="0" to-layer="293" to-port="0" /> + <edge from-layer="289" from-port="0" to-layer="291" to-port="1" /> + <edge from-layer="290" from-port="1" to-layer="294" to-port="1" /> + <edge from-layer="290" from-port="1" to-layer="327" to-port="1" /> + <edge from-layer="290" from-port="1" to-layer="291" to-port="0" /> + <edge from-layer="291" from-port="2" to-layer="293" to-port="1" /> + <edge from-layer="292" from-port="0" to-layer="293" to-port="2" /> + <edge from-layer="293" from-port="3" to-layer="294" to-port="0" /> + <edge from-layer="294" from-port="2" to-layer="303" to-port="0" /> + <edge from-layer="295" from-port="0" to-layer="303" to-port="1" /> + <edge from-layer="296" from-port="0" to-layer="301" to-port="0" /> + <edge from-layer="297" from-port="0" to-layer="301" to-port="1" /> + <edge from-layer="298" from-port="0" to-layer="299" to-port="1" /> + <edge from-layer="299" from-port="2" to-layer="301" to-port="2" /> + <edge from-layer="300" from-port="0" to-layer="301" to-port="3" /> + <edge from-layer="301" from-port="4" to-layer="303" to-port="2" /> + <edge from-layer="301" from-port="4" to-layer="308" to-port="2" /> + <edge from-layer="302" from-port="0" to-layer="303" to-port="3" /> + <edge from-layer="303" from-port="4" to-layer="305" to-port="0" /> + <edge from-layer="304" from-port="0" to-layer="305" to-port="1" /> + <edge from-layer="305" from-port="2" to-layer="326" to-port="1" /> + <edge from-layer="306" from-port="0" to-layer="308" to-port="1" /> + <edge from-layer="307" from-port="0" to-layer="308" to-port="3" /> + <edge from-layer="308" from-port="4" to-layer="311" to-port="0" /> + <edge from-layer="308" from-port="4" to-layer="322" to-port="0" /> + <edge from-layer="308" from-port="4" to-layer="321" to-port="2" /> + <edge from-layer="309" from-port="0" to-layer="310" to-port="0" /> + <edge from-layer="310" from-port="1" to-layer="311" to-port="1" /> + <edge from-layer="311" from-port="2" to-layer="318" to-port="0" /> + <edge from-layer="312" from-port="2" to-layer="313" to-port="0" /> + <edge from-layer="313" from-port="2" to-layer="314" to-port="0" /> + <edge from-layer="314" from-port="1" to-layer="317" to-port="0" /> + <edge from-layer="315" from-port="0" to-layer="316" to-port="0" /> + <edge from-layer="316" from-port="1" to-layer="317" to-port="1" /> + <edge from-layer="317" from-port="2" to-layer="318" to-port="1" /> + <edge from-layer="318" from-port="2" to-layer="321" to-port="0" /> + <edge from-layer="319" from-port="0" to-layer="320" to-port="0" /> + <edge from-layer="320" from-port="1" to-layer="321" to-port="1" /> + <edge from-layer="321" from-port="3" to-layer="323" to-port="0" /> + <edge from-layer="322" from-port="1" to-layer="323" to-port="1" /> + <edge from-layer="323" from-port="2" to-layer="325" to-port="0" /> + <edge from-layer="324" from-port="0" to-layer="325" to-port="1" /> + <edge from-layer="325" from-port="2" to-layer="326" to-port="2" /> + <edge from-layer="326" from-port="3" to-layer="327" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="5373" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="2096" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="4871" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="3114" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="4369" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="3365" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="5644" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="2612" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="1092" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="4620" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="2863" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="4118" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="590" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="3867" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="2347" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="1594" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="841" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="3616" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="1343" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="1845" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="5122" to-port="0" /> + <edge from-layer="327" from-port="2" to-layer="338" to-port="0" /> + <edge from-layer="328" from-port="0" to-layer="338" to-port="1" /> + <edge from-layer="329" from-port="0" to-layer="336" to-port="0" /> + <edge from-layer="330" from-port="0" to-layer="336" to-port="1" /> + <edge from-layer="331" from-port="1" to-layer="334" to-port="0" /> + <edge from-layer="332" from-port="0" to-layer="334" to-port="1" /> + <edge from-layer="333" from-port="0" to-layer="334" to-port="2" /> + <edge from-layer="334" from-port="3" to-layer="336" to-port="2" /> + <edge from-layer="335" from-port="0" to-layer="336" to-port="3" /> + <edge from-layer="336" from-port="4" to-layer="338" to-port="2" /> + <edge from-layer="337" from-port="0" to-layer="338" to-port="3" /> + <edge from-layer="338" from-port="4" to-layer="339" to-port="3" /> + <edge from-layer="339" from-port="4" to-layer="341" to-port="0" /> + <edge from-layer="340" from-port="0" to-layer="341" to-port="1" /> + <edge from-layer="341" from-port="2" to-layer="344" to-port="0" /> + <edge from-layer="342" from-port="0" to-layer="4373" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="2616" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="2867" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="1096" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="4122" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="4624" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="594" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="5648" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="4875" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="343" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="5377" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="3118" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="3369" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="5126" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="2100" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="1849" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="845" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="1347" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="3620" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="1598" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="2351" to-port="1" /> + <edge from-layer="342" from-port="0" to-layer="3871" to-port="1" /> + <edge from-layer="343" from-port="2" to-layer="344" to-port="1" /> + <edge from-layer="344" from-port="2" to-layer="353" to-port="0" /> + <edge from-layer="345" from-port="0" to-layer="346" to-port="0" /> + <edge from-layer="346" from-port="1" to-layer="349" to-port="0" /> + <edge from-layer="347" from-port="0" to-layer="348" to-port="0" /> + <edge from-layer="348" from-port="1" to-layer="349" to-port="1" /> + <edge from-layer="349" from-port="2" to-layer="351" to-port="0" /> + <edge from-layer="350" from-port="0" to-layer="351" to-port="1" /> + <edge from-layer="351" from-port="2" to-layer="352" to-port="0" /> + <edge from-layer="352" from-port="1" to-layer="353" to-port="1" /> + <edge from-layer="353" from-port="2" to-layer="354" to-port="1" /> + <edge from-layer="354" from-port="2" to-layer="400" to-port="0" /> + <edge from-layer="354" from-port="2" to-layer="361" to-port="0" /> + <edge from-layer="354" from-port="2" to-layer="369" to-port="0" /> + <edge from-layer="355" from-port="0" to-layer="356" to-port="0" /> + <edge from-layer="356" from-port="1" to-layer="370" to-port="0" /> + <edge from-layer="357" from-port="0" to-layer="358" to-port="0" /> + <edge from-layer="358" from-port="1" to-layer="368" to-port="0" /> + <edge from-layer="359" from-port="0" to-layer="360" to-port="0" /> + <edge from-layer="360" from-port="1" to-layer="361" to-port="1" /> + <edge from-layer="361" from-port="2" to-layer="363" to-port="0" /> + <edge from-layer="362" from-port="0" to-layer="363" to-port="1" /> + <edge from-layer="363" from-port="2" to-layer="366" to-port="0" /> + <edge from-layer="364" from-port="0" to-layer="365" to-port="0" /> + <edge from-layer="365" from-port="1" to-layer="366" to-port="1" /> + <edge from-layer="366" from-port="2" to-layer="367" to-port="0" /> + <edge from-layer="367" from-port="1" to-layer="368" to-port="1" /> + <edge from-layer="368" from-port="2" to-layer="369" to-port="1" /> + <edge from-layer="369" from-port="2" to-layer="370" to-port="1" /> + <edge from-layer="370" from-port="2" to-layer="379" to-port="0" /> + <edge from-layer="370" from-port="2" to-layer="389" to-port="0" /> + <edge from-layer="371" from-port="0" to-layer="372" to-port="0" /> + <edge from-layer="372" from-port="1" to-layer="375" to-port="0" /> + <edge from-layer="373" from-port="0" to-layer="374" to-port="0" /> + <edge from-layer="374" from-port="1" to-layer="375" to-port="1" /> + <edge from-layer="375" from-port="2" to-layer="377" to-port="0" /> + <edge from-layer="376" from-port="0" to-layer="377" to-port="1" /> + <edge from-layer="377" from-port="2" to-layer="378" to-port="0" /> + <edge from-layer="378" from-port="1" to-layer="379" to-port="1" /> + <edge from-layer="379" from-port="2" to-layer="380" to-port="0" /> + <edge from-layer="380" from-port="1" to-layer="390" to-port="0" /> + <edge from-layer="381" from-port="0" to-layer="382" to-port="0" /> + <edge from-layer="382" from-port="1" to-layer="385" to-port="0" /> + <edge from-layer="383" from-port="0" to-layer="384" to-port="0" /> + <edge from-layer="384" from-port="1" to-layer="385" to-port="1" /> + <edge from-layer="385" from-port="2" to-layer="387" to-port="0" /> + <edge from-layer="386" from-port="0" to-layer="387" to-port="1" /> + <edge from-layer="387" from-port="2" to-layer="388" to-port="0" /> + <edge from-layer="388" from-port="1" to-layer="389" to-port="1" /> + <edge from-layer="389" from-port="2" to-layer="390" to-port="1" /> + <edge from-layer="390" from-port="2" to-layer="399" to-port="0" /> + <edge from-layer="391" from-port="0" to-layer="392" to-port="0" /> + <edge from-layer="392" from-port="1" to-layer="395" to-port="0" /> + <edge from-layer="393" from-port="0" to-layer="394" to-port="0" /> + <edge from-layer="394" from-port="1" to-layer="395" to-port="1" /> + <edge from-layer="395" from-port="2" to-layer="397" to-port="0" /> + <edge from-layer="396" from-port="0" to-layer="397" to-port="1" /> + <edge from-layer="397" from-port="2" to-layer="398" to-port="0" /> + <edge from-layer="398" from-port="1" to-layer="399" to-port="1" /> + <edge from-layer="399" from-port="2" to-layer="400" to-port="1" /> + <edge from-layer="400" from-port="2" to-layer="605" to-port="0" /> + <edge from-layer="400" from-port="2" to-layer="407" to-port="0" /> + <edge from-layer="400" from-port="2" to-layer="415" to-port="0" /> + <edge from-layer="401" from-port="0" to-layer="402" to-port="0" /> + <edge from-layer="402" from-port="1" to-layer="416" to-port="0" /> + <edge from-layer="403" from-port="0" to-layer="404" to-port="0" /> + <edge from-layer="404" from-port="1" to-layer="414" to-port="0" /> + <edge from-layer="405" from-port="0" to-layer="406" to-port="0" /> + <edge from-layer="406" from-port="1" to-layer="407" to-port="1" /> + <edge from-layer="407" from-port="2" to-layer="409" to-port="0" /> + <edge from-layer="408" from-port="0" to-layer="409" to-port="1" /> + <edge from-layer="409" from-port="2" to-layer="412" to-port="0" /> + <edge from-layer="410" from-port="0" to-layer="411" to-port="0" /> + <edge from-layer="411" from-port="1" to-layer="412" to-port="1" /> + <edge from-layer="412" from-port="2" to-layer="413" to-port="0" /> + <edge from-layer="413" from-port="1" to-layer="414" to-port="1" /> + <edge from-layer="414" from-port="2" to-layer="415" to-port="1" /> + <edge from-layer="415" from-port="2" to-layer="416" to-port="1" /> + <edge from-layer="416" from-port="2" to-layer="487" to-port="0" /> + <edge from-layer="416" from-port="2" to-layer="425" to-port="0" /> + <edge from-layer="416" from-port="2" to-layer="426" to-port="0" /> + <edge from-layer="416" from-port="2" to-layer="555" to-port="0" /> + <edge from-layer="417" from-port="0" to-layer="418" to-port="0" /> + <edge from-layer="418" from-port="1" to-layer="421" to-port="0" /> + <edge from-layer="419" from-port="0" to-layer="420" to-port="0" /> + <edge from-layer="420" from-port="1" to-layer="421" to-port="1" /> + <edge from-layer="421" from-port="2" to-layer="423" to-port="0" /> + <edge from-layer="422" from-port="0" to-layer="423" to-port="1" /> + <edge from-layer="423" from-port="2" to-layer="424" to-port="0" /> + <edge from-layer="424" from-port="1" to-layer="425" to-port="1" /> + <edge from-layer="425" from-port="2" to-layer="431" to-port="0" /> + <edge from-layer="426" from-port="1" to-layer="429" to-port="0" /> + <edge from-layer="427" from-port="0" to-layer="429" to-port="1" /> + <edge from-layer="428" from-port="0" to-layer="429" to-port="2" /> + <edge from-layer="429" from-port="3" to-layer="488" to-port="0" /> + <edge from-layer="429" from-port="3" to-layer="430" to-port="0" /> + <edge from-layer="429" from-port="3" to-layer="594" to-port="0" /> + <edge from-layer="430" from-port="3" to-layer="431" to-port="1" /> + <edge from-layer="431" from-port="2" to-layer="433" to-port="0" /> + <edge from-layer="432" from-port="0" to-layer="433" to-port="1" /> + <edge from-layer="433" from-port="2" to-layer="463" to-port="0" /> + <edge from-layer="433" from-port="2" to-layer="443" to-port="0" /> + <edge from-layer="433" from-port="2" to-layer="457" to-port="0" /> + <edge from-layer="433" from-port="2" to-layer="446" to-port="0" /> + <edge from-layer="434" from-port="0" to-layer="435" to-port="0" /> + <edge from-layer="435" from-port="1" to-layer="436" to-port="0" /> + <edge from-layer="436" from-port="2" to-layer="437" to-port="0" /> + <edge from-layer="437" from-port="2" to-layer="439" to-port="0" /> + <edge from-layer="438" from-port="0" to-layer="439" to-port="1" /> + <edge from-layer="439" from-port="2" to-layer="440" to-port="0" /> + <edge from-layer="439" from-port="2" to-layer="440" to-port="1" /> + <edge from-layer="440" from-port="2" to-layer="465" to-port="0" /> + <edge from-layer="440" from-port="2" to-layer="441" to-port="0" /> + <edge from-layer="441" from-port="1" to-layer="442" to-port="0" /> + <edge from-layer="442" from-port="2" to-layer="443" to-port="1" /> + <edge from-layer="442" from-port="2" to-layer="492" to-port="1" /> + <edge from-layer="443" from-port="2" to-layer="468" to-port="0" /> + <edge from-layer="444" from-port="0" to-layer="454" to-port="0" /> + <edge from-layer="445" from-port="0" to-layer="454" to-port="1" /> + <edge from-layer="446" from-port="1" to-layer="448" to-port="0" /> + <edge from-layer="447" from-port="0" to-layer="448" to-port="2" /> + <edge from-layer="448" from-port="3" to-layer="449" to-port="0" /> + <edge from-layer="449" from-port="2" to-layer="450" to-port="0" /> + <edge from-layer="450" from-port="1" to-layer="452" to-port="0" /> + <edge from-layer="451" from-port="0" to-layer="452" to-port="1" /> + <edge from-layer="452" from-port="2" to-layer="454" to-port="2" /> + <edge from-layer="453" from-port="0" to-layer="454" to-port="3" /> + <edge from-layer="454" from-port="4" to-layer="463" to-port="2" /> + <edge from-layer="454" from-port="4" to-layer="457" to-port="1" /> + <edge from-layer="455" from-port="0" to-layer="457" to-port="2" /> + <edge from-layer="456" from-port="0" to-layer="457" to-port="3" /> + <edge from-layer="457" from-port="4" to-layer="460" to-port="0" /> + <edge from-layer="458" from-port="0" to-layer="459" to-port="0" /> + <edge from-layer="459" from-port="1" to-layer="460" to-port="1" /> + <edge from-layer="460" from-port="2" to-layer="464" to-port="0" /> + <edge from-layer="461" from-port="0" to-layer="463" to-port="1" /> + <edge from-layer="462" from-port="0" to-layer="463" to-port="3" /> + <edge from-layer="463" from-port="4" to-layer="464" to-port="1" /> + <edge from-layer="464" from-port="2" to-layer="467" to-port="0" /> + <edge from-layer="465" from-port="1" to-layer="466" to-port="0" /> + <edge from-layer="466" from-port="2" to-layer="514" to-port="1" /> + <edge from-layer="466" from-port="2" to-layer="467" to-port="1" /> + <edge from-layer="467" from-port="2" to-layer="468" to-port="1" /> + <edge from-layer="468" from-port="2" to-layer="591" to-port="0" /> + <edge from-layer="469" from-port="0" to-layer="470" to-port="0" /> + <edge from-layer="470" from-port="1" to-layer="475" to-port="0" /> + <edge from-layer="471" from-port="0" to-layer="474" to-port="1" /> + <edge from-layer="472" from-port="0" to-layer="474" to-port="2" /> + <edge from-layer="473" from-port="0" to-layer="474" to-port="3" /> + <edge from-layer="474" from-port="4" to-layer="475" to-port="1" /> + <edge from-layer="475" from-port="2" to-layer="476" to-port="0" /> + <edge from-layer="476" from-port="1" to-layer="478" to-port="0" /> + <edge from-layer="477" from-port="0" to-layer="478" to-port="2" /> + <edge from-layer="478" from-port="3" to-layer="516" to-port="0" /> + <edge from-layer="479" from-port="0" to-layer="480" to-port="0" /> + <edge from-layer="480" from-port="1" to-layer="483" to-port="0" /> + <edge from-layer="481" from-port="0" to-layer="482" to-port="0" /> + <edge from-layer="482" from-port="1" to-layer="483" to-port="1" /> + <edge from-layer="483" from-port="2" to-layer="485" to-port="0" /> + <edge from-layer="484" from-port="0" to-layer="485" to-port="1" /> + <edge from-layer="485" from-port="2" to-layer="486" to-port="0" /> + <edge from-layer="486" from-port="1" to-layer="487" to-port="1" /> + <edge from-layer="487" from-port="2" to-layer="489" to-port="0" /> + <edge from-layer="488" from-port="3" to-layer="556" to-port="1" /> + <edge from-layer="488" from-port="3" to-layer="489" to-port="1" /> + <edge from-layer="489" from-port="2" to-layer="491" to-port="0" /> + <edge from-layer="490" from-port="0" to-layer="491" to-port="1" /> + <edge from-layer="491" from-port="2" to-layer="495" to-port="0" /> + <edge from-layer="491" from-port="2" to-layer="492" to-port="0" /> + <edge from-layer="491" from-port="2" to-layer="512" to-port="0" /> + <edge from-layer="491" from-port="2" to-layer="506" to-port="0" /> + <edge from-layer="492" from-port="2" to-layer="515" to-port="0" /> + <edge from-layer="493" from-port="0" to-layer="503" to-port="0" /> + <edge from-layer="494" from-port="0" to-layer="503" to-port="1" /> + <edge from-layer="495" from-port="1" to-layer="497" to-port="0" /> + <edge from-layer="496" from-port="0" to-layer="497" to-port="2" /> + <edge from-layer="497" from-port="3" to-layer="498" to-port="0" /> + <edge from-layer="498" from-port="2" to-layer="499" to-port="0" /> + <edge from-layer="499" from-port="1" to-layer="501" to-port="0" /> + <edge from-layer="500" from-port="0" to-layer="501" to-port="1" /> + <edge from-layer="501" from-port="2" to-layer="503" to-port="2" /> + <edge from-layer="502" from-port="0" to-layer="503" to-port="3" /> + <edge from-layer="503" from-port="4" to-layer="506" to-port="1" /> + <edge from-layer="503" from-port="4" to-layer="512" to-port="2" /> + <edge from-layer="504" from-port="0" to-layer="506" to-port="2" /> + <edge from-layer="505" from-port="0" to-layer="506" to-port="3" /> + <edge from-layer="506" from-port="4" to-layer="509" to-port="0" /> + <edge from-layer="507" from-port="0" to-layer="508" to-port="0" /> + <edge from-layer="508" from-port="1" to-layer="509" to-port="1" /> + <edge from-layer="509" from-port="2" to-layer="513" to-port="0" /> + <edge from-layer="510" from-port="0" to-layer="512" to-port="1" /> + <edge from-layer="511" from-port="0" to-layer="512" to-port="3" /> + <edge from-layer="512" from-port="4" to-layer="513" to-port="1" /> + <edge from-layer="513" from-port="2" to-layer="514" to-port="0" /> + <edge from-layer="514" from-port="2" to-layer="515" to-port="1" /> + <edge from-layer="515" from-port="2" to-layer="516" to-port="1" /> + <edge from-layer="516" from-port="2" to-layer="517" to-port="0" /> + <edge from-layer="516" from-port="2" to-layer="518" to-port="0" /> + <edge from-layer="516" from-port="2" to-layer="5552" to-port="0" /> + <edge from-layer="517" from-port="2" to-layer="531" to-port="0" /> + <edge from-layer="518" from-port="1" to-layer="521" to-port="0" /> + <edge from-layer="518" from-port="1" to-layer="523" to-port="0" /> + <edge from-layer="518" from-port="1" to-layer="528" to-port="0" /> + <edge from-layer="519" from-port="0" to-layer="521" to-port="1" /> + <edge from-layer="520" from-port="0" to-layer="521" to-port="2" /> + <edge from-layer="521" from-port="3" to-layer="529" to-port="0" /> + <edge from-layer="521" from-port="3" to-layer="535" to-port="0" /> + <edge from-layer="522" from-port="0" to-layer="523" to-port="2" /> + <edge from-layer="523" from-port="3" to-layer="525" to-port="0" /> + <edge from-layer="523" from-port="3" to-layer="532" to-port="0" /> + <edge from-layer="524" from-port="0" to-layer="525" to-port="1" /> + <edge from-layer="525" from-port="2" to-layer="529" to-port="1" /> + <edge from-layer="526" from-port="0" to-layer="528" to-port="1" /> + <edge from-layer="527" from-port="0" to-layer="528" to-port="2" /> + <edge from-layer="528" from-port="3" to-layer="535" to-port="2" /> + <edge from-layer="528" from-port="3" to-layer="529" to-port="3" /> + <edge from-layer="529" from-port="4" to-layer="530" to-port="0" /> + <edge from-layer="530" from-port="1" to-layer="531" to-port="1" /> + <edge from-layer="531" from-port="2" to-layer="536" to-port="0" /> + <edge from-layer="532" from-port="2" to-layer="534" to-port="0" /> + <edge from-layer="533" from-port="0" to-layer="534" to-port="1" /> + <edge from-layer="534" from-port="2" to-layer="535" to-port="1" /> + <edge from-layer="535" from-port="3" to-layer="536" to-port="1" /> + <edge from-layer="536" from-port="2" to-layer="583" to-port="0" /> + <edge from-layer="536" from-port="2" to-layer="591" to-port="1" /> + <edge from-layer="537" from-port="0" to-layer="538" to-port="0" /> + <edge from-layer="538" from-port="1" to-layer="543" to-port="0" /> + <edge from-layer="539" from-port="0" to-layer="542" to-port="1" /> + <edge from-layer="540" from-port="0" to-layer="542" to-port="2" /> + <edge from-layer="541" from-port="0" to-layer="542" to-port="3" /> + <edge from-layer="542" from-port="4" to-layer="543" to-port="1" /> + <edge from-layer="543" from-port="2" to-layer="544" to-port="0" /> + <edge from-layer="544" from-port="1" to-layer="546" to-port="0" /> + <edge from-layer="545" from-port="0" to-layer="546" to-port="2" /> + <edge from-layer="546" from-port="3" to-layer="559" to-port="0" /> + <edge from-layer="547" from-port="0" to-layer="548" to-port="0" /> + <edge from-layer="548" from-port="1" to-layer="551" to-port="0" /> + <edge from-layer="549" from-port="0" to-layer="550" to-port="0" /> + <edge from-layer="550" from-port="1" to-layer="551" to-port="1" /> + <edge from-layer="551" from-port="2" to-layer="553" to-port="0" /> + <edge from-layer="552" from-port="0" to-layer="553" to-port="1" /> + <edge from-layer="553" from-port="2" to-layer="554" to-port="0" /> + <edge from-layer="554" from-port="1" to-layer="555" to-port="1" /> + <edge from-layer="555" from-port="2" to-layer="556" to-port="0" /> + <edge from-layer="556" from-port="2" to-layer="558" to-port="0" /> + <edge from-layer="557" from-port="0" to-layer="558" to-port="1" /> + <edge from-layer="558" from-port="2" to-layer="559" to-port="1" /> + <edge from-layer="559" from-port="2" to-layer="561" to-port="0" /> + <edge from-layer="559" from-port="2" to-layer="560" to-port="0" /> + <edge from-layer="559" from-port="2" to-layer="5551" to-port="0" /> + <edge from-layer="560" from-port="2" to-layer="574" to-port="0" /> + <edge from-layer="561" from-port="1" to-layer="566" to-port="0" /> + <edge from-layer="561" from-port="1" to-layer="564" to-port="0" /> + <edge from-layer="561" from-port="1" to-layer="571" to-port="0" /> + <edge from-layer="562" from-port="0" to-layer="564" to-port="1" /> + <edge from-layer="563" from-port="0" to-layer="564" to-port="2" /> + <edge from-layer="564" from-port="3" to-layer="578" to-port="0" /> + <edge from-layer="564" from-port="3" to-layer="572" to-port="0" /> + <edge from-layer="565" from-port="0" to-layer="566" to-port="2" /> + <edge from-layer="566" from-port="3" to-layer="568" to-port="0" /> + <edge from-layer="566" from-port="3" to-layer="575" to-port="0" /> + <edge from-layer="567" from-port="0" to-layer="568" to-port="1" /> + <edge from-layer="568" from-port="2" to-layer="572" to-port="1" /> + <edge from-layer="569" from-port="0" to-layer="571" to-port="1" /> + <edge from-layer="570" from-port="0" to-layer="571" to-port="2" /> + <edge from-layer="571" from-port="3" to-layer="572" to-port="3" /> + <edge from-layer="571" from-port="3" to-layer="578" to-port="2" /> + <edge from-layer="572" from-port="4" to-layer="573" to-port="0" /> + <edge from-layer="573" from-port="1" to-layer="574" to-port="1" /> + <edge from-layer="574" from-port="2" to-layer="579" to-port="0" /> + <edge from-layer="575" from-port="2" to-layer="577" to-port="0" /> + <edge from-layer="576" from-port="0" to-layer="577" to-port="1" /> + <edge from-layer="577" from-port="2" to-layer="578" to-port="1" /> + <edge from-layer="578" from-port="3" to-layer="579" to-port="1" /> + <edge from-layer="579" from-port="2" to-layer="591" to-port="2" /> + <edge from-layer="580" from-port="0" to-layer="590" to-port="1" /> + <edge from-layer="581" from-port="0" to-layer="588" to-port="0" /> + <edge from-layer="582" from-port="0" to-layer="588" to-port="1" /> + <edge from-layer="583" from-port="1" to-layer="586" to-port="0" /> + <edge from-layer="584" from-port="0" to-layer="586" to-port="1" /> + <edge from-layer="585" from-port="0" to-layer="586" to-port="2" /> + <edge from-layer="586" from-port="3" to-layer="588" to-port="2" /> + <edge from-layer="587" from-port="0" to-layer="588" to-port="3" /> + <edge from-layer="588" from-port="4" to-layer="590" to-port="2" /> + <edge from-layer="589" from-port="0" to-layer="590" to-port="3" /> + <edge from-layer="590" from-port="4" to-layer="591" to-port="3" /> + <edge from-layer="591" from-port="4" to-layer="593" to-port="0" /> + <edge from-layer="592" from-port="0" to-layer="593" to-port="1" /> + <edge from-layer="593" from-port="2" to-layer="595" to-port="0" /> + <edge from-layer="594" from-port="2" to-layer="595" to-port="1" /> + <edge from-layer="595" from-port="2" to-layer="604" to-port="0" /> + <edge from-layer="596" from-port="0" to-layer="597" to-port="0" /> + <edge from-layer="597" from-port="1" to-layer="600" to-port="0" /> + <edge from-layer="598" from-port="0" to-layer="599" to-port="0" /> + <edge from-layer="599" from-port="1" to-layer="600" to-port="1" /> + <edge from-layer="600" from-port="2" to-layer="602" to-port="0" /> + <edge from-layer="601" from-port="0" to-layer="602" to-port="1" /> + <edge from-layer="602" from-port="2" to-layer="603" to-port="0" /> + <edge from-layer="603" from-port="1" to-layer="604" to-port="1" /> + <edge from-layer="604" from-port="2" to-layer="605" to-port="1" /> + <edge from-layer="605" from-port="2" to-layer="651" to-port="0" /> + <edge from-layer="605" from-port="2" to-layer="620" to-port="0" /> + <edge from-layer="605" from-port="2" to-layer="612" to-port="0" /> + <edge from-layer="606" from-port="0" to-layer="607" to-port="0" /> + <edge from-layer="607" from-port="1" to-layer="621" to-port="0" /> + <edge from-layer="608" from-port="0" to-layer="609" to-port="0" /> + <edge from-layer="609" from-port="1" to-layer="619" to-port="0" /> + <edge from-layer="610" from-port="0" to-layer="611" to-port="0" /> + <edge from-layer="611" from-port="1" to-layer="612" to-port="1" /> + <edge from-layer="612" from-port="2" to-layer="614" to-port="0" /> + <edge from-layer="613" from-port="0" to-layer="614" to-port="1" /> + <edge from-layer="614" from-port="2" to-layer="617" to-port="0" /> + <edge from-layer="615" from-port="0" to-layer="616" to-port="0" /> + <edge from-layer="616" from-port="1" to-layer="617" to-port="1" /> + <edge from-layer="617" from-port="2" to-layer="618" to-port="0" /> + <edge from-layer="618" from-port="1" to-layer="619" to-port="1" /> + <edge from-layer="619" from-port="2" to-layer="620" to-port="1" /> + <edge from-layer="620" from-port="2" to-layer="621" to-port="1" /> + <edge from-layer="621" from-port="2" to-layer="630" to-port="0" /> + <edge from-layer="621" from-port="2" to-layer="640" to-port="0" /> + <edge from-layer="622" from-port="0" to-layer="623" to-port="0" /> + <edge from-layer="623" from-port="1" to-layer="626" to-port="0" /> + <edge from-layer="624" from-port="0" to-layer="625" to-port="0" /> + <edge from-layer="625" from-port="1" to-layer="626" to-port="1" /> + <edge from-layer="626" from-port="2" to-layer="628" to-port="0" /> + <edge from-layer="627" from-port="0" to-layer="628" to-port="1" /> + <edge from-layer="628" from-port="2" to-layer="629" to-port="0" /> + <edge from-layer="629" from-port="1" to-layer="630" to-port="1" /> + <edge from-layer="630" from-port="2" to-layer="631" to-port="0" /> + <edge from-layer="631" from-port="1" to-layer="641" to-port="0" /> + <edge from-layer="632" from-port="0" to-layer="633" to-port="0" /> + <edge from-layer="633" from-port="1" to-layer="636" to-port="0" /> + <edge from-layer="634" from-port="0" to-layer="635" to-port="0" /> + <edge from-layer="635" from-port="1" to-layer="636" to-port="1" /> + <edge from-layer="636" from-port="2" to-layer="638" to-port="0" /> + <edge from-layer="637" from-port="0" to-layer="638" to-port="1" /> + <edge from-layer="638" from-port="2" to-layer="639" to-port="0" /> + <edge from-layer="639" from-port="1" to-layer="640" to-port="1" /> + <edge from-layer="640" from-port="2" to-layer="641" to-port="1" /> + <edge from-layer="641" from-port="2" to-layer="650" to-port="0" /> + <edge from-layer="642" from-port="0" to-layer="643" to-port="0" /> + <edge from-layer="643" from-port="1" to-layer="646" to-port="0" /> + <edge from-layer="644" from-port="0" to-layer="645" to-port="0" /> + <edge from-layer="645" from-port="1" to-layer="646" to-port="1" /> + <edge from-layer="646" from-port="2" to-layer="648" to-port="0" /> + <edge from-layer="647" from-port="0" to-layer="648" to-port="1" /> + <edge from-layer="648" from-port="2" to-layer="649" to-port="0" /> + <edge from-layer="649" from-port="1" to-layer="650" to-port="1" /> + <edge from-layer="650" from-port="2" to-layer="651" to-port="1" /> + <edge from-layer="651" from-port="2" to-layer="658" to-port="0" /> + <edge from-layer="651" from-port="2" to-layer="666" to-port="0" /> + <edge from-layer="651" from-port="2" to-layer="856" to-port="0" /> + <edge from-layer="652" from-port="0" to-layer="653" to-port="0" /> + <edge from-layer="653" from-port="1" to-layer="667" to-port="0" /> + <edge from-layer="654" from-port="0" to-layer="655" to-port="0" /> + <edge from-layer="655" from-port="1" to-layer="665" to-port="0" /> + <edge from-layer="656" from-port="0" to-layer="657" to-port="0" /> + <edge from-layer="657" from-port="1" to-layer="658" to-port="1" /> + <edge from-layer="658" from-port="2" to-layer="660" to-port="0" /> + <edge from-layer="659" from-port="0" to-layer="660" to-port="1" /> + <edge from-layer="660" from-port="2" to-layer="663" to-port="0" /> + <edge from-layer="661" from-port="0" to-layer="662" to-port="0" /> + <edge from-layer="662" from-port="1" to-layer="663" to-port="1" /> + <edge from-layer="663" from-port="2" to-layer="664" to-port="0" /> + <edge from-layer="664" from-port="1" to-layer="665" to-port="1" /> + <edge from-layer="665" from-port="2" to-layer="666" to-port="1" /> + <edge from-layer="666" from-port="2" to-layer="667" to-port="1" /> + <edge from-layer="667" from-port="2" to-layer="676" to-port="0" /> + <edge from-layer="667" from-port="2" to-layer="738" to-port="0" /> + <edge from-layer="667" from-port="2" to-layer="806" to-port="0" /> + <edge from-layer="667" from-port="2" to-layer="677" to-port="0" /> + <edge from-layer="668" from-port="0" to-layer="669" to-port="0" /> + <edge from-layer="669" from-port="1" to-layer="672" to-port="0" /> + <edge from-layer="670" from-port="0" to-layer="671" to-port="0" /> + <edge from-layer="671" from-port="1" to-layer="672" to-port="1" /> + <edge from-layer="672" from-port="2" to-layer="674" to-port="0" /> + <edge from-layer="673" from-port="0" to-layer="674" to-port="1" /> + <edge from-layer="674" from-port="2" to-layer="675" to-port="0" /> + <edge from-layer="675" from-port="1" to-layer="676" to-port="1" /> + <edge from-layer="676" from-port="2" to-layer="682" to-port="0" /> + <edge from-layer="677" from-port="1" to-layer="680" to-port="0" /> + <edge from-layer="678" from-port="0" to-layer="680" to-port="1" /> + <edge from-layer="679" from-port="0" to-layer="680" to-port="2" /> + <edge from-layer="680" from-port="3" to-layer="845" to-port="0" /> + <edge from-layer="680" from-port="3" to-layer="681" to-port="0" /> + <edge from-layer="680" from-port="3" to-layer="739" to-port="0" /> + <edge from-layer="681" from-port="3" to-layer="682" to-port="1" /> + <edge from-layer="682" from-port="2" to-layer="684" to-port="0" /> + <edge from-layer="683" from-port="0" to-layer="684" to-port="1" /> + <edge from-layer="684" from-port="2" to-layer="708" to-port="0" /> + <edge from-layer="684" from-port="2" to-layer="714" to-port="0" /> + <edge from-layer="684" from-port="2" to-layer="697" to-port="0" /> + <edge from-layer="684" from-port="2" to-layer="694" to-port="0" /> + <edge from-layer="685" from-port="0" to-layer="686" to-port="0" /> + <edge from-layer="686" from-port="1" to-layer="687" to-port="0" /> + <edge from-layer="687" from-port="2" to-layer="688" to-port="0" /> + <edge from-layer="688" from-port="2" to-layer="690" to-port="0" /> + <edge from-layer="689" from-port="0" to-layer="690" to-port="1" /> + <edge from-layer="690" from-port="2" to-layer="691" to-port="0" /> + <edge from-layer="690" from-port="2" to-layer="691" to-port="1" /> + <edge from-layer="691" from-port="2" to-layer="692" to-port="0" /> + <edge from-layer="691" from-port="2" to-layer="716" to-port="0" /> + <edge from-layer="692" from-port="1" to-layer="693" to-port="0" /> + <edge from-layer="693" from-port="2" to-layer="694" to-port="1" /> + <edge from-layer="693" from-port="2" to-layer="743" to-port="1" /> + <edge from-layer="694" from-port="2" to-layer="719" to-port="0" /> + <edge from-layer="695" from-port="0" to-layer="705" to-port="0" /> + <edge from-layer="696" from-port="0" to-layer="705" to-port="1" /> + <edge from-layer="697" from-port="1" to-layer="699" to-port="0" /> + <edge from-layer="698" from-port="0" to-layer="699" to-port="2" /> + <edge from-layer="699" from-port="3" to-layer="700" to-port="0" /> + <edge from-layer="700" from-port="2" to-layer="701" to-port="0" /> + <edge from-layer="701" from-port="1" to-layer="703" to-port="0" /> + <edge from-layer="702" from-port="0" to-layer="703" to-port="1" /> + <edge from-layer="703" from-port="2" to-layer="705" to-port="2" /> + <edge from-layer="704" from-port="0" to-layer="705" to-port="3" /> + <edge from-layer="705" from-port="4" to-layer="708" to-port="1" /> + <edge from-layer="705" from-port="4" to-layer="714" to-port="2" /> + <edge from-layer="706" from-port="0" to-layer="708" to-port="2" /> + <edge from-layer="707" from-port="0" to-layer="708" to-port="3" /> + <edge from-layer="708" from-port="4" to-layer="711" to-port="0" /> + <edge from-layer="709" from-port="0" to-layer="710" to-port="0" /> + <edge from-layer="710" from-port="1" to-layer="711" to-port="1" /> + <edge from-layer="711" from-port="2" to-layer="715" to-port="0" /> + <edge from-layer="712" from-port="0" to-layer="714" to-port="1" /> + <edge from-layer="713" from-port="0" to-layer="714" to-port="3" /> + <edge from-layer="714" from-port="4" to-layer="715" to-port="1" /> + <edge from-layer="715" from-port="2" to-layer="718" to-port="0" /> + <edge from-layer="716" from-port="1" to-layer="717" to-port="0" /> + <edge from-layer="717" from-port="2" to-layer="718" to-port="1" /> + <edge from-layer="717" from-port="2" to-layer="765" to-port="1" /> + <edge from-layer="718" from-port="2" to-layer="719" to-port="1" /> + <edge from-layer="719" from-port="2" to-layer="842" to-port="0" /> + <edge from-layer="720" from-port="0" to-layer="721" to-port="0" /> + <edge from-layer="721" from-port="1" to-layer="726" to-port="0" /> + <edge from-layer="722" from-port="0" to-layer="725" to-port="1" /> + <edge from-layer="723" from-port="0" to-layer="725" to-port="2" /> + <edge from-layer="724" from-port="0" to-layer="725" to-port="3" /> + <edge from-layer="725" from-port="4" to-layer="726" to-port="1" /> + <edge from-layer="726" from-port="2" to-layer="727" to-port="0" /> + <edge from-layer="727" from-port="1" to-layer="729" to-port="0" /> + <edge from-layer="728" from-port="0" to-layer="729" to-port="2" /> + <edge from-layer="729" from-port="3" to-layer="767" to-port="0" /> + <edge from-layer="730" from-port="0" to-layer="731" to-port="0" /> + <edge from-layer="731" from-port="1" to-layer="734" to-port="0" /> + <edge from-layer="732" from-port="0" to-layer="733" to-port="0" /> + <edge from-layer="733" from-port="1" to-layer="734" to-port="1" /> + <edge from-layer="734" from-port="2" to-layer="736" to-port="0" /> + <edge from-layer="735" from-port="0" to-layer="736" to-port="1" /> + <edge from-layer="736" from-port="2" to-layer="737" to-port="0" /> + <edge from-layer="737" from-port="1" to-layer="738" to-port="1" /> + <edge from-layer="738" from-port="2" to-layer="740" to-port="0" /> + <edge from-layer="739" from-port="3" to-layer="740" to-port="1" /> + <edge from-layer="739" from-port="3" to-layer="807" to-port="1" /> + <edge from-layer="740" from-port="2" to-layer="742" to-port="0" /> + <edge from-layer="741" from-port="0" to-layer="742" to-port="1" /> + <edge from-layer="742" from-port="2" to-layer="746" to-port="0" /> + <edge from-layer="742" from-port="2" to-layer="743" to-port="0" /> + <edge from-layer="742" from-port="2" to-layer="763" to-port="0" /> + <edge from-layer="742" from-port="2" to-layer="757" to-port="0" /> + <edge from-layer="743" from-port="2" to-layer="766" to-port="0" /> + <edge from-layer="744" from-port="0" to-layer="754" to-port="0" /> + <edge from-layer="745" from-port="0" to-layer="754" to-port="1" /> + <edge from-layer="746" from-port="1" to-layer="748" to-port="0" /> + <edge from-layer="747" from-port="0" to-layer="748" to-port="2" /> + <edge from-layer="748" from-port="3" to-layer="749" to-port="0" /> + <edge from-layer="749" from-port="2" to-layer="750" to-port="0" /> + <edge from-layer="750" from-port="1" to-layer="752" to-port="0" /> + <edge from-layer="751" from-port="0" to-layer="752" to-port="1" /> + <edge from-layer="752" from-port="2" to-layer="754" to-port="2" /> + <edge from-layer="753" from-port="0" to-layer="754" to-port="3" /> + <edge from-layer="754" from-port="4" to-layer="757" to-port="1" /> + <edge from-layer="754" from-port="4" to-layer="763" to-port="2" /> + <edge from-layer="755" from-port="0" to-layer="757" to-port="2" /> + <edge from-layer="756" from-port="0" to-layer="757" to-port="3" /> + <edge from-layer="757" from-port="4" to-layer="760" to-port="0" /> + <edge from-layer="758" from-port="0" to-layer="759" to-port="0" /> + <edge from-layer="759" from-port="1" to-layer="760" to-port="1" /> + <edge from-layer="760" from-port="2" to-layer="764" to-port="0" /> + <edge from-layer="761" from-port="0" to-layer="763" to-port="1" /> + <edge from-layer="762" from-port="0" to-layer="763" to-port="3" /> + <edge from-layer="763" from-port="4" to-layer="764" to-port="1" /> + <edge from-layer="764" from-port="2" to-layer="765" to-port="0" /> + <edge from-layer="765" from-port="2" to-layer="766" to-port="1" /> + <edge from-layer="766" from-port="2" to-layer="767" to-port="1" /> + <edge from-layer="767" from-port="2" to-layer="768" to-port="0" /> + <edge from-layer="767" from-port="2" to-layer="5530" to-port="0" /> + <edge from-layer="767" from-port="2" to-layer="769" to-port="0" /> + <edge from-layer="768" from-port="2" to-layer="782" to-port="0" /> + <edge from-layer="769" from-port="1" to-layer="774" to-port="0" /> + <edge from-layer="769" from-port="1" to-layer="772" to-port="0" /> + <edge from-layer="769" from-port="1" to-layer="779" to-port="0" /> + <edge from-layer="770" from-port="0" to-layer="772" to-port="1" /> + <edge from-layer="771" from-port="0" to-layer="772" to-port="2" /> + <edge from-layer="772" from-port="3" to-layer="786" to-port="0" /> + <edge from-layer="772" from-port="3" to-layer="780" to-port="0" /> + <edge from-layer="773" from-port="0" to-layer="774" to-port="2" /> + <edge from-layer="774" from-port="3" to-layer="776" to-port="0" /> + <edge from-layer="774" from-port="3" to-layer="783" to-port="0" /> + <edge from-layer="775" from-port="0" to-layer="776" to-port="1" /> + <edge from-layer="776" from-port="2" to-layer="780" to-port="1" /> + <edge from-layer="777" from-port="0" to-layer="779" to-port="1" /> + <edge from-layer="778" from-port="0" to-layer="779" to-port="2" /> + <edge from-layer="779" from-port="3" to-layer="780" to-port="3" /> + <edge from-layer="779" from-port="3" to-layer="786" to-port="2" /> + <edge from-layer="780" from-port="4" to-layer="781" to-port="0" /> + <edge from-layer="781" from-port="1" to-layer="782" to-port="1" /> + <edge from-layer="782" from-port="2" to-layer="787" to-port="0" /> + <edge from-layer="783" from-port="2" to-layer="785" to-port="0" /> + <edge from-layer="784" from-port="0" to-layer="785" to-port="1" /> + <edge from-layer="785" from-port="2" to-layer="786" to-port="1" /> + <edge from-layer="786" from-port="3" to-layer="787" to-port="1" /> + <edge from-layer="787" from-port="2" to-layer="842" to-port="1" /> + <edge from-layer="787" from-port="2" to-layer="834" to-port="0" /> + <edge from-layer="788" from-port="0" to-layer="789" to-port="0" /> + <edge from-layer="789" from-port="1" to-layer="794" to-port="0" /> + <edge from-layer="790" from-port="0" to-layer="793" to-port="1" /> + <edge from-layer="791" from-port="0" to-layer="793" to-port="2" /> + <edge from-layer="792" from-port="0" to-layer="793" to-port="3" /> + <edge from-layer="793" from-port="4" to-layer="794" to-port="1" /> + <edge from-layer="794" from-port="2" to-layer="795" to-port="0" /> + <edge from-layer="795" from-port="1" to-layer="797" to-port="0" /> + <edge from-layer="796" from-port="0" to-layer="797" to-port="2" /> + <edge from-layer="797" from-port="3" to-layer="810" to-port="0" /> + <edge from-layer="798" from-port="0" to-layer="799" to-port="0" /> + <edge from-layer="799" from-port="1" to-layer="802" to-port="0" /> + <edge from-layer="800" from-port="0" to-layer="801" to-port="0" /> + <edge from-layer="801" from-port="1" to-layer="802" to-port="1" /> + <edge from-layer="802" from-port="2" to-layer="804" to-port="0" /> + <edge from-layer="803" from-port="0" to-layer="804" to-port="1" /> + <edge from-layer="804" from-port="2" to-layer="805" to-port="0" /> + <edge from-layer="805" from-port="1" to-layer="806" to-port="1" /> + <edge from-layer="806" from-port="2" to-layer="807" to-port="0" /> + <edge from-layer="807" from-port="2" to-layer="809" to-port="0" /> + <edge from-layer="808" from-port="0" to-layer="809" to-port="1" /> + <edge from-layer="809" from-port="2" to-layer="810" to-port="1" /> + <edge from-layer="810" from-port="2" to-layer="811" to-port="0" /> + <edge from-layer="810" from-port="2" to-layer="812" to-port="0" /> + <edge from-layer="810" from-port="2" to-layer="5529" to-port="0" /> + <edge from-layer="811" from-port="2" to-layer="825" to-port="0" /> + <edge from-layer="812" from-port="1" to-layer="815" to-port="0" /> + <edge from-layer="812" from-port="1" to-layer="822" to-port="0" /> + <edge from-layer="812" from-port="1" to-layer="817" to-port="0" /> + <edge from-layer="813" from-port="0" to-layer="815" to-port="1" /> + <edge from-layer="814" from-port="0" to-layer="815" to-port="2" /> + <edge from-layer="815" from-port="3" to-layer="829" to-port="0" /> + <edge from-layer="815" from-port="3" to-layer="823" to-port="0" /> + <edge from-layer="816" from-port="0" to-layer="817" to-port="2" /> + <edge from-layer="817" from-port="3" to-layer="826" to-port="0" /> + <edge from-layer="817" from-port="3" to-layer="819" to-port="0" /> + <edge from-layer="818" from-port="0" to-layer="819" to-port="1" /> + <edge from-layer="819" from-port="2" to-layer="823" to-port="1" /> + <edge from-layer="820" from-port="0" to-layer="822" to-port="1" /> + <edge from-layer="821" from-port="0" to-layer="822" to-port="2" /> + <edge from-layer="822" from-port="3" to-layer="823" to-port="3" /> + <edge from-layer="822" from-port="3" to-layer="829" to-port="2" /> + <edge from-layer="823" from-port="4" to-layer="824" to-port="0" /> + <edge from-layer="824" from-port="1" to-layer="825" to-port="1" /> + <edge from-layer="825" from-port="2" to-layer="830" to-port="0" /> + <edge from-layer="826" from-port="2" to-layer="828" to-port="0" /> + <edge from-layer="827" from-port="0" to-layer="828" to-port="1" /> + <edge from-layer="828" from-port="2" to-layer="829" to-port="1" /> + <edge from-layer="829" from-port="3" to-layer="830" to-port="1" /> + <edge from-layer="830" from-port="2" to-layer="842" to-port="2" /> + <edge from-layer="831" from-port="0" to-layer="841" to-port="1" /> + <edge from-layer="832" from-port="0" to-layer="839" to-port="0" /> + <edge from-layer="833" from-port="0" to-layer="839" to-port="1" /> + <edge from-layer="834" from-port="1" to-layer="837" to-port="0" /> + <edge from-layer="835" from-port="0" to-layer="837" to-port="1" /> + <edge from-layer="836" from-port="0" to-layer="837" to-port="2" /> + <edge from-layer="837" from-port="3" to-layer="839" to-port="2" /> + <edge from-layer="838" from-port="0" to-layer="839" to-port="3" /> + <edge from-layer="839" from-port="4" to-layer="841" to-port="2" /> + <edge from-layer="840" from-port="0" to-layer="841" to-port="3" /> + <edge from-layer="841" from-port="4" to-layer="842" to-port="3" /> + <edge from-layer="842" from-port="4" to-layer="844" to-port="0" /> + <edge from-layer="843" from-port="0" to-layer="844" to-port="1" /> + <edge from-layer="844" from-port="2" to-layer="846" to-port="0" /> + <edge from-layer="845" from-port="2" to-layer="846" to-port="1" /> + <edge from-layer="846" from-port="2" to-layer="855" to-port="0" /> + <edge from-layer="847" from-port="0" to-layer="848" to-port="0" /> + <edge from-layer="848" from-port="1" to-layer="851" to-port="0" /> + <edge from-layer="849" from-port="0" to-layer="850" to-port="0" /> + <edge from-layer="850" from-port="1" to-layer="851" to-port="1" /> + <edge from-layer="851" from-port="2" to-layer="853" to-port="0" /> + <edge from-layer="852" from-port="0" to-layer="853" to-port="1" /> + <edge from-layer="853" from-port="2" to-layer="854" to-port="0" /> + <edge from-layer="854" from-port="1" to-layer="855" to-port="1" /> + <edge from-layer="855" from-port="2" to-layer="856" to-port="1" /> + <edge from-layer="856" from-port="2" to-layer="902" to-port="0" /> + <edge from-layer="856" from-port="2" to-layer="871" to-port="0" /> + <edge from-layer="856" from-port="2" to-layer="863" to-port="0" /> + <edge from-layer="857" from-port="0" to-layer="858" to-port="0" /> + <edge from-layer="858" from-port="1" to-layer="872" to-port="0" /> + <edge from-layer="859" from-port="0" to-layer="860" to-port="0" /> + <edge from-layer="860" from-port="1" to-layer="870" to-port="0" /> + <edge from-layer="861" from-port="0" to-layer="862" to-port="0" /> + <edge from-layer="862" from-port="1" to-layer="863" to-port="1" /> + <edge from-layer="863" from-port="2" to-layer="865" to-port="0" /> + <edge from-layer="864" from-port="0" to-layer="865" to-port="1" /> + <edge from-layer="865" from-port="2" to-layer="868" to-port="0" /> + <edge from-layer="866" from-port="0" to-layer="867" to-port="0" /> + <edge from-layer="867" from-port="1" to-layer="868" to-port="1" /> + <edge from-layer="868" from-port="2" to-layer="869" to-port="0" /> + <edge from-layer="869" from-port="1" to-layer="870" to-port="1" /> + <edge from-layer="870" from-port="2" to-layer="871" to-port="1" /> + <edge from-layer="871" from-port="2" to-layer="872" to-port="1" /> + <edge from-layer="872" from-port="2" to-layer="891" to-port="0" /> + <edge from-layer="872" from-port="2" to-layer="881" to-port="0" /> + <edge from-layer="873" from-port="0" to-layer="874" to-port="0" /> + <edge from-layer="874" from-port="1" to-layer="877" to-port="0" /> + <edge from-layer="875" from-port="0" to-layer="876" to-port="0" /> + <edge from-layer="876" from-port="1" to-layer="877" to-port="1" /> + <edge from-layer="877" from-port="2" to-layer="879" to-port="0" /> + <edge from-layer="878" from-port="0" to-layer="879" to-port="1" /> + <edge from-layer="879" from-port="2" to-layer="880" to-port="0" /> + <edge from-layer="880" from-port="1" to-layer="881" to-port="1" /> + <edge from-layer="881" from-port="2" to-layer="882" to-port="0" /> + <edge from-layer="882" from-port="1" to-layer="892" to-port="0" /> + <edge from-layer="883" from-port="0" to-layer="884" to-port="0" /> + <edge from-layer="884" from-port="1" to-layer="887" to-port="0" /> + <edge from-layer="885" from-port="0" to-layer="886" to-port="0" /> + <edge from-layer="886" from-port="1" to-layer="887" to-port="1" /> + <edge from-layer="887" from-port="2" to-layer="889" to-port="0" /> + <edge from-layer="888" from-port="0" to-layer="889" to-port="1" /> + <edge from-layer="889" from-port="2" to-layer="890" to-port="0" /> + <edge from-layer="890" from-port="1" to-layer="891" to-port="1" /> + <edge from-layer="891" from-port="2" to-layer="892" to-port="1" /> + <edge from-layer="892" from-port="2" to-layer="901" to-port="0" /> + <edge from-layer="893" from-port="0" to-layer="894" to-port="0" /> + <edge from-layer="894" from-port="1" to-layer="897" to-port="0" /> + <edge from-layer="895" from-port="0" to-layer="896" to-port="0" /> + <edge from-layer="896" from-port="1" to-layer="897" to-port="1" /> + <edge from-layer="897" from-port="2" to-layer="899" to-port="0" /> + <edge from-layer="898" from-port="0" to-layer="899" to-port="1" /> + <edge from-layer="899" from-port="2" to-layer="900" to-port="0" /> + <edge from-layer="900" from-port="1" to-layer="901" to-port="1" /> + <edge from-layer="901" from-port="2" to-layer="902" to-port="1" /> + <edge from-layer="902" from-port="2" to-layer="917" to-port="0" /> + <edge from-layer="902" from-port="2" to-layer="1107" to-port="0" /> + <edge from-layer="902" from-port="2" to-layer="909" to-port="0" /> + <edge from-layer="903" from-port="0" to-layer="904" to-port="0" /> + <edge from-layer="904" from-port="1" to-layer="918" to-port="0" /> + <edge from-layer="905" from-port="0" to-layer="906" to-port="0" /> + <edge from-layer="906" from-port="1" to-layer="916" to-port="0" /> + <edge from-layer="907" from-port="0" to-layer="908" to-port="0" /> + <edge from-layer="908" from-port="1" to-layer="909" to-port="1" /> + <edge from-layer="909" from-port="2" to-layer="911" to-port="0" /> + <edge from-layer="910" from-port="0" to-layer="911" to-port="1" /> + <edge from-layer="911" from-port="2" to-layer="914" to-port="0" /> + <edge from-layer="912" from-port="0" to-layer="913" to-port="0" /> + <edge from-layer="913" from-port="1" to-layer="914" to-port="1" /> + <edge from-layer="914" from-port="2" to-layer="915" to-port="0" /> + <edge from-layer="915" from-port="1" to-layer="916" to-port="1" /> + <edge from-layer="916" from-port="2" to-layer="917" to-port="1" /> + <edge from-layer="917" from-port="2" to-layer="918" to-port="1" /> + <edge from-layer="918" from-port="2" to-layer="927" to-port="0" /> + <edge from-layer="918" from-port="2" to-layer="1057" to-port="0" /> + <edge from-layer="918" from-port="2" to-layer="928" to-port="0" /> + <edge from-layer="918" from-port="2" to-layer="989" to-port="0" /> + <edge from-layer="919" from-port="0" to-layer="920" to-port="0" /> + <edge from-layer="920" from-port="1" to-layer="923" to-port="0" /> + <edge from-layer="921" from-port="0" to-layer="922" to-port="0" /> + <edge from-layer="922" from-port="1" to-layer="923" to-port="1" /> + <edge from-layer="923" from-port="2" to-layer="925" to-port="0" /> + <edge from-layer="924" from-port="0" to-layer="925" to-port="1" /> + <edge from-layer="925" from-port="2" to-layer="926" to-port="0" /> + <edge from-layer="926" from-port="1" to-layer="927" to-port="1" /> + <edge from-layer="927" from-port="2" to-layer="933" to-port="0" /> + <edge from-layer="928" from-port="1" to-layer="931" to-port="0" /> + <edge from-layer="929" from-port="0" to-layer="931" to-port="1" /> + <edge from-layer="930" from-port="0" to-layer="931" to-port="2" /> + <edge from-layer="931" from-port="3" to-layer="1096" to-port="0" /> + <edge from-layer="931" from-port="3" to-layer="990" to-port="0" /> + <edge from-layer="931" from-port="3" to-layer="932" to-port="0" /> + <edge from-layer="932" from-port="3" to-layer="933" to-port="1" /> + <edge from-layer="933" from-port="2" to-layer="935" to-port="0" /> + <edge from-layer="934" from-port="0" to-layer="935" to-port="1" /> + <edge from-layer="935" from-port="2" to-layer="945" to-port="0" /> + <edge from-layer="935" from-port="2" to-layer="948" to-port="0" /> + <edge from-layer="935" from-port="2" to-layer="959" to-port="0" /> + <edge from-layer="935" from-port="2" to-layer="965" to-port="0" /> + <edge from-layer="936" from-port="0" to-layer="937" to-port="0" /> + <edge from-layer="937" from-port="1" to-layer="938" to-port="0" /> + <edge from-layer="938" from-port="2" to-layer="939" to-port="0" /> + <edge from-layer="939" from-port="2" to-layer="941" to-port="0" /> + <edge from-layer="940" from-port="0" to-layer="941" to-port="1" /> + <edge from-layer="941" from-port="2" to-layer="942" to-port="1" /> + <edge from-layer="941" from-port="2" to-layer="942" to-port="0" /> + <edge from-layer="942" from-port="2" to-layer="943" to-port="0" /> + <edge from-layer="942" from-port="2" to-layer="967" to-port="0" /> + <edge from-layer="943" from-port="1" to-layer="944" to-port="0" /> + <edge from-layer="944" from-port="2" to-layer="945" to-port="1" /> + <edge from-layer="944" from-port="2" to-layer="994" to-port="1" /> + <edge from-layer="945" from-port="2" to-layer="970" to-port="0" /> + <edge from-layer="946" from-port="0" to-layer="956" to-port="0" /> + <edge from-layer="947" from-port="0" to-layer="956" to-port="1" /> + <edge from-layer="948" from-port="1" to-layer="950" to-port="0" /> + <edge from-layer="949" from-port="0" to-layer="950" to-port="2" /> + <edge from-layer="950" from-port="3" to-layer="951" to-port="0" /> + <edge from-layer="951" from-port="2" to-layer="952" to-port="0" /> + <edge from-layer="952" from-port="1" to-layer="954" to-port="0" /> + <edge from-layer="953" from-port="0" to-layer="954" to-port="1" /> + <edge from-layer="954" from-port="2" to-layer="956" to-port="2" /> + <edge from-layer="955" from-port="0" to-layer="956" to-port="3" /> + <edge from-layer="956" from-port="4" to-layer="965" to-port="2" /> + <edge from-layer="956" from-port="4" to-layer="959" to-port="1" /> + <edge from-layer="957" from-port="0" to-layer="959" to-port="2" /> + <edge from-layer="958" from-port="0" to-layer="959" to-port="3" /> + <edge from-layer="959" from-port="4" to-layer="962" to-port="0" /> + <edge from-layer="960" from-port="0" to-layer="961" to-port="0" /> + <edge from-layer="961" from-port="1" to-layer="962" to-port="1" /> + <edge from-layer="962" from-port="2" to-layer="966" to-port="0" /> + <edge from-layer="963" from-port="0" to-layer="965" to-port="1" /> + <edge from-layer="964" from-port="0" to-layer="965" to-port="3" /> + <edge from-layer="965" from-port="4" to-layer="966" to-port="1" /> + <edge from-layer="966" from-port="2" to-layer="969" to-port="0" /> + <edge from-layer="967" from-port="1" to-layer="968" to-port="0" /> + <edge from-layer="968" from-port="2" to-layer="969" to-port="1" /> + <edge from-layer="968" from-port="2" to-layer="1016" to-port="1" /> + <edge from-layer="969" from-port="2" to-layer="970" to-port="1" /> + <edge from-layer="970" from-port="2" to-layer="1093" to-port="0" /> + <edge from-layer="971" from-port="0" to-layer="972" to-port="0" /> + <edge from-layer="972" from-port="1" to-layer="977" to-port="0" /> + <edge from-layer="973" from-port="0" to-layer="976" to-port="1" /> + <edge from-layer="974" from-port="0" to-layer="976" to-port="2" /> + <edge from-layer="975" from-port="0" to-layer="976" to-port="3" /> + <edge from-layer="976" from-port="4" to-layer="977" to-port="1" /> + <edge from-layer="977" from-port="2" to-layer="978" to-port="0" /> + <edge from-layer="978" from-port="1" to-layer="980" to-port="0" /> + <edge from-layer="979" from-port="0" to-layer="980" to-port="2" /> + <edge from-layer="980" from-port="3" to-layer="1018" to-port="0" /> + <edge from-layer="981" from-port="0" to-layer="982" to-port="0" /> + <edge from-layer="982" from-port="1" to-layer="985" to-port="0" /> + <edge from-layer="983" from-port="0" to-layer="984" to-port="0" /> + <edge from-layer="984" from-port="1" to-layer="985" to-port="1" /> + <edge from-layer="985" from-port="2" to-layer="987" to-port="0" /> + <edge from-layer="986" from-port="0" to-layer="987" to-port="1" /> + <edge from-layer="987" from-port="2" to-layer="988" to-port="0" /> + <edge from-layer="988" from-port="1" to-layer="989" to-port="1" /> + <edge from-layer="989" from-port="2" to-layer="991" to-port="0" /> + <edge from-layer="990" from-port="3" to-layer="991" to-port="1" /> + <edge from-layer="990" from-port="3" to-layer="1058" to-port="1" /> + <edge from-layer="991" from-port="2" to-layer="993" to-port="0" /> + <edge from-layer="992" from-port="0" to-layer="993" to-port="1" /> + <edge from-layer="993" from-port="2" to-layer="997" to-port="0" /> + <edge from-layer="993" from-port="2" to-layer="994" to-port="0" /> + <edge from-layer="993" from-port="2" to-layer="1008" to-port="0" /> + <edge from-layer="993" from-port="2" to-layer="1014" to-port="0" /> + <edge from-layer="994" from-port="2" to-layer="1017" to-port="0" /> + <edge from-layer="995" from-port="0" to-layer="1005" to-port="0" /> + <edge from-layer="996" from-port="0" to-layer="1005" to-port="1" /> + <edge from-layer="997" from-port="1" to-layer="999" to-port="0" /> + <edge from-layer="998" from-port="0" to-layer="999" to-port="2" /> + <edge from-layer="999" from-port="3" to-layer="1000" to-port="0" /> + <edge from-layer="1000" from-port="2" to-layer="1001" to-port="0" /> + <edge from-layer="1001" from-port="1" to-layer="1003" to-port="0" /> + <edge from-layer="1002" from-port="0" to-layer="1003" to-port="1" /> + <edge from-layer="1003" from-port="2" to-layer="1005" to-port="2" /> + <edge from-layer="1004" from-port="0" to-layer="1005" to-port="3" /> + <edge from-layer="1005" from-port="4" to-layer="1014" to-port="2" /> + <edge from-layer="1005" from-port="4" to-layer="1008" to-port="1" /> + <edge from-layer="1006" from-port="0" to-layer="1008" to-port="2" /> + <edge from-layer="1007" from-port="0" to-layer="1008" to-port="3" /> + <edge from-layer="1008" from-port="4" to-layer="1011" to-port="0" /> + <edge from-layer="1009" from-port="0" to-layer="1010" to-port="0" /> + <edge from-layer="1010" from-port="1" to-layer="1011" to-port="1" /> + <edge from-layer="1011" from-port="2" to-layer="1015" to-port="0" /> + <edge from-layer="1012" from-port="0" to-layer="1014" to-port="1" /> + <edge from-layer="1013" from-port="0" to-layer="1014" to-port="3" /> + <edge from-layer="1014" from-port="4" to-layer="1015" to-port="1" /> + <edge from-layer="1015" from-port="2" to-layer="1016" to-port="0" /> + <edge from-layer="1016" from-port="2" to-layer="1017" to-port="1" /> + <edge from-layer="1017" from-port="2" to-layer="1018" to-port="1" /> + <edge from-layer="1018" from-port="2" to-layer="1019" to-port="0" /> + <edge from-layer="1018" from-port="2" to-layer="2512" to-port="0" /> + <edge from-layer="1018" from-port="2" to-layer="1020" to-port="0" /> + <edge from-layer="1019" from-port="2" to-layer="1033" to-port="0" /> + <edge from-layer="1020" from-port="1" to-layer="1023" to-port="0" /> + <edge from-layer="1020" from-port="1" to-layer="1025" to-port="0" /> + <edge from-layer="1020" from-port="1" to-layer="1030" to-port="0" /> + <edge from-layer="1021" from-port="0" to-layer="1023" to-port="1" /> + <edge from-layer="1022" from-port="0" to-layer="1023" to-port="2" /> + <edge from-layer="1023" from-port="3" to-layer="1037" to-port="0" /> + <edge from-layer="1023" from-port="3" to-layer="1031" to-port="0" /> + <edge from-layer="1024" from-port="0" to-layer="1025" to-port="2" /> + <edge from-layer="1025" from-port="3" to-layer="1027" to-port="0" /> + <edge from-layer="1025" from-port="3" to-layer="1034" to-port="0" /> + <edge from-layer="1026" from-port="0" to-layer="1027" to-port="1" /> + <edge from-layer="1027" from-port="2" to-layer="1031" to-port="1" /> + <edge from-layer="1028" from-port="0" to-layer="1030" to-port="1" /> + <edge from-layer="1029" from-port="0" to-layer="1030" to-port="2" /> + <edge from-layer="1030" from-port="3" to-layer="1031" to-port="3" /> + <edge from-layer="1030" from-port="3" to-layer="1037" to-port="2" /> + <edge from-layer="1031" from-port="4" to-layer="1032" to-port="0" /> + <edge from-layer="1032" from-port="1" to-layer="1033" to-port="1" /> + <edge from-layer="1033" from-port="2" to-layer="1038" to-port="0" /> + <edge from-layer="1034" from-port="2" to-layer="1036" to-port="0" /> + <edge from-layer="1035" from-port="0" to-layer="1036" to-port="1" /> + <edge from-layer="1036" from-port="2" to-layer="1037" to-port="1" /> + <edge from-layer="1037" from-port="3" to-layer="1038" to-port="1" /> + <edge from-layer="1038" from-port="2" to-layer="1093" to-port="1" /> + <edge from-layer="1038" from-port="2" to-layer="1085" to-port="0" /> + <edge from-layer="1039" from-port="0" to-layer="1040" to-port="0" /> + <edge from-layer="1040" from-port="1" to-layer="1045" to-port="0" /> + <edge from-layer="1041" from-port="0" to-layer="1044" to-port="1" /> + <edge from-layer="1042" from-port="0" to-layer="1044" to-port="2" /> + <edge from-layer="1043" from-port="0" to-layer="1044" to-port="3" /> + <edge from-layer="1044" from-port="4" to-layer="1045" to-port="1" /> + <edge from-layer="1045" from-port="2" to-layer="1046" to-port="0" /> + <edge from-layer="1046" from-port="1" to-layer="1048" to-port="0" /> + <edge from-layer="1047" from-port="0" to-layer="1048" to-port="2" /> + <edge from-layer="1048" from-port="3" to-layer="1061" to-port="0" /> + <edge from-layer="1049" from-port="0" to-layer="1050" to-port="0" /> + <edge from-layer="1050" from-port="1" to-layer="1053" to-port="0" /> + <edge from-layer="1051" from-port="0" to-layer="1052" to-port="0" /> + <edge from-layer="1052" from-port="1" to-layer="1053" to-port="1" /> + <edge from-layer="1053" from-port="2" to-layer="1055" to-port="0" /> + <edge from-layer="1054" from-port="0" to-layer="1055" to-port="1" /> + <edge from-layer="1055" from-port="2" to-layer="1056" to-port="0" /> + <edge from-layer="1056" from-port="1" to-layer="1057" to-port="1" /> + <edge from-layer="1057" from-port="2" to-layer="1058" to-port="0" /> + <edge from-layer="1058" from-port="2" to-layer="1060" to-port="0" /> + <edge from-layer="1059" from-port="0" to-layer="1060" to-port="1" /> + <edge from-layer="1060" from-port="2" to-layer="1061" to-port="1" /> + <edge from-layer="1061" from-port="2" to-layer="1062" to-port="0" /> + <edge from-layer="1061" from-port="2" to-layer="2511" to-port="0" /> + <edge from-layer="1061" from-port="2" to-layer="1063" to-port="0" /> + <edge from-layer="1062" from-port="2" to-layer="1076" to-port="0" /> + <edge from-layer="1063" from-port="1" to-layer="1068" to-port="0" /> + <edge from-layer="1063" from-port="1" to-layer="1073" to-port="0" /> + <edge from-layer="1063" from-port="1" to-layer="1066" to-port="0" /> + <edge from-layer="1064" from-port="0" to-layer="1066" to-port="1" /> + <edge from-layer="1065" from-port="0" to-layer="1066" to-port="2" /> + <edge from-layer="1066" from-port="3" to-layer="1080" to-port="0" /> + <edge from-layer="1066" from-port="3" to-layer="1074" to-port="0" /> + <edge from-layer="1067" from-port="0" to-layer="1068" to-port="2" /> + <edge from-layer="1068" from-port="3" to-layer="1070" to-port="0" /> + <edge from-layer="1068" from-port="3" to-layer="1077" to-port="0" /> + <edge from-layer="1069" from-port="0" to-layer="1070" to-port="1" /> + <edge from-layer="1070" from-port="2" to-layer="1074" to-port="1" /> + <edge from-layer="1071" from-port="0" to-layer="1073" to-port="1" /> + <edge from-layer="1072" from-port="0" to-layer="1073" to-port="2" /> + <edge from-layer="1073" from-port="3" to-layer="1080" to-port="2" /> + <edge from-layer="1073" from-port="3" to-layer="1074" to-port="3" /> + <edge from-layer="1074" from-port="4" to-layer="1075" to-port="0" /> + <edge from-layer="1075" from-port="1" to-layer="1076" to-port="1" /> + <edge from-layer="1076" from-port="2" to-layer="1081" to-port="0" /> + <edge from-layer="1077" from-port="2" to-layer="1079" to-port="0" /> + <edge from-layer="1078" from-port="0" to-layer="1079" to-port="1" /> + <edge from-layer="1079" from-port="2" to-layer="1080" to-port="1" /> + <edge from-layer="1080" from-port="3" to-layer="1081" to-port="1" /> + <edge from-layer="1081" from-port="2" to-layer="1093" to-port="2" /> + <edge from-layer="1082" from-port="0" to-layer="1092" to-port="1" /> + <edge from-layer="1083" from-port="0" to-layer="1090" to-port="0" /> + <edge from-layer="1084" from-port="0" to-layer="1090" to-port="1" /> + <edge from-layer="1085" from-port="1" to-layer="1088" to-port="0" /> + <edge from-layer="1086" from-port="0" to-layer="1088" to-port="1" /> + <edge from-layer="1087" from-port="0" to-layer="1088" to-port="2" /> + <edge from-layer="1088" from-port="3" to-layer="1090" to-port="2" /> + <edge from-layer="1089" from-port="0" to-layer="1090" to-port="3" /> + <edge from-layer="1090" from-port="4" to-layer="1092" to-port="2" /> + <edge from-layer="1091" from-port="0" to-layer="1092" to-port="3" /> + <edge from-layer="1092" from-port="4" to-layer="1093" to-port="3" /> + <edge from-layer="1093" from-port="4" to-layer="1095" to-port="0" /> + <edge from-layer="1094" from-port="0" to-layer="1095" to-port="1" /> + <edge from-layer="1095" from-port="2" to-layer="1097" to-port="0" /> + <edge from-layer="1096" from-port="2" to-layer="1097" to-port="1" /> + <edge from-layer="1097" from-port="2" to-layer="1106" to-port="0" /> + <edge from-layer="1098" from-port="0" to-layer="1099" to-port="0" /> + <edge from-layer="1099" from-port="1" to-layer="1102" to-port="0" /> + <edge from-layer="1100" from-port="0" to-layer="1101" to-port="0" /> + <edge from-layer="1101" from-port="1" to-layer="1102" to-port="1" /> + <edge from-layer="1102" from-port="2" to-layer="1104" to-port="0" /> + <edge from-layer="1103" from-port="0" to-layer="1104" to-port="1" /> + <edge from-layer="1104" from-port="2" to-layer="1105" to-port="0" /> + <edge from-layer="1105" from-port="1" to-layer="1106" to-port="1" /> + <edge from-layer="1106" from-port="2" to-layer="1107" to-port="1" /> + <edge from-layer="1107" from-port="2" to-layer="1114" to-port="0" /> + <edge from-layer="1107" from-port="2" to-layer="1153" to-port="0" /> + <edge from-layer="1107" from-port="2" to-layer="1122" to-port="0" /> + <edge from-layer="1108" from-port="0" to-layer="1109" to-port="0" /> + <edge from-layer="1109" from-port="1" to-layer="1123" to-port="0" /> + <edge from-layer="1110" from-port="0" to-layer="1111" to-port="0" /> + <edge from-layer="1111" from-port="1" to-layer="1121" to-port="0" /> + <edge from-layer="1112" from-port="0" to-layer="1113" to-port="0" /> + <edge from-layer="1113" from-port="1" to-layer="1114" to-port="1" /> + <edge from-layer="1114" from-port="2" to-layer="1116" to-port="0" /> + <edge from-layer="1115" from-port="0" to-layer="1116" to-port="1" /> + <edge from-layer="1116" from-port="2" to-layer="1119" to-port="0" /> + <edge from-layer="1117" from-port="0" to-layer="1118" to-port="0" /> + <edge from-layer="1118" from-port="1" to-layer="1119" to-port="1" /> + <edge from-layer="1119" from-port="2" to-layer="1120" to-port="0" /> + <edge from-layer="1120" from-port="1" to-layer="1121" to-port="1" /> + <edge from-layer="1121" from-port="2" to-layer="1122" to-port="1" /> + <edge from-layer="1122" from-port="2" to-layer="1123" to-port="1" /> + <edge from-layer="1123" from-port="2" to-layer="1142" to-port="0" /> + <edge from-layer="1123" from-port="2" to-layer="1132" to-port="0" /> + <edge from-layer="1124" from-port="0" to-layer="1125" to-port="0" /> + <edge from-layer="1125" from-port="1" to-layer="1128" to-port="0" /> + <edge from-layer="1126" from-port="0" to-layer="1127" to-port="0" /> + <edge from-layer="1127" from-port="1" to-layer="1128" to-port="1" /> + <edge from-layer="1128" from-port="2" to-layer="1130" to-port="0" /> + <edge from-layer="1129" from-port="0" to-layer="1130" to-port="1" /> + <edge from-layer="1130" from-port="2" to-layer="1131" to-port="0" /> + <edge from-layer="1131" from-port="1" to-layer="1132" to-port="1" /> + <edge from-layer="1132" from-port="2" to-layer="1133" to-port="0" /> + <edge from-layer="1133" from-port="1" to-layer="1143" to-port="0" /> + <edge from-layer="1134" from-port="0" to-layer="1135" to-port="0" /> + <edge from-layer="1135" from-port="1" to-layer="1138" to-port="0" /> + <edge from-layer="1136" from-port="0" to-layer="1137" to-port="0" /> + <edge from-layer="1137" from-port="1" to-layer="1138" to-port="1" /> + <edge from-layer="1138" from-port="2" to-layer="1140" to-port="0" /> + <edge from-layer="1139" from-port="0" to-layer="1140" to-port="1" /> + <edge from-layer="1140" from-port="2" to-layer="1141" to-port="0" /> + <edge from-layer="1141" from-port="1" to-layer="1142" to-port="1" /> + <edge from-layer="1142" from-port="2" to-layer="1143" to-port="1" /> + <edge from-layer="1143" from-port="2" to-layer="1152" to-port="0" /> + <edge from-layer="1144" from-port="0" to-layer="1145" to-port="0" /> + <edge from-layer="1145" from-port="1" to-layer="1148" to-port="0" /> + <edge from-layer="1146" from-port="0" to-layer="1147" to-port="0" /> + <edge from-layer="1147" from-port="1" to-layer="1148" to-port="1" /> + <edge from-layer="1148" from-port="2" to-layer="1150" to-port="0" /> + <edge from-layer="1149" from-port="0" to-layer="1150" to-port="1" /> + <edge from-layer="1150" from-port="2" to-layer="1151" to-port="0" /> + <edge from-layer="1151" from-port="1" to-layer="1152" to-port="1" /> + <edge from-layer="1152" from-port="2" to-layer="1153" to-port="1" /> + <edge from-layer="1153" from-port="2" to-layer="1160" to-port="0" /> + <edge from-layer="1153" from-port="2" to-layer="1358" to-port="0" /> + <edge from-layer="1153" from-port="2" to-layer="1168" to-port="0" /> + <edge from-layer="1154" from-port="0" to-layer="1155" to-port="0" /> + <edge from-layer="1155" from-port="1" to-layer="1169" to-port="0" /> + <edge from-layer="1156" from-port="0" to-layer="1157" to-port="0" /> + <edge from-layer="1157" from-port="1" to-layer="1167" to-port="0" /> + <edge from-layer="1158" from-port="0" to-layer="1159" to-port="0" /> + <edge from-layer="1159" from-port="1" to-layer="1160" to-port="1" /> + <edge from-layer="1160" from-port="2" to-layer="1162" to-port="0" /> + <edge from-layer="1161" from-port="0" to-layer="1162" to-port="1" /> + <edge from-layer="1162" from-port="2" to-layer="1165" to-port="0" /> + <edge from-layer="1163" from-port="0" to-layer="1164" to-port="0" /> + <edge from-layer="1164" from-port="1" to-layer="1165" to-port="1" /> + <edge from-layer="1165" from-port="2" to-layer="1166" to-port="0" /> + <edge from-layer="1166" from-port="1" to-layer="1167" to-port="1" /> + <edge from-layer="1167" from-port="2" to-layer="1168" to-port="1" /> + <edge from-layer="1168" from-port="2" to-layer="1169" to-port="1" /> + <edge from-layer="1169" from-port="2" to-layer="1240" to-port="0" /> + <edge from-layer="1169" from-port="2" to-layer="1308" to-port="0" /> + <edge from-layer="1169" from-port="2" to-layer="1178" to-port="0" /> + <edge from-layer="1169" from-port="2" to-layer="1179" to-port="0" /> + <edge from-layer="1170" from-port="0" to-layer="1171" to-port="0" /> + <edge from-layer="1171" from-port="1" to-layer="1174" to-port="0" /> + <edge from-layer="1172" from-port="0" to-layer="1173" to-port="0" /> + <edge from-layer="1173" from-port="1" to-layer="1174" to-port="1" /> + <edge from-layer="1174" from-port="2" to-layer="1176" to-port="0" /> + <edge from-layer="1175" from-port="0" to-layer="1176" to-port="1" /> + <edge from-layer="1176" from-port="2" to-layer="1177" to-port="0" /> + <edge from-layer="1177" from-port="1" to-layer="1178" to-port="1" /> + <edge from-layer="1178" from-port="2" to-layer="1184" to-port="0" /> + <edge from-layer="1179" from-port="1" to-layer="1182" to-port="0" /> + <edge from-layer="1180" from-port="0" to-layer="1182" to-port="1" /> + <edge from-layer="1181" from-port="0" to-layer="1182" to-port="2" /> + <edge from-layer="1182" from-port="3" to-layer="1183" to-port="0" /> + <edge from-layer="1182" from-port="3" to-layer="1241" to-port="0" /> + <edge from-layer="1182" from-port="3" to-layer="1347" to-port="0" /> + <edge from-layer="1183" from-port="3" to-layer="1184" to-port="1" /> + <edge from-layer="1184" from-port="2" to-layer="1186" to-port="0" /> + <edge from-layer="1185" from-port="0" to-layer="1186" to-port="1" /> + <edge from-layer="1186" from-port="2" to-layer="1216" to-port="0" /> + <edge from-layer="1186" from-port="2" to-layer="1199" to-port="0" /> + <edge from-layer="1186" from-port="2" to-layer="1196" to-port="0" /> + <edge from-layer="1186" from-port="2" to-layer="1210" to-port="0" /> + <edge from-layer="1187" from-port="0" to-layer="1188" to-port="0" /> + <edge from-layer="1188" from-port="1" to-layer="1189" to-port="0" /> + <edge from-layer="1189" from-port="2" to-layer="1190" to-port="0" /> + <edge from-layer="1190" from-port="2" to-layer="1192" to-port="0" /> + <edge from-layer="1191" from-port="0" to-layer="1192" to-port="1" /> + <edge from-layer="1192" from-port="2" to-layer="1193" to-port="0" /> + <edge from-layer="1192" from-port="2" to-layer="1193" to-port="1" /> + <edge from-layer="1193" from-port="2" to-layer="1194" to-port="0" /> + <edge from-layer="1193" from-port="2" to-layer="1218" to-port="0" /> + <edge from-layer="1194" from-port="1" to-layer="1195" to-port="0" /> + <edge from-layer="1195" from-port="2" to-layer="1196" to-port="1" /> + <edge from-layer="1195" from-port="2" to-layer="1245" to-port="1" /> + <edge from-layer="1196" from-port="2" to-layer="1221" to-port="0" /> + <edge from-layer="1197" from-port="0" to-layer="1207" to-port="0" /> + <edge from-layer="1198" from-port="0" to-layer="1207" to-port="1" /> + <edge from-layer="1199" from-port="1" to-layer="1201" to-port="0" /> + <edge from-layer="1200" from-port="0" to-layer="1201" to-port="2" /> + <edge from-layer="1201" from-port="3" to-layer="1202" to-port="0" /> + <edge from-layer="1202" from-port="2" to-layer="1203" to-port="0" /> + <edge from-layer="1203" from-port="1" to-layer="1205" to-port="0" /> + <edge from-layer="1204" from-port="0" to-layer="1205" to-port="1" /> + <edge from-layer="1205" from-port="2" to-layer="1207" to-port="2" /> + <edge from-layer="1206" from-port="0" to-layer="1207" to-port="3" /> + <edge from-layer="1207" from-port="4" to-layer="1216" to-port="2" /> + <edge from-layer="1207" from-port="4" to-layer="1210" to-port="1" /> + <edge from-layer="1208" from-port="0" to-layer="1210" to-port="2" /> + <edge from-layer="1209" from-port="0" to-layer="1210" to-port="3" /> + <edge from-layer="1210" from-port="4" to-layer="1213" to-port="0" /> + <edge from-layer="1211" from-port="0" to-layer="1212" to-port="0" /> + <edge from-layer="1212" from-port="1" to-layer="1213" to-port="1" /> + <edge from-layer="1213" from-port="2" to-layer="1217" to-port="0" /> + <edge from-layer="1214" from-port="0" to-layer="1216" to-port="1" /> + <edge from-layer="1215" from-port="0" to-layer="1216" to-port="3" /> + <edge from-layer="1216" from-port="4" to-layer="1217" to-port="1" /> + <edge from-layer="1217" from-port="2" to-layer="1220" to-port="0" /> + <edge from-layer="1218" from-port="1" to-layer="1219" to-port="0" /> + <edge from-layer="1219" from-port="2" to-layer="1220" to-port="1" /> + <edge from-layer="1219" from-port="2" to-layer="1267" to-port="1" /> + <edge from-layer="1220" from-port="2" to-layer="1221" to-port="1" /> + <edge from-layer="1221" from-port="2" to-layer="1344" to-port="0" /> + <edge from-layer="1222" from-port="0" to-layer="1223" to-port="0" /> + <edge from-layer="1223" from-port="1" to-layer="1228" to-port="0" /> + <edge from-layer="1224" from-port="0" to-layer="1227" to-port="1" /> + <edge from-layer="1225" from-port="0" to-layer="1227" to-port="2" /> + <edge from-layer="1226" from-port="0" to-layer="1227" to-port="3" /> + <edge from-layer="1227" from-port="4" to-layer="1228" to-port="1" /> + <edge from-layer="1228" from-port="2" to-layer="1229" to-port="0" /> + <edge from-layer="1229" from-port="1" to-layer="1231" to-port="0" /> + <edge from-layer="1230" from-port="0" to-layer="1231" to-port="2" /> + <edge from-layer="1231" from-port="3" to-layer="1269" to-port="0" /> + <edge from-layer="1232" from-port="0" to-layer="1233" to-port="0" /> + <edge from-layer="1233" from-port="1" to-layer="1236" to-port="0" /> + <edge from-layer="1234" from-port="0" to-layer="1235" to-port="0" /> + <edge from-layer="1235" from-port="1" to-layer="1236" to-port="1" /> + <edge from-layer="1236" from-port="2" to-layer="1238" to-port="0" /> + <edge from-layer="1237" from-port="0" to-layer="1238" to-port="1" /> + <edge from-layer="1238" from-port="2" to-layer="1239" to-port="0" /> + <edge from-layer="1239" from-port="1" to-layer="1240" to-port="1" /> + <edge from-layer="1240" from-port="2" to-layer="1242" to-port="0" /> + <edge from-layer="1241" from-port="3" to-layer="1309" to-port="1" /> + <edge from-layer="1241" from-port="3" to-layer="1242" to-port="1" /> + <edge from-layer="1242" from-port="2" to-layer="1244" to-port="0" /> + <edge from-layer="1243" from-port="0" to-layer="1244" to-port="1" /> + <edge from-layer="1244" from-port="2" to-layer="1245" to-port="0" /> + <edge from-layer="1244" from-port="2" to-layer="1248" to-port="0" /> + <edge from-layer="1244" from-port="2" to-layer="1259" to-port="0" /> + <edge from-layer="1244" from-port="2" to-layer="1265" to-port="0" /> + <edge from-layer="1245" from-port="2" to-layer="1268" to-port="0" /> + <edge from-layer="1246" from-port="0" to-layer="1256" to-port="0" /> + <edge from-layer="1247" from-port="0" to-layer="1256" to-port="1" /> + <edge from-layer="1248" from-port="1" to-layer="1250" to-port="0" /> + <edge from-layer="1249" from-port="0" to-layer="1250" to-port="2" /> + <edge from-layer="1250" from-port="3" to-layer="1251" to-port="0" /> + <edge from-layer="1251" from-port="2" to-layer="1252" to-port="0" /> + <edge from-layer="1252" from-port="1" to-layer="1254" to-port="0" /> + <edge from-layer="1253" from-port="0" to-layer="1254" to-port="1" /> + <edge from-layer="1254" from-port="2" to-layer="1256" to-port="2" /> + <edge from-layer="1255" from-port="0" to-layer="1256" to-port="3" /> + <edge from-layer="1256" from-port="4" to-layer="1265" to-port="2" /> + <edge from-layer="1256" from-port="4" to-layer="1259" to-port="1" /> + <edge from-layer="1257" from-port="0" to-layer="1259" to-port="2" /> + <edge from-layer="1258" from-port="0" to-layer="1259" to-port="3" /> + <edge from-layer="1259" from-port="4" to-layer="1262" to-port="0" /> + <edge from-layer="1260" from-port="0" to-layer="1261" to-port="0" /> + <edge from-layer="1261" from-port="1" to-layer="1262" to-port="1" /> + <edge from-layer="1262" from-port="2" to-layer="1266" to-port="0" /> + <edge from-layer="1263" from-port="0" to-layer="1265" to-port="1" /> + <edge from-layer="1264" from-port="0" to-layer="1265" to-port="3" /> + <edge from-layer="1265" from-port="4" to-layer="1266" to-port="1" /> + <edge from-layer="1266" from-port="2" to-layer="1267" to-port="0" /> + <edge from-layer="1267" from-port="2" to-layer="1268" to-port="1" /> + <edge from-layer="1268" from-port="2" to-layer="1269" to-port="1" /> + <edge from-layer="1269" from-port="2" to-layer="1270" to-port="0" /> + <edge from-layer="1269" from-port="2" to-layer="2510" to-port="0" /> + <edge from-layer="1269" from-port="2" to-layer="1271" to-port="0" /> + <edge from-layer="1270" from-port="2" to-layer="1284" to-port="0" /> + <edge from-layer="1271" from-port="1" to-layer="1281" to-port="0" /> + <edge from-layer="1271" from-port="1" to-layer="1274" to-port="0" /> + <edge from-layer="1271" from-port="1" to-layer="1276" to-port="0" /> + <edge from-layer="1272" from-port="0" to-layer="1274" to-port="1" /> + <edge from-layer="1273" from-port="0" to-layer="1274" to-port="2" /> + <edge from-layer="1274" from-port="3" to-layer="1288" to-port="0" /> + <edge from-layer="1274" from-port="3" to-layer="1282" to-port="0" /> + <edge from-layer="1275" from-port="0" to-layer="1276" to-port="2" /> + <edge from-layer="1276" from-port="3" to-layer="1278" to-port="0" /> + <edge from-layer="1276" from-port="3" to-layer="1285" to-port="0" /> + <edge from-layer="1277" from-port="0" to-layer="1278" to-port="1" /> + <edge from-layer="1278" from-port="2" to-layer="1282" to-port="1" /> + <edge from-layer="1279" from-port="0" to-layer="1281" to-port="1" /> + <edge from-layer="1280" from-port="0" to-layer="1281" to-port="2" /> + <edge from-layer="1281" from-port="3" to-layer="1288" to-port="2" /> + <edge from-layer="1281" from-port="3" to-layer="1282" to-port="3" /> + <edge from-layer="1282" from-port="4" to-layer="1283" to-port="0" /> + <edge from-layer="1283" from-port="1" to-layer="1284" to-port="1" /> + <edge from-layer="1284" from-port="2" to-layer="1289" to-port="0" /> + <edge from-layer="1285" from-port="2" to-layer="1287" to-port="0" /> + <edge from-layer="1286" from-port="0" to-layer="1287" to-port="1" /> + <edge from-layer="1287" from-port="2" to-layer="1288" to-port="1" /> + <edge from-layer="1288" from-port="3" to-layer="1289" to-port="1" /> + <edge from-layer="1289" from-port="2" to-layer="1336" to-port="0" /> + <edge from-layer="1289" from-port="2" to-layer="1344" to-port="1" /> + <edge from-layer="1290" from-port="0" to-layer="1291" to-port="0" /> + <edge from-layer="1291" from-port="1" to-layer="1296" to-port="0" /> + <edge from-layer="1292" from-port="0" to-layer="1295" to-port="1" /> + <edge from-layer="1293" from-port="0" to-layer="1295" to-port="2" /> + <edge from-layer="1294" from-port="0" to-layer="1295" to-port="3" /> + <edge from-layer="1295" from-port="4" to-layer="1296" to-port="1" /> + <edge from-layer="1296" from-port="2" to-layer="1297" to-port="0" /> + <edge from-layer="1297" from-port="1" to-layer="1299" to-port="0" /> + <edge from-layer="1298" from-port="0" to-layer="1299" to-port="2" /> + <edge from-layer="1299" from-port="3" to-layer="1312" to-port="0" /> + <edge from-layer="1300" from-port="0" to-layer="1301" to-port="0" /> + <edge from-layer="1301" from-port="1" to-layer="1304" to-port="0" /> + <edge from-layer="1302" from-port="0" to-layer="1303" to-port="0" /> + <edge from-layer="1303" from-port="1" to-layer="1304" to-port="1" /> + <edge from-layer="1304" from-port="2" to-layer="1306" to-port="0" /> + <edge from-layer="1305" from-port="0" to-layer="1306" to-port="1" /> + <edge from-layer="1306" from-port="2" to-layer="1307" to-port="0" /> + <edge from-layer="1307" from-port="1" to-layer="1308" to-port="1" /> + <edge from-layer="1308" from-port="2" to-layer="1309" to-port="0" /> + <edge from-layer="1309" from-port="2" to-layer="1311" to-port="0" /> + <edge from-layer="1310" from-port="0" to-layer="1311" to-port="1" /> + <edge from-layer="1311" from-port="2" to-layer="1312" to-port="1" /> + <edge from-layer="1312" from-port="2" to-layer="2509" to-port="0" /> + <edge from-layer="1312" from-port="2" to-layer="1313" to-port="0" /> + <edge from-layer="1312" from-port="2" to-layer="1314" to-port="0" /> + <edge from-layer="1313" from-port="2" to-layer="1327" to-port="0" /> + <edge from-layer="1314" from-port="1" to-layer="1324" to-port="0" /> + <edge from-layer="1314" from-port="1" to-layer="1317" to-port="0" /> + <edge from-layer="1314" from-port="1" to-layer="1319" to-port="0" /> + <edge from-layer="1315" from-port="0" to-layer="1317" to-port="1" /> + <edge from-layer="1316" from-port="0" to-layer="1317" to-port="2" /> + <edge from-layer="1317" from-port="3" to-layer="1331" to-port="0" /> + <edge from-layer="1317" from-port="3" to-layer="1325" to-port="0" /> + <edge from-layer="1318" from-port="0" to-layer="1319" to-port="2" /> + <edge from-layer="1319" from-port="3" to-layer="1328" to-port="0" /> + <edge from-layer="1319" from-port="3" to-layer="1321" to-port="0" /> + <edge from-layer="1320" from-port="0" to-layer="1321" to-port="1" /> + <edge from-layer="1321" from-port="2" to-layer="1325" to-port="1" /> + <edge from-layer="1322" from-port="0" to-layer="1324" to-port="1" /> + <edge from-layer="1323" from-port="0" to-layer="1324" to-port="2" /> + <edge from-layer="1324" from-port="3" to-layer="1331" to-port="2" /> + <edge from-layer="1324" from-port="3" to-layer="1325" to-port="3" /> + <edge from-layer="1325" from-port="4" to-layer="1326" to-port="0" /> + <edge from-layer="1326" from-port="1" to-layer="1327" to-port="1" /> + <edge from-layer="1327" from-port="2" to-layer="1332" to-port="0" /> + <edge from-layer="1328" from-port="2" to-layer="1330" to-port="0" /> + <edge from-layer="1329" from-port="0" to-layer="1330" to-port="1" /> + <edge from-layer="1330" from-port="2" to-layer="1331" to-port="1" /> + <edge from-layer="1331" from-port="3" to-layer="1332" to-port="1" /> + <edge from-layer="1332" from-port="2" to-layer="1344" to-port="2" /> + <edge from-layer="1333" from-port="0" to-layer="1343" to-port="1" /> + <edge from-layer="1334" from-port="0" to-layer="1341" to-port="0" /> + <edge from-layer="1335" from-port="0" to-layer="1341" to-port="1" /> + <edge from-layer="1336" from-port="1" to-layer="1339" to-port="0" /> + <edge from-layer="1337" from-port="0" to-layer="1339" to-port="1" /> + <edge from-layer="1338" from-port="0" to-layer="1339" to-port="2" /> + <edge from-layer="1339" from-port="3" to-layer="1341" to-port="2" /> + <edge from-layer="1340" from-port="0" to-layer="1341" to-port="3" /> + <edge from-layer="1341" from-port="4" to-layer="1343" to-port="2" /> + <edge from-layer="1342" from-port="0" to-layer="1343" to-port="3" /> + <edge from-layer="1343" from-port="4" to-layer="1344" to-port="3" /> + <edge from-layer="1344" from-port="4" to-layer="1346" to-port="0" /> + <edge from-layer="1345" from-port="0" to-layer="1346" to-port="1" /> + <edge from-layer="1346" from-port="2" to-layer="1348" to-port="0" /> + <edge from-layer="1347" from-port="2" to-layer="1348" to-port="1" /> + <edge from-layer="1348" from-port="2" to-layer="1357" to-port="0" /> + <edge from-layer="1349" from-port="0" to-layer="1350" to-port="0" /> + <edge from-layer="1350" from-port="1" to-layer="1353" to-port="0" /> + <edge from-layer="1351" from-port="0" to-layer="1352" to-port="0" /> + <edge from-layer="1352" from-port="1" to-layer="1353" to-port="1" /> + <edge from-layer="1353" from-port="2" to-layer="1355" to-port="0" /> + <edge from-layer="1354" from-port="0" to-layer="1355" to-port="1" /> + <edge from-layer="1355" from-port="2" to-layer="1356" to-port="0" /> + <edge from-layer="1356" from-port="1" to-layer="1357" to-port="1" /> + <edge from-layer="1357" from-port="2" to-layer="1358" to-port="1" /> + <edge from-layer="1358" from-port="2" to-layer="1404" to-port="0" /> + <edge from-layer="1358" from-port="2" to-layer="1373" to-port="0" /> + <edge from-layer="1358" from-port="2" to-layer="1365" to-port="0" /> + <edge from-layer="1359" from-port="0" to-layer="1360" to-port="0" /> + <edge from-layer="1360" from-port="1" to-layer="1374" to-port="0" /> + <edge from-layer="1361" from-port="0" to-layer="1362" to-port="0" /> + <edge from-layer="1362" from-port="1" to-layer="1372" to-port="0" /> + <edge from-layer="1363" from-port="0" to-layer="1364" to-port="0" /> + <edge from-layer="1364" from-port="1" to-layer="1365" to-port="1" /> + <edge from-layer="1365" from-port="2" to-layer="1367" to-port="0" /> + <edge from-layer="1366" from-port="0" to-layer="1367" to-port="1" /> + <edge from-layer="1367" from-port="2" to-layer="1370" to-port="0" /> + <edge from-layer="1368" from-port="0" to-layer="1369" to-port="0" /> + <edge from-layer="1369" from-port="1" to-layer="1370" to-port="1" /> + <edge from-layer="1370" from-port="2" to-layer="1371" to-port="0" /> + <edge from-layer="1371" from-port="1" to-layer="1372" to-port="1" /> + <edge from-layer="1372" from-port="2" to-layer="1373" to-port="1" /> + <edge from-layer="1373" from-port="2" to-layer="1374" to-port="1" /> + <edge from-layer="1374" from-port="2" to-layer="1383" to-port="0" /> + <edge from-layer="1374" from-port="2" to-layer="1393" to-port="0" /> + <edge from-layer="1375" from-port="0" to-layer="1376" to-port="0" /> + <edge from-layer="1376" from-port="1" to-layer="1379" to-port="0" /> + <edge from-layer="1377" from-port="0" to-layer="1378" to-port="0" /> + <edge from-layer="1378" from-port="1" to-layer="1379" to-port="1" /> + <edge from-layer="1379" from-port="2" to-layer="1381" to-port="0" /> + <edge from-layer="1380" from-port="0" to-layer="1381" to-port="1" /> + <edge from-layer="1381" from-port="2" to-layer="1382" to-port="0" /> + <edge from-layer="1382" from-port="1" to-layer="1383" to-port="1" /> + <edge from-layer="1383" from-port="2" to-layer="1384" to-port="0" /> + <edge from-layer="1384" from-port="1" to-layer="1394" to-port="0" /> + <edge from-layer="1385" from-port="0" to-layer="1386" to-port="0" /> + <edge from-layer="1386" from-port="1" to-layer="1389" to-port="0" /> + <edge from-layer="1387" from-port="0" to-layer="1388" to-port="0" /> + <edge from-layer="1388" from-port="1" to-layer="1389" to-port="1" /> + <edge from-layer="1389" from-port="2" to-layer="1391" to-port="0" /> + <edge from-layer="1390" from-port="0" to-layer="1391" to-port="1" /> + <edge from-layer="1391" from-port="2" to-layer="1392" to-port="0" /> + <edge from-layer="1392" from-port="1" to-layer="1393" to-port="1" /> + <edge from-layer="1393" from-port="2" to-layer="1394" to-port="1" /> + <edge from-layer="1394" from-port="2" to-layer="1403" to-port="0" /> + <edge from-layer="1395" from-port="0" to-layer="1396" to-port="0" /> + <edge from-layer="1396" from-port="1" to-layer="1399" to-port="0" /> + <edge from-layer="1397" from-port="0" to-layer="1398" to-port="0" /> + <edge from-layer="1398" from-port="1" to-layer="1399" to-port="1" /> + <edge from-layer="1399" from-port="2" to-layer="1401" to-port="0" /> + <edge from-layer="1400" from-port="0" to-layer="1401" to-port="1" /> + <edge from-layer="1401" from-port="2" to-layer="1402" to-port="0" /> + <edge from-layer="1402" from-port="1" to-layer="1403" to-port="1" /> + <edge from-layer="1403" from-port="2" to-layer="1404" to-port="1" /> + <edge from-layer="1404" from-port="2" to-layer="1419" to-port="0" /> + <edge from-layer="1404" from-port="2" to-layer="1609" to-port="0" /> + <edge from-layer="1404" from-port="2" to-layer="1411" to-port="0" /> + <edge from-layer="1405" from-port="0" to-layer="1406" to-port="0" /> + <edge from-layer="1406" from-port="1" to-layer="1420" to-port="0" /> + <edge from-layer="1407" from-port="0" to-layer="1408" to-port="0" /> + <edge from-layer="1408" from-port="1" to-layer="1418" to-port="0" /> + <edge from-layer="1409" from-port="0" to-layer="1410" to-port="0" /> + <edge from-layer="1410" from-port="1" to-layer="1411" to-port="1" /> + <edge from-layer="1411" from-port="2" to-layer="1413" to-port="0" /> + <edge from-layer="1412" from-port="0" to-layer="1413" to-port="1" /> + <edge from-layer="1413" from-port="2" to-layer="1416" to-port="0" /> + <edge from-layer="1414" from-port="0" to-layer="1415" to-port="0" /> + <edge from-layer="1415" from-port="1" to-layer="1416" to-port="1" /> + <edge from-layer="1416" from-port="2" to-layer="1417" to-port="0" /> + <edge from-layer="1417" from-port="1" to-layer="1418" to-port="1" /> + <edge from-layer="1418" from-port="2" to-layer="1419" to-port="1" /> + <edge from-layer="1419" from-port="2" to-layer="1420" to-port="1" /> + <edge from-layer="1420" from-port="2" to-layer="1430" to-port="0" /> + <edge from-layer="1420" from-port="2" to-layer="1559" to-port="0" /> + <edge from-layer="1420" from-port="2" to-layer="1491" to-port="0" /> + <edge from-layer="1420" from-port="2" to-layer="1429" to-port="0" /> + <edge from-layer="1421" from-port="0" to-layer="1422" to-port="0" /> + <edge from-layer="1422" from-port="1" to-layer="1425" to-port="0" /> + <edge from-layer="1423" from-port="0" to-layer="1424" to-port="0" /> + <edge from-layer="1424" from-port="1" to-layer="1425" to-port="1" /> + <edge from-layer="1425" from-port="2" to-layer="1427" to-port="0" /> + <edge from-layer="1426" from-port="0" to-layer="1427" to-port="1" /> + <edge from-layer="1427" from-port="2" to-layer="1428" to-port="0" /> + <edge from-layer="1428" from-port="1" to-layer="1429" to-port="1" /> + <edge from-layer="1429" from-port="2" to-layer="1435" to-port="0" /> + <edge from-layer="1430" from-port="1" to-layer="1433" to-port="0" /> + <edge from-layer="1431" from-port="0" to-layer="1433" to-port="1" /> + <edge from-layer="1432" from-port="0" to-layer="1433" to-port="2" /> + <edge from-layer="1433" from-port="3" to-layer="1492" to-port="0" /> + <edge from-layer="1433" from-port="3" to-layer="1434" to-port="0" /> + <edge from-layer="1433" from-port="3" to-layer="1598" to-port="0" /> + <edge from-layer="1434" from-port="3" to-layer="1435" to-port="1" /> + <edge from-layer="1435" from-port="2" to-layer="1437" to-port="0" /> + <edge from-layer="1436" from-port="0" to-layer="1437" to-port="1" /> + <edge from-layer="1437" from-port="2" to-layer="1450" to-port="0" /> + <edge from-layer="1437" from-port="2" to-layer="1461" to-port="0" /> + <edge from-layer="1437" from-port="2" to-layer="1447" to-port="0" /> + <edge from-layer="1437" from-port="2" to-layer="1467" to-port="0" /> + <edge from-layer="1438" from-port="0" to-layer="1439" to-port="0" /> + <edge from-layer="1439" from-port="1" to-layer="1440" to-port="0" /> + <edge from-layer="1440" from-port="2" to-layer="1441" to-port="0" /> + <edge from-layer="1441" from-port="2" to-layer="1443" to-port="0" /> + <edge from-layer="1442" from-port="0" to-layer="1443" to-port="1" /> + <edge from-layer="1443" from-port="2" to-layer="1444" to-port="1" /> + <edge from-layer="1443" from-port="2" to-layer="1444" to-port="0" /> + <edge from-layer="1444" from-port="2" to-layer="1445" to-port="0" /> + <edge from-layer="1444" from-port="2" to-layer="1469" to-port="0" /> + <edge from-layer="1445" from-port="1" to-layer="1446" to-port="0" /> + <edge from-layer="1446" from-port="2" to-layer="1496" to-port="1" /> + <edge from-layer="1446" from-port="2" to-layer="1447" to-port="1" /> + <edge from-layer="1447" from-port="2" to-layer="1472" to-port="0" /> + <edge from-layer="1448" from-port="0" to-layer="1458" to-port="0" /> + <edge from-layer="1449" from-port="0" to-layer="1458" to-port="1" /> + <edge from-layer="1450" from-port="1" to-layer="1452" to-port="0" /> + <edge from-layer="1451" from-port="0" to-layer="1452" to-port="2" /> + <edge from-layer="1452" from-port="3" to-layer="1453" to-port="0" /> + <edge from-layer="1453" from-port="2" to-layer="1454" to-port="0" /> + <edge from-layer="1454" from-port="1" to-layer="1456" to-port="0" /> + <edge from-layer="1455" from-port="0" to-layer="1456" to-port="1" /> + <edge from-layer="1456" from-port="2" to-layer="1458" to-port="2" /> + <edge from-layer="1457" from-port="0" to-layer="1458" to-port="3" /> + <edge from-layer="1458" from-port="4" to-layer="1461" to-port="1" /> + <edge from-layer="1458" from-port="4" to-layer="1467" to-port="2" /> + <edge from-layer="1459" from-port="0" to-layer="1461" to-port="2" /> + <edge from-layer="1460" from-port="0" to-layer="1461" to-port="3" /> + <edge from-layer="1461" from-port="4" to-layer="1464" to-port="0" /> + <edge from-layer="1462" from-port="0" to-layer="1463" to-port="0" /> + <edge from-layer="1463" from-port="1" to-layer="1464" to-port="1" /> + <edge from-layer="1464" from-port="2" to-layer="1468" to-port="0" /> + <edge from-layer="1465" from-port="0" to-layer="1467" to-port="1" /> + <edge from-layer="1466" from-port="0" to-layer="1467" to-port="3" /> + <edge from-layer="1467" from-port="4" to-layer="1468" to-port="1" /> + <edge from-layer="1468" from-port="2" to-layer="1471" to-port="0" /> + <edge from-layer="1469" from-port="1" to-layer="1470" to-port="0" /> + <edge from-layer="1470" from-port="2" to-layer="1471" to-port="1" /> + <edge from-layer="1470" from-port="2" to-layer="1518" to-port="1" /> + <edge from-layer="1471" from-port="2" to-layer="1472" to-port="1" /> + <edge from-layer="1472" from-port="2" to-layer="1595" to-port="0" /> + <edge from-layer="1473" from-port="0" to-layer="1474" to-port="0" /> + <edge from-layer="1474" from-port="1" to-layer="1479" to-port="0" /> + <edge from-layer="1475" from-port="0" to-layer="1478" to-port="1" /> + <edge from-layer="1476" from-port="0" to-layer="1478" to-port="2" /> + <edge from-layer="1477" from-port="0" to-layer="1478" to-port="3" /> + <edge from-layer="1478" from-port="4" to-layer="1479" to-port="1" /> + <edge from-layer="1479" from-port="2" to-layer="1480" to-port="0" /> + <edge from-layer="1480" from-port="1" to-layer="1482" to-port="0" /> + <edge from-layer="1481" from-port="0" to-layer="1482" to-port="2" /> + <edge from-layer="1482" from-port="3" to-layer="1520" to-port="0" /> + <edge from-layer="1483" from-port="0" to-layer="1484" to-port="0" /> + <edge from-layer="1484" from-port="1" to-layer="1487" to-port="0" /> + <edge from-layer="1485" from-port="0" to-layer="1486" to-port="0" /> + <edge from-layer="1486" from-port="1" to-layer="1487" to-port="1" /> + <edge from-layer="1487" from-port="2" to-layer="1489" to-port="0" /> + <edge from-layer="1488" from-port="0" to-layer="1489" to-port="1" /> + <edge from-layer="1489" from-port="2" to-layer="1490" to-port="0" /> + <edge from-layer="1490" from-port="1" to-layer="1491" to-port="1" /> + <edge from-layer="1491" from-port="2" to-layer="1493" to-port="0" /> + <edge from-layer="1492" from-port="3" to-layer="1493" to-port="1" /> + <edge from-layer="1492" from-port="3" to-layer="1560" to-port="1" /> + <edge from-layer="1493" from-port="2" to-layer="1495" to-port="0" /> + <edge from-layer="1494" from-port="0" to-layer="1495" to-port="1" /> + <edge from-layer="1495" from-port="2" to-layer="1496" to-port="0" /> + <edge from-layer="1495" from-port="2" to-layer="1499" to-port="0" /> + <edge from-layer="1495" from-port="2" to-layer="1516" to-port="0" /> + <edge from-layer="1495" from-port="2" to-layer="1510" to-port="0" /> + <edge from-layer="1496" from-port="2" to-layer="1519" to-port="0" /> + <edge from-layer="1497" from-port="0" to-layer="1507" to-port="0" /> + <edge from-layer="1498" from-port="0" to-layer="1507" to-port="1" /> + <edge from-layer="1499" from-port="1" to-layer="1501" to-port="0" /> + <edge from-layer="1500" from-port="0" to-layer="1501" to-port="2" /> + <edge from-layer="1501" from-port="3" to-layer="1502" to-port="0" /> + <edge from-layer="1502" from-port="2" to-layer="1503" to-port="0" /> + <edge from-layer="1503" from-port="1" to-layer="1505" to-port="0" /> + <edge from-layer="1504" from-port="0" to-layer="1505" to-port="1" /> + <edge from-layer="1505" from-port="2" to-layer="1507" to-port="2" /> + <edge from-layer="1506" from-port="0" to-layer="1507" to-port="3" /> + <edge from-layer="1507" from-port="4" to-layer="1510" to-port="1" /> + <edge from-layer="1507" from-port="4" to-layer="1516" to-port="2" /> + <edge from-layer="1508" from-port="0" to-layer="1510" to-port="2" /> + <edge from-layer="1509" from-port="0" to-layer="1510" to-port="3" /> + <edge from-layer="1510" from-port="4" to-layer="1513" to-port="0" /> + <edge from-layer="1511" from-port="0" to-layer="1512" to-port="0" /> + <edge from-layer="1512" from-port="1" to-layer="1513" to-port="1" /> + <edge from-layer="1513" from-port="2" to-layer="1517" to-port="0" /> + <edge from-layer="1514" from-port="0" to-layer="1516" to-port="1" /> + <edge from-layer="1515" from-port="0" to-layer="1516" to-port="3" /> + <edge from-layer="1516" from-port="4" to-layer="1517" to-port="1" /> + <edge from-layer="1517" from-port="2" to-layer="1518" to-port="0" /> + <edge from-layer="1518" from-port="2" to-layer="1519" to-port="1" /> + <edge from-layer="1519" from-port="2" to-layer="1520" to-port="1" /> + <edge from-layer="1520" from-port="2" to-layer="2508" to-port="0" /> + <edge from-layer="1520" from-port="2" to-layer="1522" to-port="0" /> + <edge from-layer="1520" from-port="2" to-layer="1521" to-port="0" /> + <edge from-layer="1521" from-port="2" to-layer="1535" to-port="0" /> + <edge from-layer="1522" from-port="1" to-layer="1525" to-port="0" /> + <edge from-layer="1522" from-port="1" to-layer="1532" to-port="0" /> + <edge from-layer="1522" from-port="1" to-layer="1527" to-port="0" /> + <edge from-layer="1523" from-port="0" to-layer="1525" to-port="1" /> + <edge from-layer="1524" from-port="0" to-layer="1525" to-port="2" /> + <edge from-layer="1525" from-port="3" to-layer="1533" to-port="0" /> + <edge from-layer="1525" from-port="3" to-layer="1539" to-port="0" /> + <edge from-layer="1526" from-port="0" to-layer="1527" to-port="2" /> + <edge from-layer="1527" from-port="3" to-layer="1536" to-port="0" /> + <edge from-layer="1527" from-port="3" to-layer="1529" to-port="0" /> + <edge from-layer="1528" from-port="0" to-layer="1529" to-port="1" /> + <edge from-layer="1529" from-port="2" to-layer="1533" to-port="1" /> + <edge from-layer="1530" from-port="0" to-layer="1532" to-port="1" /> + <edge from-layer="1531" from-port="0" to-layer="1532" to-port="2" /> + <edge from-layer="1532" from-port="3" to-layer="1539" to-port="2" /> + <edge from-layer="1532" from-port="3" to-layer="1533" to-port="3" /> + <edge from-layer="1533" from-port="4" to-layer="1534" to-port="0" /> + <edge from-layer="1534" from-port="1" to-layer="1535" to-port="1" /> + <edge from-layer="1535" from-port="2" to-layer="1540" to-port="0" /> + <edge from-layer="1536" from-port="2" to-layer="1538" to-port="0" /> + <edge from-layer="1537" from-port="0" to-layer="1538" to-port="1" /> + <edge from-layer="1538" from-port="2" to-layer="1539" to-port="1" /> + <edge from-layer="1539" from-port="3" to-layer="1540" to-port="1" /> + <edge from-layer="1540" from-port="2" to-layer="1595" to-port="1" /> + <edge from-layer="1540" from-port="2" to-layer="1587" to-port="0" /> + <edge from-layer="1541" from-port="0" to-layer="1542" to-port="0" /> + <edge from-layer="1542" from-port="1" to-layer="1547" to-port="0" /> + <edge from-layer="1543" from-port="0" to-layer="1546" to-port="1" /> + <edge from-layer="1544" from-port="0" to-layer="1546" to-port="2" /> + <edge from-layer="1545" from-port="0" to-layer="1546" to-port="3" /> + <edge from-layer="1546" from-port="4" to-layer="1547" to-port="1" /> + <edge from-layer="1547" from-port="2" to-layer="1548" to-port="0" /> + <edge from-layer="1548" from-port="1" to-layer="1550" to-port="0" /> + <edge from-layer="1549" from-port="0" to-layer="1550" to-port="2" /> + <edge from-layer="1550" from-port="3" to-layer="1563" to-port="0" /> + <edge from-layer="1551" from-port="0" to-layer="1552" to-port="0" /> + <edge from-layer="1552" from-port="1" to-layer="1555" to-port="0" /> + <edge from-layer="1553" from-port="0" to-layer="1554" to-port="0" /> + <edge from-layer="1554" from-port="1" to-layer="1555" to-port="1" /> + <edge from-layer="1555" from-port="2" to-layer="1557" to-port="0" /> + <edge from-layer="1556" from-port="0" to-layer="1557" to-port="1" /> + <edge from-layer="1557" from-port="2" to-layer="1558" to-port="0" /> + <edge from-layer="1558" from-port="1" to-layer="1559" to-port="1" /> + <edge from-layer="1559" from-port="2" to-layer="1560" to-port="0" /> + <edge from-layer="1560" from-port="2" to-layer="1562" to-port="0" /> + <edge from-layer="1561" from-port="0" to-layer="1562" to-port="1" /> + <edge from-layer="1562" from-port="2" to-layer="1563" to-port="1" /> + <edge from-layer="1563" from-port="2" to-layer="1565" to-port="0" /> + <edge from-layer="1563" from-port="2" to-layer="1564" to-port="0" /> + <edge from-layer="1563" from-port="2" to-layer="2507" to-port="0" /> + <edge from-layer="1564" from-port="2" to-layer="1578" to-port="0" /> + <edge from-layer="1565" from-port="1" to-layer="1568" to-port="0" /> + <edge from-layer="1565" from-port="1" to-layer="1570" to-port="0" /> + <edge from-layer="1565" from-port="1" to-layer="1575" to-port="0" /> + <edge from-layer="1566" from-port="0" to-layer="1568" to-port="1" /> + <edge from-layer="1567" from-port="0" to-layer="1568" to-port="2" /> + <edge from-layer="1568" from-port="3" to-layer="1582" to-port="0" /> + <edge from-layer="1568" from-port="3" to-layer="1576" to-port="0" /> + <edge from-layer="1569" from-port="0" to-layer="1570" to-port="2" /> + <edge from-layer="1570" from-port="3" to-layer="1572" to-port="0" /> + <edge from-layer="1570" from-port="3" to-layer="1579" to-port="0" /> + <edge from-layer="1571" from-port="0" to-layer="1572" to-port="1" /> + <edge from-layer="1572" from-port="2" to-layer="1576" to-port="1" /> + <edge from-layer="1573" from-port="0" to-layer="1575" to-port="1" /> + <edge from-layer="1574" from-port="0" to-layer="1575" to-port="2" /> + <edge from-layer="1575" from-port="3" to-layer="1576" to-port="3" /> + <edge from-layer="1575" from-port="3" to-layer="1582" to-port="2" /> + <edge from-layer="1576" from-port="4" to-layer="1577" to-port="0" /> + <edge from-layer="1577" from-port="1" to-layer="1578" to-port="1" /> + <edge from-layer="1578" from-port="2" to-layer="1583" to-port="0" /> + <edge from-layer="1579" from-port="2" to-layer="1581" to-port="0" /> + <edge from-layer="1580" from-port="0" to-layer="1581" to-port="1" /> + <edge from-layer="1581" from-port="2" to-layer="1582" to-port="1" /> + <edge from-layer="1582" from-port="3" to-layer="1583" to-port="1" /> + <edge from-layer="1583" from-port="2" to-layer="1595" to-port="2" /> + <edge from-layer="1584" from-port="0" to-layer="1594" to-port="1" /> + <edge from-layer="1585" from-port="0" to-layer="1592" to-port="0" /> + <edge from-layer="1586" from-port="0" to-layer="1592" to-port="1" /> + <edge from-layer="1587" from-port="1" to-layer="1590" to-port="0" /> + <edge from-layer="1588" from-port="0" to-layer="1590" to-port="1" /> + <edge from-layer="1589" from-port="0" to-layer="1590" to-port="2" /> + <edge from-layer="1590" from-port="3" to-layer="1592" to-port="2" /> + <edge from-layer="1591" from-port="0" to-layer="1592" to-port="3" /> + <edge from-layer="1592" from-port="4" to-layer="1594" to-port="2" /> + <edge from-layer="1593" from-port="0" to-layer="1594" to-port="3" /> + <edge from-layer="1594" from-port="4" to-layer="1595" to-port="3" /> + <edge from-layer="1595" from-port="4" to-layer="1597" to-port="0" /> + <edge from-layer="1596" from-port="0" to-layer="1597" to-port="1" /> + <edge from-layer="1597" from-port="2" to-layer="1599" to-port="0" /> + <edge from-layer="1598" from-port="2" to-layer="1599" to-port="1" /> + <edge from-layer="1599" from-port="2" to-layer="1608" to-port="0" /> + <edge from-layer="1600" from-port="0" to-layer="1601" to-port="0" /> + <edge from-layer="1601" from-port="1" to-layer="1604" to-port="0" /> + <edge from-layer="1602" from-port="0" to-layer="1603" to-port="0" /> + <edge from-layer="1603" from-port="1" to-layer="1604" to-port="1" /> + <edge from-layer="1604" from-port="2" to-layer="1606" to-port="0" /> + <edge from-layer="1605" from-port="0" to-layer="1606" to-port="1" /> + <edge from-layer="1606" from-port="2" to-layer="1607" to-port="0" /> + <edge from-layer="1607" from-port="1" to-layer="1608" to-port="1" /> + <edge from-layer="1608" from-port="2" to-layer="1609" to-port="1" /> + <edge from-layer="1609" from-port="2" to-layer="1624" to-port="0" /> + <edge from-layer="1609" from-port="2" to-layer="1616" to-port="0" /> + <edge from-layer="1609" from-port="2" to-layer="1655" to-port="0" /> + <edge from-layer="1610" from-port="0" to-layer="1611" to-port="0" /> + <edge from-layer="1611" from-port="1" to-layer="1625" to-port="0" /> + <edge from-layer="1612" from-port="0" to-layer="1613" to-port="0" /> + <edge from-layer="1613" from-port="1" to-layer="1623" to-port="0" /> + <edge from-layer="1614" from-port="0" to-layer="1615" to-port="0" /> + <edge from-layer="1615" from-port="1" to-layer="1616" to-port="1" /> + <edge from-layer="1616" from-port="2" to-layer="1618" to-port="0" /> + <edge from-layer="1617" from-port="0" to-layer="1618" to-port="1" /> + <edge from-layer="1618" from-port="2" to-layer="1621" to-port="0" /> + <edge from-layer="1619" from-port="0" to-layer="1620" to-port="0" /> + <edge from-layer="1620" from-port="1" to-layer="1621" to-port="1" /> + <edge from-layer="1621" from-port="2" to-layer="1622" to-port="0" /> + <edge from-layer="1622" from-port="1" to-layer="1623" to-port="1" /> + <edge from-layer="1623" from-port="2" to-layer="1624" to-port="1" /> + <edge from-layer="1624" from-port="2" to-layer="1625" to-port="1" /> + <edge from-layer="1625" from-port="2" to-layer="1644" to-port="0" /> + <edge from-layer="1625" from-port="2" to-layer="1634" to-port="0" /> + <edge from-layer="1626" from-port="0" to-layer="1627" to-port="0" /> + <edge from-layer="1627" from-port="1" to-layer="1630" to-port="0" /> + <edge from-layer="1628" from-port="0" to-layer="1629" to-port="0" /> + <edge from-layer="1629" from-port="1" to-layer="1630" to-port="1" /> + <edge from-layer="1630" from-port="2" to-layer="1632" to-port="0" /> + <edge from-layer="1631" from-port="0" to-layer="1632" to-port="1" /> + <edge from-layer="1632" from-port="2" to-layer="1633" to-port="0" /> + <edge from-layer="1633" from-port="1" to-layer="1634" to-port="1" /> + <edge from-layer="1634" from-port="2" to-layer="1635" to-port="0" /> + <edge from-layer="1635" from-port="1" to-layer="1645" to-port="0" /> + <edge from-layer="1636" from-port="0" to-layer="1637" to-port="0" /> + <edge from-layer="1637" from-port="1" to-layer="1640" to-port="0" /> + <edge from-layer="1638" from-port="0" to-layer="1639" to-port="0" /> + <edge from-layer="1639" from-port="1" to-layer="1640" to-port="1" /> + <edge from-layer="1640" from-port="2" to-layer="1642" to-port="0" /> + <edge from-layer="1641" from-port="0" to-layer="1642" to-port="1" /> + <edge from-layer="1642" from-port="2" to-layer="1643" to-port="0" /> + <edge from-layer="1643" from-port="1" to-layer="1644" to-port="1" /> + <edge from-layer="1644" from-port="2" to-layer="1645" to-port="1" /> + <edge from-layer="1645" from-port="2" to-layer="1654" to-port="0" /> + <edge from-layer="1646" from-port="0" to-layer="1647" to-port="0" /> + <edge from-layer="1647" from-port="1" to-layer="1650" to-port="0" /> + <edge from-layer="1648" from-port="0" to-layer="1649" to-port="0" /> + <edge from-layer="1649" from-port="1" to-layer="1650" to-port="1" /> + <edge from-layer="1650" from-port="2" to-layer="1652" to-port="0" /> + <edge from-layer="1651" from-port="0" to-layer="1652" to-port="1" /> + <edge from-layer="1652" from-port="2" to-layer="1653" to-port="0" /> + <edge from-layer="1653" from-port="1" to-layer="1654" to-port="1" /> + <edge from-layer="1654" from-port="2" to-layer="1655" to-port="1" /> + <edge from-layer="1655" from-port="2" to-layer="1860" to-port="0" /> + <edge from-layer="1655" from-port="2" to-layer="1670" to-port="0" /> + <edge from-layer="1655" from-port="2" to-layer="1662" to-port="0" /> + <edge from-layer="1656" from-port="0" to-layer="1657" to-port="0" /> + <edge from-layer="1657" from-port="1" to-layer="1671" to-port="0" /> + <edge from-layer="1658" from-port="0" to-layer="1659" to-port="0" /> + <edge from-layer="1659" from-port="1" to-layer="1669" to-port="0" /> + <edge from-layer="1660" from-port="0" to-layer="1661" to-port="0" /> + <edge from-layer="1661" from-port="1" to-layer="1662" to-port="1" /> + <edge from-layer="1662" from-port="2" to-layer="1664" to-port="0" /> + <edge from-layer="1663" from-port="0" to-layer="1664" to-port="1" /> + <edge from-layer="1664" from-port="2" to-layer="1667" to-port="0" /> + <edge from-layer="1665" from-port="0" to-layer="1666" to-port="0" /> + <edge from-layer="1666" from-port="1" to-layer="1667" to-port="1" /> + <edge from-layer="1667" from-port="2" to-layer="1668" to-port="0" /> + <edge from-layer="1668" from-port="1" to-layer="1669" to-port="1" /> + <edge from-layer="1669" from-port="2" to-layer="1670" to-port="1" /> + <edge from-layer="1670" from-port="2" to-layer="1671" to-port="1" /> + <edge from-layer="1671" from-port="2" to-layer="1810" to-port="0" /> + <edge from-layer="1671" from-port="2" to-layer="1742" to-port="0" /> + <edge from-layer="1671" from-port="2" to-layer="1681" to-port="0" /> + <edge from-layer="1671" from-port="2" to-layer="1680" to-port="0" /> + <edge from-layer="1672" from-port="0" to-layer="1673" to-port="0" /> + <edge from-layer="1673" from-port="1" to-layer="1676" to-port="0" /> + <edge from-layer="1674" from-port="0" to-layer="1675" to-port="0" /> + <edge from-layer="1675" from-port="1" to-layer="1676" to-port="1" /> + <edge from-layer="1676" from-port="2" to-layer="1678" to-port="0" /> + <edge from-layer="1677" from-port="0" to-layer="1678" to-port="1" /> + <edge from-layer="1678" from-port="2" to-layer="1679" to-port="0" /> + <edge from-layer="1679" from-port="1" to-layer="1680" to-port="1" /> + <edge from-layer="1680" from-port="2" to-layer="1686" to-port="0" /> + <edge from-layer="1681" from-port="1" to-layer="1684" to-port="0" /> + <edge from-layer="1682" from-port="0" to-layer="1684" to-port="1" /> + <edge from-layer="1683" from-port="0" to-layer="1684" to-port="2" /> + <edge from-layer="1684" from-port="3" to-layer="1743" to-port="0" /> + <edge from-layer="1684" from-port="3" to-layer="1849" to-port="0" /> + <edge from-layer="1684" from-port="3" to-layer="1685" to-port="0" /> + <edge from-layer="1685" from-port="3" to-layer="1686" to-port="1" /> + <edge from-layer="1686" from-port="2" to-layer="1688" to-port="0" /> + <edge from-layer="1687" from-port="0" to-layer="1688" to-port="1" /> + <edge from-layer="1688" from-port="2" to-layer="1718" to-port="0" /> + <edge from-layer="1688" from-port="2" to-layer="1712" to-port="0" /> + <edge from-layer="1688" from-port="2" to-layer="1701" to-port="0" /> + <edge from-layer="1688" from-port="2" to-layer="1698" to-port="0" /> + <edge from-layer="1689" from-port="0" to-layer="1690" to-port="0" /> + <edge from-layer="1690" from-port="1" to-layer="1691" to-port="0" /> + <edge from-layer="1691" from-port="2" to-layer="1692" to-port="0" /> + <edge from-layer="1692" from-port="2" to-layer="1694" to-port="0" /> + <edge from-layer="1693" from-port="0" to-layer="1694" to-port="1" /> + <edge from-layer="1694" from-port="2" to-layer="1695" to-port="1" /> + <edge from-layer="1694" from-port="2" to-layer="1695" to-port="0" /> + <edge from-layer="1695" from-port="2" to-layer="1720" to-port="0" /> + <edge from-layer="1695" from-port="2" to-layer="1696" to-port="0" /> + <edge from-layer="1696" from-port="1" to-layer="1697" to-port="0" /> + <edge from-layer="1697" from-port="2" to-layer="1698" to-port="1" /> + <edge from-layer="1697" from-port="2" to-layer="1747" to-port="1" /> + <edge from-layer="1698" from-port="2" to-layer="1723" to-port="0" /> + <edge from-layer="1699" from-port="0" to-layer="1709" to-port="0" /> + <edge from-layer="1700" from-port="0" to-layer="1709" to-port="1" /> + <edge from-layer="1701" from-port="1" to-layer="1703" to-port="0" /> + <edge from-layer="1702" from-port="0" to-layer="1703" to-port="2" /> + <edge from-layer="1703" from-port="3" to-layer="1704" to-port="0" /> + <edge from-layer="1704" from-port="2" to-layer="1705" to-port="0" /> + <edge from-layer="1705" from-port="1" to-layer="1707" to-port="0" /> + <edge from-layer="1706" from-port="0" to-layer="1707" to-port="1" /> + <edge from-layer="1707" from-port="2" to-layer="1709" to-port="2" /> + <edge from-layer="1708" from-port="0" to-layer="1709" to-port="3" /> + <edge from-layer="1709" from-port="4" to-layer="1718" to-port="2" /> + <edge from-layer="1709" from-port="4" to-layer="1712" to-port="1" /> + <edge from-layer="1710" from-port="0" to-layer="1712" to-port="2" /> + <edge from-layer="1711" from-port="0" to-layer="1712" to-port="3" /> + <edge from-layer="1712" from-port="4" to-layer="1715" to-port="0" /> + <edge from-layer="1713" from-port="0" to-layer="1714" to-port="0" /> + <edge from-layer="1714" from-port="1" to-layer="1715" to-port="1" /> + <edge from-layer="1715" from-port="2" to-layer="1719" to-port="0" /> + <edge from-layer="1716" from-port="0" to-layer="1718" to-port="1" /> + <edge from-layer="1717" from-port="0" to-layer="1718" to-port="3" /> + <edge from-layer="1718" from-port="4" to-layer="1719" to-port="1" /> + <edge from-layer="1719" from-port="2" to-layer="1722" to-port="0" /> + <edge from-layer="1720" from-port="1" to-layer="1721" to-port="0" /> + <edge from-layer="1721" from-port="2" to-layer="1769" to-port="1" /> + <edge from-layer="1721" from-port="2" to-layer="1722" to-port="1" /> + <edge from-layer="1722" from-port="2" to-layer="1723" to-port="1" /> + <edge from-layer="1723" from-port="2" to-layer="1846" to-port="0" /> + <edge from-layer="1724" from-port="0" to-layer="1725" to-port="0" /> + <edge from-layer="1725" from-port="1" to-layer="1730" to-port="0" /> + <edge from-layer="1726" from-port="0" to-layer="1729" to-port="1" /> + <edge from-layer="1727" from-port="0" to-layer="1729" to-port="2" /> + <edge from-layer="1728" from-port="0" to-layer="1729" to-port="3" /> + <edge from-layer="1729" from-port="4" to-layer="1730" to-port="1" /> + <edge from-layer="1730" from-port="2" to-layer="1731" to-port="0" /> + <edge from-layer="1731" from-port="1" to-layer="1733" to-port="0" /> + <edge from-layer="1732" from-port="0" to-layer="1733" to-port="2" /> + <edge from-layer="1733" from-port="3" to-layer="1771" to-port="0" /> + <edge from-layer="1734" from-port="0" to-layer="1735" to-port="0" /> + <edge from-layer="1735" from-port="1" to-layer="1738" to-port="0" /> + <edge from-layer="1736" from-port="0" to-layer="1737" to-port="0" /> + <edge from-layer="1737" from-port="1" to-layer="1738" to-port="1" /> + <edge from-layer="1738" from-port="2" to-layer="1740" to-port="0" /> + <edge from-layer="1739" from-port="0" to-layer="1740" to-port="1" /> + <edge from-layer="1740" from-port="2" to-layer="1741" to-port="0" /> + <edge from-layer="1741" from-port="1" to-layer="1742" to-port="1" /> + <edge from-layer="1742" from-port="2" to-layer="1744" to-port="0" /> + <edge from-layer="1743" from-port="3" to-layer="1811" to-port="1" /> + <edge from-layer="1743" from-port="3" to-layer="1744" to-port="1" /> + <edge from-layer="1744" from-port="2" to-layer="1746" to-port="0" /> + <edge from-layer="1745" from-port="0" to-layer="1746" to-port="1" /> + <edge from-layer="1746" from-port="2" to-layer="1747" to-port="0" /> + <edge from-layer="1746" from-port="2" to-layer="1750" to-port="0" /> + <edge from-layer="1746" from-port="2" to-layer="1767" to-port="0" /> + <edge from-layer="1746" from-port="2" to-layer="1761" to-port="0" /> + <edge from-layer="1747" from-port="2" to-layer="1770" to-port="0" /> + <edge from-layer="1748" from-port="0" to-layer="1758" to-port="0" /> + <edge from-layer="1749" from-port="0" to-layer="1758" to-port="1" /> + <edge from-layer="1750" from-port="1" to-layer="1752" to-port="0" /> + <edge from-layer="1751" from-port="0" to-layer="1752" to-port="2" /> + <edge from-layer="1752" from-port="3" to-layer="1753" to-port="0" /> + <edge from-layer="1753" from-port="2" to-layer="1754" to-port="0" /> + <edge from-layer="1754" from-port="1" to-layer="1756" to-port="0" /> + <edge from-layer="1755" from-port="0" to-layer="1756" to-port="1" /> + <edge from-layer="1756" from-port="2" to-layer="1758" to-port="2" /> + <edge from-layer="1757" from-port="0" to-layer="1758" to-port="3" /> + <edge from-layer="1758" from-port="4" to-layer="1761" to-port="1" /> + <edge from-layer="1758" from-port="4" to-layer="1767" to-port="2" /> + <edge from-layer="1759" from-port="0" to-layer="1761" to-port="2" /> + <edge from-layer="1760" from-port="0" to-layer="1761" to-port="3" /> + <edge from-layer="1761" from-port="4" to-layer="1764" to-port="0" /> + <edge from-layer="1762" from-port="0" to-layer="1763" to-port="0" /> + <edge from-layer="1763" from-port="1" to-layer="1764" to-port="1" /> + <edge from-layer="1764" from-port="2" to-layer="1768" to-port="0" /> + <edge from-layer="1765" from-port="0" to-layer="1767" to-port="1" /> + <edge from-layer="1766" from-port="0" to-layer="1767" to-port="3" /> + <edge from-layer="1767" from-port="4" to-layer="1768" to-port="1" /> + <edge from-layer="1768" from-port="2" to-layer="1769" to-port="0" /> + <edge from-layer="1769" from-port="2" to-layer="1770" to-port="1" /> + <edge from-layer="1770" from-port="2" to-layer="1771" to-port="1" /> + <edge from-layer="1771" from-port="2" to-layer="1772" to-port="0" /> + <edge from-layer="1771" from-port="2" to-layer="1773" to-port="0" /> + <edge from-layer="1771" from-port="2" to-layer="2506" to-port="0" /> + <edge from-layer="1772" from-port="2" to-layer="1786" to-port="0" /> + <edge from-layer="1773" from-port="1" to-layer="1778" to-port="0" /> + <edge from-layer="1773" from-port="1" to-layer="1783" to-port="0" /> + <edge from-layer="1773" from-port="1" to-layer="1776" to-port="0" /> + <edge from-layer="1774" from-port="0" to-layer="1776" to-port="1" /> + <edge from-layer="1775" from-port="0" to-layer="1776" to-port="2" /> + <edge from-layer="1776" from-port="3" to-layer="1790" to-port="0" /> + <edge from-layer="1776" from-port="3" to-layer="1784" to-port="0" /> + <edge from-layer="1777" from-port="0" to-layer="1778" to-port="2" /> + <edge from-layer="1778" from-port="3" to-layer="1780" to-port="0" /> + <edge from-layer="1778" from-port="3" to-layer="1787" to-port="0" /> + <edge from-layer="1779" from-port="0" to-layer="1780" to-port="1" /> + <edge from-layer="1780" from-port="2" to-layer="1784" to-port="1" /> + <edge from-layer="1781" from-port="0" to-layer="1783" to-port="1" /> + <edge from-layer="1782" from-port="0" to-layer="1783" to-port="2" /> + <edge from-layer="1783" from-port="3" to-layer="1784" to-port="3" /> + <edge from-layer="1783" from-port="3" to-layer="1790" to-port="2" /> + <edge from-layer="1784" from-port="4" to-layer="1785" to-port="0" /> + <edge from-layer="1785" from-port="1" to-layer="1786" to-port="1" /> + <edge from-layer="1786" from-port="2" to-layer="1791" to-port="0" /> + <edge from-layer="1787" from-port="2" to-layer="1789" to-port="0" /> + <edge from-layer="1788" from-port="0" to-layer="1789" to-port="1" /> + <edge from-layer="1789" from-port="2" to-layer="1790" to-port="1" /> + <edge from-layer="1790" from-port="3" to-layer="1791" to-port="1" /> + <edge from-layer="1791" from-port="2" to-layer="1838" to-port="0" /> + <edge from-layer="1791" from-port="2" to-layer="1846" to-port="1" /> + <edge from-layer="1792" from-port="0" to-layer="1793" to-port="0" /> + <edge from-layer="1793" from-port="1" to-layer="1798" to-port="0" /> + <edge from-layer="1794" from-port="0" to-layer="1797" to-port="1" /> + <edge from-layer="1795" from-port="0" to-layer="1797" to-port="2" /> + <edge from-layer="1796" from-port="0" to-layer="1797" to-port="3" /> + <edge from-layer="1797" from-port="4" to-layer="1798" to-port="1" /> + <edge from-layer="1798" from-port="2" to-layer="1799" to-port="0" /> + <edge from-layer="1799" from-port="1" to-layer="1801" to-port="0" /> + <edge from-layer="1800" from-port="0" to-layer="1801" to-port="2" /> + <edge from-layer="1801" from-port="3" to-layer="1814" to-port="0" /> + <edge from-layer="1802" from-port="0" to-layer="1803" to-port="0" /> + <edge from-layer="1803" from-port="1" to-layer="1806" to-port="0" /> + <edge from-layer="1804" from-port="0" to-layer="1805" to-port="0" /> + <edge from-layer="1805" from-port="1" to-layer="1806" to-port="1" /> + <edge from-layer="1806" from-port="2" to-layer="1808" to-port="0" /> + <edge from-layer="1807" from-port="0" to-layer="1808" to-port="1" /> + <edge from-layer="1808" from-port="2" to-layer="1809" to-port="0" /> + <edge from-layer="1809" from-port="1" to-layer="1810" to-port="1" /> + <edge from-layer="1810" from-port="2" to-layer="1811" to-port="0" /> + <edge from-layer="1811" from-port="2" to-layer="1813" to-port="0" /> + <edge from-layer="1812" from-port="0" to-layer="1813" to-port="1" /> + <edge from-layer="1813" from-port="2" to-layer="1814" to-port="1" /> + <edge from-layer="1814" from-port="2" to-layer="2505" to-port="0" /> + <edge from-layer="1814" from-port="2" to-layer="1815" to-port="0" /> + <edge from-layer="1814" from-port="2" to-layer="1816" to-port="0" /> + <edge from-layer="1815" from-port="2" to-layer="1829" to-port="0" /> + <edge from-layer="1816" from-port="1" to-layer="1819" to-port="0" /> + <edge from-layer="1816" from-port="1" to-layer="1821" to-port="0" /> + <edge from-layer="1816" from-port="1" to-layer="1826" to-port="0" /> + <edge from-layer="1817" from-port="0" to-layer="1819" to-port="1" /> + <edge from-layer="1818" from-port="0" to-layer="1819" to-port="2" /> + <edge from-layer="1819" from-port="3" to-layer="1833" to-port="0" /> + <edge from-layer="1819" from-port="3" to-layer="1827" to-port="0" /> + <edge from-layer="1820" from-port="0" to-layer="1821" to-port="2" /> + <edge from-layer="1821" from-port="3" to-layer="1830" to-port="0" /> + <edge from-layer="1821" from-port="3" to-layer="1823" to-port="0" /> + <edge from-layer="1822" from-port="0" to-layer="1823" to-port="1" /> + <edge from-layer="1823" from-port="2" to-layer="1827" to-port="1" /> + <edge from-layer="1824" from-port="0" to-layer="1826" to-port="1" /> + <edge from-layer="1825" from-port="0" to-layer="1826" to-port="2" /> + <edge from-layer="1826" from-port="3" to-layer="1827" to-port="3" /> + <edge from-layer="1826" from-port="3" to-layer="1833" to-port="2" /> + <edge from-layer="1827" from-port="4" to-layer="1828" to-port="0" /> + <edge from-layer="1828" from-port="1" to-layer="1829" to-port="1" /> + <edge from-layer="1829" from-port="2" to-layer="1834" to-port="0" /> + <edge from-layer="1830" from-port="2" to-layer="1832" to-port="0" /> + <edge from-layer="1831" from-port="0" to-layer="1832" to-port="1" /> + <edge from-layer="1832" from-port="2" to-layer="1833" to-port="1" /> + <edge from-layer="1833" from-port="3" to-layer="1834" to-port="1" /> + <edge from-layer="1834" from-port="2" to-layer="1846" to-port="2" /> + <edge from-layer="1835" from-port="0" to-layer="1845" to-port="1" /> + <edge from-layer="1836" from-port="0" to-layer="1843" to-port="0" /> + <edge from-layer="1837" from-port="0" to-layer="1843" to-port="1" /> + <edge from-layer="1838" from-port="1" to-layer="1841" to-port="0" /> + <edge from-layer="1839" from-port="0" to-layer="1841" to-port="1" /> + <edge from-layer="1840" from-port="0" to-layer="1841" to-port="2" /> + <edge from-layer="1841" from-port="3" to-layer="1843" to-port="2" /> + <edge from-layer="1842" from-port="0" to-layer="1843" to-port="3" /> + <edge from-layer="1843" from-port="4" to-layer="1845" to-port="2" /> + <edge from-layer="1844" from-port="0" to-layer="1845" to-port="3" /> + <edge from-layer="1845" from-port="4" to-layer="1846" to-port="3" /> + <edge from-layer="1846" from-port="4" to-layer="1848" to-port="0" /> + <edge from-layer="1847" from-port="0" to-layer="1848" to-port="1" /> + <edge from-layer="1848" from-port="2" to-layer="1850" to-port="0" /> + <edge from-layer="1849" from-port="2" to-layer="1850" to-port="1" /> + <edge from-layer="1850" from-port="2" to-layer="1859" to-port="0" /> + <edge from-layer="1851" from-port="0" to-layer="1852" to-port="0" /> + <edge from-layer="1852" from-port="1" to-layer="1855" to-port="0" /> + <edge from-layer="1853" from-port="0" to-layer="1854" to-port="0" /> + <edge from-layer="1854" from-port="1" to-layer="1855" to-port="1" /> + <edge from-layer="1855" from-port="2" to-layer="1857" to-port="0" /> + <edge from-layer="1856" from-port="0" to-layer="1857" to-port="1" /> + <edge from-layer="1857" from-port="2" to-layer="1858" to-port="0" /> + <edge from-layer="1858" from-port="1" to-layer="1859" to-port="1" /> + <edge from-layer="1859" from-port="2" to-layer="1860" to-port="1" /> + <edge from-layer="1860" from-port="2" to-layer="1906" to-port="0" /> + <edge from-layer="1860" from-port="2" to-layer="1867" to-port="0" /> + <edge from-layer="1860" from-port="2" to-layer="1875" to-port="0" /> + <edge from-layer="1861" from-port="0" to-layer="1862" to-port="0" /> + <edge from-layer="1862" from-port="1" to-layer="1876" to-port="0" /> + <edge from-layer="1863" from-port="0" to-layer="1864" to-port="0" /> + <edge from-layer="1864" from-port="1" to-layer="1874" to-port="0" /> + <edge from-layer="1865" from-port="0" to-layer="1866" to-port="0" /> + <edge from-layer="1866" from-port="1" to-layer="1867" to-port="1" /> + <edge from-layer="1867" from-port="2" to-layer="1869" to-port="0" /> + <edge from-layer="1868" from-port="0" to-layer="1869" to-port="1" /> + <edge from-layer="1869" from-port="2" to-layer="1872" to-port="0" /> + <edge from-layer="1870" from-port="0" to-layer="1871" to-port="0" /> + <edge from-layer="1871" from-port="1" to-layer="1872" to-port="1" /> + <edge from-layer="1872" from-port="2" to-layer="1873" to-port="0" /> + <edge from-layer="1873" from-port="1" to-layer="1874" to-port="1" /> + <edge from-layer="1874" from-port="2" to-layer="1875" to-port="1" /> + <edge from-layer="1875" from-port="2" to-layer="1876" to-port="1" /> + <edge from-layer="1876" from-port="2" to-layer="1885" to-port="0" /> + <edge from-layer="1876" from-port="2" to-layer="1895" to-port="0" /> + <edge from-layer="1877" from-port="0" to-layer="1878" to-port="0" /> + <edge from-layer="1878" from-port="1" to-layer="1881" to-port="0" /> + <edge from-layer="1879" from-port="0" to-layer="1880" to-port="0" /> + <edge from-layer="1880" from-port="1" to-layer="1881" to-port="1" /> + <edge from-layer="1881" from-port="2" to-layer="1883" to-port="0" /> + <edge from-layer="1882" from-port="0" to-layer="1883" to-port="1" /> + <edge from-layer="1883" from-port="2" to-layer="1884" to-port="0" /> + <edge from-layer="1884" from-port="1" to-layer="1885" to-port="1" /> + <edge from-layer="1885" from-port="2" to-layer="1886" to-port="0" /> + <edge from-layer="1886" from-port="1" to-layer="1896" to-port="0" /> + <edge from-layer="1887" from-port="0" to-layer="1888" to-port="0" /> + <edge from-layer="1888" from-port="1" to-layer="1891" to-port="0" /> + <edge from-layer="1889" from-port="0" to-layer="1890" to-port="0" /> + <edge from-layer="1890" from-port="1" to-layer="1891" to-port="1" /> + <edge from-layer="1891" from-port="2" to-layer="1893" to-port="0" /> + <edge from-layer="1892" from-port="0" to-layer="1893" to-port="1" /> + <edge from-layer="1893" from-port="2" to-layer="1894" to-port="0" /> + <edge from-layer="1894" from-port="1" to-layer="1895" to-port="1" /> + <edge from-layer="1895" from-port="2" to-layer="1896" to-port="1" /> + <edge from-layer="1896" from-port="2" to-layer="1905" to-port="0" /> + <edge from-layer="1897" from-port="0" to-layer="1898" to-port="0" /> + <edge from-layer="1898" from-port="1" to-layer="1901" to-port="0" /> + <edge from-layer="1899" from-port="0" to-layer="1900" to-port="0" /> + <edge from-layer="1900" from-port="1" to-layer="1901" to-port="1" /> + <edge from-layer="1901" from-port="2" to-layer="1903" to-port="0" /> + <edge from-layer="1902" from-port="0" to-layer="1903" to-port="1" /> + <edge from-layer="1903" from-port="2" to-layer="1904" to-port="0" /> + <edge from-layer="1904" from-port="1" to-layer="1905" to-port="1" /> + <edge from-layer="1905" from-port="2" to-layer="1906" to-port="1" /> + <edge from-layer="1906" from-port="2" to-layer="1921" to-port="0" /> + <edge from-layer="1906" from-port="2" to-layer="1913" to-port="0" /> + <edge from-layer="1906" from-port="2" to-layer="2111" to-port="0" /> + <edge from-layer="1907" from-port="0" to-layer="1908" to-port="0" /> + <edge from-layer="1908" from-port="1" to-layer="1922" to-port="0" /> + <edge from-layer="1909" from-port="0" to-layer="1910" to-port="0" /> + <edge from-layer="1910" from-port="1" to-layer="1920" to-port="0" /> + <edge from-layer="1911" from-port="0" to-layer="1912" to-port="0" /> + <edge from-layer="1912" from-port="1" to-layer="1913" to-port="1" /> + <edge from-layer="1913" from-port="2" to-layer="1915" to-port="0" /> + <edge from-layer="1914" from-port="0" to-layer="1915" to-port="1" /> + <edge from-layer="1915" from-port="2" to-layer="1918" to-port="0" /> + <edge from-layer="1916" from-port="0" to-layer="1917" to-port="0" /> + <edge from-layer="1917" from-port="1" to-layer="1918" to-port="1" /> + <edge from-layer="1918" from-port="2" to-layer="1919" to-port="0" /> + <edge from-layer="1919" from-port="1" to-layer="1920" to-port="1" /> + <edge from-layer="1920" from-port="2" to-layer="1921" to-port="1" /> + <edge from-layer="1921" from-port="2" to-layer="1922" to-port="1" /> + <edge from-layer="1922" from-port="2" to-layer="1993" to-port="0" /> + <edge from-layer="1922" from-port="2" to-layer="2061" to-port="0" /> + <edge from-layer="1922" from-port="2" to-layer="1932" to-port="0" /> + <edge from-layer="1922" from-port="2" to-layer="1931" to-port="0" /> + <edge from-layer="1923" from-port="0" to-layer="1924" to-port="0" /> + <edge from-layer="1924" from-port="1" to-layer="1927" to-port="0" /> + <edge from-layer="1925" from-port="0" to-layer="1926" to-port="0" /> + <edge from-layer="1926" from-port="1" to-layer="1927" to-port="1" /> + <edge from-layer="1927" from-port="2" to-layer="1929" to-port="0" /> + <edge from-layer="1928" from-port="0" to-layer="1929" to-port="1" /> + <edge from-layer="1929" from-port="2" to-layer="1930" to-port="0" /> + <edge from-layer="1930" from-port="1" to-layer="1931" to-port="1" /> + <edge from-layer="1931" from-port="2" to-layer="1937" to-port="0" /> + <edge from-layer="1932" from-port="1" to-layer="1935" to-port="0" /> + <edge from-layer="1933" from-port="0" to-layer="1935" to-port="1" /> + <edge from-layer="1934" from-port="0" to-layer="1935" to-port="2" /> + <edge from-layer="1935" from-port="3" to-layer="2100" to-port="0" /> + <edge from-layer="1935" from-port="3" to-layer="1994" to-port="0" /> + <edge from-layer="1935" from-port="3" to-layer="1936" to-port="0" /> + <edge from-layer="1936" from-port="3" to-layer="1937" to-port="1" /> + <edge from-layer="1937" from-port="2" to-layer="1939" to-port="0" /> + <edge from-layer="1938" from-port="0" to-layer="1939" to-port="1" /> + <edge from-layer="1939" from-port="2" to-layer="1952" to-port="0" /> + <edge from-layer="1939" from-port="2" to-layer="1949" to-port="0" /> + <edge from-layer="1939" from-port="2" to-layer="1963" to-port="0" /> + <edge from-layer="1939" from-port="2" to-layer="1969" to-port="0" /> + <edge from-layer="1940" from-port="0" to-layer="1941" to-port="0" /> + <edge from-layer="1941" from-port="1" to-layer="1942" to-port="0" /> + <edge from-layer="1942" from-port="2" to-layer="1943" to-port="0" /> + <edge from-layer="1943" from-port="2" to-layer="1945" to-port="0" /> + <edge from-layer="1944" from-port="0" to-layer="1945" to-port="1" /> + <edge from-layer="1945" from-port="2" to-layer="1946" to-port="0" /> + <edge from-layer="1945" from-port="2" to-layer="1946" to-port="1" /> + <edge from-layer="1946" from-port="2" to-layer="1947" to-port="0" /> + <edge from-layer="1946" from-port="2" to-layer="1971" to-port="0" /> + <edge from-layer="1947" from-port="1" to-layer="1948" to-port="0" /> + <edge from-layer="1948" from-port="2" to-layer="1998" to-port="1" /> + <edge from-layer="1948" from-port="2" to-layer="1949" to-port="1" /> + <edge from-layer="1949" from-port="2" to-layer="1974" to-port="0" /> + <edge from-layer="1950" from-port="0" to-layer="1960" to-port="0" /> + <edge from-layer="1951" from-port="0" to-layer="1960" to-port="1" /> + <edge from-layer="1952" from-port="1" to-layer="1954" to-port="0" /> + <edge from-layer="1953" from-port="0" to-layer="1954" to-port="2" /> + <edge from-layer="1954" from-port="3" to-layer="1955" to-port="0" /> + <edge from-layer="1955" from-port="2" to-layer="1956" to-port="0" /> + <edge from-layer="1956" from-port="1" to-layer="1958" to-port="0" /> + <edge from-layer="1957" from-port="0" to-layer="1958" to-port="1" /> + <edge from-layer="1958" from-port="2" to-layer="1960" to-port="2" /> + <edge from-layer="1959" from-port="0" to-layer="1960" to-port="3" /> + <edge from-layer="1960" from-port="4" to-layer="1969" to-port="2" /> + <edge from-layer="1960" from-port="4" to-layer="1963" to-port="1" /> + <edge from-layer="1961" from-port="0" to-layer="1963" to-port="2" /> + <edge from-layer="1962" from-port="0" to-layer="1963" to-port="3" /> + <edge from-layer="1963" from-port="4" to-layer="1966" to-port="0" /> + <edge from-layer="1964" from-port="0" to-layer="1965" to-port="0" /> + <edge from-layer="1965" from-port="1" to-layer="1966" to-port="1" /> + <edge from-layer="1966" from-port="2" to-layer="1970" to-port="0" /> + <edge from-layer="1967" from-port="0" to-layer="1969" to-port="1" /> + <edge from-layer="1968" from-port="0" to-layer="1969" to-port="3" /> + <edge from-layer="1969" from-port="4" to-layer="1970" to-port="1" /> + <edge from-layer="1970" from-port="2" to-layer="1973" to-port="0" /> + <edge from-layer="1971" from-port="1" to-layer="1972" to-port="0" /> + <edge from-layer="1972" from-port="2" to-layer="1973" to-port="1" /> + <edge from-layer="1972" from-port="2" to-layer="2020" to-port="1" /> + <edge from-layer="1973" from-port="2" to-layer="1974" to-port="1" /> + <edge from-layer="1974" from-port="2" to-layer="2097" to-port="0" /> + <edge from-layer="1975" from-port="0" to-layer="1976" to-port="0" /> + <edge from-layer="1976" from-port="1" to-layer="1981" to-port="0" /> + <edge from-layer="1977" from-port="0" to-layer="1980" to-port="1" /> + <edge from-layer="1978" from-port="0" to-layer="1980" to-port="2" /> + <edge from-layer="1979" from-port="0" to-layer="1980" to-port="3" /> + <edge from-layer="1980" from-port="4" to-layer="1981" to-port="1" /> + <edge from-layer="1981" from-port="2" to-layer="1982" to-port="0" /> + <edge from-layer="1982" from-port="1" to-layer="1984" to-port="0" /> + <edge from-layer="1983" from-port="0" to-layer="1984" to-port="2" /> + <edge from-layer="1984" from-port="3" to-layer="2022" to-port="0" /> + <edge from-layer="1985" from-port="0" to-layer="1986" to-port="0" /> + <edge from-layer="1986" from-port="1" to-layer="1989" to-port="0" /> + <edge from-layer="1987" from-port="0" to-layer="1988" to-port="0" /> + <edge from-layer="1988" from-port="1" to-layer="1989" to-port="1" /> + <edge from-layer="1989" from-port="2" to-layer="1991" to-port="0" /> + <edge from-layer="1990" from-port="0" to-layer="1991" to-port="1" /> + <edge from-layer="1991" from-port="2" to-layer="1992" to-port="0" /> + <edge from-layer="1992" from-port="1" to-layer="1993" to-port="1" /> + <edge from-layer="1993" from-port="2" to-layer="1995" to-port="0" /> + <edge from-layer="1994" from-port="3" to-layer="2062" to-port="1" /> + <edge from-layer="1994" from-port="3" to-layer="1995" to-port="1" /> + <edge from-layer="1995" from-port="2" to-layer="1997" to-port="0" /> + <edge from-layer="1996" from-port="0" to-layer="1997" to-port="1" /> + <edge from-layer="1997" from-port="2" to-layer="2012" to-port="0" /> + <edge from-layer="1997" from-port="2" to-layer="1998" to-port="0" /> + <edge from-layer="1997" from-port="2" to-layer="2001" to-port="0" /> + <edge from-layer="1997" from-port="2" to-layer="2018" to-port="0" /> + <edge from-layer="1998" from-port="2" to-layer="2021" to-port="0" /> + <edge from-layer="1999" from-port="0" to-layer="2009" to-port="0" /> + <edge from-layer="2000" from-port="0" to-layer="2009" to-port="1" /> + <edge from-layer="2001" from-port="1" to-layer="2003" to-port="0" /> + <edge from-layer="2002" from-port="0" to-layer="2003" to-port="2" /> + <edge from-layer="2003" from-port="3" to-layer="2004" to-port="0" /> + <edge from-layer="2004" from-port="2" to-layer="2005" to-port="0" /> + <edge from-layer="2005" from-port="1" to-layer="2007" to-port="0" /> + <edge from-layer="2006" from-port="0" to-layer="2007" to-port="1" /> + <edge from-layer="2007" from-port="2" to-layer="2009" to-port="2" /> + <edge from-layer="2008" from-port="0" to-layer="2009" to-port="3" /> + <edge from-layer="2009" from-port="4" to-layer="2012" to-port="1" /> + <edge from-layer="2009" from-port="4" to-layer="2018" to-port="2" /> + <edge from-layer="2010" from-port="0" to-layer="2012" to-port="2" /> + <edge from-layer="2011" from-port="0" to-layer="2012" to-port="3" /> + <edge from-layer="2012" from-port="4" to-layer="2015" to-port="0" /> + <edge from-layer="2013" from-port="0" to-layer="2014" to-port="0" /> + <edge from-layer="2014" from-port="1" to-layer="2015" to-port="1" /> + <edge from-layer="2015" from-port="2" to-layer="2019" to-port="0" /> + <edge from-layer="2016" from-port="0" to-layer="2018" to-port="1" /> + <edge from-layer="2017" from-port="0" to-layer="2018" to-port="3" /> + <edge from-layer="2018" from-port="4" to-layer="2019" to-port="1" /> + <edge from-layer="2019" from-port="2" to-layer="2020" to-port="0" /> + <edge from-layer="2020" from-port="2" to-layer="2021" to-port="1" /> + <edge from-layer="2021" from-port="2" to-layer="2022" to-port="1" /> + <edge from-layer="2022" from-port="2" to-layer="2504" to-port="0" /> + <edge from-layer="2022" from-port="2" to-layer="2023" to-port="0" /> + <edge from-layer="2022" from-port="2" to-layer="2024" to-port="0" /> + <edge from-layer="2023" from-port="2" to-layer="2037" to-port="0" /> + <edge from-layer="2024" from-port="1" to-layer="2034" to-port="0" /> + <edge from-layer="2024" from-port="1" to-layer="2029" to-port="0" /> + <edge from-layer="2024" from-port="1" to-layer="2027" to-port="0" /> + <edge from-layer="2025" from-port="0" to-layer="2027" to-port="1" /> + <edge from-layer="2026" from-port="0" to-layer="2027" to-port="2" /> + <edge from-layer="2027" from-port="3" to-layer="2035" to-port="0" /> + <edge from-layer="2027" from-port="3" to-layer="2041" to-port="0" /> + <edge from-layer="2028" from-port="0" to-layer="2029" to-port="2" /> + <edge from-layer="2029" from-port="3" to-layer="2031" to-port="0" /> + <edge from-layer="2029" from-port="3" to-layer="2038" to-port="0" /> + <edge from-layer="2030" from-port="0" to-layer="2031" to-port="1" /> + <edge from-layer="2031" from-port="2" to-layer="2035" to-port="1" /> + <edge from-layer="2032" from-port="0" to-layer="2034" to-port="1" /> + <edge from-layer="2033" from-port="0" to-layer="2034" to-port="2" /> + <edge from-layer="2034" from-port="3" to-layer="2041" to-port="2" /> + <edge from-layer="2034" from-port="3" to-layer="2035" to-port="3" /> + <edge from-layer="2035" from-port="4" to-layer="2036" to-port="0" /> + <edge from-layer="2036" from-port="1" to-layer="2037" to-port="1" /> + <edge from-layer="2037" from-port="2" to-layer="2042" to-port="0" /> + <edge from-layer="2038" from-port="2" to-layer="2040" to-port="0" /> + <edge from-layer="2039" from-port="0" to-layer="2040" to-port="1" /> + <edge from-layer="2040" from-port="2" to-layer="2041" to-port="1" /> + <edge from-layer="2041" from-port="3" to-layer="2042" to-port="1" /> + <edge from-layer="2042" from-port="2" to-layer="2097" to-port="1" /> + <edge from-layer="2042" from-port="2" to-layer="2089" to-port="0" /> + <edge from-layer="2043" from-port="0" to-layer="2044" to-port="0" /> + <edge from-layer="2044" from-port="1" to-layer="2049" to-port="0" /> + <edge from-layer="2045" from-port="0" to-layer="2048" to-port="1" /> + <edge from-layer="2046" from-port="0" to-layer="2048" to-port="2" /> + <edge from-layer="2047" from-port="0" to-layer="2048" to-port="3" /> + <edge from-layer="2048" from-port="4" to-layer="2049" to-port="1" /> + <edge from-layer="2049" from-port="2" to-layer="2050" to-port="0" /> + <edge from-layer="2050" from-port="1" to-layer="2052" to-port="0" /> + <edge from-layer="2051" from-port="0" to-layer="2052" to-port="2" /> + <edge from-layer="2052" from-port="3" to-layer="2065" to-port="0" /> + <edge from-layer="2053" from-port="0" to-layer="2054" to-port="0" /> + <edge from-layer="2054" from-port="1" to-layer="2057" to-port="0" /> + <edge from-layer="2055" from-port="0" to-layer="2056" to-port="0" /> + <edge from-layer="2056" from-port="1" to-layer="2057" to-port="1" /> + <edge from-layer="2057" from-port="2" to-layer="2059" to-port="0" /> + <edge from-layer="2058" from-port="0" to-layer="2059" to-port="1" /> + <edge from-layer="2059" from-port="2" to-layer="2060" to-port="0" /> + <edge from-layer="2060" from-port="1" to-layer="2061" to-port="1" /> + <edge from-layer="2061" from-port="2" to-layer="2062" to-port="0" /> + <edge from-layer="2062" from-port="2" to-layer="2064" to-port="0" /> + <edge from-layer="2063" from-port="0" to-layer="2064" to-port="1" /> + <edge from-layer="2064" from-port="2" to-layer="2065" to-port="1" /> + <edge from-layer="2065" from-port="2" to-layer="2066" to-port="0" /> + <edge from-layer="2065" from-port="2" to-layer="2067" to-port="0" /> + <edge from-layer="2065" from-port="2" to-layer="2503" to-port="0" /> + <edge from-layer="2066" from-port="2" to-layer="2080" to-port="0" /> + <edge from-layer="2067" from-port="1" to-layer="2070" to-port="0" /> + <edge from-layer="2067" from-port="1" to-layer="2072" to-port="0" /> + <edge from-layer="2067" from-port="1" to-layer="2077" to-port="0" /> + <edge from-layer="2068" from-port="0" to-layer="2070" to-port="1" /> + <edge from-layer="2069" from-port="0" to-layer="2070" to-port="2" /> + <edge from-layer="2070" from-port="3" to-layer="2084" to-port="0" /> + <edge from-layer="2070" from-port="3" to-layer="2078" to-port="0" /> + <edge from-layer="2071" from-port="0" to-layer="2072" to-port="2" /> + <edge from-layer="2072" from-port="3" to-layer="2074" to-port="0" /> + <edge from-layer="2072" from-port="3" to-layer="2081" to-port="0" /> + <edge from-layer="2073" from-port="0" to-layer="2074" to-port="1" /> + <edge from-layer="2074" from-port="2" to-layer="2078" to-port="1" /> + <edge from-layer="2075" from-port="0" to-layer="2077" to-port="1" /> + <edge from-layer="2076" from-port="0" to-layer="2077" to-port="2" /> + <edge from-layer="2077" from-port="3" to-layer="2078" to-port="3" /> + <edge from-layer="2077" from-port="3" to-layer="2084" to-port="2" /> + <edge from-layer="2078" from-port="4" to-layer="2079" to-port="0" /> + <edge from-layer="2079" from-port="1" to-layer="2080" to-port="1" /> + <edge from-layer="2080" from-port="2" to-layer="2085" to-port="0" /> + <edge from-layer="2081" from-port="2" to-layer="2083" to-port="0" /> + <edge from-layer="2082" from-port="0" to-layer="2083" to-port="1" /> + <edge from-layer="2083" from-port="2" to-layer="2084" to-port="1" /> + <edge from-layer="2084" from-port="3" to-layer="2085" to-port="1" /> + <edge from-layer="2085" from-port="2" to-layer="2097" to-port="2" /> + <edge from-layer="2086" from-port="0" to-layer="2096" to-port="1" /> + <edge from-layer="2087" from-port="0" to-layer="2094" to-port="0" /> + <edge from-layer="2088" from-port="0" to-layer="2094" to-port="1" /> + <edge from-layer="2089" from-port="1" to-layer="2092" to-port="0" /> + <edge from-layer="2090" from-port="0" to-layer="2092" to-port="1" /> + <edge from-layer="2091" from-port="0" to-layer="2092" to-port="2" /> + <edge from-layer="2092" from-port="3" to-layer="2094" to-port="2" /> + <edge from-layer="2093" from-port="0" to-layer="2094" to-port="3" /> + <edge from-layer="2094" from-port="4" to-layer="2096" to-port="2" /> + <edge from-layer="2095" from-port="0" to-layer="2096" to-port="3" /> + <edge from-layer="2096" from-port="4" to-layer="2097" to-port="3" /> + <edge from-layer="2097" from-port="4" to-layer="2099" to-port="0" /> + <edge from-layer="2098" from-port="0" to-layer="2099" to-port="1" /> + <edge from-layer="2099" from-port="2" to-layer="2101" to-port="0" /> + <edge from-layer="2100" from-port="2" to-layer="2101" to-port="1" /> + <edge from-layer="2101" from-port="2" to-layer="2110" to-port="0" /> + <edge from-layer="2102" from-port="0" to-layer="2103" to-port="0" /> + <edge from-layer="2103" from-port="1" to-layer="2106" to-port="0" /> + <edge from-layer="2104" from-port="0" to-layer="2105" to-port="0" /> + <edge from-layer="2105" from-port="1" to-layer="2106" to-port="1" /> + <edge from-layer="2106" from-port="2" to-layer="2108" to-port="0" /> + <edge from-layer="2107" from-port="0" to-layer="2108" to-port="1" /> + <edge from-layer="2108" from-port="2" to-layer="2109" to-port="0" /> + <edge from-layer="2109" from-port="1" to-layer="2110" to-port="1" /> + <edge from-layer="2110" from-port="2" to-layer="2111" to-port="1" /> + <edge from-layer="2111" from-port="2" to-layer="2126" to-port="0" /> + <edge from-layer="2111" from-port="2" to-layer="2157" to-port="0" /> + <edge from-layer="2111" from-port="2" to-layer="2118" to-port="0" /> + <edge from-layer="2112" from-port="0" to-layer="2113" to-port="0" /> + <edge from-layer="2113" from-port="1" to-layer="2127" to-port="0" /> + <edge from-layer="2114" from-port="0" to-layer="2115" to-port="0" /> + <edge from-layer="2115" from-port="1" to-layer="2125" to-port="0" /> + <edge from-layer="2116" from-port="0" to-layer="2117" to-port="0" /> + <edge from-layer="2117" from-port="1" to-layer="2118" to-port="1" /> + <edge from-layer="2118" from-port="2" to-layer="2120" to-port="0" /> + <edge from-layer="2119" from-port="0" to-layer="2120" to-port="1" /> + <edge from-layer="2120" from-port="2" to-layer="2123" to-port="0" /> + <edge from-layer="2121" from-port="0" to-layer="2122" to-port="0" /> + <edge from-layer="2122" from-port="1" to-layer="2123" to-port="1" /> + <edge from-layer="2123" from-port="2" to-layer="2124" to-port="0" /> + <edge from-layer="2124" from-port="1" to-layer="2125" to-port="1" /> + <edge from-layer="2125" from-port="2" to-layer="2126" to-port="1" /> + <edge from-layer="2126" from-port="2" to-layer="2127" to-port="1" /> + <edge from-layer="2127" from-port="2" to-layer="2136" to-port="0" /> + <edge from-layer="2127" from-port="2" to-layer="2146" to-port="0" /> + <edge from-layer="2128" from-port="0" to-layer="2129" to-port="0" /> + <edge from-layer="2129" from-port="1" to-layer="2132" to-port="0" /> + <edge from-layer="2130" from-port="0" to-layer="2131" to-port="0" /> + <edge from-layer="2131" from-port="1" to-layer="2132" to-port="1" /> + <edge from-layer="2132" from-port="2" to-layer="2134" to-port="0" /> + <edge from-layer="2133" from-port="0" to-layer="2134" to-port="1" /> + <edge from-layer="2134" from-port="2" to-layer="2135" to-port="0" /> + <edge from-layer="2135" from-port="1" to-layer="2136" to-port="1" /> + <edge from-layer="2136" from-port="2" to-layer="2137" to-port="0" /> + <edge from-layer="2137" from-port="1" to-layer="2147" to-port="0" /> + <edge from-layer="2138" from-port="0" to-layer="2139" to-port="0" /> + <edge from-layer="2139" from-port="1" to-layer="2142" to-port="0" /> + <edge from-layer="2140" from-port="0" to-layer="2141" to-port="0" /> + <edge from-layer="2141" from-port="1" to-layer="2142" to-port="1" /> + <edge from-layer="2142" from-port="2" to-layer="2144" to-port="0" /> + <edge from-layer="2143" from-port="0" to-layer="2144" to-port="1" /> + <edge from-layer="2144" from-port="2" to-layer="2145" to-port="0" /> + <edge from-layer="2145" from-port="1" to-layer="2146" to-port="1" /> + <edge from-layer="2146" from-port="2" to-layer="2147" to-port="1" /> + <edge from-layer="2147" from-port="2" to-layer="2156" to-port="0" /> + <edge from-layer="2148" from-port="0" to-layer="2149" to-port="0" /> + <edge from-layer="2149" from-port="1" to-layer="2152" to-port="0" /> + <edge from-layer="2150" from-port="0" to-layer="2151" to-port="0" /> + <edge from-layer="2151" from-port="1" to-layer="2152" to-port="1" /> + <edge from-layer="2152" from-port="2" to-layer="2154" to-port="0" /> + <edge from-layer="2153" from-port="0" to-layer="2154" to-port="1" /> + <edge from-layer="2154" from-port="2" to-layer="2155" to-port="0" /> + <edge from-layer="2155" from-port="1" to-layer="2156" to-port="1" /> + <edge from-layer="2156" from-port="2" to-layer="2157" to-port="1" /> + <edge from-layer="2157" from-port="2" to-layer="2362" to-port="0" /> + <edge from-layer="2157" from-port="2" to-layer="2164" to-port="0" /> + <edge from-layer="2157" from-port="2" to-layer="2172" to-port="0" /> + <edge from-layer="2158" from-port="0" to-layer="2159" to-port="0" /> + <edge from-layer="2159" from-port="1" to-layer="2173" to-port="0" /> + <edge from-layer="2160" from-port="0" to-layer="2161" to-port="0" /> + <edge from-layer="2161" from-port="1" to-layer="2171" to-port="0" /> + <edge from-layer="2162" from-port="0" to-layer="2163" to-port="0" /> + <edge from-layer="2163" from-port="1" to-layer="2164" to-port="1" /> + <edge from-layer="2164" from-port="2" to-layer="2166" to-port="0" /> + <edge from-layer="2165" from-port="0" to-layer="2166" to-port="1" /> + <edge from-layer="2166" from-port="2" to-layer="2169" to-port="0" /> + <edge from-layer="2167" from-port="0" to-layer="2168" to-port="0" /> + <edge from-layer="2168" from-port="1" to-layer="2169" to-port="1" /> + <edge from-layer="2169" from-port="2" to-layer="2170" to-port="0" /> + <edge from-layer="2170" from-port="1" to-layer="2171" to-port="1" /> + <edge from-layer="2171" from-port="2" to-layer="2172" to-port="1" /> + <edge from-layer="2172" from-port="2" to-layer="2173" to-port="1" /> + <edge from-layer="2173" from-port="2" to-layer="2244" to-port="0" /> + <edge from-layer="2173" from-port="2" to-layer="2183" to-port="0" /> + <edge from-layer="2173" from-port="2" to-layer="2182" to-port="0" /> + <edge from-layer="2173" from-port="2" to-layer="2312" to-port="0" /> + <edge from-layer="2174" from-port="0" to-layer="2175" to-port="0" /> + <edge from-layer="2175" from-port="1" to-layer="2178" to-port="0" /> + <edge from-layer="2176" from-port="0" to-layer="2177" to-port="0" /> + <edge from-layer="2177" from-port="1" to-layer="2178" to-port="1" /> + <edge from-layer="2178" from-port="2" to-layer="2180" to-port="0" /> + <edge from-layer="2179" from-port="0" to-layer="2180" to-port="1" /> + <edge from-layer="2180" from-port="2" to-layer="2181" to-port="0" /> + <edge from-layer="2181" from-port="1" to-layer="2182" to-port="1" /> + <edge from-layer="2182" from-port="2" to-layer="2188" to-port="0" /> + <edge from-layer="2183" from-port="1" to-layer="2186" to-port="0" /> + <edge from-layer="2184" from-port="0" to-layer="2186" to-port="1" /> + <edge from-layer="2185" from-port="0" to-layer="2186" to-port="2" /> + <edge from-layer="2186" from-port="3" to-layer="2187" to-port="0" /> + <edge from-layer="2186" from-port="3" to-layer="2351" to-port="0" /> + <edge from-layer="2186" from-port="3" to-layer="2245" to-port="0" /> + <edge from-layer="2187" from-port="3" to-layer="2188" to-port="1" /> + <edge from-layer="2188" from-port="2" to-layer="2190" to-port="0" /> + <edge from-layer="2189" from-port="0" to-layer="2190" to-port="1" /> + <edge from-layer="2190" from-port="2" to-layer="2203" to-port="0" /> + <edge from-layer="2190" from-port="2" to-layer="2214" to-port="0" /> + <edge from-layer="2190" from-port="2" to-layer="2200" to-port="0" /> + <edge from-layer="2190" from-port="2" to-layer="2220" to-port="0" /> + <edge from-layer="2191" from-port="0" to-layer="2192" to-port="0" /> + <edge from-layer="2192" from-port="1" to-layer="2193" to-port="0" /> + <edge from-layer="2193" from-port="2" to-layer="2194" to-port="0" /> + <edge from-layer="2194" from-port="2" to-layer="2196" to-port="0" /> + <edge from-layer="2195" from-port="0" to-layer="2196" to-port="1" /> + <edge from-layer="2196" from-port="2" to-layer="2197" to-port="1" /> + <edge from-layer="2196" from-port="2" to-layer="2197" to-port="0" /> + <edge from-layer="2197" from-port="2" to-layer="2198" to-port="0" /> + <edge from-layer="2197" from-port="2" to-layer="2222" to-port="0" /> + <edge from-layer="2198" from-port="1" to-layer="2199" to-port="0" /> + <edge from-layer="2199" from-port="2" to-layer="2200" to-port="1" /> + <edge from-layer="2199" from-port="2" to-layer="2249" to-port="1" /> + <edge from-layer="2200" from-port="2" to-layer="2225" to-port="0" /> + <edge from-layer="2201" from-port="0" to-layer="2211" to-port="0" /> + <edge from-layer="2202" from-port="0" to-layer="2211" to-port="1" /> + <edge from-layer="2203" from-port="1" to-layer="2205" to-port="0" /> + <edge from-layer="2204" from-port="0" to-layer="2205" to-port="2" /> + <edge from-layer="2205" from-port="3" to-layer="2206" to-port="0" /> + <edge from-layer="2206" from-port="2" to-layer="2207" to-port="0" /> + <edge from-layer="2207" from-port="1" to-layer="2209" to-port="0" /> + <edge from-layer="2208" from-port="0" to-layer="2209" to-port="1" /> + <edge from-layer="2209" from-port="2" to-layer="2211" to-port="2" /> + <edge from-layer="2210" from-port="0" to-layer="2211" to-port="3" /> + <edge from-layer="2211" from-port="4" to-layer="2214" to-port="1" /> + <edge from-layer="2211" from-port="4" to-layer="2220" to-port="2" /> + <edge from-layer="2212" from-port="0" to-layer="2214" to-port="2" /> + <edge from-layer="2213" from-port="0" to-layer="2214" to-port="3" /> + <edge from-layer="2214" from-port="4" to-layer="2217" to-port="0" /> + <edge from-layer="2215" from-port="0" to-layer="2216" to-port="0" /> + <edge from-layer="2216" from-port="1" to-layer="2217" to-port="1" /> + <edge from-layer="2217" from-port="2" to-layer="2221" to-port="0" /> + <edge from-layer="2218" from-port="0" to-layer="2220" to-port="1" /> + <edge from-layer="2219" from-port="0" to-layer="2220" to-port="3" /> + <edge from-layer="2220" from-port="4" to-layer="2221" to-port="1" /> + <edge from-layer="2221" from-port="2" to-layer="2224" to-port="0" /> + <edge from-layer="2222" from-port="1" to-layer="2223" to-port="0" /> + <edge from-layer="2223" from-port="2" to-layer="2224" to-port="1" /> + <edge from-layer="2223" from-port="2" to-layer="2271" to-port="1" /> + <edge from-layer="2224" from-port="2" to-layer="2225" to-port="1" /> + <edge from-layer="2225" from-port="2" to-layer="2348" to-port="0" /> + <edge from-layer="2226" from-port="0" to-layer="2227" to-port="0" /> + <edge from-layer="2227" from-port="1" to-layer="2232" to-port="0" /> + <edge from-layer="2228" from-port="0" to-layer="2231" to-port="1" /> + <edge from-layer="2229" from-port="0" to-layer="2231" to-port="2" /> + <edge from-layer="2230" from-port="0" to-layer="2231" to-port="3" /> + <edge from-layer="2231" from-port="4" to-layer="2232" to-port="1" /> + <edge from-layer="2232" from-port="2" to-layer="2233" to-port="0" /> + <edge from-layer="2233" from-port="1" to-layer="2235" to-port="0" /> + <edge from-layer="2234" from-port="0" to-layer="2235" to-port="2" /> + <edge from-layer="2235" from-port="3" to-layer="2273" to-port="0" /> + <edge from-layer="2236" from-port="0" to-layer="2237" to-port="0" /> + <edge from-layer="2237" from-port="1" to-layer="2240" to-port="0" /> + <edge from-layer="2238" from-port="0" to-layer="2239" to-port="0" /> + <edge from-layer="2239" from-port="1" to-layer="2240" to-port="1" /> + <edge from-layer="2240" from-port="2" to-layer="2242" to-port="0" /> + <edge from-layer="2241" from-port="0" to-layer="2242" to-port="1" /> + <edge from-layer="2242" from-port="2" to-layer="2243" to-port="0" /> + <edge from-layer="2243" from-port="1" to-layer="2244" to-port="1" /> + <edge from-layer="2244" from-port="2" to-layer="2246" to-port="0" /> + <edge from-layer="2245" from-port="3" to-layer="2246" to-port="1" /> + <edge from-layer="2245" from-port="3" to-layer="2313" to-port="1" /> + <edge from-layer="2246" from-port="2" to-layer="2248" to-port="0" /> + <edge from-layer="2247" from-port="0" to-layer="2248" to-port="1" /> + <edge from-layer="2248" from-port="2" to-layer="2252" to-port="0" /> + <edge from-layer="2248" from-port="2" to-layer="2269" to-port="0" /> + <edge from-layer="2248" from-port="2" to-layer="2263" to-port="0" /> + <edge from-layer="2248" from-port="2" to-layer="2249" to-port="0" /> + <edge from-layer="2249" from-port="2" to-layer="2272" to-port="0" /> + <edge from-layer="2250" from-port="0" to-layer="2260" to-port="0" /> + <edge from-layer="2251" from-port="0" to-layer="2260" to-port="1" /> + <edge from-layer="2252" from-port="1" to-layer="2254" to-port="0" /> + <edge from-layer="2253" from-port="0" to-layer="2254" to-port="2" /> + <edge from-layer="2254" from-port="3" to-layer="2255" to-port="0" /> + <edge from-layer="2255" from-port="2" to-layer="2256" to-port="0" /> + <edge from-layer="2256" from-port="1" to-layer="2258" to-port="0" /> + <edge from-layer="2257" from-port="0" to-layer="2258" to-port="1" /> + <edge from-layer="2258" from-port="2" to-layer="2260" to-port="2" /> + <edge from-layer="2259" from-port="0" to-layer="2260" to-port="3" /> + <edge from-layer="2260" from-port="4" to-layer="2263" to-port="1" /> + <edge from-layer="2260" from-port="4" to-layer="2269" to-port="2" /> + <edge from-layer="2261" from-port="0" to-layer="2263" to-port="2" /> + <edge from-layer="2262" from-port="0" to-layer="2263" to-port="3" /> + <edge from-layer="2263" from-port="4" to-layer="2266" to-port="0" /> + <edge from-layer="2264" from-port="0" to-layer="2265" to-port="0" /> + <edge from-layer="2265" from-port="1" to-layer="2266" to-port="1" /> + <edge from-layer="2266" from-port="2" to-layer="2270" to-port="0" /> + <edge from-layer="2267" from-port="0" to-layer="2269" to-port="1" /> + <edge from-layer="2268" from-port="0" to-layer="2269" to-port="3" /> + <edge from-layer="2269" from-port="4" to-layer="2270" to-port="1" /> + <edge from-layer="2270" from-port="2" to-layer="2271" to-port="0" /> + <edge from-layer="2271" from-port="2" to-layer="2272" to-port="1" /> + <edge from-layer="2272" from-port="2" to-layer="2273" to-port="1" /> + <edge from-layer="2273" from-port="2" to-layer="2502" to-port="0" /> + <edge from-layer="2273" from-port="2" to-layer="2275" to-port="0" /> + <edge from-layer="2273" from-port="2" to-layer="2274" to-port="0" /> + <edge from-layer="2274" from-port="2" to-layer="2288" to-port="0" /> + <edge from-layer="2275" from-port="1" to-layer="2278" to-port="0" /> + <edge from-layer="2275" from-port="1" to-layer="2280" to-port="0" /> + <edge from-layer="2275" from-port="1" to-layer="2285" to-port="0" /> + <edge from-layer="2276" from-port="0" to-layer="2278" to-port="1" /> + <edge from-layer="2277" from-port="0" to-layer="2278" to-port="2" /> + <edge from-layer="2278" from-port="3" to-layer="2292" to-port="0" /> + <edge from-layer="2278" from-port="3" to-layer="2286" to-port="0" /> + <edge from-layer="2279" from-port="0" to-layer="2280" to-port="2" /> + <edge from-layer="2280" from-port="3" to-layer="2282" to-port="0" /> + <edge from-layer="2280" from-port="3" to-layer="2289" to-port="0" /> + <edge from-layer="2281" from-port="0" to-layer="2282" to-port="1" /> + <edge from-layer="2282" from-port="2" to-layer="2286" to-port="1" /> + <edge from-layer="2283" from-port="0" to-layer="2285" to-port="1" /> + <edge from-layer="2284" from-port="0" to-layer="2285" to-port="2" /> + <edge from-layer="2285" from-port="3" to-layer="2292" to-port="2" /> + <edge from-layer="2285" from-port="3" to-layer="2286" to-port="3" /> + <edge from-layer="2286" from-port="4" to-layer="2287" to-port="0" /> + <edge from-layer="2287" from-port="1" to-layer="2288" to-port="1" /> + <edge from-layer="2288" from-port="2" to-layer="2293" to-port="0" /> + <edge from-layer="2289" from-port="2" to-layer="2291" to-port="0" /> + <edge from-layer="2290" from-port="0" to-layer="2291" to-port="1" /> + <edge from-layer="2291" from-port="2" to-layer="2292" to-port="1" /> + <edge from-layer="2292" from-port="3" to-layer="2293" to-port="1" /> + <edge from-layer="2293" from-port="2" to-layer="2340" to-port="0" /> + <edge from-layer="2293" from-port="2" to-layer="2348" to-port="1" /> + <edge from-layer="2294" from-port="0" to-layer="2295" to-port="0" /> + <edge from-layer="2295" from-port="1" to-layer="2300" to-port="0" /> + <edge from-layer="2296" from-port="0" to-layer="2299" to-port="1" /> + <edge from-layer="2297" from-port="0" to-layer="2299" to-port="2" /> + <edge from-layer="2298" from-port="0" to-layer="2299" to-port="3" /> + <edge from-layer="2299" from-port="4" to-layer="2300" to-port="1" /> + <edge from-layer="2300" from-port="2" to-layer="2301" to-port="0" /> + <edge from-layer="2301" from-port="1" to-layer="2303" to-port="0" /> + <edge from-layer="2302" from-port="0" to-layer="2303" to-port="2" /> + <edge from-layer="2303" from-port="3" to-layer="2316" to-port="0" /> + <edge from-layer="2304" from-port="0" to-layer="2305" to-port="0" /> + <edge from-layer="2305" from-port="1" to-layer="2308" to-port="0" /> + <edge from-layer="2306" from-port="0" to-layer="2307" to-port="0" /> + <edge from-layer="2307" from-port="1" to-layer="2308" to-port="1" /> + <edge from-layer="2308" from-port="2" to-layer="2310" to-port="0" /> + <edge from-layer="2309" from-port="0" to-layer="2310" to-port="1" /> + <edge from-layer="2310" from-port="2" to-layer="2311" to-port="0" /> + <edge from-layer="2311" from-port="1" to-layer="2312" to-port="1" /> + <edge from-layer="2312" from-port="2" to-layer="2313" to-port="0" /> + <edge from-layer="2313" from-port="2" to-layer="2315" to-port="0" /> + <edge from-layer="2314" from-port="0" to-layer="2315" to-port="1" /> + <edge from-layer="2315" from-port="2" to-layer="2316" to-port="1" /> + <edge from-layer="2316" from-port="2" to-layer="2318" to-port="0" /> + <edge from-layer="2316" from-port="2" to-layer="2501" to-port="0" /> + <edge from-layer="2316" from-port="2" to-layer="2317" to-port="0" /> + <edge from-layer="2317" from-port="2" to-layer="2331" to-port="0" /> + <edge from-layer="2318" from-port="1" to-layer="2323" to-port="0" /> + <edge from-layer="2318" from-port="1" to-layer="2321" to-port="0" /> + <edge from-layer="2318" from-port="1" to-layer="2328" to-port="0" /> + <edge from-layer="2319" from-port="0" to-layer="2321" to-port="1" /> + <edge from-layer="2320" from-port="0" to-layer="2321" to-port="2" /> + <edge from-layer="2321" from-port="3" to-layer="2329" to-port="0" /> + <edge from-layer="2321" from-port="3" to-layer="2335" to-port="0" /> + <edge from-layer="2322" from-port="0" to-layer="2323" to-port="2" /> + <edge from-layer="2323" from-port="3" to-layer="2325" to-port="0" /> + <edge from-layer="2323" from-port="3" to-layer="2332" to-port="0" /> + <edge from-layer="2324" from-port="0" to-layer="2325" to-port="1" /> + <edge from-layer="2325" from-port="2" to-layer="2329" to-port="1" /> + <edge from-layer="2326" from-port="0" to-layer="2328" to-port="1" /> + <edge from-layer="2327" from-port="0" to-layer="2328" to-port="2" /> + <edge from-layer="2328" from-port="3" to-layer="2329" to-port="3" /> + <edge from-layer="2328" from-port="3" to-layer="2335" to-port="2" /> + <edge from-layer="2329" from-port="4" to-layer="2330" to-port="0" /> + <edge from-layer="2330" from-port="1" to-layer="2331" to-port="1" /> + <edge from-layer="2331" from-port="2" to-layer="2336" to-port="0" /> + <edge from-layer="2332" from-port="2" to-layer="2334" to-port="0" /> + <edge from-layer="2333" from-port="0" to-layer="2334" to-port="1" /> + <edge from-layer="2334" from-port="2" to-layer="2335" to-port="1" /> + <edge from-layer="2335" from-port="3" to-layer="2336" to-port="1" /> + <edge from-layer="2336" from-port="2" to-layer="2348" to-port="2" /> + <edge from-layer="2337" from-port="0" to-layer="2347" to-port="1" /> + <edge from-layer="2338" from-port="0" to-layer="2345" to-port="0" /> + <edge from-layer="2339" from-port="0" to-layer="2345" to-port="1" /> + <edge from-layer="2340" from-port="1" to-layer="2343" to-port="0" /> + <edge from-layer="2341" from-port="0" to-layer="2343" to-port="1" /> + <edge from-layer="2342" from-port="0" to-layer="2343" to-port="2" /> + <edge from-layer="2343" from-port="3" to-layer="2345" to-port="2" /> + <edge from-layer="2344" from-port="0" to-layer="2345" to-port="3" /> + <edge from-layer="2345" from-port="4" to-layer="2347" to-port="2" /> + <edge from-layer="2346" from-port="0" to-layer="2347" to-port="3" /> + <edge from-layer="2347" from-port="4" to-layer="2348" to-port="3" /> + <edge from-layer="2348" from-port="4" to-layer="2350" to-port="0" /> + <edge from-layer="2349" from-port="0" to-layer="2350" to-port="1" /> + <edge from-layer="2350" from-port="2" to-layer="2352" to-port="0" /> + <edge from-layer="2351" from-port="2" to-layer="2352" to-port="1" /> + <edge from-layer="2352" from-port="2" to-layer="2361" to-port="0" /> + <edge from-layer="2353" from-port="0" to-layer="2354" to-port="0" /> + <edge from-layer="2354" from-port="1" to-layer="2357" to-port="0" /> + <edge from-layer="2355" from-port="0" to-layer="2356" to-port="0" /> + <edge from-layer="2356" from-port="1" to-layer="2357" to-port="1" /> + <edge from-layer="2357" from-port="2" to-layer="2359" to-port="0" /> + <edge from-layer="2358" from-port="0" to-layer="2359" to-port="1" /> + <edge from-layer="2359" from-port="2" to-layer="2360" to-port="0" /> + <edge from-layer="2360" from-port="1" to-layer="2361" to-port="1" /> + <edge from-layer="2361" from-port="2" to-layer="2362" to-port="1" /> + <edge from-layer="2362" from-port="2" to-layer="2369" to-port="0" /> + <edge from-layer="2362" from-port="2" to-layer="2377" to-port="0" /> + <edge from-layer="2362" from-port="2" to-layer="2408" to-port="0" /> + <edge from-layer="2363" from-port="0" to-layer="2364" to-port="0" /> + <edge from-layer="2364" from-port="1" to-layer="2378" to-port="0" /> + <edge from-layer="2365" from-port="0" to-layer="2366" to-port="0" /> + <edge from-layer="2366" from-port="1" to-layer="2376" to-port="0" /> + <edge from-layer="2367" from-port="0" to-layer="2368" to-port="0" /> + <edge from-layer="2368" from-port="1" to-layer="2369" to-port="1" /> + <edge from-layer="2369" from-port="2" to-layer="2371" to-port="0" /> + <edge from-layer="2370" from-port="0" to-layer="2371" to-port="1" /> + <edge from-layer="2371" from-port="2" to-layer="2374" to-port="0" /> + <edge from-layer="2372" from-port="0" to-layer="2373" to-port="0" /> + <edge from-layer="2373" from-port="1" to-layer="2374" to-port="1" /> + <edge from-layer="2374" from-port="2" to-layer="2375" to-port="0" /> + <edge from-layer="2375" from-port="1" to-layer="2376" to-port="1" /> + <edge from-layer="2376" from-port="2" to-layer="2377" to-port="1" /> + <edge from-layer="2377" from-port="2" to-layer="2378" to-port="1" /> + <edge from-layer="2378" from-port="2" to-layer="2387" to-port="0" /> + <edge from-layer="2378" from-port="2" to-layer="2397" to-port="0" /> + <edge from-layer="2379" from-port="0" to-layer="2380" to-port="0" /> + <edge from-layer="2380" from-port="1" to-layer="2383" to-port="0" /> + <edge from-layer="2381" from-port="0" to-layer="2382" to-port="0" /> + <edge from-layer="2382" from-port="1" to-layer="2383" to-port="1" /> + <edge from-layer="2383" from-port="2" to-layer="2385" to-port="0" /> + <edge from-layer="2384" from-port="0" to-layer="2385" to-port="1" /> + <edge from-layer="2385" from-port="2" to-layer="2386" to-port="0" /> + <edge from-layer="2386" from-port="1" to-layer="2387" to-port="1" /> + <edge from-layer="2387" from-port="2" to-layer="2388" to-port="0" /> + <edge from-layer="2388" from-port="1" to-layer="2398" to-port="0" /> + <edge from-layer="2389" from-port="0" to-layer="2390" to-port="0" /> + <edge from-layer="2390" from-port="1" to-layer="2393" to-port="0" /> + <edge from-layer="2391" from-port="0" to-layer="2392" to-port="0" /> + <edge from-layer="2392" from-port="1" to-layer="2393" to-port="1" /> + <edge from-layer="2393" from-port="2" to-layer="2395" to-port="0" /> + <edge from-layer="2394" from-port="0" to-layer="2395" to-port="1" /> + <edge from-layer="2395" from-port="2" to-layer="2396" to-port="0" /> + <edge from-layer="2396" from-port="1" to-layer="2397" to-port="1" /> + <edge from-layer="2397" from-port="2" to-layer="2398" to-port="1" /> + <edge from-layer="2398" from-port="2" to-layer="2407" to-port="0" /> + <edge from-layer="2399" from-port="0" to-layer="2400" to-port="0" /> + <edge from-layer="2400" from-port="1" to-layer="2403" to-port="0" /> + <edge from-layer="2401" from-port="0" to-layer="2402" to-port="0" /> + <edge from-layer="2402" from-port="1" to-layer="2403" to-port="1" /> + <edge from-layer="2403" from-port="2" to-layer="2405" to-port="0" /> + <edge from-layer="2404" from-port="0" to-layer="2405" to-port="1" /> + <edge from-layer="2405" from-port="2" to-layer="2406" to-port="0" /> + <edge from-layer="2406" from-port="1" to-layer="2407" to-port="1" /> + <edge from-layer="2407" from-port="2" to-layer="2408" to-port="1" /> + <edge from-layer="2408" from-port="2" to-layer="2421" to-port="0" /> + <edge from-layer="2408" from-port="2" to-layer="2413" to-port="0" /> + <edge from-layer="2408" from-port="2" to-layer="2627" to-port="0" /> + <edge from-layer="2409" from-port="0" to-layer="2410" to-port="0" /> + <edge from-layer="2410" from-port="1" to-layer="2420" to-port="0" /> + <edge from-layer="2411" from-port="0" to-layer="2412" to-port="0" /> + <edge from-layer="2412" from-port="1" to-layer="2413" to-port="1" /> + <edge from-layer="2413" from-port="2" to-layer="2415" to-port="0" /> + <edge from-layer="2414" from-port="0" to-layer="2415" to-port="1" /> + <edge from-layer="2415" from-port="2" to-layer="2418" to-port="0" /> + <edge from-layer="2416" from-port="0" to-layer="2417" to-port="0" /> + <edge from-layer="2417" from-port="1" to-layer="2418" to-port="1" /> + <edge from-layer="2418" from-port="2" to-layer="2419" to-port="0" /> + <edge from-layer="2419" from-port="1" to-layer="2420" to-port="1" /> + <edge from-layer="2420" from-port="2" to-layer="2421" to-port="1" /> + <edge from-layer="2421" from-port="2" to-layer="2422" to-port="1" /> + <edge from-layer="2422" from-port="2" to-layer="2432" to-port="0" /> + <edge from-layer="2422" from-port="2" to-layer="2533" to-port="0" /> + <edge from-layer="2422" from-port="2" to-layer="2460" to-port="0" /> + <edge from-layer="2422" from-port="2" to-layer="2431" to-port="0" /> + <edge from-layer="2423" from-port="0" to-layer="2424" to-port="0" /> + <edge from-layer="2424" from-port="1" to-layer="2427" to-port="0" /> + <edge from-layer="2425" from-port="0" to-layer="2426" to-port="0" /> + <edge from-layer="2426" from-port="1" to-layer="2427" to-port="1" /> + <edge from-layer="2427" from-port="2" to-layer="2429" to-port="0" /> + <edge from-layer="2428" from-port="0" to-layer="2429" to-port="1" /> + <edge from-layer="2429" from-port="2" to-layer="2430" to-port="0" /> + <edge from-layer="2430" from-port="1" to-layer="2431" to-port="1" /> + <edge from-layer="2431" from-port="2" to-layer="2437" to-port="0" /> + <edge from-layer="2432" from-port="1" to-layer="2435" to-port="0" /> + <edge from-layer="2433" from-port="0" to-layer="2435" to-port="1" /> + <edge from-layer="2434" from-port="0" to-layer="2435" to-port="2" /> + <edge from-layer="2435" from-port="3" to-layer="2534" to-port="0" /> + <edge from-layer="2435" from-port="3" to-layer="2436" to-port="0" /> + <edge from-layer="2435" from-port="3" to-layer="2616" to-port="0" /> + <edge from-layer="2436" from-port="3" to-layer="2437" to-port="1" /> + <edge from-layer="2436" from-port="3" to-layer="2461" to-port="1" /> + <edge from-layer="2437" from-port="2" to-layer="2439" to-port="0" /> + <edge from-layer="2438" from-port="0" to-layer="2439" to-port="1" /> + <edge from-layer="2439" from-port="2" to-layer="2440" to-port="1" /> + <edge from-layer="2440" from-port="2" to-layer="2441" to-port="0" /> + <edge from-layer="2440" from-port="2" to-layer="2583" to-port="0" /> + <edge from-layer="2440" from-port="2" to-layer="2582" to-port="0" /> + <edge from-layer="2442" from-port="0" to-layer="2443" to-port="0" /> + <edge from-layer="2443" from-port="1" to-layer="2448" to-port="0" /> + <edge from-layer="2444" from-port="0" to-layer="2447" to-port="1" /> + <edge from-layer="2445" from-port="0" to-layer="2447" to-port="2" /> + <edge from-layer="2446" from-port="0" to-layer="2447" to-port="3" /> + <edge from-layer="2447" from-port="4" to-layer="2448" to-port="1" /> + <edge from-layer="2448" from-port="2" to-layer="2449" to-port="0" /> + <edge from-layer="2449" from-port="1" to-layer="2451" to-port="0" /> + <edge from-layer="2450" from-port="0" to-layer="2451" to-port="2" /> + <edge from-layer="2451" from-port="3" to-layer="2499" to-port="0" /> + <edge from-layer="2452" from-port="0" to-layer="2453" to-port="0" /> + <edge from-layer="2453" from-port="1" to-layer="2456" to-port="0" /> + <edge from-layer="2454" from-port="0" to-layer="2455" to-port="0" /> + <edge from-layer="2455" from-port="1" to-layer="2456" to-port="1" /> + <edge from-layer="2456" from-port="2" to-layer="2458" to-port="0" /> + <edge from-layer="2457" from-port="0" to-layer="2458" to-port="1" /> + <edge from-layer="2458" from-port="2" to-layer="2459" to-port="0" /> + <edge from-layer="2459" from-port="1" to-layer="2460" to-port="1" /> + <edge from-layer="2460" from-port="2" to-layer="2461" to-port="0" /> + <edge from-layer="2461" from-port="2" to-layer="2463" to-port="0" /> + <edge from-layer="2462" from-port="0" to-layer="2463" to-port="1" /> + <edge from-layer="2463" from-port="2" to-layer="2487" to-port="0" /> + <edge from-layer="2463" from-port="2" to-layer="2493" to-port="0" /> + <edge from-layer="2463" from-port="2" to-layer="2476" to-port="0" /> + <edge from-layer="2463" from-port="2" to-layer="2473" to-port="0" /> + <edge from-layer="2464" from-port="0" to-layer="2465" to-port="0" /> + <edge from-layer="2465" from-port="1" to-layer="2466" to-port="0" /> + <edge from-layer="2466" from-port="2" to-layer="2467" to-port="0" /> + <edge from-layer="2467" from-port="2" to-layer="2469" to-port="0" /> + <edge from-layer="2468" from-port="0" to-layer="2469" to-port="1" /> + <edge from-layer="2469" from-port="2" to-layer="2470" to-port="1" /> + <edge from-layer="2469" from-port="2" to-layer="2470" to-port="0" /> + <edge from-layer="2470" from-port="2" to-layer="2471" to-port="0" /> + <edge from-layer="2470" from-port="2" to-layer="2495" to-port="0" /> + <edge from-layer="2471" from-port="1" to-layer="2472" to-port="0" /> + <edge from-layer="2472" from-port="2" to-layer="2538" to-port="1" /> + <edge from-layer="2472" from-port="2" to-layer="2473" to-port="1" /> + <edge from-layer="2473" from-port="2" to-layer="2498" to-port="0" /> + <edge from-layer="2474" from-port="0" to-layer="2484" to-port="0" /> + <edge from-layer="2475" from-port="0" to-layer="2484" to-port="1" /> + <edge from-layer="2476" from-port="1" to-layer="2478" to-port="0" /> + <edge from-layer="2477" from-port="0" to-layer="2478" to-port="2" /> + <edge from-layer="2478" from-port="3" to-layer="2479" to-port="0" /> + <edge from-layer="2479" from-port="2" to-layer="2480" to-port="0" /> + <edge from-layer="2480" from-port="1" to-layer="2482" to-port="0" /> + <edge from-layer="2481" from-port="0" to-layer="2482" to-port="1" /> + <edge from-layer="2482" from-port="2" to-layer="2484" to-port="2" /> + <edge from-layer="2483" from-port="0" to-layer="2484" to-port="3" /> + <edge from-layer="2484" from-port="4" to-layer="2487" to-port="1" /> + <edge from-layer="2484" from-port="4" to-layer="2493" to-port="2" /> + <edge from-layer="2485" from-port="0" to-layer="2487" to-port="2" /> + <edge from-layer="2486" from-port="0" to-layer="2487" to-port="3" /> + <edge from-layer="2487" from-port="4" to-layer="2490" to-port="0" /> + <edge from-layer="2488" from-port="0" to-layer="2489" to-port="0" /> + <edge from-layer="2489" from-port="1" to-layer="2490" to-port="1" /> + <edge from-layer="2490" from-port="2" to-layer="2494" to-port="0" /> + <edge from-layer="2491" from-port="0" to-layer="2493" to-port="1" /> + <edge from-layer="2492" from-port="0" to-layer="2493" to-port="3" /> + <edge from-layer="2493" from-port="4" to-layer="2494" to-port="1" /> + <edge from-layer="2494" from-port="2" to-layer="2497" to-port="0" /> + <edge from-layer="2495" from-port="1" to-layer="2496" to-port="0" /> + <edge from-layer="2496" from-port="2" to-layer="2497" to-port="1" /> + <edge from-layer="2496" from-port="2" to-layer="2560" to-port="1" /> + <edge from-layer="2497" from-port="2" to-layer="2498" to-port="1" /> + <edge from-layer="2498" from-port="2" to-layer="2499" to-port="1" /> + <edge from-layer="2499" from-port="2" to-layer="2500" to-port="0" /> + <edge from-layer="2499" from-port="2" to-layer="2562" to-port="0" /> + <edge from-layer="2499" from-port="2" to-layer="2563" to-port="0" /> + <edge from-layer="2513" from-port="0" to-layer="2514" to-port="0" /> + <edge from-layer="2514" from-port="1" to-layer="2519" to-port="0" /> + <edge from-layer="2515" from-port="0" to-layer="2518" to-port="1" /> + <edge from-layer="2516" from-port="0" to-layer="2518" to-port="2" /> + <edge from-layer="2517" from-port="0" to-layer="2518" to-port="3" /> + <edge from-layer="2518" from-port="4" to-layer="2519" to-port="1" /> + <edge from-layer="2519" from-port="2" to-layer="2520" to-port="0" /> + <edge from-layer="2520" from-port="1" to-layer="2522" to-port="0" /> + <edge from-layer="2521" from-port="0" to-layer="2522" to-port="2" /> + <edge from-layer="2522" from-port="3" to-layer="5466" to-port="0" /> + <edge from-layer="2523" from-port="0" to-layer="2524" to-port="0" /> + <edge from-layer="2524" from-port="1" to-layer="5448" to-port="0" /> + <edge from-layer="2525" from-port="0" to-layer="2526" to-port="0" /> + <edge from-layer="2526" from-port="1" to-layer="2529" to-port="0" /> + <edge from-layer="2527" from-port="0" to-layer="2528" to-port="0" /> + <edge from-layer="2528" from-port="1" to-layer="2529" to-port="1" /> + <edge from-layer="2529" from-port="2" to-layer="2531" to-port="0" /> + <edge from-layer="2530" from-port="0" to-layer="2531" to-port="1" /> + <edge from-layer="2531" from-port="2" to-layer="2532" to-port="0" /> + <edge from-layer="2532" from-port="1" to-layer="2533" to-port="1" /> + <edge from-layer="2533" from-port="2" to-layer="2535" to-port="0" /> + <edge from-layer="2534" from-port="3" to-layer="2535" to-port="1" /> + <edge from-layer="2535" from-port="2" to-layer="2537" to-port="0" /> + <edge from-layer="2536" from-port="0" to-layer="2537" to-port="1" /> + <edge from-layer="2537" from-port="2" to-layer="2541" to-port="0" /> + <edge from-layer="2537" from-port="2" to-layer="2538" to-port="0" /> + <edge from-layer="2537" from-port="2" to-layer="2558" to-port="0" /> + <edge from-layer="2537" from-port="2" to-layer="2552" to-port="0" /> + <edge from-layer="2538" from-port="2" to-layer="2561" to-port="0" /> + <edge from-layer="2539" from-port="0" to-layer="2549" to-port="0" /> + <edge from-layer="2540" from-port="0" to-layer="2549" to-port="1" /> + <edge from-layer="2541" from-port="1" to-layer="2543" to-port="0" /> + <edge from-layer="2542" from-port="0" to-layer="2543" to-port="2" /> + <edge from-layer="2543" from-port="3" to-layer="2544" to-port="0" /> + <edge from-layer="2544" from-port="2" to-layer="2545" to-port="0" /> + <edge from-layer="2545" from-port="1" to-layer="2547" to-port="0" /> + <edge from-layer="2546" from-port="0" to-layer="2547" to-port="1" /> + <edge from-layer="2547" from-port="2" to-layer="2549" to-port="2" /> + <edge from-layer="2548" from-port="0" to-layer="2549" to-port="3" /> + <edge from-layer="2549" from-port="4" to-layer="2558" to-port="2" /> + <edge from-layer="2549" from-port="4" to-layer="2552" to-port="1" /> + <edge from-layer="2550" from-port="0" to-layer="2552" to-port="2" /> + <edge from-layer="2551" from-port="0" to-layer="2552" to-port="3" /> + <edge from-layer="2552" from-port="4" to-layer="2555" to-port="0" /> + <edge from-layer="2553" from-port="0" to-layer="2554" to-port="0" /> + <edge from-layer="2554" from-port="1" to-layer="2555" to-port="1" /> + <edge from-layer="2555" from-port="2" to-layer="2559" to-port="0" /> + <edge from-layer="2556" from-port="0" to-layer="2558" to-port="1" /> + <edge from-layer="2557" from-port="0" to-layer="2558" to-port="3" /> + <edge from-layer="2558" from-port="4" to-layer="2559" to-port="1" /> + <edge from-layer="2559" from-port="2" to-layer="2560" to-port="0" /> + <edge from-layer="2560" from-port="2" to-layer="2561" to-port="1" /> + <edge from-layer="2561" from-port="2" to-layer="2613" to-port="0" /> + <edge from-layer="2562" from-port="2" to-layer="2576" to-port="0" /> + <edge from-layer="2563" from-port="1" to-layer="2566" to-port="0" /> + <edge from-layer="2563" from-port="1" to-layer="2568" to-port="0" /> + <edge from-layer="2563" from-port="1" to-layer="2573" to-port="0" /> + <edge from-layer="2564" from-port="0" to-layer="2566" to-port="1" /> + <edge from-layer="2565" from-port="0" to-layer="2566" to-port="2" /> + <edge from-layer="2566" from-port="3" to-layer="2580" to-port="0" /> + <edge from-layer="2566" from-port="3" to-layer="2574" to-port="0" /> + <edge from-layer="2567" from-port="0" to-layer="2568" to-port="2" /> + <edge from-layer="2568" from-port="3" to-layer="2577" to-port="0" /> + <edge from-layer="2568" from-port="3" to-layer="2570" to-port="0" /> + <edge from-layer="2569" from-port="0" to-layer="2570" to-port="1" /> + <edge from-layer="2570" from-port="2" to-layer="2574" to-port="1" /> + <edge from-layer="2571" from-port="0" to-layer="2573" to-port="1" /> + <edge from-layer="2572" from-port="0" to-layer="2573" to-port="2" /> + <edge from-layer="2573" from-port="3" to-layer="2574" to-port="3" /> + <edge from-layer="2573" from-port="3" to-layer="2580" to-port="2" /> + <edge from-layer="2574" from-port="4" to-layer="2575" to-port="0" /> + <edge from-layer="2575" from-port="1" to-layer="2576" to-port="1" /> + <edge from-layer="2576" from-port="2" to-layer="2581" to-port="0" /> + <edge from-layer="2577" from-port="2" to-layer="2579" to-port="0" /> + <edge from-layer="2578" from-port="0" to-layer="2579" to-port="1" /> + <edge from-layer="2579" from-port="2" to-layer="2580" to-port="1" /> + <edge from-layer="2580" from-port="3" to-layer="2581" to-port="1" /> + <edge from-layer="2581" from-port="2" to-layer="2613" to-port="1" /> + <edge from-layer="2581" from-port="2" to-layer="2605" to-port="0" /> + <edge from-layer="2582" from-port="2" to-layer="2596" to-port="0" /> + <edge from-layer="2583" from-port="1" to-layer="2593" to-port="0" /> + <edge from-layer="2583" from-port="1" to-layer="2588" to-port="0" /> + <edge from-layer="2583" from-port="1" to-layer="2586" to-port="0" /> + <edge from-layer="2584" from-port="0" to-layer="2586" to-port="1" /> + <edge from-layer="2585" from-port="0" to-layer="2586" to-port="2" /> + <edge from-layer="2586" from-port="3" to-layer="2594" to-port="0" /> + <edge from-layer="2586" from-port="3" to-layer="2600" to-port="0" /> + <edge from-layer="2587" from-port="0" to-layer="2588" to-port="2" /> + <edge from-layer="2588" from-port="3" to-layer="2597" to-port="0" /> + <edge from-layer="2588" from-port="3" to-layer="2590" to-port="0" /> + <edge from-layer="2589" from-port="0" to-layer="2590" to-port="1" /> + <edge from-layer="2590" from-port="2" to-layer="2594" to-port="1" /> + <edge from-layer="2591" from-port="0" to-layer="2593" to-port="1" /> + <edge from-layer="2592" from-port="0" to-layer="2593" to-port="2" /> + <edge from-layer="2593" from-port="3" to-layer="2600" to-port="2" /> + <edge from-layer="2593" from-port="3" to-layer="2594" to-port="3" /> + <edge from-layer="2594" from-port="4" to-layer="2595" to-port="0" /> + <edge from-layer="2595" from-port="1" to-layer="2596" to-port="1" /> + <edge from-layer="2596" from-port="2" to-layer="2601" to-port="0" /> + <edge from-layer="2597" from-port="2" to-layer="2599" to-port="0" /> + <edge from-layer="2598" from-port="0" to-layer="2599" to-port="1" /> + <edge from-layer="2599" from-port="2" to-layer="2600" to-port="1" /> + <edge from-layer="2600" from-port="3" to-layer="2601" to-port="1" /> + <edge from-layer="2601" from-port="2" to-layer="2613" to-port="2" /> + <edge from-layer="2602" from-port="0" to-layer="2612" to-port="1" /> + <edge from-layer="2603" from-port="0" to-layer="2610" to-port="0" /> + <edge from-layer="2604" from-port="0" to-layer="2610" to-port="1" /> + <edge from-layer="2605" from-port="1" to-layer="2608" to-port="0" /> + <edge from-layer="2606" from-port="0" to-layer="2608" to-port="1" /> + <edge from-layer="2607" from-port="0" to-layer="2608" to-port="2" /> + <edge from-layer="2608" from-port="3" to-layer="2610" to-port="2" /> + <edge from-layer="2609" from-port="0" to-layer="2610" to-port="3" /> + <edge from-layer="2610" from-port="4" to-layer="2612" to-port="2" /> + <edge from-layer="2611" from-port="0" to-layer="2612" to-port="3" /> + <edge from-layer="2612" from-port="4" to-layer="2613" to-port="3" /> + <edge from-layer="2613" from-port="4" to-layer="2615" to-port="0" /> + <edge from-layer="2614" from-port="0" to-layer="2615" to-port="1" /> + <edge from-layer="2615" from-port="2" to-layer="2617" to-port="0" /> + <edge from-layer="2616" from-port="2" to-layer="2617" to-port="1" /> + <edge from-layer="2617" from-port="2" to-layer="2626" to-port="0" /> + <edge from-layer="2618" from-port="0" to-layer="2619" to-port="0" /> + <edge from-layer="2619" from-port="1" to-layer="2622" to-port="0" /> + <edge from-layer="2620" from-port="0" to-layer="2621" to-port="0" /> + <edge from-layer="2621" from-port="1" to-layer="2622" to-port="1" /> + <edge from-layer="2622" from-port="2" to-layer="2624" to-port="0" /> + <edge from-layer="2623" from-port="0" to-layer="2624" to-port="1" /> + <edge from-layer="2624" from-port="2" to-layer="2625" to-port="0" /> + <edge from-layer="2625" from-port="1" to-layer="2626" to-port="1" /> + <edge from-layer="2626" from-port="2" to-layer="2627" to-port="1" /> + <edge from-layer="2627" from-port="2" to-layer="2634" to-port="0" /> + <edge from-layer="2627" from-port="2" to-layer="2673" to-port="0" /> + <edge from-layer="2627" from-port="2" to-layer="2642" to-port="0" /> + <edge from-layer="2628" from-port="0" to-layer="2629" to-port="0" /> + <edge from-layer="2629" from-port="1" to-layer="2643" to-port="0" /> + <edge from-layer="2630" from-port="0" to-layer="2631" to-port="0" /> + <edge from-layer="2631" from-port="1" to-layer="2641" to-port="0" /> + <edge from-layer="2632" from-port="0" to-layer="2633" to-port="0" /> + <edge from-layer="2633" from-port="1" to-layer="2634" to-port="1" /> + <edge from-layer="2634" from-port="2" to-layer="2636" to-port="0" /> + <edge from-layer="2635" from-port="0" to-layer="2636" to-port="1" /> + <edge from-layer="2636" from-port="2" to-layer="2639" to-port="0" /> + <edge from-layer="2637" from-port="0" to-layer="2638" to-port="0" /> + <edge from-layer="2638" from-port="1" to-layer="2639" to-port="1" /> + <edge from-layer="2639" from-port="2" to-layer="2640" to-port="0" /> + <edge from-layer="2640" from-port="1" to-layer="2641" to-port="1" /> + <edge from-layer="2641" from-port="2" to-layer="2642" to-port="1" /> + <edge from-layer="2642" from-port="2" to-layer="2643" to-port="1" /> + <edge from-layer="2643" from-port="2" to-layer="2652" to-port="0" /> + <edge from-layer="2643" from-port="2" to-layer="2662" to-port="0" /> + <edge from-layer="2644" from-port="0" to-layer="2645" to-port="0" /> + <edge from-layer="2645" from-port="1" to-layer="2648" to-port="0" /> + <edge from-layer="2646" from-port="0" to-layer="2647" to-port="0" /> + <edge from-layer="2647" from-port="1" to-layer="2648" to-port="1" /> + <edge from-layer="2648" from-port="2" to-layer="2650" to-port="0" /> + <edge from-layer="2649" from-port="0" to-layer="2650" to-port="1" /> + <edge from-layer="2650" from-port="2" to-layer="2651" to-port="0" /> + <edge from-layer="2651" from-port="1" to-layer="2652" to-port="1" /> + <edge from-layer="2652" from-port="2" to-layer="2653" to-port="0" /> + <edge from-layer="2653" from-port="1" to-layer="2663" to-port="0" /> + <edge from-layer="2654" from-port="0" to-layer="2655" to-port="0" /> + <edge from-layer="2655" from-port="1" to-layer="2658" to-port="0" /> + <edge from-layer="2656" from-port="0" to-layer="2657" to-port="0" /> + <edge from-layer="2657" from-port="1" to-layer="2658" to-port="1" /> + <edge from-layer="2658" from-port="2" to-layer="2660" to-port="0" /> + <edge from-layer="2659" from-port="0" to-layer="2660" to-port="1" /> + <edge from-layer="2660" from-port="2" to-layer="2661" to-port="0" /> + <edge from-layer="2661" from-port="1" to-layer="2662" to-port="1" /> + <edge from-layer="2662" from-port="2" to-layer="2663" to-port="1" /> + <edge from-layer="2663" from-port="2" to-layer="2672" to-port="0" /> + <edge from-layer="2664" from-port="0" to-layer="2665" to-port="0" /> + <edge from-layer="2665" from-port="1" to-layer="2668" to-port="0" /> + <edge from-layer="2666" from-port="0" to-layer="2667" to-port="0" /> + <edge from-layer="2667" from-port="1" to-layer="2668" to-port="1" /> + <edge from-layer="2668" from-port="2" to-layer="2670" to-port="0" /> + <edge from-layer="2669" from-port="0" to-layer="2670" to-port="1" /> + <edge from-layer="2670" from-port="2" to-layer="2671" to-port="0" /> + <edge from-layer="2671" from-port="1" to-layer="2672" to-port="1" /> + <edge from-layer="2672" from-port="2" to-layer="2673" to-port="1" /> + <edge from-layer="2673" from-port="2" to-layer="2680" to-port="0" /> + <edge from-layer="2673" from-port="2" to-layer="2688" to-port="0" /> + <edge from-layer="2673" from-port="2" to-layer="2878" to-port="0" /> + <edge from-layer="2674" from-port="0" to-layer="2675" to-port="0" /> + <edge from-layer="2675" from-port="1" to-layer="2689" to-port="0" /> + <edge from-layer="2676" from-port="0" to-layer="2677" to-port="0" /> + <edge from-layer="2677" from-port="1" to-layer="2687" to-port="0" /> + <edge from-layer="2678" from-port="0" to-layer="2679" to-port="0" /> + <edge from-layer="2679" from-port="1" to-layer="2680" to-port="1" /> + <edge from-layer="2680" from-port="2" to-layer="2682" to-port="0" /> + <edge from-layer="2681" from-port="0" to-layer="2682" to-port="1" /> + <edge from-layer="2682" from-port="2" to-layer="2685" to-port="0" /> + <edge from-layer="2683" from-port="0" to-layer="2684" to-port="0" /> + <edge from-layer="2684" from-port="1" to-layer="2685" to-port="1" /> + <edge from-layer="2685" from-port="2" to-layer="2686" to-port="0" /> + <edge from-layer="2686" from-port="1" to-layer="2687" to-port="1" /> + <edge from-layer="2687" from-port="2" to-layer="2688" to-port="1" /> + <edge from-layer="2688" from-port="2" to-layer="2689" to-port="1" /> + <edge from-layer="2689" from-port="2" to-layer="2698" to-port="0" /> + <edge from-layer="2689" from-port="2" to-layer="2699" to-port="0" /> + <edge from-layer="2689" from-port="2" to-layer="2828" to-port="0" /> + <edge from-layer="2689" from-port="2" to-layer="2760" to-port="0" /> + <edge from-layer="2690" from-port="0" to-layer="2691" to-port="0" /> + <edge from-layer="2691" from-port="1" to-layer="2694" to-port="0" /> + <edge from-layer="2692" from-port="0" to-layer="2693" to-port="0" /> + <edge from-layer="2693" from-port="1" to-layer="2694" to-port="1" /> + <edge from-layer="2694" from-port="2" to-layer="2696" to-port="0" /> + <edge from-layer="2695" from-port="0" to-layer="2696" to-port="1" /> + <edge from-layer="2696" from-port="2" to-layer="2697" to-port="0" /> + <edge from-layer="2697" from-port="1" to-layer="2698" to-port="1" /> + <edge from-layer="2698" from-port="2" to-layer="2704" to-port="0" /> + <edge from-layer="2699" from-port="1" to-layer="2702" to-port="0" /> + <edge from-layer="2700" from-port="0" to-layer="2702" to-port="1" /> + <edge from-layer="2701" from-port="0" to-layer="2702" to-port="2" /> + <edge from-layer="2702" from-port="3" to-layer="2703" to-port="0" /> + <edge from-layer="2702" from-port="3" to-layer="2867" to-port="0" /> + <edge from-layer="2702" from-port="3" to-layer="2761" to-port="0" /> + <edge from-layer="2703" from-port="3" to-layer="2704" to-port="1" /> + <edge from-layer="2704" from-port="2" to-layer="2706" to-port="0" /> + <edge from-layer="2705" from-port="0" to-layer="2706" to-port="1" /> + <edge from-layer="2706" from-port="2" to-layer="2736" to-port="0" /> + <edge from-layer="2706" from-port="2" to-layer="2730" to-port="0" /> + <edge from-layer="2706" from-port="2" to-layer="2719" to-port="0" /> + <edge from-layer="2706" from-port="2" to-layer="2716" to-port="0" /> + <edge from-layer="2707" from-port="0" to-layer="2708" to-port="0" /> + <edge from-layer="2708" from-port="1" to-layer="2709" to-port="0" /> + <edge from-layer="2709" from-port="2" to-layer="2710" to-port="0" /> + <edge from-layer="2710" from-port="2" to-layer="2712" to-port="0" /> + <edge from-layer="2711" from-port="0" to-layer="2712" to-port="1" /> + <edge from-layer="2712" from-port="2" to-layer="2713" to-port="0" /> + <edge from-layer="2712" from-port="2" to-layer="2713" to-port="1" /> + <edge from-layer="2713" from-port="2" to-layer="2714" to-port="0" /> + <edge from-layer="2713" from-port="2" to-layer="2738" to-port="0" /> + <edge from-layer="2714" from-port="1" to-layer="2715" to-port="0" /> + <edge from-layer="2715" from-port="2" to-layer="2765" to-port="1" /> + <edge from-layer="2715" from-port="2" to-layer="2716" to-port="1" /> + <edge from-layer="2716" from-port="2" to-layer="2741" to-port="0" /> + <edge from-layer="2717" from-port="0" to-layer="2727" to-port="0" /> + <edge from-layer="2718" from-port="0" to-layer="2727" to-port="1" /> + <edge from-layer="2719" from-port="1" to-layer="2721" to-port="0" /> + <edge from-layer="2720" from-port="0" to-layer="2721" to-port="2" /> + <edge from-layer="2721" from-port="3" to-layer="2722" to-port="0" /> + <edge from-layer="2722" from-port="2" to-layer="2723" to-port="0" /> + <edge from-layer="2723" from-port="1" to-layer="2725" to-port="0" /> + <edge from-layer="2724" from-port="0" to-layer="2725" to-port="1" /> + <edge from-layer="2725" from-port="2" to-layer="2727" to-port="2" /> + <edge from-layer="2726" from-port="0" to-layer="2727" to-port="3" /> + <edge from-layer="2727" from-port="4" to-layer="2730" to-port="1" /> + <edge from-layer="2727" from-port="4" to-layer="2736" to-port="2" /> + <edge from-layer="2728" from-port="0" to-layer="2730" to-port="2" /> + <edge from-layer="2729" from-port="0" to-layer="2730" to-port="3" /> + <edge from-layer="2730" from-port="4" to-layer="2733" to-port="0" /> + <edge from-layer="2731" from-port="0" to-layer="2732" to-port="0" /> + <edge from-layer="2732" from-port="1" to-layer="2733" to-port="1" /> + <edge from-layer="2733" from-port="2" to-layer="2737" to-port="0" /> + <edge from-layer="2734" from-port="0" to-layer="2736" to-port="1" /> + <edge from-layer="2735" from-port="0" to-layer="2736" to-port="3" /> + <edge from-layer="2736" from-port="4" to-layer="2737" to-port="1" /> + <edge from-layer="2737" from-port="2" to-layer="2740" to-port="0" /> + <edge from-layer="2738" from-port="1" to-layer="2739" to-port="0" /> + <edge from-layer="2739" from-port="2" to-layer="2787" to-port="1" /> + <edge from-layer="2739" from-port="2" to-layer="2740" to-port="1" /> + <edge from-layer="2740" from-port="2" to-layer="2741" to-port="1" /> + <edge from-layer="2741" from-port="2" to-layer="2864" to-port="0" /> + <edge from-layer="2742" from-port="0" to-layer="2743" to-port="0" /> + <edge from-layer="2743" from-port="1" to-layer="2748" to-port="0" /> + <edge from-layer="2744" from-port="0" to-layer="2747" to-port="1" /> + <edge from-layer="2745" from-port="0" to-layer="2747" to-port="2" /> + <edge from-layer="2746" from-port="0" to-layer="2747" to-port="3" /> + <edge from-layer="2747" from-port="4" to-layer="2748" to-port="1" /> + <edge from-layer="2748" from-port="2" to-layer="2749" to-port="0" /> + <edge from-layer="2749" from-port="1" to-layer="2751" to-port="0" /> + <edge from-layer="2750" from-port="0" to-layer="2751" to-port="2" /> + <edge from-layer="2751" from-port="3" to-layer="2789" to-port="0" /> + <edge from-layer="2752" from-port="0" to-layer="2753" to-port="0" /> + <edge from-layer="2753" from-port="1" to-layer="2756" to-port="0" /> + <edge from-layer="2754" from-port="0" to-layer="2755" to-port="0" /> + <edge from-layer="2755" from-port="1" to-layer="2756" to-port="1" /> + <edge from-layer="2756" from-port="2" to-layer="2758" to-port="0" /> + <edge from-layer="2757" from-port="0" to-layer="2758" to-port="1" /> + <edge from-layer="2758" from-port="2" to-layer="2759" to-port="0" /> + <edge from-layer="2759" from-port="1" to-layer="2760" to-port="1" /> + <edge from-layer="2760" from-port="2" to-layer="2762" to-port="0" /> + <edge from-layer="2761" from-port="3" to-layer="2762" to-port="1" /> + <edge from-layer="2761" from-port="3" to-layer="2829" to-port="1" /> + <edge from-layer="2762" from-port="2" to-layer="2764" to-port="0" /> + <edge from-layer="2763" from-port="0" to-layer="2764" to-port="1" /> + <edge from-layer="2764" from-port="2" to-layer="2779" to-port="0" /> + <edge from-layer="2764" from-port="2" to-layer="2785" to-port="0" /> + <edge from-layer="2764" from-port="2" to-layer="2765" to-port="0" /> + <edge from-layer="2764" from-port="2" to-layer="2768" to-port="0" /> + <edge from-layer="2765" from-port="2" to-layer="2788" to-port="0" /> + <edge from-layer="2766" from-port="0" to-layer="2776" to-port="0" /> + <edge from-layer="2767" from-port="0" to-layer="2776" to-port="1" /> + <edge from-layer="2768" from-port="1" to-layer="2770" to-port="0" /> + <edge from-layer="2769" from-port="0" to-layer="2770" to-port="2" /> + <edge from-layer="2770" from-port="3" to-layer="2771" to-port="0" /> + <edge from-layer="2771" from-port="2" to-layer="2772" to-port="0" /> + <edge from-layer="2772" from-port="1" to-layer="2774" to-port="0" /> + <edge from-layer="2773" from-port="0" to-layer="2774" to-port="1" /> + <edge from-layer="2774" from-port="2" to-layer="2776" to-port="2" /> + <edge from-layer="2775" from-port="0" to-layer="2776" to-port="3" /> + <edge from-layer="2776" from-port="4" to-layer="2785" to-port="2" /> + <edge from-layer="2776" from-port="4" to-layer="2779" to-port="1" /> + <edge from-layer="2777" from-port="0" to-layer="2779" to-port="2" /> + <edge from-layer="2778" from-port="0" to-layer="2779" to-port="3" /> + <edge from-layer="2779" from-port="4" to-layer="2782" to-port="0" /> + <edge from-layer="2780" from-port="0" to-layer="2781" to-port="0" /> + <edge from-layer="2781" from-port="1" to-layer="2782" to-port="1" /> + <edge from-layer="2782" from-port="2" to-layer="2786" to-port="0" /> + <edge from-layer="2783" from-port="0" to-layer="2785" to-port="1" /> + <edge from-layer="2784" from-port="0" to-layer="2785" to-port="3" /> + <edge from-layer="2785" from-port="4" to-layer="2786" to-port="1" /> + <edge from-layer="2786" from-port="2" to-layer="2787" to-port="0" /> + <edge from-layer="2787" from-port="2" to-layer="2788" to-port="1" /> + <edge from-layer="2788" from-port="2" to-layer="2789" to-port="1" /> + <edge from-layer="2789" from-port="2" to-layer="2790" to-port="0" /> + <edge from-layer="2789" from-port="2" to-layer="2791" to-port="0" /> + <edge from-layer="2789" from-port="2" to-layer="5550" to-port="0" /> + <edge from-layer="2790" from-port="2" to-layer="2804" to-port="0" /> + <edge from-layer="2791" from-port="1" to-layer="2794" to-port="0" /> + <edge from-layer="2791" from-port="1" to-layer="2796" to-port="0" /> + <edge from-layer="2791" from-port="1" to-layer="2801" to-port="0" /> + <edge from-layer="2792" from-port="0" to-layer="2794" to-port="1" /> + <edge from-layer="2793" from-port="0" to-layer="2794" to-port="2" /> + <edge from-layer="2794" from-port="3" to-layer="2802" to-port="0" /> + <edge from-layer="2794" from-port="3" to-layer="2808" to-port="0" /> + <edge from-layer="2795" from-port="0" to-layer="2796" to-port="2" /> + <edge from-layer="2796" from-port="3" to-layer="2805" to-port="0" /> + <edge from-layer="2796" from-port="3" to-layer="2798" to-port="0" /> + <edge from-layer="2797" from-port="0" to-layer="2798" to-port="1" /> + <edge from-layer="2798" from-port="2" to-layer="2802" to-port="1" /> + <edge from-layer="2799" from-port="0" to-layer="2801" to-port="1" /> + <edge from-layer="2800" from-port="0" to-layer="2801" to-port="2" /> + <edge from-layer="2801" from-port="3" to-layer="2808" to-port="2" /> + <edge from-layer="2801" from-port="3" to-layer="2802" to-port="3" /> + <edge from-layer="2802" from-port="4" to-layer="2803" to-port="0" /> + <edge from-layer="2803" from-port="1" to-layer="2804" to-port="1" /> + <edge from-layer="2804" from-port="2" to-layer="2809" to-port="0" /> + <edge from-layer="2805" from-port="2" to-layer="2807" to-port="0" /> + <edge from-layer="2806" from-port="0" to-layer="2807" to-port="1" /> + <edge from-layer="2807" from-port="2" to-layer="2808" to-port="1" /> + <edge from-layer="2808" from-port="3" to-layer="2809" to-port="1" /> + <edge from-layer="2809" from-port="2" to-layer="2856" to-port="0" /> + <edge from-layer="2809" from-port="2" to-layer="2864" to-port="1" /> + <edge from-layer="2810" from-port="0" to-layer="2811" to-port="0" /> + <edge from-layer="2811" from-port="1" to-layer="2816" to-port="0" /> + <edge from-layer="2812" from-port="0" to-layer="2815" to-port="1" /> + <edge from-layer="2813" from-port="0" to-layer="2815" to-port="2" /> + <edge from-layer="2814" from-port="0" to-layer="2815" to-port="3" /> + <edge from-layer="2815" from-port="4" to-layer="2816" to-port="1" /> + <edge from-layer="2816" from-port="2" to-layer="2817" to-port="0" /> + <edge from-layer="2817" from-port="1" to-layer="2819" to-port="0" /> + <edge from-layer="2818" from-port="0" to-layer="2819" to-port="2" /> + <edge from-layer="2819" from-port="3" to-layer="2832" to-port="0" /> + <edge from-layer="2820" from-port="0" to-layer="2821" to-port="0" /> + <edge from-layer="2821" from-port="1" to-layer="2824" to-port="0" /> + <edge from-layer="2822" from-port="0" to-layer="2823" to-port="0" /> + <edge from-layer="2823" from-port="1" to-layer="2824" to-port="1" /> + <edge from-layer="2824" from-port="2" to-layer="2826" to-port="0" /> + <edge from-layer="2825" from-port="0" to-layer="2826" to-port="1" /> + <edge from-layer="2826" from-port="2" to-layer="2827" to-port="0" /> + <edge from-layer="2827" from-port="1" to-layer="2828" to-port="1" /> + <edge from-layer="2828" from-port="2" to-layer="2829" to-port="0" /> + <edge from-layer="2829" from-port="2" to-layer="2831" to-port="0" /> + <edge from-layer="2830" from-port="0" to-layer="2831" to-port="1" /> + <edge from-layer="2831" from-port="2" to-layer="2832" to-port="1" /> + <edge from-layer="2832" from-port="2" to-layer="2833" to-port="0" /> + <edge from-layer="2832" from-port="2" to-layer="2834" to-port="0" /> + <edge from-layer="2832" from-port="2" to-layer="5549" to-port="0" /> + <edge from-layer="2833" from-port="2" to-layer="2847" to-port="0" /> + <edge from-layer="2834" from-port="1" to-layer="2837" to-port="0" /> + <edge from-layer="2834" from-port="1" to-layer="2844" to-port="0" /> + <edge from-layer="2834" from-port="1" to-layer="2839" to-port="0" /> + <edge from-layer="2835" from-port="0" to-layer="2837" to-port="1" /> + <edge from-layer="2836" from-port="0" to-layer="2837" to-port="2" /> + <edge from-layer="2837" from-port="3" to-layer="2845" to-port="0" /> + <edge from-layer="2837" from-port="3" to-layer="2851" to-port="0" /> + <edge from-layer="2838" from-port="0" to-layer="2839" to-port="2" /> + <edge from-layer="2839" from-port="3" to-layer="2848" to-port="0" /> + <edge from-layer="2839" from-port="3" to-layer="2841" to-port="0" /> + <edge from-layer="2840" from-port="0" to-layer="2841" to-port="1" /> + <edge from-layer="2841" from-port="2" to-layer="2845" to-port="1" /> + <edge from-layer="2842" from-port="0" to-layer="2844" to-port="1" /> + <edge from-layer="2843" from-port="0" to-layer="2844" to-port="2" /> + <edge from-layer="2844" from-port="3" to-layer="2851" to-port="2" /> + <edge from-layer="2844" from-port="3" to-layer="2845" to-port="3" /> + <edge from-layer="2845" from-port="4" to-layer="2846" to-port="0" /> + <edge from-layer="2846" from-port="1" to-layer="2847" to-port="1" /> + <edge from-layer="2847" from-port="2" to-layer="2852" to-port="0" /> + <edge from-layer="2848" from-port="2" to-layer="2850" to-port="0" /> + <edge from-layer="2849" from-port="0" to-layer="2850" to-port="1" /> + <edge from-layer="2850" from-port="2" to-layer="2851" to-port="1" /> + <edge from-layer="2851" from-port="3" to-layer="2852" to-port="1" /> + <edge from-layer="2852" from-port="2" to-layer="2864" to-port="2" /> + <edge from-layer="2853" from-port="0" to-layer="2863" to-port="1" /> + <edge from-layer="2854" from-port="0" to-layer="2861" to-port="0" /> + <edge from-layer="2855" from-port="0" to-layer="2861" to-port="1" /> + <edge from-layer="2856" from-port="1" to-layer="2859" to-port="0" /> + <edge from-layer="2857" from-port="0" to-layer="2859" to-port="1" /> + <edge from-layer="2858" from-port="0" to-layer="2859" to-port="2" /> + <edge from-layer="2859" from-port="3" to-layer="2861" to-port="2" /> + <edge from-layer="2860" from-port="0" to-layer="2861" to-port="3" /> + <edge from-layer="2861" from-port="4" to-layer="2863" to-port="2" /> + <edge from-layer="2862" from-port="0" to-layer="2863" to-port="3" /> + <edge from-layer="2863" from-port="4" to-layer="2864" to-port="3" /> + <edge from-layer="2864" from-port="4" to-layer="2866" to-port="0" /> + <edge from-layer="2865" from-port="0" to-layer="2866" to-port="1" /> + <edge from-layer="2866" from-port="2" to-layer="2868" to-port="0" /> + <edge from-layer="2867" from-port="2" to-layer="2868" to-port="1" /> + <edge from-layer="2868" from-port="2" to-layer="2877" to-port="0" /> + <edge from-layer="2869" from-port="0" to-layer="2870" to-port="0" /> + <edge from-layer="2870" from-port="1" to-layer="2873" to-port="0" /> + <edge from-layer="2871" from-port="0" to-layer="2872" to-port="0" /> + <edge from-layer="2872" from-port="1" to-layer="2873" to-port="1" /> + <edge from-layer="2873" from-port="2" to-layer="2875" to-port="0" /> + <edge from-layer="2874" from-port="0" to-layer="2875" to-port="1" /> + <edge from-layer="2875" from-port="2" to-layer="2876" to-port="0" /> + <edge from-layer="2876" from-port="1" to-layer="2877" to-port="1" /> + <edge from-layer="2877" from-port="2" to-layer="2878" to-port="1" /> + <edge from-layer="2878" from-port="2" to-layer="2885" to-port="0" /> + <edge from-layer="2878" from-port="2" to-layer="2924" to-port="0" /> + <edge from-layer="2878" from-port="2" to-layer="2893" to-port="0" /> + <edge from-layer="2879" from-port="0" to-layer="2880" to-port="0" /> + <edge from-layer="2880" from-port="1" to-layer="2894" to-port="0" /> + <edge from-layer="2881" from-port="0" to-layer="2882" to-port="0" /> + <edge from-layer="2882" from-port="1" to-layer="2892" to-port="0" /> + <edge from-layer="2883" from-port="0" to-layer="2884" to-port="0" /> + <edge from-layer="2884" from-port="1" to-layer="2885" to-port="1" /> + <edge from-layer="2885" from-port="2" to-layer="2887" to-port="0" /> + <edge from-layer="2886" from-port="0" to-layer="2887" to-port="1" /> + <edge from-layer="2887" from-port="2" to-layer="2890" to-port="0" /> + <edge from-layer="2888" from-port="0" to-layer="2889" to-port="0" /> + <edge from-layer="2889" from-port="1" to-layer="2890" to-port="1" /> + <edge from-layer="2890" from-port="2" to-layer="2891" to-port="0" /> + <edge from-layer="2891" from-port="1" to-layer="2892" to-port="1" /> + <edge from-layer="2892" from-port="2" to-layer="2893" to-port="1" /> + <edge from-layer="2893" from-port="2" to-layer="2894" to-port="1" /> + <edge from-layer="2894" from-port="2" to-layer="2903" to-port="0" /> + <edge from-layer="2894" from-port="2" to-layer="2913" to-port="0" /> + <edge from-layer="2895" from-port="0" to-layer="2896" to-port="0" /> + <edge from-layer="2896" from-port="1" to-layer="2899" to-port="0" /> + <edge from-layer="2897" from-port="0" to-layer="2898" to-port="0" /> + <edge from-layer="2898" from-port="1" to-layer="2899" to-port="1" /> + <edge from-layer="2899" from-port="2" to-layer="2901" to-port="0" /> + <edge from-layer="2900" from-port="0" to-layer="2901" to-port="1" /> + <edge from-layer="2901" from-port="2" to-layer="2902" to-port="0" /> + <edge from-layer="2902" from-port="1" to-layer="2903" to-port="1" /> + <edge from-layer="2903" from-port="2" to-layer="2904" to-port="0" /> + <edge from-layer="2904" from-port="1" to-layer="2914" to-port="0" /> + <edge from-layer="2905" from-port="0" to-layer="2906" to-port="0" /> + <edge from-layer="2906" from-port="1" to-layer="2909" to-port="0" /> + <edge from-layer="2907" from-port="0" to-layer="2908" to-port="0" /> + <edge from-layer="2908" from-port="1" to-layer="2909" to-port="1" /> + <edge from-layer="2909" from-port="2" to-layer="2911" to-port="0" /> + <edge from-layer="2910" from-port="0" to-layer="2911" to-port="1" /> + <edge from-layer="2911" from-port="2" to-layer="2912" to-port="0" /> + <edge from-layer="2912" from-port="1" to-layer="2913" to-port="1" /> + <edge from-layer="2913" from-port="2" to-layer="2914" to-port="1" /> + <edge from-layer="2914" from-port="2" to-layer="2923" to-port="0" /> + <edge from-layer="2915" from-port="0" to-layer="2916" to-port="0" /> + <edge from-layer="2916" from-port="1" to-layer="2919" to-port="0" /> + <edge from-layer="2917" from-port="0" to-layer="2918" to-port="0" /> + <edge from-layer="2918" from-port="1" to-layer="2919" to-port="1" /> + <edge from-layer="2919" from-port="2" to-layer="2921" to-port="0" /> + <edge from-layer="2920" from-port="0" to-layer="2921" to-port="1" /> + <edge from-layer="2921" from-port="2" to-layer="2922" to-port="0" /> + <edge from-layer="2922" from-port="1" to-layer="2923" to-port="1" /> + <edge from-layer="2923" from-port="2" to-layer="2924" to-port="1" /> + <edge from-layer="2924" from-port="2" to-layer="2939" to-port="0" /> + <edge from-layer="2924" from-port="2" to-layer="2931" to-port="0" /> + <edge from-layer="2924" from-port="2" to-layer="3129" to-port="0" /> + <edge from-layer="2925" from-port="0" to-layer="2926" to-port="0" /> + <edge from-layer="2926" from-port="1" to-layer="2940" to-port="0" /> + <edge from-layer="2927" from-port="0" to-layer="2928" to-port="0" /> + <edge from-layer="2928" from-port="1" to-layer="2938" to-port="0" /> + <edge from-layer="2929" from-port="0" to-layer="2930" to-port="0" /> + <edge from-layer="2930" from-port="1" to-layer="2931" to-port="1" /> + <edge from-layer="2931" from-port="2" to-layer="2933" to-port="0" /> + <edge from-layer="2932" from-port="0" to-layer="2933" to-port="1" /> + <edge from-layer="2933" from-port="2" to-layer="2936" to-port="0" /> + <edge from-layer="2934" from-port="0" to-layer="2935" to-port="0" /> + <edge from-layer="2935" from-port="1" to-layer="2936" to-port="1" /> + <edge from-layer="2936" from-port="2" to-layer="2937" to-port="0" /> + <edge from-layer="2937" from-port="1" to-layer="2938" to-port="1" /> + <edge from-layer="2938" from-port="2" to-layer="2939" to-port="1" /> + <edge from-layer="2939" from-port="2" to-layer="2940" to-port="1" /> + <edge from-layer="2940" from-port="2" to-layer="3079" to-port="0" /> + <edge from-layer="2940" from-port="2" to-layer="2949" to-port="0" /> + <edge from-layer="2940" from-port="2" to-layer="3011" to-port="0" /> + <edge from-layer="2940" from-port="2" to-layer="2950" to-port="0" /> + <edge from-layer="2941" from-port="0" to-layer="2942" to-port="0" /> + <edge from-layer="2942" from-port="1" to-layer="2945" to-port="0" /> + <edge from-layer="2943" from-port="0" to-layer="2944" to-port="0" /> + <edge from-layer="2944" from-port="1" to-layer="2945" to-port="1" /> + <edge from-layer="2945" from-port="2" to-layer="2947" to-port="0" /> + <edge from-layer="2946" from-port="0" to-layer="2947" to-port="1" /> + <edge from-layer="2947" from-port="2" to-layer="2948" to-port="0" /> + <edge from-layer="2948" from-port="1" to-layer="2949" to-port="1" /> + <edge from-layer="2949" from-port="2" to-layer="2955" to-port="0" /> + <edge from-layer="2950" from-port="1" to-layer="2953" to-port="0" /> + <edge from-layer="2951" from-port="0" to-layer="2953" to-port="1" /> + <edge from-layer="2952" from-port="0" to-layer="2953" to-port="2" /> + <edge from-layer="2953" from-port="3" to-layer="2954" to-port="0" /> + <edge from-layer="2953" from-port="3" to-layer="3012" to-port="0" /> + <edge from-layer="2953" from-port="3" to-layer="3118" to-port="0" /> + <edge from-layer="2954" from-port="3" to-layer="2955" to-port="1" /> + <edge from-layer="2955" from-port="2" to-layer="2957" to-port="0" /> + <edge from-layer="2956" from-port="0" to-layer="2957" to-port="1" /> + <edge from-layer="2957" from-port="2" to-layer="2981" to-port="0" /> + <edge from-layer="2957" from-port="2" to-layer="2987" to-port="0" /> + <edge from-layer="2957" from-port="2" to-layer="2970" to-port="0" /> + <edge from-layer="2957" from-port="2" to-layer="2967" to-port="0" /> + <edge from-layer="2958" from-port="0" to-layer="2959" to-port="0" /> + <edge from-layer="2959" from-port="1" to-layer="2960" to-port="0" /> + <edge from-layer="2960" from-port="2" to-layer="2961" to-port="0" /> + <edge from-layer="2961" from-port="2" to-layer="2963" to-port="0" /> + <edge from-layer="2962" from-port="0" to-layer="2963" to-port="1" /> + <edge from-layer="2963" from-port="2" to-layer="2964" to-port="1" /> + <edge from-layer="2963" from-port="2" to-layer="2964" to-port="0" /> + <edge from-layer="2964" from-port="2" to-layer="2989" to-port="0" /> + <edge from-layer="2964" from-port="2" to-layer="2965" to-port="0" /> + <edge from-layer="2965" from-port="1" to-layer="2966" to-port="0" /> + <edge from-layer="2966" from-port="2" to-layer="2967" to-port="1" /> + <edge from-layer="2966" from-port="2" to-layer="3016" to-port="1" /> + <edge from-layer="2967" from-port="2" to-layer="2992" to-port="0" /> + <edge from-layer="2968" from-port="0" to-layer="2978" to-port="0" /> + <edge from-layer="2969" from-port="0" to-layer="2978" to-port="1" /> + <edge from-layer="2970" from-port="1" to-layer="2972" to-port="0" /> + <edge from-layer="2971" from-port="0" to-layer="2972" to-port="2" /> + <edge from-layer="2972" from-port="3" to-layer="2973" to-port="0" /> + <edge from-layer="2973" from-port="2" to-layer="2974" to-port="0" /> + <edge from-layer="2974" from-port="1" to-layer="2976" to-port="0" /> + <edge from-layer="2975" from-port="0" to-layer="2976" to-port="1" /> + <edge from-layer="2976" from-port="2" to-layer="2978" to-port="2" /> + <edge from-layer="2977" from-port="0" to-layer="2978" to-port="3" /> + <edge from-layer="2978" from-port="4" to-layer="2981" to-port="1" /> + <edge from-layer="2978" from-port="4" to-layer="2987" to-port="2" /> + <edge from-layer="2979" from-port="0" to-layer="2981" to-port="2" /> + <edge from-layer="2980" from-port="0" to-layer="2981" to-port="3" /> + <edge from-layer="2981" from-port="4" to-layer="2984" to-port="0" /> + <edge from-layer="2982" from-port="0" to-layer="2983" to-port="0" /> + <edge from-layer="2983" from-port="1" to-layer="2984" to-port="1" /> + <edge from-layer="2984" from-port="2" to-layer="2988" to-port="0" /> + <edge from-layer="2985" from-port="0" to-layer="2987" to-port="1" /> + <edge from-layer="2986" from-port="0" to-layer="2987" to-port="3" /> + <edge from-layer="2987" from-port="4" to-layer="2988" to-port="1" /> + <edge from-layer="2988" from-port="2" to-layer="2991" to-port="0" /> + <edge from-layer="2989" from-port="1" to-layer="2990" to-port="0" /> + <edge from-layer="2990" from-port="2" to-layer="2991" to-port="1" /> + <edge from-layer="2990" from-port="2" to-layer="3038" to-port="1" /> + <edge from-layer="2991" from-port="2" to-layer="2992" to-port="1" /> + <edge from-layer="2992" from-port="2" to-layer="3115" to-port="0" /> + <edge from-layer="2993" from-port="0" to-layer="2994" to-port="0" /> + <edge from-layer="2994" from-port="1" to-layer="2999" to-port="0" /> + <edge from-layer="2995" from-port="0" to-layer="2998" to-port="1" /> + <edge from-layer="2996" from-port="0" to-layer="2998" to-port="2" /> + <edge from-layer="2997" from-port="0" to-layer="2998" to-port="3" /> + <edge from-layer="2998" from-port="4" to-layer="2999" to-port="1" /> + <edge from-layer="2999" from-port="2" to-layer="3000" to-port="0" /> + <edge from-layer="3000" from-port="1" to-layer="3002" to-port="0" /> + <edge from-layer="3001" from-port="0" to-layer="3002" to-port="2" /> + <edge from-layer="3002" from-port="3" to-layer="3040" to-port="0" /> + <edge from-layer="3003" from-port="0" to-layer="3004" to-port="0" /> + <edge from-layer="3004" from-port="1" to-layer="3007" to-port="0" /> + <edge from-layer="3005" from-port="0" to-layer="3006" to-port="0" /> + <edge from-layer="3006" from-port="1" to-layer="3007" to-port="1" /> + <edge from-layer="3007" from-port="2" to-layer="3009" to-port="0" /> + <edge from-layer="3008" from-port="0" to-layer="3009" to-port="1" /> + <edge from-layer="3009" from-port="2" to-layer="3010" to-port="0" /> + <edge from-layer="3010" from-port="1" to-layer="3011" to-port="1" /> + <edge from-layer="3011" from-port="2" to-layer="3013" to-port="0" /> + <edge from-layer="3012" from-port="3" to-layer="3013" to-port="1" /> + <edge from-layer="3012" from-port="3" to-layer="3080" to-port="1" /> + <edge from-layer="3013" from-port="2" to-layer="3015" to-port="0" /> + <edge from-layer="3014" from-port="0" to-layer="3015" to-port="1" /> + <edge from-layer="3015" from-port="2" to-layer="3019" to-port="0" /> + <edge from-layer="3015" from-port="2" to-layer="3036" to-port="0" /> + <edge from-layer="3015" from-port="2" to-layer="3016" to-port="0" /> + <edge from-layer="3015" from-port="2" to-layer="3030" to-port="0" /> + <edge from-layer="3016" from-port="2" to-layer="3039" to-port="0" /> + <edge from-layer="3017" from-port="0" to-layer="3027" to-port="0" /> + <edge from-layer="3018" from-port="0" to-layer="3027" to-port="1" /> + <edge from-layer="3019" from-port="1" to-layer="3021" to-port="0" /> + <edge from-layer="3020" from-port="0" to-layer="3021" to-port="2" /> + <edge from-layer="3021" from-port="3" to-layer="3022" to-port="0" /> + <edge from-layer="3022" from-port="2" to-layer="3023" to-port="0" /> + <edge from-layer="3023" from-port="1" to-layer="3025" to-port="0" /> + <edge from-layer="3024" from-port="0" to-layer="3025" to-port="1" /> + <edge from-layer="3025" from-port="2" to-layer="3027" to-port="2" /> + <edge from-layer="3026" from-port="0" to-layer="3027" to-port="3" /> + <edge from-layer="3027" from-port="4" to-layer="3030" to-port="1" /> + <edge from-layer="3027" from-port="4" to-layer="3036" to-port="2" /> + <edge from-layer="3028" from-port="0" to-layer="3030" to-port="2" /> + <edge from-layer="3029" from-port="0" to-layer="3030" to-port="3" /> + <edge from-layer="3030" from-port="4" to-layer="3033" to-port="0" /> + <edge from-layer="3031" from-port="0" to-layer="3032" to-port="0" /> + <edge from-layer="3032" from-port="1" to-layer="3033" to-port="1" /> + <edge from-layer="3033" from-port="2" to-layer="3037" to-port="0" /> + <edge from-layer="3034" from-port="0" to-layer="3036" to-port="1" /> + <edge from-layer="3035" from-port="0" to-layer="3036" to-port="3" /> + <edge from-layer="3036" from-port="4" to-layer="3037" to-port="1" /> + <edge from-layer="3037" from-port="2" to-layer="3038" to-port="0" /> + <edge from-layer="3038" from-port="2" to-layer="3039" to-port="1" /> + <edge from-layer="3039" from-port="2" to-layer="3040" to-port="1" /> + <edge from-layer="3040" from-port="2" to-layer="3042" to-port="0" /> + <edge from-layer="3040" from-port="2" to-layer="3041" to-port="0" /> + <edge from-layer="3040" from-port="2" to-layer="5548" to-port="0" /> + <edge from-layer="3041" from-port="2" to-layer="3055" to-port="0" /> + <edge from-layer="3042" from-port="1" to-layer="3047" to-port="0" /> + <edge from-layer="3042" from-port="1" to-layer="3045" to-port="0" /> + <edge from-layer="3042" from-port="1" to-layer="3052" to-port="0" /> + <edge from-layer="3043" from-port="0" to-layer="3045" to-port="1" /> + <edge from-layer="3044" from-port="0" to-layer="3045" to-port="2" /> + <edge from-layer="3045" from-port="3" to-layer="3059" to-port="0" /> + <edge from-layer="3045" from-port="3" to-layer="3053" to-port="0" /> + <edge from-layer="3046" from-port="0" to-layer="3047" to-port="2" /> + <edge from-layer="3047" from-port="3" to-layer="3049" to-port="0" /> + <edge from-layer="3047" from-port="3" to-layer="3056" to-port="0" /> + <edge from-layer="3048" from-port="0" to-layer="3049" to-port="1" /> + <edge from-layer="3049" from-port="2" to-layer="3053" to-port="1" /> + <edge from-layer="3050" from-port="0" to-layer="3052" to-port="1" /> + <edge from-layer="3051" from-port="0" to-layer="3052" to-port="2" /> + <edge from-layer="3052" from-port="3" to-layer="3053" to-port="3" /> + <edge from-layer="3052" from-port="3" to-layer="3059" to-port="2" /> + <edge from-layer="3053" from-port="4" to-layer="3054" to-port="0" /> + <edge from-layer="3054" from-port="1" to-layer="3055" to-port="1" /> + <edge from-layer="3055" from-port="2" to-layer="3060" to-port="0" /> + <edge from-layer="3056" from-port="2" to-layer="3058" to-port="0" /> + <edge from-layer="3057" from-port="0" to-layer="3058" to-port="1" /> + <edge from-layer="3058" from-port="2" to-layer="3059" to-port="1" /> + <edge from-layer="3059" from-port="3" to-layer="3060" to-port="1" /> + <edge from-layer="3060" from-port="2" to-layer="3107" to-port="0" /> + <edge from-layer="3060" from-port="2" to-layer="3115" to-port="1" /> + <edge from-layer="3061" from-port="0" to-layer="3062" to-port="0" /> + <edge from-layer="3062" from-port="1" to-layer="3067" to-port="0" /> + <edge from-layer="3063" from-port="0" to-layer="3066" to-port="1" /> + <edge from-layer="3064" from-port="0" to-layer="3066" to-port="2" /> + <edge from-layer="3065" from-port="0" to-layer="3066" to-port="3" /> + <edge from-layer="3066" from-port="4" to-layer="3067" to-port="1" /> + <edge from-layer="3067" from-port="2" to-layer="3068" to-port="0" /> + <edge from-layer="3068" from-port="1" to-layer="3070" to-port="0" /> + <edge from-layer="3069" from-port="0" to-layer="3070" to-port="2" /> + <edge from-layer="3070" from-port="3" to-layer="3083" to-port="0" /> + <edge from-layer="3071" from-port="0" to-layer="3072" to-port="0" /> + <edge from-layer="3072" from-port="1" to-layer="3075" to-port="0" /> + <edge from-layer="3073" from-port="0" to-layer="3074" to-port="0" /> + <edge from-layer="3074" from-port="1" to-layer="3075" to-port="1" /> + <edge from-layer="3075" from-port="2" to-layer="3077" to-port="0" /> + <edge from-layer="3076" from-port="0" to-layer="3077" to-port="1" /> + <edge from-layer="3077" from-port="2" to-layer="3078" to-port="0" /> + <edge from-layer="3078" from-port="1" to-layer="3079" to-port="1" /> + <edge from-layer="3079" from-port="2" to-layer="3080" to-port="0" /> + <edge from-layer="3080" from-port="2" to-layer="3082" to-port="0" /> + <edge from-layer="3081" from-port="0" to-layer="3082" to-port="1" /> + <edge from-layer="3082" from-port="2" to-layer="3083" to-port="1" /> + <edge from-layer="3083" from-port="2" to-layer="5547" to-port="0" /> + <edge from-layer="3083" from-port="2" to-layer="3084" to-port="0" /> + <edge from-layer="3083" from-port="2" to-layer="3085" to-port="0" /> + <edge from-layer="3084" from-port="2" to-layer="3098" to-port="0" /> + <edge from-layer="3085" from-port="1" to-layer="3088" to-port="0" /> + <edge from-layer="3085" from-port="1" to-layer="3090" to-port="0" /> + <edge from-layer="3085" from-port="1" to-layer="3095" to-port="0" /> + <edge from-layer="3086" from-port="0" to-layer="3088" to-port="1" /> + <edge from-layer="3087" from-port="0" to-layer="3088" to-port="2" /> + <edge from-layer="3088" from-port="3" to-layer="3102" to-port="0" /> + <edge from-layer="3088" from-port="3" to-layer="3096" to-port="0" /> + <edge from-layer="3089" from-port="0" to-layer="3090" to-port="2" /> + <edge from-layer="3090" from-port="3" to-layer="3099" to-port="0" /> + <edge from-layer="3090" from-port="3" to-layer="3092" to-port="0" /> + <edge from-layer="3091" from-port="0" to-layer="3092" to-port="1" /> + <edge from-layer="3092" from-port="2" to-layer="3096" to-port="1" /> + <edge from-layer="3093" from-port="0" to-layer="3095" to-port="1" /> + <edge from-layer="3094" from-port="0" to-layer="3095" to-port="2" /> + <edge from-layer="3095" from-port="3" to-layer="3102" to-port="2" /> + <edge from-layer="3095" from-port="3" to-layer="3096" to-port="3" /> + <edge from-layer="3096" from-port="4" to-layer="3097" to-port="0" /> + <edge from-layer="3097" from-port="1" to-layer="3098" to-port="1" /> + <edge from-layer="3098" from-port="2" to-layer="3103" to-port="0" /> + <edge from-layer="3099" from-port="2" to-layer="3101" to-port="0" /> + <edge from-layer="3100" from-port="0" to-layer="3101" to-port="1" /> + <edge from-layer="3101" from-port="2" to-layer="3102" to-port="1" /> + <edge from-layer="3102" from-port="3" to-layer="3103" to-port="1" /> + <edge from-layer="3103" from-port="2" to-layer="3115" to-port="2" /> + <edge from-layer="3104" from-port="0" to-layer="3114" to-port="1" /> + <edge from-layer="3105" from-port="0" to-layer="3112" to-port="0" /> + <edge from-layer="3106" from-port="0" to-layer="3112" to-port="1" /> + <edge from-layer="3107" from-port="1" to-layer="3110" to-port="0" /> + <edge from-layer="3108" from-port="0" to-layer="3110" to-port="1" /> + <edge from-layer="3109" from-port="0" to-layer="3110" to-port="2" /> + <edge from-layer="3110" from-port="3" to-layer="3112" to-port="2" /> + <edge from-layer="3111" from-port="0" to-layer="3112" to-port="3" /> + <edge from-layer="3112" from-port="4" to-layer="3114" to-port="2" /> + <edge from-layer="3113" from-port="0" to-layer="3114" to-port="3" /> + <edge from-layer="3114" from-port="4" to-layer="3115" to-port="3" /> + <edge from-layer="3115" from-port="4" to-layer="3117" to-port="0" /> + <edge from-layer="3116" from-port="0" to-layer="3117" to-port="1" /> + <edge from-layer="3117" from-port="2" to-layer="3119" to-port="0" /> + <edge from-layer="3118" from-port="2" to-layer="3119" to-port="1" /> + <edge from-layer="3119" from-port="2" to-layer="3128" to-port="0" /> + <edge from-layer="3120" from-port="0" to-layer="3121" to-port="0" /> + <edge from-layer="3121" from-port="1" to-layer="3124" to-port="0" /> + <edge from-layer="3122" from-port="0" to-layer="3123" to-port="0" /> + <edge from-layer="3123" from-port="1" to-layer="3124" to-port="1" /> + <edge from-layer="3124" from-port="2" to-layer="3126" to-port="0" /> + <edge from-layer="3125" from-port="0" to-layer="3126" to-port="1" /> + <edge from-layer="3126" from-port="2" to-layer="3127" to-port="0" /> + <edge from-layer="3127" from-port="1" to-layer="3128" to-port="1" /> + <edge from-layer="3128" from-port="2" to-layer="3129" to-port="1" /> + <edge from-layer="3129" from-port="2" to-layer="3175" to-port="0" /> + <edge from-layer="3129" from-port="2" to-layer="3144" to-port="0" /> + <edge from-layer="3129" from-port="2" to-layer="3136" to-port="0" /> + <edge from-layer="3130" from-port="0" to-layer="3131" to-port="0" /> + <edge from-layer="3131" from-port="1" to-layer="3145" to-port="0" /> + <edge from-layer="3132" from-port="0" to-layer="3133" to-port="0" /> + <edge from-layer="3133" from-port="1" to-layer="3143" to-port="0" /> + <edge from-layer="3134" from-port="0" to-layer="3135" to-port="0" /> + <edge from-layer="3135" from-port="1" to-layer="3136" to-port="1" /> + <edge from-layer="3136" from-port="2" to-layer="3138" to-port="0" /> + <edge from-layer="3137" from-port="0" to-layer="3138" to-port="1" /> + <edge from-layer="3138" from-port="2" to-layer="3141" to-port="0" /> + <edge from-layer="3139" from-port="0" to-layer="3140" to-port="0" /> + <edge from-layer="3140" from-port="1" to-layer="3141" to-port="1" /> + <edge from-layer="3141" from-port="2" to-layer="3142" to-port="0" /> + <edge from-layer="3142" from-port="1" to-layer="3143" to-port="1" /> + <edge from-layer="3143" from-port="2" to-layer="3144" to-port="1" /> + <edge from-layer="3144" from-port="2" to-layer="3145" to-port="1" /> + <edge from-layer="3145" from-port="2" to-layer="3154" to-port="0" /> + <edge from-layer="3145" from-port="2" to-layer="3164" to-port="0" /> + <edge from-layer="3146" from-port="0" to-layer="3147" to-port="0" /> + <edge from-layer="3147" from-port="1" to-layer="3150" to-port="0" /> + <edge from-layer="3148" from-port="0" to-layer="3149" to-port="0" /> + <edge from-layer="3149" from-port="1" to-layer="3150" to-port="1" /> + <edge from-layer="3150" from-port="2" to-layer="3152" to-port="0" /> + <edge from-layer="3151" from-port="0" to-layer="3152" to-port="1" /> + <edge from-layer="3152" from-port="2" to-layer="3153" to-port="0" /> + <edge from-layer="3153" from-port="1" to-layer="3154" to-port="1" /> + <edge from-layer="3154" from-port="2" to-layer="3155" to-port="0" /> + <edge from-layer="3155" from-port="1" to-layer="3165" to-port="0" /> + <edge from-layer="3156" from-port="0" to-layer="3157" to-port="0" /> + <edge from-layer="3157" from-port="1" to-layer="3160" to-port="0" /> + <edge from-layer="3158" from-port="0" to-layer="3159" to-port="0" /> + <edge from-layer="3159" from-port="1" to-layer="3160" to-port="1" /> + <edge from-layer="3160" from-port="2" to-layer="3162" to-port="0" /> + <edge from-layer="3161" from-port="0" to-layer="3162" to-port="1" /> + <edge from-layer="3162" from-port="2" to-layer="3163" to-port="0" /> + <edge from-layer="3163" from-port="1" to-layer="3164" to-port="1" /> + <edge from-layer="3164" from-port="2" to-layer="3165" to-port="1" /> + <edge from-layer="3165" from-port="2" to-layer="3174" to-port="0" /> + <edge from-layer="3166" from-port="0" to-layer="3167" to-port="0" /> + <edge from-layer="3167" from-port="1" to-layer="3170" to-port="0" /> + <edge from-layer="3168" from-port="0" to-layer="3169" to-port="0" /> + <edge from-layer="3169" from-port="1" to-layer="3170" to-port="1" /> + <edge from-layer="3170" from-port="2" to-layer="3172" to-port="0" /> + <edge from-layer="3171" from-port="0" to-layer="3172" to-port="1" /> + <edge from-layer="3172" from-port="2" to-layer="3173" to-port="0" /> + <edge from-layer="3173" from-port="1" to-layer="3174" to-port="1" /> + <edge from-layer="3174" from-port="2" to-layer="3175" to-port="1" /> + <edge from-layer="3175" from-port="2" to-layer="3380" to-port="0" /> + <edge from-layer="3175" from-port="2" to-layer="3190" to-port="0" /> + <edge from-layer="3175" from-port="2" to-layer="3182" to-port="0" /> + <edge from-layer="3176" from-port="0" to-layer="3177" to-port="0" /> + <edge from-layer="3177" from-port="1" to-layer="3191" to-port="0" /> + <edge from-layer="3178" from-port="0" to-layer="3179" to-port="0" /> + <edge from-layer="3179" from-port="1" to-layer="3189" to-port="0" /> + <edge from-layer="3180" from-port="0" to-layer="3181" to-port="0" /> + <edge from-layer="3181" from-port="1" to-layer="3182" to-port="1" /> + <edge from-layer="3182" from-port="2" to-layer="3184" to-port="0" /> + <edge from-layer="3183" from-port="0" to-layer="3184" to-port="1" /> + <edge from-layer="3184" from-port="2" to-layer="3187" to-port="0" /> + <edge from-layer="3185" from-port="0" to-layer="3186" to-port="0" /> + <edge from-layer="3186" from-port="1" to-layer="3187" to-port="1" /> + <edge from-layer="3187" from-port="2" to-layer="3188" to-port="0" /> + <edge from-layer="3188" from-port="1" to-layer="3189" to-port="1" /> + <edge from-layer="3189" from-port="2" to-layer="3190" to-port="1" /> + <edge from-layer="3190" from-port="2" to-layer="3191" to-port="1" /> + <edge from-layer="3191" from-port="2" to-layer="3200" to-port="0" /> + <edge from-layer="3191" from-port="2" to-layer="3330" to-port="0" /> + <edge from-layer="3191" from-port="2" to-layer="3201" to-port="0" /> + <edge from-layer="3191" from-port="2" to-layer="3262" to-port="0" /> + <edge from-layer="3192" from-port="0" to-layer="3193" to-port="0" /> + <edge from-layer="3193" from-port="1" to-layer="3196" to-port="0" /> + <edge from-layer="3194" from-port="0" to-layer="3195" to-port="0" /> + <edge from-layer="3195" from-port="1" to-layer="3196" to-port="1" /> + <edge from-layer="3196" from-port="2" to-layer="3198" to-port="0" /> + <edge from-layer="3197" from-port="0" to-layer="3198" to-port="1" /> + <edge from-layer="3198" from-port="2" to-layer="3199" to-port="0" /> + <edge from-layer="3199" from-port="1" to-layer="3200" to-port="1" /> + <edge from-layer="3200" from-port="2" to-layer="3206" to-port="0" /> + <edge from-layer="3201" from-port="1" to-layer="3204" to-port="0" /> + <edge from-layer="3202" from-port="0" to-layer="3204" to-port="1" /> + <edge from-layer="3203" from-port="0" to-layer="3204" to-port="2" /> + <edge from-layer="3204" from-port="3" to-layer="3369" to-port="0" /> + <edge from-layer="3204" from-port="3" to-layer="3205" to-port="0" /> + <edge from-layer="3204" from-port="3" to-layer="3263" to-port="0" /> + <edge from-layer="3205" from-port="3" to-layer="3206" to-port="1" /> + <edge from-layer="3206" from-port="2" to-layer="3208" to-port="0" /> + <edge from-layer="3207" from-port="0" to-layer="3208" to-port="1" /> + <edge from-layer="3208" from-port="2" to-layer="3232" to-port="0" /> + <edge from-layer="3208" from-port="2" to-layer="3221" to-port="0" /> + <edge from-layer="3208" from-port="2" to-layer="3218" to-port="0" /> + <edge from-layer="3208" from-port="2" to-layer="3238" to-port="0" /> + <edge from-layer="3209" from-port="0" to-layer="3210" to-port="0" /> + <edge from-layer="3210" from-port="1" to-layer="3211" to-port="0" /> + <edge from-layer="3211" from-port="2" to-layer="3212" to-port="0" /> + <edge from-layer="3212" from-port="2" to-layer="3214" to-port="0" /> + <edge from-layer="3213" from-port="0" to-layer="3214" to-port="1" /> + <edge from-layer="3214" from-port="2" to-layer="3215" to-port="1" /> + <edge from-layer="3214" from-port="2" to-layer="3215" to-port="0" /> + <edge from-layer="3215" from-port="2" to-layer="3240" to-port="0" /> + <edge from-layer="3215" from-port="2" to-layer="3216" to-port="0" /> + <edge from-layer="3216" from-port="1" to-layer="3217" to-port="0" /> + <edge from-layer="3217" from-port="2" to-layer="3218" to-port="1" /> + <edge from-layer="3217" from-port="2" to-layer="3267" to-port="1" /> + <edge from-layer="3218" from-port="2" to-layer="3243" to-port="0" /> + <edge from-layer="3219" from-port="0" to-layer="3229" to-port="0" /> + <edge from-layer="3220" from-port="0" to-layer="3229" to-port="1" /> + <edge from-layer="3221" from-port="1" to-layer="3223" to-port="0" /> + <edge from-layer="3222" from-port="0" to-layer="3223" to-port="2" /> + <edge from-layer="3223" from-port="3" to-layer="3224" to-port="0" /> + <edge from-layer="3224" from-port="2" to-layer="3225" to-port="0" /> + <edge from-layer="3225" from-port="1" to-layer="3227" to-port="0" /> + <edge from-layer="3226" from-port="0" to-layer="3227" to-port="1" /> + <edge from-layer="3227" from-port="2" to-layer="3229" to-port="2" /> + <edge from-layer="3228" from-port="0" to-layer="3229" to-port="3" /> + <edge from-layer="3229" from-port="4" to-layer="3232" to-port="1" /> + <edge from-layer="3229" from-port="4" to-layer="3238" to-port="2" /> + <edge from-layer="3230" from-port="0" to-layer="3232" to-port="2" /> + <edge from-layer="3231" from-port="0" to-layer="3232" to-port="3" /> + <edge from-layer="3232" from-port="4" to-layer="3235" to-port="0" /> + <edge from-layer="3233" from-port="0" to-layer="3234" to-port="0" /> + <edge from-layer="3234" from-port="1" to-layer="3235" to-port="1" /> + <edge from-layer="3235" from-port="2" to-layer="3239" to-port="0" /> + <edge from-layer="3236" from-port="0" to-layer="3238" to-port="1" /> + <edge from-layer="3237" from-port="0" to-layer="3238" to-port="3" /> + <edge from-layer="3238" from-port="4" to-layer="3239" to-port="1" /> + <edge from-layer="3239" from-port="2" to-layer="3242" to-port="0" /> + <edge from-layer="3240" from-port="1" to-layer="3241" to-port="0" /> + <edge from-layer="3241" from-port="2" to-layer="3242" to-port="1" /> + <edge from-layer="3241" from-port="2" to-layer="3289" to-port="1" /> + <edge from-layer="3242" from-port="2" to-layer="3243" to-port="1" /> + <edge from-layer="3243" from-port="2" to-layer="3366" to-port="0" /> + <edge from-layer="3244" from-port="0" to-layer="3245" to-port="0" /> + <edge from-layer="3245" from-port="1" to-layer="3250" to-port="0" /> + <edge from-layer="3246" from-port="0" to-layer="3249" to-port="1" /> + <edge from-layer="3247" from-port="0" to-layer="3249" to-port="2" /> + <edge from-layer="3248" from-port="0" to-layer="3249" to-port="3" /> + <edge from-layer="3249" from-port="4" to-layer="3250" to-port="1" /> + <edge from-layer="3250" from-port="2" to-layer="3251" to-port="0" /> + <edge from-layer="3251" from-port="1" to-layer="3253" to-port="0" /> + <edge from-layer="3252" from-port="0" to-layer="3253" to-port="2" /> + <edge from-layer="3253" from-port="3" to-layer="3291" to-port="0" /> + <edge from-layer="3254" from-port="0" to-layer="3255" to-port="0" /> + <edge from-layer="3255" from-port="1" to-layer="3258" to-port="0" /> + <edge from-layer="3256" from-port="0" to-layer="3257" to-port="0" /> + <edge from-layer="3257" from-port="1" to-layer="3258" to-port="1" /> + <edge from-layer="3258" from-port="2" to-layer="3260" to-port="0" /> + <edge from-layer="3259" from-port="0" to-layer="3260" to-port="1" /> + <edge from-layer="3260" from-port="2" to-layer="3261" to-port="0" /> + <edge from-layer="3261" from-port="1" to-layer="3262" to-port="1" /> + <edge from-layer="3262" from-port="2" to-layer="3264" to-port="0" /> + <edge from-layer="3263" from-port="3" to-layer="3331" to-port="1" /> + <edge from-layer="3263" from-port="3" to-layer="3264" to-port="1" /> + <edge from-layer="3264" from-port="2" to-layer="3266" to-port="0" /> + <edge from-layer="3265" from-port="0" to-layer="3266" to-port="1" /> + <edge from-layer="3266" from-port="2" to-layer="3287" to-port="0" /> + <edge from-layer="3266" from-port="2" to-layer="3281" to-port="0" /> + <edge from-layer="3266" from-port="2" to-layer="3270" to-port="0" /> + <edge from-layer="3266" from-port="2" to-layer="3267" to-port="0" /> + <edge from-layer="3267" from-port="2" to-layer="3290" to-port="0" /> + <edge from-layer="3268" from-port="0" to-layer="3278" to-port="0" /> + <edge from-layer="3269" from-port="0" to-layer="3278" to-port="1" /> + <edge from-layer="3270" from-port="1" to-layer="3272" to-port="0" /> + <edge from-layer="3271" from-port="0" to-layer="3272" to-port="2" /> + <edge from-layer="3272" from-port="3" to-layer="3273" to-port="0" /> + <edge from-layer="3273" from-port="2" to-layer="3274" to-port="0" /> + <edge from-layer="3274" from-port="1" to-layer="3276" to-port="0" /> + <edge from-layer="3275" from-port="0" to-layer="3276" to-port="1" /> + <edge from-layer="3276" from-port="2" to-layer="3278" to-port="2" /> + <edge from-layer="3277" from-port="0" to-layer="3278" to-port="3" /> + <edge from-layer="3278" from-port="4" to-layer="3281" to-port="1" /> + <edge from-layer="3278" from-port="4" to-layer="3287" to-port="2" /> + <edge from-layer="3279" from-port="0" to-layer="3281" to-port="2" /> + <edge from-layer="3280" from-port="0" to-layer="3281" to-port="3" /> + <edge from-layer="3281" from-port="4" to-layer="3284" to-port="0" /> + <edge from-layer="3282" from-port="0" to-layer="3283" to-port="0" /> + <edge from-layer="3283" from-port="1" to-layer="3284" to-port="1" /> + <edge from-layer="3284" from-port="2" to-layer="3288" to-port="0" /> + <edge from-layer="3285" from-port="0" to-layer="3287" to-port="1" /> + <edge from-layer="3286" from-port="0" to-layer="3287" to-port="3" /> + <edge from-layer="3287" from-port="4" to-layer="3288" to-port="1" /> + <edge from-layer="3288" from-port="2" to-layer="3289" to-port="0" /> + <edge from-layer="3289" from-port="2" to-layer="3290" to-port="1" /> + <edge from-layer="3290" from-port="2" to-layer="3291" to-port="1" /> + <edge from-layer="3291" from-port="2" to-layer="5546" to-port="0" /> + <edge from-layer="3291" from-port="2" to-layer="3292" to-port="0" /> + <edge from-layer="3291" from-port="2" to-layer="3293" to-port="0" /> + <edge from-layer="3292" from-port="2" to-layer="3306" to-port="0" /> + <edge from-layer="3293" from-port="1" to-layer="3303" to-port="0" /> + <edge from-layer="3293" from-port="1" to-layer="3298" to-port="0" /> + <edge from-layer="3293" from-port="1" to-layer="3296" to-port="0" /> + <edge from-layer="3294" from-port="0" to-layer="3296" to-port="1" /> + <edge from-layer="3295" from-port="0" to-layer="3296" to-port="2" /> + <edge from-layer="3296" from-port="3" to-layer="3304" to-port="0" /> + <edge from-layer="3296" from-port="3" to-layer="3310" to-port="0" /> + <edge from-layer="3297" from-port="0" to-layer="3298" to-port="2" /> + <edge from-layer="3298" from-port="3" to-layer="3307" to-port="0" /> + <edge from-layer="3298" from-port="3" to-layer="3300" to-port="0" /> + <edge from-layer="3299" from-port="0" to-layer="3300" to-port="1" /> + <edge from-layer="3300" from-port="2" to-layer="3304" to-port="1" /> + <edge from-layer="3301" from-port="0" to-layer="3303" to-port="1" /> + <edge from-layer="3302" from-port="0" to-layer="3303" to-port="2" /> + <edge from-layer="3303" from-port="3" to-layer="3304" to-port="3" /> + <edge from-layer="3303" from-port="3" to-layer="3310" to-port="2" /> + <edge from-layer="3304" from-port="4" to-layer="3305" to-port="0" /> + <edge from-layer="3305" from-port="1" to-layer="3306" to-port="1" /> + <edge from-layer="3306" from-port="2" to-layer="3311" to-port="0" /> + <edge from-layer="3307" from-port="2" to-layer="3309" to-port="0" /> + <edge from-layer="3308" from-port="0" to-layer="3309" to-port="1" /> + <edge from-layer="3309" from-port="2" to-layer="3310" to-port="1" /> + <edge from-layer="3310" from-port="3" to-layer="3311" to-port="1" /> + <edge from-layer="3311" from-port="2" to-layer="3366" to-port="1" /> + <edge from-layer="3311" from-port="2" to-layer="3358" to-port="0" /> + <edge from-layer="3312" from-port="0" to-layer="3313" to-port="0" /> + <edge from-layer="3313" from-port="1" to-layer="3318" to-port="0" /> + <edge from-layer="3314" from-port="0" to-layer="3317" to-port="1" /> + <edge from-layer="3315" from-port="0" to-layer="3317" to-port="2" /> + <edge from-layer="3316" from-port="0" to-layer="3317" to-port="3" /> + <edge from-layer="3317" from-port="4" to-layer="3318" to-port="1" /> + <edge from-layer="3318" from-port="2" to-layer="3319" to-port="0" /> + <edge from-layer="3319" from-port="1" to-layer="3321" to-port="0" /> + <edge from-layer="3320" from-port="0" to-layer="3321" to-port="2" /> + <edge from-layer="3321" from-port="3" to-layer="3334" to-port="0" /> + <edge from-layer="3322" from-port="0" to-layer="3323" to-port="0" /> + <edge from-layer="3323" from-port="1" to-layer="3326" to-port="0" /> + <edge from-layer="3324" from-port="0" to-layer="3325" to-port="0" /> + <edge from-layer="3325" from-port="1" to-layer="3326" to-port="1" /> + <edge from-layer="3326" from-port="2" to-layer="3328" to-port="0" /> + <edge from-layer="3327" from-port="0" to-layer="3328" to-port="1" /> + <edge from-layer="3328" from-port="2" to-layer="3329" to-port="0" /> + <edge from-layer="3329" from-port="1" to-layer="3330" to-port="1" /> + <edge from-layer="3330" from-port="2" to-layer="3331" to-port="0" /> + <edge from-layer="3331" from-port="2" to-layer="3333" to-port="0" /> + <edge from-layer="3332" from-port="0" to-layer="3333" to-port="1" /> + <edge from-layer="3333" from-port="2" to-layer="3334" to-port="1" /> + <edge from-layer="3334" from-port="2" to-layer="3335" to-port="0" /> + <edge from-layer="3334" from-port="2" to-layer="5545" to-port="0" /> + <edge from-layer="3334" from-port="2" to-layer="3336" to-port="0" /> + <edge from-layer="3335" from-port="2" to-layer="3349" to-port="0" /> + <edge from-layer="3336" from-port="1" to-layer="3339" to-port="0" /> + <edge from-layer="3336" from-port="1" to-layer="3341" to-port="0" /> + <edge from-layer="3336" from-port="1" to-layer="3346" to-port="0" /> + <edge from-layer="3337" from-port="0" to-layer="3339" to-port="1" /> + <edge from-layer="3338" from-port="0" to-layer="3339" to-port="2" /> + <edge from-layer="3339" from-port="3" to-layer="3347" to-port="0" /> + <edge from-layer="3339" from-port="3" to-layer="3353" to-port="0" /> + <edge from-layer="3340" from-port="0" to-layer="3341" to-port="2" /> + <edge from-layer="3341" from-port="3" to-layer="3343" to-port="0" /> + <edge from-layer="3341" from-port="3" to-layer="3350" to-port="0" /> + <edge from-layer="3342" from-port="0" to-layer="3343" to-port="1" /> + <edge from-layer="3343" from-port="2" to-layer="3347" to-port="1" /> + <edge from-layer="3344" from-port="0" to-layer="3346" to-port="1" /> + <edge from-layer="3345" from-port="0" to-layer="3346" to-port="2" /> + <edge from-layer="3346" from-port="3" to-layer="3353" to-port="2" /> + <edge from-layer="3346" from-port="3" to-layer="3347" to-port="3" /> + <edge from-layer="3347" from-port="4" to-layer="3348" to-port="0" /> + <edge from-layer="3348" from-port="1" to-layer="3349" to-port="1" /> + <edge from-layer="3349" from-port="2" to-layer="3354" to-port="0" /> + <edge from-layer="3350" from-port="2" to-layer="3352" to-port="0" /> + <edge from-layer="3351" from-port="0" to-layer="3352" to-port="1" /> + <edge from-layer="3352" from-port="2" to-layer="3353" to-port="1" /> + <edge from-layer="3353" from-port="3" to-layer="3354" to-port="1" /> + <edge from-layer="3354" from-port="2" to-layer="3366" to-port="2" /> + <edge from-layer="3355" from-port="0" to-layer="3365" to-port="1" /> + <edge from-layer="3356" from-port="0" to-layer="3363" to-port="0" /> + <edge from-layer="3357" from-port="0" to-layer="3363" to-port="1" /> + <edge from-layer="3358" from-port="1" to-layer="3361" to-port="0" /> + <edge from-layer="3359" from-port="0" to-layer="3361" to-port="1" /> + <edge from-layer="3360" from-port="0" to-layer="3361" to-port="2" /> + <edge from-layer="3361" from-port="3" to-layer="3363" to-port="2" /> + <edge from-layer="3362" from-port="0" to-layer="3363" to-port="3" /> + <edge from-layer="3363" from-port="4" to-layer="3365" to-port="2" /> + <edge from-layer="3364" from-port="0" to-layer="3365" to-port="3" /> + <edge from-layer="3365" from-port="4" to-layer="3366" to-port="3" /> + <edge from-layer="3366" from-port="4" to-layer="3368" to-port="0" /> + <edge from-layer="3367" from-port="0" to-layer="3368" to-port="1" /> + <edge from-layer="3368" from-port="2" to-layer="3370" to-port="0" /> + <edge from-layer="3369" from-port="2" to-layer="3370" to-port="1" /> + <edge from-layer="3370" from-port="2" to-layer="3379" to-port="0" /> + <edge from-layer="3371" from-port="0" to-layer="3372" to-port="0" /> + <edge from-layer="3372" from-port="1" to-layer="3375" to-port="0" /> + <edge from-layer="3373" from-port="0" to-layer="3374" to-port="0" /> + <edge from-layer="3374" from-port="1" to-layer="3375" to-port="1" /> + <edge from-layer="3375" from-port="2" to-layer="3377" to-port="0" /> + <edge from-layer="3376" from-port="0" to-layer="3377" to-port="1" /> + <edge from-layer="3377" from-port="2" to-layer="3378" to-port="0" /> + <edge from-layer="3378" from-port="1" to-layer="3379" to-port="1" /> + <edge from-layer="3379" from-port="2" to-layer="3380" to-port="1" /> + <edge from-layer="3380" from-port="2" to-layer="3387" to-port="0" /> + <edge from-layer="3380" from-port="2" to-layer="3426" to-port="0" /> + <edge from-layer="3380" from-port="2" to-layer="3395" to-port="0" /> + <edge from-layer="3381" from-port="0" to-layer="3382" to-port="0" /> + <edge from-layer="3382" from-port="1" to-layer="3396" to-port="0" /> + <edge from-layer="3383" from-port="0" to-layer="3384" to-port="0" /> + <edge from-layer="3384" from-port="1" to-layer="3394" to-port="0" /> + <edge from-layer="3385" from-port="0" to-layer="3386" to-port="0" /> + <edge from-layer="3386" from-port="1" to-layer="3387" to-port="1" /> + <edge from-layer="3387" from-port="2" to-layer="3389" to-port="0" /> + <edge from-layer="3388" from-port="0" to-layer="3389" to-port="1" /> + <edge from-layer="3389" from-port="2" to-layer="3392" to-port="0" /> + <edge from-layer="3390" from-port="0" to-layer="3391" to-port="0" /> + <edge from-layer="3391" from-port="1" to-layer="3392" to-port="1" /> + <edge from-layer="3392" from-port="2" to-layer="3393" to-port="0" /> + <edge from-layer="3393" from-port="1" to-layer="3394" to-port="1" /> + <edge from-layer="3394" from-port="2" to-layer="3395" to-port="1" /> + <edge from-layer="3395" from-port="2" to-layer="3396" to-port="1" /> + <edge from-layer="3396" from-port="2" to-layer="3405" to-port="0" /> + <edge from-layer="3396" from-port="2" to-layer="3415" to-port="0" /> + <edge from-layer="3397" from-port="0" to-layer="3398" to-port="0" /> + <edge from-layer="3398" from-port="1" to-layer="3401" to-port="0" /> + <edge from-layer="3399" from-port="0" to-layer="3400" to-port="0" /> + <edge from-layer="3400" from-port="1" to-layer="3401" to-port="1" /> + <edge from-layer="3401" from-port="2" to-layer="3403" to-port="0" /> + <edge from-layer="3402" from-port="0" to-layer="3403" to-port="1" /> + <edge from-layer="3403" from-port="2" to-layer="3404" to-port="0" /> + <edge from-layer="3404" from-port="1" to-layer="3405" to-port="1" /> + <edge from-layer="3405" from-port="2" to-layer="3406" to-port="0" /> + <edge from-layer="3406" from-port="1" to-layer="3416" to-port="0" /> + <edge from-layer="3407" from-port="0" to-layer="3408" to-port="0" /> + <edge from-layer="3408" from-port="1" to-layer="3411" to-port="0" /> + <edge from-layer="3409" from-port="0" to-layer="3410" to-port="0" /> + <edge from-layer="3410" from-port="1" to-layer="3411" to-port="1" /> + <edge from-layer="3411" from-port="2" to-layer="3413" to-port="0" /> + <edge from-layer="3412" from-port="0" to-layer="3413" to-port="1" /> + <edge from-layer="3413" from-port="2" to-layer="3414" to-port="0" /> + <edge from-layer="3414" from-port="1" to-layer="3415" to-port="1" /> + <edge from-layer="3415" from-port="2" to-layer="3416" to-port="1" /> + <edge from-layer="3416" from-port="2" to-layer="3425" to-port="0" /> + <edge from-layer="3417" from-port="0" to-layer="3418" to-port="0" /> + <edge from-layer="3418" from-port="1" to-layer="3421" to-port="0" /> + <edge from-layer="3419" from-port="0" to-layer="3420" to-port="0" /> + <edge from-layer="3420" from-port="1" to-layer="3421" to-port="1" /> + <edge from-layer="3421" from-port="2" to-layer="3423" to-port="0" /> + <edge from-layer="3422" from-port="0" to-layer="3423" to-port="1" /> + <edge from-layer="3423" from-port="2" to-layer="3424" to-port="0" /> + <edge from-layer="3424" from-port="1" to-layer="3425" to-port="1" /> + <edge from-layer="3425" from-port="2" to-layer="3426" to-port="1" /> + <edge from-layer="3426" from-port="2" to-layer="3433" to-port="0" /> + <edge from-layer="3426" from-port="2" to-layer="3441" to-port="0" /> + <edge from-layer="3426" from-port="2" to-layer="3631" to-port="0" /> + <edge from-layer="3427" from-port="0" to-layer="3428" to-port="0" /> + <edge from-layer="3428" from-port="1" to-layer="3442" to-port="0" /> + <edge from-layer="3429" from-port="0" to-layer="3430" to-port="0" /> + <edge from-layer="3430" from-port="1" to-layer="3440" to-port="0" /> + <edge from-layer="3431" from-port="0" to-layer="3432" to-port="0" /> + <edge from-layer="3432" from-port="1" to-layer="3433" to-port="1" /> + <edge from-layer="3433" from-port="2" to-layer="3435" to-port="0" /> + <edge from-layer="3434" from-port="0" to-layer="3435" to-port="1" /> + <edge from-layer="3435" from-port="2" to-layer="3438" to-port="0" /> + <edge from-layer="3436" from-port="0" to-layer="3437" to-port="0" /> + <edge from-layer="3437" from-port="1" to-layer="3438" to-port="1" /> + <edge from-layer="3438" from-port="2" to-layer="3439" to-port="0" /> + <edge from-layer="3439" from-port="1" to-layer="3440" to-port="1" /> + <edge from-layer="3440" from-port="2" to-layer="3441" to-port="1" /> + <edge from-layer="3441" from-port="2" to-layer="3442" to-port="1" /> + <edge from-layer="3442" from-port="2" to-layer="3513" to-port="0" /> + <edge from-layer="3442" from-port="2" to-layer="3451" to-port="0" /> + <edge from-layer="3442" from-port="2" to-layer="3452" to-port="0" /> + <edge from-layer="3442" from-port="2" to-layer="3581" to-port="0" /> + <edge from-layer="3443" from-port="0" to-layer="3444" to-port="0" /> + <edge from-layer="3444" from-port="1" to-layer="3447" to-port="0" /> + <edge from-layer="3445" from-port="0" to-layer="3446" to-port="0" /> + <edge from-layer="3446" from-port="1" to-layer="3447" to-port="1" /> + <edge from-layer="3447" from-port="2" to-layer="3449" to-port="0" /> + <edge from-layer="3448" from-port="0" to-layer="3449" to-port="1" /> + <edge from-layer="3449" from-port="2" to-layer="3450" to-port="0" /> + <edge from-layer="3450" from-port="1" to-layer="3451" to-port="1" /> + <edge from-layer="3451" from-port="2" to-layer="3457" to-port="0" /> + <edge from-layer="3452" from-port="1" to-layer="3455" to-port="0" /> + <edge from-layer="3453" from-port="0" to-layer="3455" to-port="1" /> + <edge from-layer="3454" from-port="0" to-layer="3455" to-port="2" /> + <edge from-layer="3455" from-port="3" to-layer="3514" to-port="0" /> + <edge from-layer="3455" from-port="3" to-layer="3456" to-port="0" /> + <edge from-layer="3455" from-port="3" to-layer="3620" to-port="0" /> + <edge from-layer="3456" from-port="3" to-layer="3457" to-port="1" /> + <edge from-layer="3457" from-port="2" to-layer="3459" to-port="0" /> + <edge from-layer="3458" from-port="0" to-layer="3459" to-port="1" /> + <edge from-layer="3459" from-port="2" to-layer="3472" to-port="0" /> + <edge from-layer="3459" from-port="2" to-layer="3483" to-port="0" /> + <edge from-layer="3459" from-port="2" to-layer="3469" to-port="0" /> + <edge from-layer="3459" from-port="2" to-layer="3489" to-port="0" /> + <edge from-layer="3460" from-port="0" to-layer="3461" to-port="0" /> + <edge from-layer="3461" from-port="1" to-layer="3462" to-port="0" /> + <edge from-layer="3462" from-port="2" to-layer="3463" to-port="0" /> + <edge from-layer="3463" from-port="2" to-layer="3465" to-port="0" /> + <edge from-layer="3464" from-port="0" to-layer="3465" to-port="1" /> + <edge from-layer="3465" from-port="2" to-layer="3466" to-port="0" /> + <edge from-layer="3465" from-port="2" to-layer="3466" to-port="1" /> + <edge from-layer="3466" from-port="2" to-layer="3491" to-port="0" /> + <edge from-layer="3466" from-port="2" to-layer="3467" to-port="0" /> + <edge from-layer="3467" from-port="1" to-layer="3468" to-port="0" /> + <edge from-layer="3468" from-port="2" to-layer="3518" to-port="1" /> + <edge from-layer="3468" from-port="2" to-layer="3469" to-port="1" /> + <edge from-layer="3469" from-port="2" to-layer="3494" to-port="0" /> + <edge from-layer="3470" from-port="0" to-layer="3480" to-port="0" /> + <edge from-layer="3471" from-port="0" to-layer="3480" to-port="1" /> + <edge from-layer="3472" from-port="1" to-layer="3474" to-port="0" /> + <edge from-layer="3473" from-port="0" to-layer="3474" to-port="2" /> + <edge from-layer="3474" from-port="3" to-layer="3475" to-port="0" /> + <edge from-layer="3475" from-port="2" to-layer="3476" to-port="0" /> + <edge from-layer="3476" from-port="1" to-layer="3478" to-port="0" /> + <edge from-layer="3477" from-port="0" to-layer="3478" to-port="1" /> + <edge from-layer="3478" from-port="2" to-layer="3480" to-port="2" /> + <edge from-layer="3479" from-port="0" to-layer="3480" to-port="3" /> + <edge from-layer="3480" from-port="4" to-layer="3483" to-port="1" /> + <edge from-layer="3480" from-port="4" to-layer="3489" to-port="2" /> + <edge from-layer="3481" from-port="0" to-layer="3483" to-port="2" /> + <edge from-layer="3482" from-port="0" to-layer="3483" to-port="3" /> + <edge from-layer="3483" from-port="4" to-layer="3486" to-port="0" /> + <edge from-layer="3484" from-port="0" to-layer="3485" to-port="0" /> + <edge from-layer="3485" from-port="1" to-layer="3486" to-port="1" /> + <edge from-layer="3486" from-port="2" to-layer="3490" to-port="0" /> + <edge from-layer="3487" from-port="0" to-layer="3489" to-port="1" /> + <edge from-layer="3488" from-port="0" to-layer="3489" to-port="3" /> + <edge from-layer="3489" from-port="4" to-layer="3490" to-port="1" /> + <edge from-layer="3490" from-port="2" to-layer="3493" to-port="0" /> + <edge from-layer="3491" from-port="1" to-layer="3492" to-port="0" /> + <edge from-layer="3492" from-port="2" to-layer="3540" to-port="1" /> + <edge from-layer="3492" from-port="2" to-layer="3493" to-port="1" /> + <edge from-layer="3493" from-port="2" to-layer="3494" to-port="1" /> + <edge from-layer="3494" from-port="2" to-layer="3617" to-port="0" /> + <edge from-layer="3495" from-port="0" to-layer="3496" to-port="0" /> + <edge from-layer="3496" from-port="1" to-layer="3501" to-port="0" /> + <edge from-layer="3497" from-port="0" to-layer="3500" to-port="1" /> + <edge from-layer="3498" from-port="0" to-layer="3500" to-port="2" /> + <edge from-layer="3499" from-port="0" to-layer="3500" to-port="3" /> + <edge from-layer="3500" from-port="4" to-layer="3501" to-port="1" /> + <edge from-layer="3501" from-port="2" to-layer="3502" to-port="0" /> + <edge from-layer="3502" from-port="1" to-layer="3504" to-port="0" /> + <edge from-layer="3503" from-port="0" to-layer="3504" to-port="2" /> + <edge from-layer="3504" from-port="3" to-layer="3542" to-port="0" /> + <edge from-layer="3505" from-port="0" to-layer="3506" to-port="0" /> + <edge from-layer="3506" from-port="1" to-layer="3509" to-port="0" /> + <edge from-layer="3507" from-port="0" to-layer="3508" to-port="0" /> + <edge from-layer="3508" from-port="1" to-layer="3509" to-port="1" /> + <edge from-layer="3509" from-port="2" to-layer="3511" to-port="0" /> + <edge from-layer="3510" from-port="0" to-layer="3511" to-port="1" /> + <edge from-layer="3511" from-port="2" to-layer="3512" to-port="0" /> + <edge from-layer="3512" from-port="1" to-layer="3513" to-port="1" /> + <edge from-layer="3513" from-port="2" to-layer="3515" to-port="0" /> + <edge from-layer="3514" from-port="3" to-layer="3582" to-port="1" /> + <edge from-layer="3514" from-port="3" to-layer="3515" to-port="1" /> + <edge from-layer="3515" from-port="2" to-layer="3517" to-port="0" /> + <edge from-layer="3516" from-port="0" to-layer="3517" to-port="1" /> + <edge from-layer="3517" from-port="2" to-layer="3532" to-port="0" /> + <edge from-layer="3517" from-port="2" to-layer="3538" to-port="0" /> + <edge from-layer="3517" from-port="2" to-layer="3521" to-port="0" /> + <edge from-layer="3517" from-port="2" to-layer="3518" to-port="0" /> + <edge from-layer="3518" from-port="2" to-layer="3541" to-port="0" /> + <edge from-layer="3519" from-port="0" to-layer="3529" to-port="0" /> + <edge from-layer="3520" from-port="0" to-layer="3529" to-port="1" /> + <edge from-layer="3521" from-port="1" to-layer="3523" to-port="0" /> + <edge from-layer="3522" from-port="0" to-layer="3523" to-port="2" /> + <edge from-layer="3523" from-port="3" to-layer="3524" to-port="0" /> + <edge from-layer="3524" from-port="2" to-layer="3525" to-port="0" /> + <edge from-layer="3525" from-port="1" to-layer="3527" to-port="0" /> + <edge from-layer="3526" from-port="0" to-layer="3527" to-port="1" /> + <edge from-layer="3527" from-port="2" to-layer="3529" to-port="2" /> + <edge from-layer="3528" from-port="0" to-layer="3529" to-port="3" /> + <edge from-layer="3529" from-port="4" to-layer="3532" to-port="1" /> + <edge from-layer="3529" from-port="4" to-layer="3538" to-port="2" /> + <edge from-layer="3530" from-port="0" to-layer="3532" to-port="2" /> + <edge from-layer="3531" from-port="0" to-layer="3532" to-port="3" /> + <edge from-layer="3532" from-port="4" to-layer="3535" to-port="0" /> + <edge from-layer="3533" from-port="0" to-layer="3534" to-port="0" /> + <edge from-layer="3534" from-port="1" to-layer="3535" to-port="1" /> + <edge from-layer="3535" from-port="2" to-layer="3539" to-port="0" /> + <edge from-layer="3536" from-port="0" to-layer="3538" to-port="1" /> + <edge from-layer="3537" from-port="0" to-layer="3538" to-port="3" /> + <edge from-layer="3538" from-port="4" to-layer="3539" to-port="1" /> + <edge from-layer="3539" from-port="2" to-layer="3540" to-port="0" /> + <edge from-layer="3540" from-port="2" to-layer="3541" to-port="1" /> + <edge from-layer="3541" from-port="2" to-layer="3542" to-port="1" /> + <edge from-layer="3542" from-port="2" to-layer="5544" to-port="0" /> + <edge from-layer="3542" from-port="2" to-layer="3543" to-port="0" /> + <edge from-layer="3542" from-port="2" to-layer="3544" to-port="0" /> + <edge from-layer="3543" from-port="2" to-layer="3557" to-port="0" /> + <edge from-layer="3544" from-port="1" to-layer="3547" to-port="0" /> + <edge from-layer="3544" from-port="1" to-layer="3549" to-port="0" /> + <edge from-layer="3544" from-port="1" to-layer="3554" to-port="0" /> + <edge from-layer="3545" from-port="0" to-layer="3547" to-port="1" /> + <edge from-layer="3546" from-port="0" to-layer="3547" to-port="2" /> + <edge from-layer="3547" from-port="3" to-layer="3555" to-port="0" /> + <edge from-layer="3547" from-port="3" to-layer="3561" to-port="0" /> + <edge from-layer="3548" from-port="0" to-layer="3549" to-port="2" /> + <edge from-layer="3549" from-port="3" to-layer="3558" to-port="0" /> + <edge from-layer="3549" from-port="3" to-layer="3551" to-port="0" /> + <edge from-layer="3550" from-port="0" to-layer="3551" to-port="1" /> + <edge from-layer="3551" from-port="2" to-layer="3555" to-port="1" /> + <edge from-layer="3552" from-port="0" to-layer="3554" to-port="1" /> + <edge from-layer="3553" from-port="0" to-layer="3554" to-port="2" /> + <edge from-layer="3554" from-port="3" to-layer="3561" to-port="2" /> + <edge from-layer="3554" from-port="3" to-layer="3555" to-port="3" /> + <edge from-layer="3555" from-port="4" to-layer="3556" to-port="0" /> + <edge from-layer="3556" from-port="1" to-layer="3557" to-port="1" /> + <edge from-layer="3557" from-port="2" to-layer="3562" to-port="0" /> + <edge from-layer="3558" from-port="2" to-layer="3560" to-port="0" /> + <edge from-layer="3559" from-port="0" to-layer="3560" to-port="1" /> + <edge from-layer="3560" from-port="2" to-layer="3561" to-port="1" /> + <edge from-layer="3561" from-port="3" to-layer="3562" to-port="1" /> + <edge from-layer="3562" from-port="2" to-layer="3617" to-port="1" /> + <edge from-layer="3562" from-port="2" to-layer="3609" to-port="0" /> + <edge from-layer="3563" from-port="0" to-layer="3564" to-port="0" /> + <edge from-layer="3564" from-port="1" to-layer="3569" to-port="0" /> + <edge from-layer="3565" from-port="0" to-layer="3568" to-port="1" /> + <edge from-layer="3566" from-port="0" to-layer="3568" to-port="2" /> + <edge from-layer="3567" from-port="0" to-layer="3568" to-port="3" /> + <edge from-layer="3568" from-port="4" to-layer="3569" to-port="1" /> + <edge from-layer="3569" from-port="2" to-layer="3570" to-port="0" /> + <edge from-layer="3570" from-port="1" to-layer="3572" to-port="0" /> + <edge from-layer="3571" from-port="0" to-layer="3572" to-port="2" /> + <edge from-layer="3572" from-port="3" to-layer="3585" to-port="0" /> + <edge from-layer="3573" from-port="0" to-layer="3574" to-port="0" /> + <edge from-layer="3574" from-port="1" to-layer="3577" to-port="0" /> + <edge from-layer="3575" from-port="0" to-layer="3576" to-port="0" /> + <edge from-layer="3576" from-port="1" to-layer="3577" to-port="1" /> + <edge from-layer="3577" from-port="2" to-layer="3579" to-port="0" /> + <edge from-layer="3578" from-port="0" to-layer="3579" to-port="1" /> + <edge from-layer="3579" from-port="2" to-layer="3580" to-port="0" /> + <edge from-layer="3580" from-port="1" to-layer="3581" to-port="1" /> + <edge from-layer="3581" from-port="2" to-layer="3582" to-port="0" /> + <edge from-layer="3582" from-port="2" to-layer="3584" to-port="0" /> + <edge from-layer="3583" from-port="0" to-layer="3584" to-port="1" /> + <edge from-layer="3584" from-port="2" to-layer="3585" to-port="1" /> + <edge from-layer="3585" from-port="2" to-layer="3587" to-port="0" /> + <edge from-layer="3585" from-port="2" to-layer="3586" to-port="0" /> + <edge from-layer="3585" from-port="2" to-layer="5543" to-port="0" /> + <edge from-layer="3586" from-port="2" to-layer="3600" to-port="0" /> + <edge from-layer="3587" from-port="1" to-layer="3590" to-port="0" /> + <edge from-layer="3587" from-port="1" to-layer="3592" to-port="0" /> + <edge from-layer="3587" from-port="1" to-layer="3597" to-port="0" /> + <edge from-layer="3588" from-port="0" to-layer="3590" to-port="1" /> + <edge from-layer="3589" from-port="0" to-layer="3590" to-port="2" /> + <edge from-layer="3590" from-port="3" to-layer="3604" to-port="0" /> + <edge from-layer="3590" from-port="3" to-layer="3598" to-port="0" /> + <edge from-layer="3591" from-port="0" to-layer="3592" to-port="2" /> + <edge from-layer="3592" from-port="3" to-layer="3601" to-port="0" /> + <edge from-layer="3592" from-port="3" to-layer="3594" to-port="0" /> + <edge from-layer="3593" from-port="0" to-layer="3594" to-port="1" /> + <edge from-layer="3594" from-port="2" to-layer="3598" to-port="1" /> + <edge from-layer="3595" from-port="0" to-layer="3597" to-port="1" /> + <edge from-layer="3596" from-port="0" to-layer="3597" to-port="2" /> + <edge from-layer="3597" from-port="3" to-layer="3598" to-port="3" /> + <edge from-layer="3597" from-port="3" to-layer="3604" to-port="2" /> + <edge from-layer="3598" from-port="4" to-layer="3599" to-port="0" /> + <edge from-layer="3599" from-port="1" to-layer="3600" to-port="1" /> + <edge from-layer="3600" from-port="2" to-layer="3605" to-port="0" /> + <edge from-layer="3601" from-port="2" to-layer="3603" to-port="0" /> + <edge from-layer="3602" from-port="0" to-layer="3603" to-port="1" /> + <edge from-layer="3603" from-port="2" to-layer="3604" to-port="1" /> + <edge from-layer="3604" from-port="3" to-layer="3605" to-port="1" /> + <edge from-layer="3605" from-port="2" to-layer="3617" to-port="2" /> + <edge from-layer="3606" from-port="0" to-layer="3616" to-port="1" /> + <edge from-layer="3607" from-port="0" to-layer="3614" to-port="0" /> + <edge from-layer="3608" from-port="0" to-layer="3614" to-port="1" /> + <edge from-layer="3609" from-port="1" to-layer="3612" to-port="0" /> + <edge from-layer="3610" from-port="0" to-layer="3612" to-port="1" /> + <edge from-layer="3611" from-port="0" to-layer="3612" to-port="2" /> + <edge from-layer="3612" from-port="3" to-layer="3614" to-port="2" /> + <edge from-layer="3613" from-port="0" to-layer="3614" to-port="3" /> + <edge from-layer="3614" from-port="4" to-layer="3616" to-port="2" /> + <edge from-layer="3615" from-port="0" to-layer="3616" to-port="3" /> + <edge from-layer="3616" from-port="4" to-layer="3617" to-port="3" /> + <edge from-layer="3617" from-port="4" to-layer="3619" to-port="0" /> + <edge from-layer="3618" from-port="0" to-layer="3619" to-port="1" /> + <edge from-layer="3619" from-port="2" to-layer="3621" to-port="0" /> + <edge from-layer="3620" from-port="2" to-layer="3621" to-port="1" /> + <edge from-layer="3621" from-port="2" to-layer="3630" to-port="0" /> + <edge from-layer="3622" from-port="0" to-layer="3623" to-port="0" /> + <edge from-layer="3623" from-port="1" to-layer="3626" to-port="0" /> + <edge from-layer="3624" from-port="0" to-layer="3625" to-port="0" /> + <edge from-layer="3625" from-port="1" to-layer="3626" to-port="1" /> + <edge from-layer="3626" from-port="2" to-layer="3628" to-port="0" /> + <edge from-layer="3627" from-port="0" to-layer="3628" to-port="1" /> + <edge from-layer="3628" from-port="2" to-layer="3629" to-port="0" /> + <edge from-layer="3629" from-port="1" to-layer="3630" to-port="1" /> + <edge from-layer="3630" from-port="2" to-layer="3631" to-port="1" /> + <edge from-layer="3631" from-port="2" to-layer="3638" to-port="0" /> + <edge from-layer="3631" from-port="2" to-layer="3646" to-port="0" /> + <edge from-layer="3631" from-port="2" to-layer="3677" to-port="0" /> + <edge from-layer="3632" from-port="0" to-layer="3633" to-port="0" /> + <edge from-layer="3633" from-port="1" to-layer="3647" to-port="0" /> + <edge from-layer="3634" from-port="0" to-layer="3635" to-port="0" /> + <edge from-layer="3635" from-port="1" to-layer="3645" to-port="0" /> + <edge from-layer="3636" from-port="0" to-layer="3637" to-port="0" /> + <edge from-layer="3637" from-port="1" to-layer="3638" to-port="1" /> + <edge from-layer="3638" from-port="2" to-layer="3640" to-port="0" /> + <edge from-layer="3639" from-port="0" to-layer="3640" to-port="1" /> + <edge from-layer="3640" from-port="2" to-layer="3643" to-port="0" /> + <edge from-layer="3641" from-port="0" to-layer="3642" to-port="0" /> + <edge from-layer="3642" from-port="1" to-layer="3643" to-port="1" /> + <edge from-layer="3643" from-port="2" to-layer="3644" to-port="0" /> + <edge from-layer="3644" from-port="1" to-layer="3645" to-port="1" /> + <edge from-layer="3645" from-port="2" to-layer="3646" to-port="1" /> + <edge from-layer="3646" from-port="2" to-layer="3647" to-port="1" /> + <edge from-layer="3647" from-port="2" to-layer="3656" to-port="0" /> + <edge from-layer="3647" from-port="2" to-layer="3666" to-port="0" /> + <edge from-layer="3648" from-port="0" to-layer="3649" to-port="0" /> + <edge from-layer="3649" from-port="1" to-layer="3652" to-port="0" /> + <edge from-layer="3650" from-port="0" to-layer="3651" to-port="0" /> + <edge from-layer="3651" from-port="1" to-layer="3652" to-port="1" /> + <edge from-layer="3652" from-port="2" to-layer="3654" to-port="0" /> + <edge from-layer="3653" from-port="0" to-layer="3654" to-port="1" /> + <edge from-layer="3654" from-port="2" to-layer="3655" to-port="0" /> + <edge from-layer="3655" from-port="1" to-layer="3656" to-port="1" /> + <edge from-layer="3656" from-port="2" to-layer="3657" to-port="0" /> + <edge from-layer="3657" from-port="1" to-layer="3667" to-port="0" /> + <edge from-layer="3658" from-port="0" to-layer="3659" to-port="0" /> + <edge from-layer="3659" from-port="1" to-layer="3662" to-port="0" /> + <edge from-layer="3660" from-port="0" to-layer="3661" to-port="0" /> + <edge from-layer="3661" from-port="1" to-layer="3662" to-port="1" /> + <edge from-layer="3662" from-port="2" to-layer="3664" to-port="0" /> + <edge from-layer="3663" from-port="0" to-layer="3664" to-port="1" /> + <edge from-layer="3664" from-port="2" to-layer="3665" to-port="0" /> + <edge from-layer="3665" from-port="1" to-layer="3666" to-port="1" /> + <edge from-layer="3666" from-port="2" to-layer="3667" to-port="1" /> + <edge from-layer="3667" from-port="2" to-layer="3676" to-port="0" /> + <edge from-layer="3668" from-port="0" to-layer="3669" to-port="0" /> + <edge from-layer="3669" from-port="1" to-layer="3672" to-port="0" /> + <edge from-layer="3670" from-port="0" to-layer="3671" to-port="0" /> + <edge from-layer="3671" from-port="1" to-layer="3672" to-port="1" /> + <edge from-layer="3672" from-port="2" to-layer="3674" to-port="0" /> + <edge from-layer="3673" from-port="0" to-layer="3674" to-port="1" /> + <edge from-layer="3674" from-port="2" to-layer="3675" to-port="0" /> + <edge from-layer="3675" from-port="1" to-layer="3676" to-port="1" /> + <edge from-layer="3676" from-port="2" to-layer="3677" to-port="1" /> + <edge from-layer="3677" from-port="2" to-layer="3684" to-port="0" /> + <edge from-layer="3677" from-port="2" to-layer="3692" to-port="0" /> + <edge from-layer="3677" from-port="2" to-layer="3882" to-port="0" /> + <edge from-layer="3678" from-port="0" to-layer="3679" to-port="0" /> + <edge from-layer="3679" from-port="1" to-layer="3693" to-port="0" /> + <edge from-layer="3680" from-port="0" to-layer="3681" to-port="0" /> + <edge from-layer="3681" from-port="1" to-layer="3691" to-port="0" /> + <edge from-layer="3682" from-port="0" to-layer="3683" to-port="0" /> + <edge from-layer="3683" from-port="1" to-layer="3684" to-port="1" /> + <edge from-layer="3684" from-port="2" to-layer="3686" to-port="0" /> + <edge from-layer="3685" from-port="0" to-layer="3686" to-port="1" /> + <edge from-layer="3686" from-port="2" to-layer="3689" to-port="0" /> + <edge from-layer="3687" from-port="0" to-layer="3688" to-port="0" /> + <edge from-layer="3688" from-port="1" to-layer="3689" to-port="1" /> + <edge from-layer="3689" from-port="2" to-layer="3690" to-port="0" /> + <edge from-layer="3690" from-port="1" to-layer="3691" to-port="1" /> + <edge from-layer="3691" from-port="2" to-layer="3692" to-port="1" /> + <edge from-layer="3692" from-port="2" to-layer="3693" to-port="1" /> + <edge from-layer="3693" from-port="2" to-layer="3764" to-port="0" /> + <edge from-layer="3693" from-port="2" to-layer="3832" to-port="0" /> + <edge from-layer="3693" from-port="2" to-layer="3703" to-port="0" /> + <edge from-layer="3693" from-port="2" to-layer="3702" to-port="0" /> + <edge from-layer="3694" from-port="0" to-layer="3695" to-port="0" /> + <edge from-layer="3695" from-port="1" to-layer="3698" to-port="0" /> + <edge from-layer="3696" from-port="0" to-layer="3697" to-port="0" /> + <edge from-layer="3697" from-port="1" to-layer="3698" to-port="1" /> + <edge from-layer="3698" from-port="2" to-layer="3700" to-port="0" /> + <edge from-layer="3699" from-port="0" to-layer="3700" to-port="1" /> + <edge from-layer="3700" from-port="2" to-layer="3701" to-port="0" /> + <edge from-layer="3701" from-port="1" to-layer="3702" to-port="1" /> + <edge from-layer="3702" from-port="2" to-layer="3708" to-port="0" /> + <edge from-layer="3703" from-port="1" to-layer="3706" to-port="0" /> + <edge from-layer="3704" from-port="0" to-layer="3706" to-port="1" /> + <edge from-layer="3705" from-port="0" to-layer="3706" to-port="2" /> + <edge from-layer="3706" from-port="3" to-layer="3871" to-port="0" /> + <edge from-layer="3706" from-port="3" to-layer="3765" to-port="0" /> + <edge from-layer="3706" from-port="3" to-layer="3707" to-port="0" /> + <edge from-layer="3707" from-port="3" to-layer="3708" to-port="1" /> + <edge from-layer="3708" from-port="2" to-layer="3710" to-port="0" /> + <edge from-layer="3709" from-port="0" to-layer="3710" to-port="1" /> + <edge from-layer="3710" from-port="2" to-layer="3720" to-port="0" /> + <edge from-layer="3710" from-port="2" to-layer="3734" to-port="0" /> + <edge from-layer="3710" from-port="2" to-layer="3723" to-port="0" /> + <edge from-layer="3710" from-port="2" to-layer="3740" to-port="0" /> + <edge from-layer="3711" from-port="0" to-layer="3712" to-port="0" /> + <edge from-layer="3712" from-port="1" to-layer="3713" to-port="0" /> + <edge from-layer="3713" from-port="2" to-layer="3714" to-port="0" /> + <edge from-layer="3714" from-port="2" to-layer="3716" to-port="0" /> + <edge from-layer="3715" from-port="0" to-layer="3716" to-port="1" /> + <edge from-layer="3716" from-port="2" to-layer="3717" to-port="0" /> + <edge from-layer="3716" from-port="2" to-layer="3717" to-port="1" /> + <edge from-layer="3717" from-port="2" to-layer="3718" to-port="0" /> + <edge from-layer="3717" from-port="2" to-layer="3742" to-port="0" /> + <edge from-layer="3718" from-port="1" to-layer="3719" to-port="0" /> + <edge from-layer="3719" from-port="2" to-layer="3720" to-port="1" /> + <edge from-layer="3719" from-port="2" to-layer="3769" to-port="1" /> + <edge from-layer="3720" from-port="2" to-layer="3745" to-port="0" /> + <edge from-layer="3721" from-port="0" to-layer="3731" to-port="0" /> + <edge from-layer="3722" from-port="0" to-layer="3731" to-port="1" /> + <edge from-layer="3723" from-port="1" to-layer="3725" to-port="0" /> + <edge from-layer="3724" from-port="0" to-layer="3725" to-port="2" /> + <edge from-layer="3725" from-port="3" to-layer="3726" to-port="0" /> + <edge from-layer="3726" from-port="2" to-layer="3727" to-port="0" /> + <edge from-layer="3727" from-port="1" to-layer="3729" to-port="0" /> + <edge from-layer="3728" from-port="0" to-layer="3729" to-port="1" /> + <edge from-layer="3729" from-port="2" to-layer="3731" to-port="2" /> + <edge from-layer="3730" from-port="0" to-layer="3731" to-port="3" /> + <edge from-layer="3731" from-port="4" to-layer="3740" to-port="2" /> + <edge from-layer="3731" from-port="4" to-layer="3734" to-port="1" /> + <edge from-layer="3732" from-port="0" to-layer="3734" to-port="2" /> + <edge from-layer="3733" from-port="0" to-layer="3734" to-port="3" /> + <edge from-layer="3734" from-port="4" to-layer="3737" to-port="0" /> + <edge from-layer="3735" from-port="0" to-layer="3736" to-port="0" /> + <edge from-layer="3736" from-port="1" to-layer="3737" to-port="1" /> + <edge from-layer="3737" from-port="2" to-layer="3741" to-port="0" /> + <edge from-layer="3738" from-port="0" to-layer="3740" to-port="1" /> + <edge from-layer="3739" from-port="0" to-layer="3740" to-port="3" /> + <edge from-layer="3740" from-port="4" to-layer="3741" to-port="1" /> + <edge from-layer="3741" from-port="2" to-layer="3744" to-port="0" /> + <edge from-layer="3742" from-port="1" to-layer="3743" to-port="0" /> + <edge from-layer="3743" from-port="2" to-layer="3744" to-port="1" /> + <edge from-layer="3743" from-port="2" to-layer="3791" to-port="1" /> + <edge from-layer="3744" from-port="2" to-layer="3745" to-port="1" /> + <edge from-layer="3745" from-port="2" to-layer="3868" to-port="0" /> + <edge from-layer="3746" from-port="0" to-layer="3747" to-port="0" /> + <edge from-layer="3747" from-port="1" to-layer="3752" to-port="0" /> + <edge from-layer="3748" from-port="0" to-layer="3751" to-port="1" /> + <edge from-layer="3749" from-port="0" to-layer="3751" to-port="2" /> + <edge from-layer="3750" from-port="0" to-layer="3751" to-port="3" /> + <edge from-layer="3751" from-port="4" to-layer="3752" to-port="1" /> + <edge from-layer="3752" from-port="2" to-layer="3753" to-port="0" /> + <edge from-layer="3753" from-port="1" to-layer="3755" to-port="0" /> + <edge from-layer="3754" from-port="0" to-layer="3755" to-port="2" /> + <edge from-layer="3755" from-port="3" to-layer="3793" to-port="0" /> + <edge from-layer="3756" from-port="0" to-layer="3757" to-port="0" /> + <edge from-layer="3757" from-port="1" to-layer="3760" to-port="0" /> + <edge from-layer="3758" from-port="0" to-layer="3759" to-port="0" /> + <edge from-layer="3759" from-port="1" to-layer="3760" to-port="1" /> + <edge from-layer="3760" from-port="2" to-layer="3762" to-port="0" /> + <edge from-layer="3761" from-port="0" to-layer="3762" to-port="1" /> + <edge from-layer="3762" from-port="2" to-layer="3763" to-port="0" /> + <edge from-layer="3763" from-port="1" to-layer="3764" to-port="1" /> + <edge from-layer="3764" from-port="2" to-layer="3766" to-port="0" /> + <edge from-layer="3765" from-port="3" to-layer="3833" to-port="1" /> + <edge from-layer="3765" from-port="3" to-layer="3766" to-port="1" /> + <edge from-layer="3766" from-port="2" to-layer="3768" to-port="0" /> + <edge from-layer="3767" from-port="0" to-layer="3768" to-port="1" /> + <edge from-layer="3768" from-port="2" to-layer="3783" to-port="0" /> + <edge from-layer="3768" from-port="2" to-layer="3769" to-port="0" /> + <edge from-layer="3768" from-port="2" to-layer="3772" to-port="0" /> + <edge from-layer="3768" from-port="2" to-layer="3789" to-port="0" /> + <edge from-layer="3769" from-port="2" to-layer="3792" to-port="0" /> + <edge from-layer="3770" from-port="0" to-layer="3780" to-port="0" /> + <edge from-layer="3771" from-port="0" to-layer="3780" to-port="1" /> + <edge from-layer="3772" from-port="1" to-layer="3774" to-port="0" /> + <edge from-layer="3773" from-port="0" to-layer="3774" to-port="2" /> + <edge from-layer="3774" from-port="3" to-layer="3775" to-port="0" /> + <edge from-layer="3775" from-port="2" to-layer="3776" to-port="0" /> + <edge from-layer="3776" from-port="1" to-layer="3778" to-port="0" /> + <edge from-layer="3777" from-port="0" to-layer="3778" to-port="1" /> + <edge from-layer="3778" from-port="2" to-layer="3780" to-port="2" /> + <edge from-layer="3779" from-port="0" to-layer="3780" to-port="3" /> + <edge from-layer="3780" from-port="4" to-layer="3789" to-port="2" /> + <edge from-layer="3780" from-port="4" to-layer="3783" to-port="1" /> + <edge from-layer="3781" from-port="0" to-layer="3783" to-port="2" /> + <edge from-layer="3782" from-port="0" to-layer="3783" to-port="3" /> + <edge from-layer="3783" from-port="4" to-layer="3786" to-port="0" /> + <edge from-layer="3784" from-port="0" to-layer="3785" to-port="0" /> + <edge from-layer="3785" from-port="1" to-layer="3786" to-port="1" /> + <edge from-layer="3786" from-port="2" to-layer="3790" to-port="0" /> + <edge from-layer="3787" from-port="0" to-layer="3789" to-port="1" /> + <edge from-layer="3788" from-port="0" to-layer="3789" to-port="3" /> + <edge from-layer="3789" from-port="4" to-layer="3790" to-port="1" /> + <edge from-layer="3790" from-port="2" to-layer="3791" to-port="0" /> + <edge from-layer="3791" from-port="2" to-layer="3792" to-port="1" /> + <edge from-layer="3792" from-port="2" to-layer="3793" to-port="1" /> + <edge from-layer="3793" from-port="2" to-layer="5542" to-port="0" /> + <edge from-layer="3793" from-port="2" to-layer="3794" to-port="0" /> + <edge from-layer="3793" from-port="2" to-layer="3795" to-port="0" /> + <edge from-layer="3794" from-port="2" to-layer="3808" to-port="0" /> + <edge from-layer="3795" from-port="1" to-layer="3798" to-port="0" /> + <edge from-layer="3795" from-port="1" to-layer="3805" to-port="0" /> + <edge from-layer="3795" from-port="1" to-layer="3800" to-port="0" /> + <edge from-layer="3796" from-port="0" to-layer="3798" to-port="1" /> + <edge from-layer="3797" from-port="0" to-layer="3798" to-port="2" /> + <edge from-layer="3798" from-port="3" to-layer="3806" to-port="0" /> + <edge from-layer="3798" from-port="3" to-layer="3812" to-port="0" /> + <edge from-layer="3799" from-port="0" to-layer="3800" to-port="2" /> + <edge from-layer="3800" from-port="3" to-layer="3802" to-port="0" /> + <edge from-layer="3800" from-port="3" to-layer="3809" to-port="0" /> + <edge from-layer="3801" from-port="0" to-layer="3802" to-port="1" /> + <edge from-layer="3802" from-port="2" to-layer="3806" to-port="1" /> + <edge from-layer="3803" from-port="0" to-layer="3805" to-port="1" /> + <edge from-layer="3804" from-port="0" to-layer="3805" to-port="2" /> + <edge from-layer="3805" from-port="3" to-layer="3812" to-port="2" /> + <edge from-layer="3805" from-port="3" to-layer="3806" to-port="3" /> + <edge from-layer="3806" from-port="4" to-layer="3807" to-port="0" /> + <edge from-layer="3807" from-port="1" to-layer="3808" to-port="1" /> + <edge from-layer="3808" from-port="2" to-layer="3813" to-port="0" /> + <edge from-layer="3809" from-port="2" to-layer="3811" to-port="0" /> + <edge from-layer="3810" from-port="0" to-layer="3811" to-port="1" /> + <edge from-layer="3811" from-port="2" to-layer="3812" to-port="1" /> + <edge from-layer="3812" from-port="3" to-layer="3813" to-port="1" /> + <edge from-layer="3813" from-port="2" to-layer="3868" to-port="1" /> + <edge from-layer="3813" from-port="2" to-layer="3860" to-port="0" /> + <edge from-layer="3814" from-port="0" to-layer="3815" to-port="0" /> + <edge from-layer="3815" from-port="1" to-layer="3820" to-port="0" /> + <edge from-layer="3816" from-port="0" to-layer="3819" to-port="1" /> + <edge from-layer="3817" from-port="0" to-layer="3819" to-port="2" /> + <edge from-layer="3818" from-port="0" to-layer="3819" to-port="3" /> + <edge from-layer="3819" from-port="4" to-layer="3820" to-port="1" /> + <edge from-layer="3820" from-port="2" to-layer="3821" to-port="0" /> + <edge from-layer="3821" from-port="1" to-layer="3823" to-port="0" /> + <edge from-layer="3822" from-port="0" to-layer="3823" to-port="2" /> + <edge from-layer="3823" from-port="3" to-layer="3836" to-port="0" /> + <edge from-layer="3824" from-port="0" to-layer="3825" to-port="0" /> + <edge from-layer="3825" from-port="1" to-layer="3828" to-port="0" /> + <edge from-layer="3826" from-port="0" to-layer="3827" to-port="0" /> + <edge from-layer="3827" from-port="1" to-layer="3828" to-port="1" /> + <edge from-layer="3828" from-port="2" to-layer="3830" to-port="0" /> + <edge from-layer="3829" from-port="0" to-layer="3830" to-port="1" /> + <edge from-layer="3830" from-port="2" to-layer="3831" to-port="0" /> + <edge from-layer="3831" from-port="1" to-layer="3832" to-port="1" /> + <edge from-layer="3832" from-port="2" to-layer="3833" to-port="0" /> + <edge from-layer="3833" from-port="2" to-layer="3835" to-port="0" /> + <edge from-layer="3834" from-port="0" to-layer="3835" to-port="1" /> + <edge from-layer="3835" from-port="2" to-layer="3836" to-port="1" /> + <edge from-layer="3836" from-port="2" to-layer="5541" to-port="0" /> + <edge from-layer="3836" from-port="2" to-layer="3837" to-port="0" /> + <edge from-layer="3836" from-port="2" to-layer="3838" to-port="0" /> + <edge from-layer="3837" from-port="2" to-layer="3851" to-port="0" /> + <edge from-layer="3838" from-port="1" to-layer="3841" to-port="0" /> + <edge from-layer="3838" from-port="1" to-layer="3843" to-port="0" /> + <edge from-layer="3838" from-port="1" to-layer="3848" to-port="0" /> + <edge from-layer="3839" from-port="0" to-layer="3841" to-port="1" /> + <edge from-layer="3840" from-port="0" to-layer="3841" to-port="2" /> + <edge from-layer="3841" from-port="3" to-layer="3849" to-port="0" /> + <edge from-layer="3841" from-port="3" to-layer="3855" to-port="0" /> + <edge from-layer="3842" from-port="0" to-layer="3843" to-port="2" /> + <edge from-layer="3843" from-port="3" to-layer="3845" to-port="0" /> + <edge from-layer="3843" from-port="3" to-layer="3852" to-port="0" /> + <edge from-layer="3844" from-port="0" to-layer="3845" to-port="1" /> + <edge from-layer="3845" from-port="2" to-layer="3849" to-port="1" /> + <edge from-layer="3846" from-port="0" to-layer="3848" to-port="1" /> + <edge from-layer="3847" from-port="0" to-layer="3848" to-port="2" /> + <edge from-layer="3848" from-port="3" to-layer="3855" to-port="2" /> + <edge from-layer="3848" from-port="3" to-layer="3849" to-port="3" /> + <edge from-layer="3849" from-port="4" to-layer="3850" to-port="0" /> + <edge from-layer="3850" from-port="1" to-layer="3851" to-port="1" /> + <edge from-layer="3851" from-port="2" to-layer="3856" to-port="0" /> + <edge from-layer="3852" from-port="2" to-layer="3854" to-port="0" /> + <edge from-layer="3853" from-port="0" to-layer="3854" to-port="1" /> + <edge from-layer="3854" from-port="2" to-layer="3855" to-port="1" /> + <edge from-layer="3855" from-port="3" to-layer="3856" to-port="1" /> + <edge from-layer="3856" from-port="2" to-layer="3868" to-port="2" /> + <edge from-layer="3857" from-port="0" to-layer="3867" to-port="1" /> + <edge from-layer="3858" from-port="0" to-layer="3865" to-port="0" /> + <edge from-layer="3859" from-port="0" to-layer="3865" to-port="1" /> + <edge from-layer="3860" from-port="1" to-layer="3863" to-port="0" /> + <edge from-layer="3861" from-port="0" to-layer="3863" to-port="1" /> + <edge from-layer="3862" from-port="0" to-layer="3863" to-port="2" /> + <edge from-layer="3863" from-port="3" to-layer="3865" to-port="2" /> + <edge from-layer="3864" from-port="0" to-layer="3865" to-port="3" /> + <edge from-layer="3865" from-port="4" to-layer="3867" to-port="2" /> + <edge from-layer="3866" from-port="0" to-layer="3867" to-port="3" /> + <edge from-layer="3867" from-port="4" to-layer="3868" to-port="3" /> + <edge from-layer="3868" from-port="4" to-layer="3870" to-port="0" /> + <edge from-layer="3869" from-port="0" to-layer="3870" to-port="1" /> + <edge from-layer="3870" from-port="2" to-layer="3872" to-port="0" /> + <edge from-layer="3871" from-port="2" to-layer="3872" to-port="1" /> + <edge from-layer="3872" from-port="2" to-layer="3881" to-port="0" /> + <edge from-layer="3873" from-port="0" to-layer="3874" to-port="0" /> + <edge from-layer="3874" from-port="1" to-layer="3877" to-port="0" /> + <edge from-layer="3875" from-port="0" to-layer="3876" to-port="0" /> + <edge from-layer="3876" from-port="1" to-layer="3877" to-port="1" /> + <edge from-layer="3877" from-port="2" to-layer="3879" to-port="0" /> + <edge from-layer="3878" from-port="0" to-layer="3879" to-port="1" /> + <edge from-layer="3879" from-port="2" to-layer="3880" to-port="0" /> + <edge from-layer="3880" from-port="1" to-layer="3881" to-port="1" /> + <edge from-layer="3881" from-port="2" to-layer="3882" to-port="1" /> + <edge from-layer="3882" from-port="2" to-layer="3928" to-port="0" /> + <edge from-layer="3882" from-port="2" to-layer="3889" to-port="0" /> + <edge from-layer="3882" from-port="2" to-layer="3897" to-port="0" /> + <edge from-layer="3883" from-port="0" to-layer="3884" to-port="0" /> + <edge from-layer="3884" from-port="1" to-layer="3898" to-port="0" /> + <edge from-layer="3885" from-port="0" to-layer="3886" to-port="0" /> + <edge from-layer="3886" from-port="1" to-layer="3896" to-port="0" /> + <edge from-layer="3887" from-port="0" to-layer="3888" to-port="0" /> + <edge from-layer="3888" from-port="1" to-layer="3889" to-port="1" /> + <edge from-layer="3889" from-port="2" to-layer="3891" to-port="0" /> + <edge from-layer="3890" from-port="0" to-layer="3891" to-port="1" /> + <edge from-layer="3891" from-port="2" to-layer="3894" to-port="0" /> + <edge from-layer="3892" from-port="0" to-layer="3893" to-port="0" /> + <edge from-layer="3893" from-port="1" to-layer="3894" to-port="1" /> + <edge from-layer="3894" from-port="2" to-layer="3895" to-port="0" /> + <edge from-layer="3895" from-port="1" to-layer="3896" to-port="1" /> + <edge from-layer="3896" from-port="2" to-layer="3897" to-port="1" /> + <edge from-layer="3897" from-port="2" to-layer="3898" to-port="1" /> + <edge from-layer="3898" from-port="2" to-layer="3917" to-port="0" /> + <edge from-layer="3898" from-port="2" to-layer="3907" to-port="0" /> + <edge from-layer="3899" from-port="0" to-layer="3900" to-port="0" /> + <edge from-layer="3900" from-port="1" to-layer="3903" to-port="0" /> + <edge from-layer="3901" from-port="0" to-layer="3902" to-port="0" /> + <edge from-layer="3902" from-port="1" to-layer="3903" to-port="1" /> + <edge from-layer="3903" from-port="2" to-layer="3905" to-port="0" /> + <edge from-layer="3904" from-port="0" to-layer="3905" to-port="1" /> + <edge from-layer="3905" from-port="2" to-layer="3906" to-port="0" /> + <edge from-layer="3906" from-port="1" to-layer="3907" to-port="1" /> + <edge from-layer="3907" from-port="2" to-layer="3908" to-port="0" /> + <edge from-layer="3908" from-port="1" to-layer="3918" to-port="0" /> + <edge from-layer="3909" from-port="0" to-layer="3910" to-port="0" /> + <edge from-layer="3910" from-port="1" to-layer="3913" to-port="0" /> + <edge from-layer="3911" from-port="0" to-layer="3912" to-port="0" /> + <edge from-layer="3912" from-port="1" to-layer="3913" to-port="1" /> + <edge from-layer="3913" from-port="2" to-layer="3915" to-port="0" /> + <edge from-layer="3914" from-port="0" to-layer="3915" to-port="1" /> + <edge from-layer="3915" from-port="2" to-layer="3916" to-port="0" /> + <edge from-layer="3916" from-port="1" to-layer="3917" to-port="1" /> + <edge from-layer="3917" from-port="2" to-layer="3918" to-port="1" /> + <edge from-layer="3918" from-port="2" to-layer="3927" to-port="0" /> + <edge from-layer="3919" from-port="0" to-layer="3920" to-port="0" /> + <edge from-layer="3920" from-port="1" to-layer="3923" to-port="0" /> + <edge from-layer="3921" from-port="0" to-layer="3922" to-port="0" /> + <edge from-layer="3922" from-port="1" to-layer="3923" to-port="1" /> + <edge from-layer="3923" from-port="2" to-layer="3925" to-port="0" /> + <edge from-layer="3924" from-port="0" to-layer="3925" to-port="1" /> + <edge from-layer="3925" from-port="2" to-layer="3926" to-port="0" /> + <edge from-layer="3926" from-port="1" to-layer="3927" to-port="1" /> + <edge from-layer="3927" from-port="2" to-layer="3928" to-port="1" /> + <edge from-layer="3928" from-port="2" to-layer="3935" to-port="0" /> + <edge from-layer="3928" from-port="2" to-layer="3943" to-port="0" /> + <edge from-layer="3928" from-port="2" to-layer="4133" to-port="0" /> + <edge from-layer="3929" from-port="0" to-layer="3930" to-port="0" /> + <edge from-layer="3930" from-port="1" to-layer="3944" to-port="0" /> + <edge from-layer="3931" from-port="0" to-layer="3932" to-port="0" /> + <edge from-layer="3932" from-port="1" to-layer="3942" to-port="0" /> + <edge from-layer="3933" from-port="0" to-layer="3934" to-port="0" /> + <edge from-layer="3934" from-port="1" to-layer="3935" to-port="1" /> + <edge from-layer="3935" from-port="2" to-layer="3937" to-port="0" /> + <edge from-layer="3936" from-port="0" to-layer="3937" to-port="1" /> + <edge from-layer="3937" from-port="2" to-layer="3940" to-port="0" /> + <edge from-layer="3938" from-port="0" to-layer="3939" to-port="0" /> + <edge from-layer="3939" from-port="1" to-layer="3940" to-port="1" /> + <edge from-layer="3940" from-port="2" to-layer="3941" to-port="0" /> + <edge from-layer="3941" from-port="1" to-layer="3942" to-port="1" /> + <edge from-layer="3942" from-port="2" to-layer="3943" to-port="1" /> + <edge from-layer="3943" from-port="2" to-layer="3944" to-port="1" /> + <edge from-layer="3944" from-port="2" to-layer="3953" to-port="0" /> + <edge from-layer="3944" from-port="2" to-layer="4083" to-port="0" /> + <edge from-layer="3944" from-port="2" to-layer="4015" to-port="0" /> + <edge from-layer="3944" from-port="2" to-layer="3954" to-port="0" /> + <edge from-layer="3945" from-port="0" to-layer="3946" to-port="0" /> + <edge from-layer="3946" from-port="1" to-layer="3949" to-port="0" /> + <edge from-layer="3947" from-port="0" to-layer="3948" to-port="0" /> + <edge from-layer="3948" from-port="1" to-layer="3949" to-port="1" /> + <edge from-layer="3949" from-port="2" to-layer="3951" to-port="0" /> + <edge from-layer="3950" from-port="0" to-layer="3951" to-port="1" /> + <edge from-layer="3951" from-port="2" to-layer="3952" to-port="0" /> + <edge from-layer="3952" from-port="1" to-layer="3953" to-port="1" /> + <edge from-layer="3953" from-port="2" to-layer="3959" to-port="0" /> + <edge from-layer="3954" from-port="1" to-layer="3957" to-port="0" /> + <edge from-layer="3955" from-port="0" to-layer="3957" to-port="1" /> + <edge from-layer="3956" from-port="0" to-layer="3957" to-port="2" /> + <edge from-layer="3957" from-port="3" to-layer="4016" to-port="0" /> + <edge from-layer="3957" from-port="3" to-layer="3958" to-port="0" /> + <edge from-layer="3957" from-port="3" to-layer="4122" to-port="0" /> + <edge from-layer="3958" from-port="3" to-layer="3959" to-port="1" /> + <edge from-layer="3959" from-port="2" to-layer="3961" to-port="0" /> + <edge from-layer="3960" from-port="0" to-layer="3961" to-port="1" /> + <edge from-layer="3961" from-port="2" to-layer="3971" to-port="0" /> + <edge from-layer="3961" from-port="2" to-layer="3991" to-port="0" /> + <edge from-layer="3961" from-port="2" to-layer="3974" to-port="0" /> + <edge from-layer="3961" from-port="2" to-layer="3985" to-port="0" /> + <edge from-layer="3962" from-port="0" to-layer="3963" to-port="0" /> + <edge from-layer="3963" from-port="1" to-layer="3964" to-port="0" /> + <edge from-layer="3964" from-port="2" to-layer="3965" to-port="0" /> + <edge from-layer="3965" from-port="2" to-layer="3967" to-port="0" /> + <edge from-layer="3966" from-port="0" to-layer="3967" to-port="1" /> + <edge from-layer="3967" from-port="2" to-layer="3968" to-port="0" /> + <edge from-layer="3967" from-port="2" to-layer="3968" to-port="1" /> + <edge from-layer="3968" from-port="2" to-layer="3969" to-port="0" /> + <edge from-layer="3968" from-port="2" to-layer="3993" to-port="0" /> + <edge from-layer="3969" from-port="1" to-layer="3970" to-port="0" /> + <edge from-layer="3970" from-port="2" to-layer="4020" to-port="1" /> + <edge from-layer="3970" from-port="2" to-layer="3971" to-port="1" /> + <edge from-layer="3971" from-port="2" to-layer="3996" to-port="0" /> + <edge from-layer="3972" from-port="0" to-layer="3982" to-port="0" /> + <edge from-layer="3973" from-port="0" to-layer="3982" to-port="1" /> + <edge from-layer="3974" from-port="1" to-layer="3976" to-port="0" /> + <edge from-layer="3975" from-port="0" to-layer="3976" to-port="2" /> + <edge from-layer="3976" from-port="3" to-layer="3977" to-port="0" /> + <edge from-layer="3977" from-port="2" to-layer="3978" to-port="0" /> + <edge from-layer="3978" from-port="1" to-layer="3980" to-port="0" /> + <edge from-layer="3979" from-port="0" to-layer="3980" to-port="1" /> + <edge from-layer="3980" from-port="2" to-layer="3982" to-port="2" /> + <edge from-layer="3981" from-port="0" to-layer="3982" to-port="3" /> + <edge from-layer="3982" from-port="4" to-layer="3985" to-port="1" /> + <edge from-layer="3982" from-port="4" to-layer="3991" to-port="2" /> + <edge from-layer="3983" from-port="0" to-layer="3985" to-port="2" /> + <edge from-layer="3984" from-port="0" to-layer="3985" to-port="3" /> + <edge from-layer="3985" from-port="4" to-layer="3988" to-port="0" /> + <edge from-layer="3986" from-port="0" to-layer="3987" to-port="0" /> + <edge from-layer="3987" from-port="1" to-layer="3988" to-port="1" /> + <edge from-layer="3988" from-port="2" to-layer="3992" to-port="0" /> + <edge from-layer="3989" from-port="0" to-layer="3991" to-port="1" /> + <edge from-layer="3990" from-port="0" to-layer="3991" to-port="3" /> + <edge from-layer="3991" from-port="4" to-layer="3992" to-port="1" /> + <edge from-layer="3992" from-port="2" to-layer="3995" to-port="0" /> + <edge from-layer="3993" from-port="1" to-layer="3994" to-port="0" /> + <edge from-layer="3994" from-port="2" to-layer="3995" to-port="1" /> + <edge from-layer="3994" from-port="2" to-layer="4042" to-port="1" /> + <edge from-layer="3995" from-port="2" to-layer="3996" to-port="1" /> + <edge from-layer="3996" from-port="2" to-layer="4119" to-port="0" /> + <edge from-layer="3997" from-port="0" to-layer="3998" to-port="0" /> + <edge from-layer="3998" from-port="1" to-layer="4003" to-port="0" /> + <edge from-layer="3999" from-port="0" to-layer="4002" to-port="1" /> + <edge from-layer="4000" from-port="0" to-layer="4002" to-port="2" /> + <edge from-layer="4001" from-port="0" to-layer="4002" to-port="3" /> + <edge from-layer="4002" from-port="4" to-layer="4003" to-port="1" /> + <edge from-layer="4003" from-port="2" to-layer="4004" to-port="0" /> + <edge from-layer="4004" from-port="1" to-layer="4006" to-port="0" /> + <edge from-layer="4005" from-port="0" to-layer="4006" to-port="2" /> + <edge from-layer="4006" from-port="3" to-layer="4044" to-port="0" /> + <edge from-layer="4007" from-port="0" to-layer="4008" to-port="0" /> + <edge from-layer="4008" from-port="1" to-layer="4011" to-port="0" /> + <edge from-layer="4009" from-port="0" to-layer="4010" to-port="0" /> + <edge from-layer="4010" from-port="1" to-layer="4011" to-port="1" /> + <edge from-layer="4011" from-port="2" to-layer="4013" to-port="0" /> + <edge from-layer="4012" from-port="0" to-layer="4013" to-port="1" /> + <edge from-layer="4013" from-port="2" to-layer="4014" to-port="0" /> + <edge from-layer="4014" from-port="1" to-layer="4015" to-port="1" /> + <edge from-layer="4015" from-port="2" to-layer="4017" to-port="0" /> + <edge from-layer="4016" from-port="3" to-layer="4017" to-port="1" /> + <edge from-layer="4016" from-port="3" to-layer="4084" to-port="1" /> + <edge from-layer="4017" from-port="2" to-layer="4019" to-port="0" /> + <edge from-layer="4018" from-port="0" to-layer="4019" to-port="1" /> + <edge from-layer="4019" from-port="2" to-layer="4020" to-port="0" /> + <edge from-layer="4019" from-port="2" to-layer="4023" to-port="0" /> + <edge from-layer="4019" from-port="2" to-layer="4034" to-port="0" /> + <edge from-layer="4019" from-port="2" to-layer="4040" to-port="0" /> + <edge from-layer="4020" from-port="2" to-layer="4043" to-port="0" /> + <edge from-layer="4021" from-port="0" to-layer="4031" to-port="0" /> + <edge from-layer="4022" from-port="0" to-layer="4031" to-port="1" /> + <edge from-layer="4023" from-port="1" to-layer="4025" to-port="0" /> + <edge from-layer="4024" from-port="0" to-layer="4025" to-port="2" /> + <edge from-layer="4025" from-port="3" to-layer="4026" to-port="0" /> + <edge from-layer="4026" from-port="2" to-layer="4027" to-port="0" /> + <edge from-layer="4027" from-port="1" to-layer="4029" to-port="0" /> + <edge from-layer="4028" from-port="0" to-layer="4029" to-port="1" /> + <edge from-layer="4029" from-port="2" to-layer="4031" to-port="2" /> + <edge from-layer="4030" from-port="0" to-layer="4031" to-port="3" /> + <edge from-layer="4031" from-port="4" to-layer="4040" to-port="2" /> + <edge from-layer="4031" from-port="4" to-layer="4034" to-port="1" /> + <edge from-layer="4032" from-port="0" to-layer="4034" to-port="2" /> + <edge from-layer="4033" from-port="0" to-layer="4034" to-port="3" /> + <edge from-layer="4034" from-port="4" to-layer="4037" to-port="0" /> + <edge from-layer="4035" from-port="0" to-layer="4036" to-port="0" /> + <edge from-layer="4036" from-port="1" to-layer="4037" to-port="1" /> + <edge from-layer="4037" from-port="2" to-layer="4041" to-port="0" /> + <edge from-layer="4038" from-port="0" to-layer="4040" to-port="1" /> + <edge from-layer="4039" from-port="0" to-layer="4040" to-port="3" /> + <edge from-layer="4040" from-port="4" to-layer="4041" to-port="1" /> + <edge from-layer="4041" from-port="2" to-layer="4042" to-port="0" /> + <edge from-layer="4042" from-port="2" to-layer="4043" to-port="1" /> + <edge from-layer="4043" from-port="2" to-layer="4044" to-port="1" /> + <edge from-layer="4044" from-port="2" to-layer="5540" to-port="0" /> + <edge from-layer="4044" from-port="2" to-layer="4045" to-port="0" /> + <edge from-layer="4044" from-port="2" to-layer="4046" to-port="0" /> + <edge from-layer="4045" from-port="2" to-layer="4059" to-port="0" /> + <edge from-layer="4046" from-port="1" to-layer="4056" to-port="0" /> + <edge from-layer="4046" from-port="1" to-layer="4051" to-port="0" /> + <edge from-layer="4046" from-port="1" to-layer="4049" to-port="0" /> + <edge from-layer="4047" from-port="0" to-layer="4049" to-port="1" /> + <edge from-layer="4048" from-port="0" to-layer="4049" to-port="2" /> + <edge from-layer="4049" from-port="3" to-layer="4057" to-port="0" /> + <edge from-layer="4049" from-port="3" to-layer="4063" to-port="0" /> + <edge from-layer="4050" from-port="0" to-layer="4051" to-port="2" /> + <edge from-layer="4051" from-port="3" to-layer="4060" to-port="0" /> + <edge from-layer="4051" from-port="3" to-layer="4053" to-port="0" /> + <edge from-layer="4052" from-port="0" to-layer="4053" to-port="1" /> + <edge from-layer="4053" from-port="2" to-layer="4057" to-port="1" /> + <edge from-layer="4054" from-port="0" to-layer="4056" to-port="1" /> + <edge from-layer="4055" from-port="0" to-layer="4056" to-port="2" /> + <edge from-layer="4056" from-port="3" to-layer="4063" to-port="2" /> + <edge from-layer="4056" from-port="3" to-layer="4057" to-port="3" /> + <edge from-layer="4057" from-port="4" to-layer="4058" to-port="0" /> + <edge from-layer="4058" from-port="1" to-layer="4059" to-port="1" /> + <edge from-layer="4059" from-port="2" to-layer="4064" to-port="0" /> + <edge from-layer="4060" from-port="2" to-layer="4062" to-port="0" /> + <edge from-layer="4061" from-port="0" to-layer="4062" to-port="1" /> + <edge from-layer="4062" from-port="2" to-layer="4063" to-port="1" /> + <edge from-layer="4063" from-port="3" to-layer="4064" to-port="1" /> + <edge from-layer="4064" from-port="2" to-layer="4111" to-port="0" /> + <edge from-layer="4064" from-port="2" to-layer="4119" to-port="1" /> + <edge from-layer="4065" from-port="0" to-layer="4066" to-port="0" /> + <edge from-layer="4066" from-port="1" to-layer="4071" to-port="0" /> + <edge from-layer="4067" from-port="0" to-layer="4070" to-port="1" /> + <edge from-layer="4068" from-port="0" to-layer="4070" to-port="2" /> + <edge from-layer="4069" from-port="0" to-layer="4070" to-port="3" /> + <edge from-layer="4070" from-port="4" to-layer="4071" to-port="1" /> + <edge from-layer="4071" from-port="2" to-layer="4072" to-port="0" /> + <edge from-layer="4072" from-port="1" to-layer="4074" to-port="0" /> + <edge from-layer="4073" from-port="0" to-layer="4074" to-port="2" /> + <edge from-layer="4074" from-port="3" to-layer="4087" to-port="0" /> + <edge from-layer="4075" from-port="0" to-layer="4076" to-port="0" /> + <edge from-layer="4076" from-port="1" to-layer="4079" to-port="0" /> + <edge from-layer="4077" from-port="0" to-layer="4078" to-port="0" /> + <edge from-layer="4078" from-port="1" to-layer="4079" to-port="1" /> + <edge from-layer="4079" from-port="2" to-layer="4081" to-port="0" /> + <edge from-layer="4080" from-port="0" to-layer="4081" to-port="1" /> + <edge from-layer="4081" from-port="2" to-layer="4082" to-port="0" /> + <edge from-layer="4082" from-port="1" to-layer="4083" to-port="1" /> + <edge from-layer="4083" from-port="2" to-layer="4084" to-port="0" /> + <edge from-layer="4084" from-port="2" to-layer="4086" to-port="0" /> + <edge from-layer="4085" from-port="0" to-layer="4086" to-port="1" /> + <edge from-layer="4086" from-port="2" to-layer="4087" to-port="1" /> + <edge from-layer="4087" from-port="2" to-layer="4088" to-port="0" /> + <edge from-layer="4087" from-port="2" to-layer="4089" to-port="0" /> + <edge from-layer="4087" from-port="2" to-layer="5539" to-port="0" /> + <edge from-layer="4088" from-port="2" to-layer="4102" to-port="0" /> + <edge from-layer="4089" from-port="1" to-layer="4099" to-port="0" /> + <edge from-layer="4089" from-port="1" to-layer="4094" to-port="0" /> + <edge from-layer="4089" from-port="1" to-layer="4092" to-port="0" /> + <edge from-layer="4090" from-port="0" to-layer="4092" to-port="1" /> + <edge from-layer="4091" from-port="0" to-layer="4092" to-port="2" /> + <edge from-layer="4092" from-port="3" to-layer="4106" to-port="0" /> + <edge from-layer="4092" from-port="3" to-layer="4100" to-port="0" /> + <edge from-layer="4093" from-port="0" to-layer="4094" to-port="2" /> + <edge from-layer="4094" from-port="3" to-layer="4103" to-port="0" /> + <edge from-layer="4094" from-port="3" to-layer="4096" to-port="0" /> + <edge from-layer="4095" from-port="0" to-layer="4096" to-port="1" /> + <edge from-layer="4096" from-port="2" to-layer="4100" to-port="1" /> + <edge from-layer="4097" from-port="0" to-layer="4099" to-port="1" /> + <edge from-layer="4098" from-port="0" to-layer="4099" to-port="2" /> + <edge from-layer="4099" from-port="3" to-layer="4106" to-port="2" /> + <edge from-layer="4099" from-port="3" to-layer="4100" to-port="3" /> + <edge from-layer="4100" from-port="4" to-layer="4101" to-port="0" /> + <edge from-layer="4101" from-port="1" to-layer="4102" to-port="1" /> + <edge from-layer="4102" from-port="2" to-layer="4107" to-port="0" /> + <edge from-layer="4103" from-port="2" to-layer="4105" to-port="0" /> + <edge from-layer="4104" from-port="0" to-layer="4105" to-port="1" /> + <edge from-layer="4105" from-port="2" to-layer="4106" to-port="1" /> + <edge from-layer="4106" from-port="3" to-layer="4107" to-port="1" /> + <edge from-layer="4107" from-port="2" to-layer="4119" to-port="2" /> + <edge from-layer="4108" from-port="0" to-layer="4118" to-port="1" /> + <edge from-layer="4109" from-port="0" to-layer="4116" to-port="0" /> + <edge from-layer="4110" from-port="0" to-layer="4116" to-port="1" /> + <edge from-layer="4111" from-port="1" to-layer="4114" to-port="0" /> + <edge from-layer="4112" from-port="0" to-layer="4114" to-port="1" /> + <edge from-layer="4113" from-port="0" to-layer="4114" to-port="2" /> + <edge from-layer="4114" from-port="3" to-layer="4116" to-port="2" /> + <edge from-layer="4115" from-port="0" to-layer="4116" to-port="3" /> + <edge from-layer="4116" from-port="4" to-layer="4118" to-port="2" /> + <edge from-layer="4117" from-port="0" to-layer="4118" to-port="3" /> + <edge from-layer="4118" from-port="4" to-layer="4119" to-port="3" /> + <edge from-layer="4119" from-port="4" to-layer="4121" to-port="0" /> + <edge from-layer="4120" from-port="0" to-layer="4121" to-port="1" /> + <edge from-layer="4121" from-port="2" to-layer="4123" to-port="0" /> + <edge from-layer="4122" from-port="2" to-layer="4123" to-port="1" /> + <edge from-layer="4123" from-port="2" to-layer="4132" to-port="0" /> + <edge from-layer="4124" from-port="0" to-layer="4125" to-port="0" /> + <edge from-layer="4125" from-port="1" to-layer="4128" to-port="0" /> + <edge from-layer="4126" from-port="0" to-layer="4127" to-port="0" /> + <edge from-layer="4127" from-port="1" to-layer="4128" to-port="1" /> + <edge from-layer="4128" from-port="2" to-layer="4130" to-port="0" /> + <edge from-layer="4129" from-port="0" to-layer="4130" to-port="1" /> + <edge from-layer="4130" from-port="2" to-layer="4131" to-port="0" /> + <edge from-layer="4131" from-port="1" to-layer="4132" to-port="1" /> + <edge from-layer="4132" from-port="2" to-layer="4133" to-port="1" /> + <edge from-layer="4133" from-port="2" to-layer="4140" to-port="0" /> + <edge from-layer="4133" from-port="2" to-layer="4148" to-port="0" /> + <edge from-layer="4133" from-port="2" to-layer="4179" to-port="0" /> + <edge from-layer="4134" from-port="0" to-layer="4135" to-port="0" /> + <edge from-layer="4135" from-port="1" to-layer="4149" to-port="0" /> + <edge from-layer="4136" from-port="0" to-layer="4137" to-port="0" /> + <edge from-layer="4137" from-port="1" to-layer="4147" to-port="0" /> + <edge from-layer="4138" from-port="0" to-layer="4139" to-port="0" /> + <edge from-layer="4139" from-port="1" to-layer="4140" to-port="1" /> + <edge from-layer="4140" from-port="2" to-layer="4142" to-port="0" /> + <edge from-layer="4141" from-port="0" to-layer="4142" to-port="1" /> + <edge from-layer="4142" from-port="2" to-layer="4145" to-port="0" /> + <edge from-layer="4143" from-port="0" to-layer="4144" to-port="0" /> + <edge from-layer="4144" from-port="1" to-layer="4145" to-port="1" /> + <edge from-layer="4145" from-port="2" to-layer="4146" to-port="0" /> + <edge from-layer="4146" from-port="1" to-layer="4147" to-port="1" /> + <edge from-layer="4147" from-port="2" to-layer="4148" to-port="1" /> + <edge from-layer="4148" from-port="2" to-layer="4149" to-port="1" /> + <edge from-layer="4149" from-port="2" to-layer="4168" to-port="0" /> + <edge from-layer="4149" from-port="2" to-layer="4158" to-port="0" /> + <edge from-layer="4150" from-port="0" to-layer="4151" to-port="0" /> + <edge from-layer="4151" from-port="1" to-layer="4154" to-port="0" /> + <edge from-layer="4152" from-port="0" to-layer="4153" to-port="0" /> + <edge from-layer="4153" from-port="1" to-layer="4154" to-port="1" /> + <edge from-layer="4154" from-port="2" to-layer="4156" to-port="0" /> + <edge from-layer="4155" from-port="0" to-layer="4156" to-port="1" /> + <edge from-layer="4156" from-port="2" to-layer="4157" to-port="0" /> + <edge from-layer="4157" from-port="1" to-layer="4158" to-port="1" /> + <edge from-layer="4158" from-port="2" to-layer="4159" to-port="0" /> + <edge from-layer="4159" from-port="1" to-layer="4169" to-port="0" /> + <edge from-layer="4160" from-port="0" to-layer="4161" to-port="0" /> + <edge from-layer="4161" from-port="1" to-layer="4164" to-port="0" /> + <edge from-layer="4162" from-port="0" to-layer="4163" to-port="0" /> + <edge from-layer="4163" from-port="1" to-layer="4164" to-port="1" /> + <edge from-layer="4164" from-port="2" to-layer="4166" to-port="0" /> + <edge from-layer="4165" from-port="0" to-layer="4166" to-port="1" /> + <edge from-layer="4166" from-port="2" to-layer="4167" to-port="0" /> + <edge from-layer="4167" from-port="1" to-layer="4168" to-port="1" /> + <edge from-layer="4168" from-port="2" to-layer="4169" to-port="1" /> + <edge from-layer="4169" from-port="2" to-layer="4178" to-port="0" /> + <edge from-layer="4170" from-port="0" to-layer="4171" to-port="0" /> + <edge from-layer="4171" from-port="1" to-layer="4174" to-port="0" /> + <edge from-layer="4172" from-port="0" to-layer="4173" to-port="0" /> + <edge from-layer="4173" from-port="1" to-layer="4174" to-port="1" /> + <edge from-layer="4174" from-port="2" to-layer="4176" to-port="0" /> + <edge from-layer="4175" from-port="0" to-layer="4176" to-port="1" /> + <edge from-layer="4176" from-port="2" to-layer="4177" to-port="0" /> + <edge from-layer="4177" from-port="1" to-layer="4178" to-port="1" /> + <edge from-layer="4178" from-port="2" to-layer="4179" to-port="1" /> + <edge from-layer="4179" from-port="2" to-layer="4186" to-port="0" /> + <edge from-layer="4179" from-port="2" to-layer="4384" to-port="0" /> + <edge from-layer="4179" from-port="2" to-layer="4194" to-port="0" /> + <edge from-layer="4180" from-port="0" to-layer="4181" to-port="0" /> + <edge from-layer="4181" from-port="1" to-layer="4195" to-port="0" /> + <edge from-layer="4182" from-port="0" to-layer="4183" to-port="0" /> + <edge from-layer="4183" from-port="1" to-layer="4193" to-port="0" /> + <edge from-layer="4184" from-port="0" to-layer="4185" to-port="0" /> + <edge from-layer="4185" from-port="1" to-layer="4186" to-port="1" /> + <edge from-layer="4186" from-port="2" to-layer="4188" to-port="0" /> + <edge from-layer="4187" from-port="0" to-layer="4188" to-port="1" /> + <edge from-layer="4188" from-port="2" to-layer="4191" to-port="0" /> + <edge from-layer="4189" from-port="0" to-layer="4190" to-port="0" /> + <edge from-layer="4190" from-port="1" to-layer="4191" to-port="1" /> + <edge from-layer="4191" from-port="2" to-layer="4192" to-port="0" /> + <edge from-layer="4192" from-port="1" to-layer="4193" to-port="1" /> + <edge from-layer="4193" from-port="2" to-layer="4194" to-port="1" /> + <edge from-layer="4194" from-port="2" to-layer="4195" to-port="1" /> + <edge from-layer="4195" from-port="2" to-layer="4204" to-port="0" /> + <edge from-layer="4195" from-port="2" to-layer="4205" to-port="0" /> + <edge from-layer="4195" from-port="2" to-layer="4334" to-port="0" /> + <edge from-layer="4195" from-port="2" to-layer="4266" to-port="0" /> + <edge from-layer="4196" from-port="0" to-layer="4197" to-port="0" /> + <edge from-layer="4197" from-port="1" to-layer="4200" to-port="0" /> + <edge from-layer="4198" from-port="0" to-layer="4199" to-port="0" /> + <edge from-layer="4199" from-port="1" to-layer="4200" to-port="1" /> + <edge from-layer="4200" from-port="2" to-layer="4202" to-port="0" /> + <edge from-layer="4201" from-port="0" to-layer="4202" to-port="1" /> + <edge from-layer="4202" from-port="2" to-layer="4203" to-port="0" /> + <edge from-layer="4203" from-port="1" to-layer="4204" to-port="1" /> + <edge from-layer="4204" from-port="2" to-layer="4210" to-port="0" /> + <edge from-layer="4205" from-port="1" to-layer="4208" to-port="0" /> + <edge from-layer="4206" from-port="0" to-layer="4208" to-port="1" /> + <edge from-layer="4207" from-port="0" to-layer="4208" to-port="2" /> + <edge from-layer="4208" from-port="3" to-layer="4209" to-port="0" /> + <edge from-layer="4208" from-port="3" to-layer="4373" to-port="0" /> + <edge from-layer="4208" from-port="3" to-layer="4267" to-port="0" /> + <edge from-layer="4209" from-port="3" to-layer="4210" to-port="1" /> + <edge from-layer="4210" from-port="2" to-layer="4212" to-port="0" /> + <edge from-layer="4211" from-port="0" to-layer="4212" to-port="1" /> + <edge from-layer="4212" from-port="2" to-layer="4242" to-port="0" /> + <edge from-layer="4212" from-port="2" to-layer="4222" to-port="0" /> + <edge from-layer="4212" from-port="2" to-layer="4225" to-port="0" /> + <edge from-layer="4212" from-port="2" to-layer="4236" to-port="0" /> + <edge from-layer="4213" from-port="0" to-layer="4214" to-port="0" /> + <edge from-layer="4214" from-port="1" to-layer="4215" to-port="0" /> + <edge from-layer="4215" from-port="2" to-layer="4216" to-port="0" /> + <edge from-layer="4216" from-port="2" to-layer="4218" to-port="0" /> + <edge from-layer="4217" from-port="0" to-layer="4218" to-port="1" /> + <edge from-layer="4218" from-port="2" to-layer="4219" to-port="1" /> + <edge from-layer="4218" from-port="2" to-layer="4219" to-port="0" /> + <edge from-layer="4219" from-port="2" to-layer="4244" to-port="0" /> + <edge from-layer="4219" from-port="2" to-layer="4220" to-port="0" /> + <edge from-layer="4220" from-port="1" to-layer="4221" to-port="0" /> + <edge from-layer="4221" from-port="2" to-layer="4222" to-port="1" /> + <edge from-layer="4221" from-port="2" to-layer="4271" to-port="1" /> + <edge from-layer="4222" from-port="2" to-layer="4247" to-port="0" /> + <edge from-layer="4223" from-port="0" to-layer="4233" to-port="0" /> + <edge from-layer="4224" from-port="0" to-layer="4233" to-port="1" /> + <edge from-layer="4225" from-port="1" to-layer="4227" to-port="0" /> + <edge from-layer="4226" from-port="0" to-layer="4227" to-port="2" /> + <edge from-layer="4227" from-port="3" to-layer="4228" to-port="0" /> + <edge from-layer="4228" from-port="2" to-layer="4229" to-port="0" /> + <edge from-layer="4229" from-port="1" to-layer="4231" to-port="0" /> + <edge from-layer="4230" from-port="0" to-layer="4231" to-port="1" /> + <edge from-layer="4231" from-port="2" to-layer="4233" to-port="2" /> + <edge from-layer="4232" from-port="0" to-layer="4233" to-port="3" /> + <edge from-layer="4233" from-port="4" to-layer="4236" to-port="1" /> + <edge from-layer="4233" from-port="4" to-layer="4242" to-port="2" /> + <edge from-layer="4234" from-port="0" to-layer="4236" to-port="2" /> + <edge from-layer="4235" from-port="0" to-layer="4236" to-port="3" /> + <edge from-layer="4236" from-port="4" to-layer="4239" to-port="0" /> + <edge from-layer="4237" from-port="0" to-layer="4238" to-port="0" /> + <edge from-layer="4238" from-port="1" to-layer="4239" to-port="1" /> + <edge from-layer="4239" from-port="2" to-layer="4243" to-port="0" /> + <edge from-layer="4240" from-port="0" to-layer="4242" to-port="1" /> + <edge from-layer="4241" from-port="0" to-layer="4242" to-port="3" /> + <edge from-layer="4242" from-port="4" to-layer="4243" to-port="1" /> + <edge from-layer="4243" from-port="2" to-layer="4246" to-port="0" /> + <edge from-layer="4244" from-port="1" to-layer="4245" to-port="0" /> + <edge from-layer="4245" from-port="2" to-layer="4293" to-port="1" /> + <edge from-layer="4245" from-port="2" to-layer="4246" to-port="1" /> + <edge from-layer="4246" from-port="2" to-layer="4247" to-port="1" /> + <edge from-layer="4247" from-port="2" to-layer="4370" to-port="0" /> + <edge from-layer="4248" from-port="0" to-layer="4249" to-port="0" /> + <edge from-layer="4249" from-port="1" to-layer="4254" to-port="0" /> + <edge from-layer="4250" from-port="0" to-layer="4253" to-port="1" /> + <edge from-layer="4251" from-port="0" to-layer="4253" to-port="2" /> + <edge from-layer="4252" from-port="0" to-layer="4253" to-port="3" /> + <edge from-layer="4253" from-port="4" to-layer="4254" to-port="1" /> + <edge from-layer="4254" from-port="2" to-layer="4255" to-port="0" /> + <edge from-layer="4255" from-port="1" to-layer="4257" to-port="0" /> + <edge from-layer="4256" from-port="0" to-layer="4257" to-port="2" /> + <edge from-layer="4257" from-port="3" to-layer="4295" to-port="0" /> + <edge from-layer="4258" from-port="0" to-layer="4259" to-port="0" /> + <edge from-layer="4259" from-port="1" to-layer="4262" to-port="0" /> + <edge from-layer="4260" from-port="0" to-layer="4261" to-port="0" /> + <edge from-layer="4261" from-port="1" to-layer="4262" to-port="1" /> + <edge from-layer="4262" from-port="2" to-layer="4264" to-port="0" /> + <edge from-layer="4263" from-port="0" to-layer="4264" to-port="1" /> + <edge from-layer="4264" from-port="2" to-layer="4265" to-port="0" /> + <edge from-layer="4265" from-port="1" to-layer="4266" to-port="1" /> + <edge from-layer="4266" from-port="2" to-layer="4268" to-port="0" /> + <edge from-layer="4267" from-port="3" to-layer="4268" to-port="1" /> + <edge from-layer="4267" from-port="3" to-layer="4335" to-port="1" /> + <edge from-layer="4268" from-port="2" to-layer="4270" to-port="0" /> + <edge from-layer="4269" from-port="0" to-layer="4270" to-port="1" /> + <edge from-layer="4270" from-port="2" to-layer="4274" to-port="0" /> + <edge from-layer="4270" from-port="2" to-layer="4285" to-port="0" /> + <edge from-layer="4270" from-port="2" to-layer="4291" to-port="0" /> + <edge from-layer="4270" from-port="2" to-layer="4271" to-port="0" /> + <edge from-layer="4271" from-port="2" to-layer="4294" to-port="0" /> + <edge from-layer="4272" from-port="0" to-layer="4282" to-port="0" /> + <edge from-layer="4273" from-port="0" to-layer="4282" to-port="1" /> + <edge from-layer="4274" from-port="1" to-layer="4276" to-port="0" /> + <edge from-layer="4275" from-port="0" to-layer="4276" to-port="2" /> + <edge from-layer="4276" from-port="3" to-layer="4277" to-port="0" /> + <edge from-layer="4277" from-port="2" to-layer="4278" to-port="0" /> + <edge from-layer="4278" from-port="1" to-layer="4280" to-port="0" /> + <edge from-layer="4279" from-port="0" to-layer="4280" to-port="1" /> + <edge from-layer="4280" from-port="2" to-layer="4282" to-port="2" /> + <edge from-layer="4281" from-port="0" to-layer="4282" to-port="3" /> + <edge from-layer="4282" from-port="4" to-layer="4291" to-port="2" /> + <edge from-layer="4282" from-port="4" to-layer="4285" to-port="1" /> + <edge from-layer="4283" from-port="0" to-layer="4285" to-port="2" /> + <edge from-layer="4284" from-port="0" to-layer="4285" to-port="3" /> + <edge from-layer="4285" from-port="4" to-layer="4288" to-port="0" /> + <edge from-layer="4286" from-port="0" to-layer="4287" to-port="0" /> + <edge from-layer="4287" from-port="1" to-layer="4288" to-port="1" /> + <edge from-layer="4288" from-port="2" to-layer="4292" to-port="0" /> + <edge from-layer="4289" from-port="0" to-layer="4291" to-port="1" /> + <edge from-layer="4290" from-port="0" to-layer="4291" to-port="3" /> + <edge from-layer="4291" from-port="4" to-layer="4292" to-port="1" /> + <edge from-layer="4292" from-port="2" to-layer="4293" to-port="0" /> + <edge from-layer="4293" from-port="2" to-layer="4294" to-port="1" /> + <edge from-layer="4294" from-port="2" to-layer="4295" to-port="1" /> + <edge from-layer="4295" from-port="2" to-layer="4296" to-port="0" /> + <edge from-layer="4295" from-port="2" to-layer="4297" to-port="0" /> + <edge from-layer="4295" from-port="2" to-layer="5538" to-port="0" /> + <edge from-layer="4296" from-port="2" to-layer="4310" to-port="0" /> + <edge from-layer="4297" from-port="1" to-layer="4307" to-port="0" /> + <edge from-layer="4297" from-port="1" to-layer="4302" to-port="0" /> + <edge from-layer="4297" from-port="1" to-layer="4300" to-port="0" /> + <edge from-layer="4298" from-port="0" to-layer="4300" to-port="1" /> + <edge from-layer="4299" from-port="0" to-layer="4300" to-port="2" /> + <edge from-layer="4300" from-port="3" to-layer="4314" to-port="0" /> + <edge from-layer="4300" from-port="3" to-layer="4308" to-port="0" /> + <edge from-layer="4301" from-port="0" to-layer="4302" to-port="2" /> + <edge from-layer="4302" from-port="3" to-layer="4311" to-port="0" /> + <edge from-layer="4302" from-port="3" to-layer="4304" to-port="0" /> + <edge from-layer="4303" from-port="0" to-layer="4304" to-port="1" /> + <edge from-layer="4304" from-port="2" to-layer="4308" to-port="1" /> + <edge from-layer="4305" from-port="0" to-layer="4307" to-port="1" /> + <edge from-layer="4306" from-port="0" to-layer="4307" to-port="2" /> + <edge from-layer="4307" from-port="3" to-layer="4308" to-port="3" /> + <edge from-layer="4307" from-port="3" to-layer="4314" to-port="2" /> + <edge from-layer="4308" from-port="4" to-layer="4309" to-port="0" /> + <edge from-layer="4309" from-port="1" to-layer="4310" to-port="1" /> + <edge from-layer="4310" from-port="2" to-layer="4315" to-port="0" /> + <edge from-layer="4311" from-port="2" to-layer="4313" to-port="0" /> + <edge from-layer="4312" from-port="0" to-layer="4313" to-port="1" /> + <edge from-layer="4313" from-port="2" to-layer="4314" to-port="1" /> + <edge from-layer="4314" from-port="3" to-layer="4315" to-port="1" /> + <edge from-layer="4315" from-port="2" to-layer="4370" to-port="1" /> + <edge from-layer="4315" from-port="2" to-layer="4362" to-port="0" /> + <edge from-layer="4316" from-port="0" to-layer="4317" to-port="0" /> + <edge from-layer="4317" from-port="1" to-layer="4322" to-port="0" /> + <edge from-layer="4318" from-port="0" to-layer="4321" to-port="1" /> + <edge from-layer="4319" from-port="0" to-layer="4321" to-port="2" /> + <edge from-layer="4320" from-port="0" to-layer="4321" to-port="3" /> + <edge from-layer="4321" from-port="4" to-layer="4322" to-port="1" /> + <edge from-layer="4322" from-port="2" to-layer="4323" to-port="0" /> + <edge from-layer="4323" from-port="1" to-layer="4325" to-port="0" /> + <edge from-layer="4324" from-port="0" to-layer="4325" to-port="2" /> + <edge from-layer="4325" from-port="3" to-layer="4338" to-port="0" /> + <edge from-layer="4326" from-port="0" to-layer="4327" to-port="0" /> + <edge from-layer="4327" from-port="1" to-layer="4330" to-port="0" /> + <edge from-layer="4328" from-port="0" to-layer="4329" to-port="0" /> + <edge from-layer="4329" from-port="1" to-layer="4330" to-port="1" /> + <edge from-layer="4330" from-port="2" to-layer="4332" to-port="0" /> + <edge from-layer="4331" from-port="0" to-layer="4332" to-port="1" /> + <edge from-layer="4332" from-port="2" to-layer="4333" to-port="0" /> + <edge from-layer="4333" from-port="1" to-layer="4334" to-port="1" /> + <edge from-layer="4334" from-port="2" to-layer="4335" to-port="0" /> + <edge from-layer="4335" from-port="2" to-layer="4337" to-port="0" /> + <edge from-layer="4336" from-port="0" to-layer="4337" to-port="1" /> + <edge from-layer="4337" from-port="2" to-layer="4338" to-port="1" /> + <edge from-layer="4338" from-port="2" to-layer="5537" to-port="0" /> + <edge from-layer="4338" from-port="2" to-layer="4339" to-port="0" /> + <edge from-layer="4338" from-port="2" to-layer="4340" to-port="0" /> + <edge from-layer="4339" from-port="2" to-layer="4353" to-port="0" /> + <edge from-layer="4340" from-port="1" to-layer="4345" to-port="0" /> + <edge from-layer="4340" from-port="1" to-layer="4343" to-port="0" /> + <edge from-layer="4340" from-port="1" to-layer="4350" to-port="0" /> + <edge from-layer="4341" from-port="0" to-layer="4343" to-port="1" /> + <edge from-layer="4342" from-port="0" to-layer="4343" to-port="2" /> + <edge from-layer="4343" from-port="3" to-layer="4351" to-port="0" /> + <edge from-layer="4343" from-port="3" to-layer="4357" to-port="0" /> + <edge from-layer="4344" from-port="0" to-layer="4345" to-port="2" /> + <edge from-layer="4345" from-port="3" to-layer="4347" to-port="0" /> + <edge from-layer="4345" from-port="3" to-layer="4354" to-port="0" /> + <edge from-layer="4346" from-port="0" to-layer="4347" to-port="1" /> + <edge from-layer="4347" from-port="2" to-layer="4351" to-port="1" /> + <edge from-layer="4348" from-port="0" to-layer="4350" to-port="1" /> + <edge from-layer="4349" from-port="0" to-layer="4350" to-port="2" /> + <edge from-layer="4350" from-port="3" to-layer="4357" to-port="2" /> + <edge from-layer="4350" from-port="3" to-layer="4351" to-port="3" /> + <edge from-layer="4351" from-port="4" to-layer="4352" to-port="0" /> + <edge from-layer="4352" from-port="1" to-layer="4353" to-port="1" /> + <edge from-layer="4353" from-port="2" to-layer="4358" to-port="0" /> + <edge from-layer="4354" from-port="2" to-layer="4356" to-port="0" /> + <edge from-layer="4355" from-port="0" to-layer="4356" to-port="1" /> + <edge from-layer="4356" from-port="2" to-layer="4357" to-port="1" /> + <edge from-layer="4357" from-port="3" to-layer="4358" to-port="1" /> + <edge from-layer="4358" from-port="2" to-layer="4370" to-port="2" /> + <edge from-layer="4359" from-port="0" to-layer="4369" to-port="1" /> + <edge from-layer="4360" from-port="0" to-layer="4367" to-port="0" /> + <edge from-layer="4361" from-port="0" to-layer="4367" to-port="1" /> + <edge from-layer="4362" from-port="1" to-layer="4365" to-port="0" /> + <edge from-layer="4363" from-port="0" to-layer="4365" to-port="1" /> + <edge from-layer="4364" from-port="0" to-layer="4365" to-port="2" /> + <edge from-layer="4365" from-port="3" to-layer="4367" to-port="2" /> + <edge from-layer="4366" from-port="0" to-layer="4367" to-port="3" /> + <edge from-layer="4367" from-port="4" to-layer="4369" to-port="2" /> + <edge from-layer="4368" from-port="0" to-layer="4369" to-port="3" /> + <edge from-layer="4369" from-port="4" to-layer="4370" to-port="3" /> + <edge from-layer="4370" from-port="4" to-layer="4372" to-port="0" /> + <edge from-layer="4371" from-port="0" to-layer="4372" to-port="1" /> + <edge from-layer="4372" from-port="2" to-layer="4374" to-port="0" /> + <edge from-layer="4373" from-port="2" to-layer="4374" to-port="1" /> + <edge from-layer="4374" from-port="2" to-layer="4383" to-port="0" /> + <edge from-layer="4375" from-port="0" to-layer="4376" to-port="0" /> + <edge from-layer="4376" from-port="1" to-layer="4379" to-port="0" /> + <edge from-layer="4377" from-port="0" to-layer="4378" to-port="0" /> + <edge from-layer="4378" from-port="1" to-layer="4379" to-port="1" /> + <edge from-layer="4379" from-port="2" to-layer="4381" to-port="0" /> + <edge from-layer="4380" from-port="0" to-layer="4381" to-port="1" /> + <edge from-layer="4381" from-port="2" to-layer="4382" to-port="0" /> + <edge from-layer="4382" from-port="1" to-layer="4383" to-port="1" /> + <edge from-layer="4383" from-port="2" to-layer="4384" to-port="1" /> + <edge from-layer="4384" from-port="2" to-layer="4399" to-port="0" /> + <edge from-layer="4384" from-port="2" to-layer="4430" to-port="0" /> + <edge from-layer="4384" from-port="2" to-layer="4391" to-port="0" /> + <edge from-layer="4385" from-port="0" to-layer="4386" to-port="0" /> + <edge from-layer="4386" from-port="1" to-layer="4400" to-port="0" /> + <edge from-layer="4387" from-port="0" to-layer="4388" to-port="0" /> + <edge from-layer="4388" from-port="1" to-layer="4398" to-port="0" /> + <edge from-layer="4389" from-port="0" to-layer="4390" to-port="0" /> + <edge from-layer="4390" from-port="1" to-layer="4391" to-port="1" /> + <edge from-layer="4391" from-port="2" to-layer="4393" to-port="0" /> + <edge from-layer="4392" from-port="0" to-layer="4393" to-port="1" /> + <edge from-layer="4393" from-port="2" to-layer="4396" to-port="0" /> + <edge from-layer="4394" from-port="0" to-layer="4395" to-port="0" /> + <edge from-layer="4395" from-port="1" to-layer="4396" to-port="1" /> + <edge from-layer="4396" from-port="2" to-layer="4397" to-port="0" /> + <edge from-layer="4397" from-port="1" to-layer="4398" to-port="1" /> + <edge from-layer="4398" from-port="2" to-layer="4399" to-port="1" /> + <edge from-layer="4399" from-port="2" to-layer="4400" to-port="1" /> + <edge from-layer="4400" from-port="2" to-layer="4419" to-port="0" /> + <edge from-layer="4400" from-port="2" to-layer="4409" to-port="0" /> + <edge from-layer="4401" from-port="0" to-layer="4402" to-port="0" /> + <edge from-layer="4402" from-port="1" to-layer="4405" to-port="0" /> + <edge from-layer="4403" from-port="0" to-layer="4404" to-port="0" /> + <edge from-layer="4404" from-port="1" to-layer="4405" to-port="1" /> + <edge from-layer="4405" from-port="2" to-layer="4407" to-port="0" /> + <edge from-layer="4406" from-port="0" to-layer="4407" to-port="1" /> + <edge from-layer="4407" from-port="2" to-layer="4408" to-port="0" /> + <edge from-layer="4408" from-port="1" to-layer="4409" to-port="1" /> + <edge from-layer="4409" from-port="2" to-layer="4410" to-port="0" /> + <edge from-layer="4410" from-port="1" to-layer="4420" to-port="0" /> + <edge from-layer="4411" from-port="0" to-layer="4412" to-port="0" /> + <edge from-layer="4412" from-port="1" to-layer="4415" to-port="0" /> + <edge from-layer="4413" from-port="0" to-layer="4414" to-port="0" /> + <edge from-layer="4414" from-port="1" to-layer="4415" to-port="1" /> + <edge from-layer="4415" from-port="2" to-layer="4417" to-port="0" /> + <edge from-layer="4416" from-port="0" to-layer="4417" to-port="1" /> + <edge from-layer="4417" from-port="2" to-layer="4418" to-port="0" /> + <edge from-layer="4418" from-port="1" to-layer="4419" to-port="1" /> + <edge from-layer="4419" from-port="2" to-layer="4420" to-port="1" /> + <edge from-layer="4420" from-port="2" to-layer="4429" to-port="0" /> + <edge from-layer="4421" from-port="0" to-layer="4422" to-port="0" /> + <edge from-layer="4422" from-port="1" to-layer="4425" to-port="0" /> + <edge from-layer="4423" from-port="0" to-layer="4424" to-port="0" /> + <edge from-layer="4424" from-port="1" to-layer="4425" to-port="1" /> + <edge from-layer="4425" from-port="2" to-layer="4427" to-port="0" /> + <edge from-layer="4426" from-port="0" to-layer="4427" to-port="1" /> + <edge from-layer="4427" from-port="2" to-layer="4428" to-port="0" /> + <edge from-layer="4428" from-port="1" to-layer="4429" to-port="1" /> + <edge from-layer="4429" from-port="2" to-layer="4430" to-port="1" /> + <edge from-layer="4430" from-port="2" to-layer="4445" to-port="0" /> + <edge from-layer="4430" from-port="2" to-layer="4635" to-port="0" /> + <edge from-layer="4430" from-port="2" to-layer="4437" to-port="0" /> + <edge from-layer="4431" from-port="0" to-layer="4432" to-port="0" /> + <edge from-layer="4432" from-port="1" to-layer="4446" to-port="0" /> + <edge from-layer="4433" from-port="0" to-layer="4434" to-port="0" /> + <edge from-layer="4434" from-port="1" to-layer="4444" to-port="0" /> + <edge from-layer="4435" from-port="0" to-layer="4436" to-port="0" /> + <edge from-layer="4436" from-port="1" to-layer="4437" to-port="1" /> + <edge from-layer="4437" from-port="2" to-layer="4439" to-port="0" /> + <edge from-layer="4438" from-port="0" to-layer="4439" to-port="1" /> + <edge from-layer="4439" from-port="2" to-layer="4442" to-port="0" /> + <edge from-layer="4440" from-port="0" to-layer="4441" to-port="0" /> + <edge from-layer="4441" from-port="1" to-layer="4442" to-port="1" /> + <edge from-layer="4442" from-port="2" to-layer="4443" to-port="0" /> + <edge from-layer="4443" from-port="1" to-layer="4444" to-port="1" /> + <edge from-layer="4444" from-port="2" to-layer="4445" to-port="1" /> + <edge from-layer="4445" from-port="2" to-layer="4446" to-port="1" /> + <edge from-layer="4446" from-port="2" to-layer="4585" to-port="0" /> + <edge from-layer="4446" from-port="2" to-layer="4517" to-port="0" /> + <edge from-layer="4446" from-port="2" to-layer="4455" to-port="0" /> + <edge from-layer="4446" from-port="2" to-layer="4456" to-port="0" /> + <edge from-layer="4447" from-port="0" to-layer="4448" to-port="0" /> + <edge from-layer="4448" from-port="1" to-layer="4451" to-port="0" /> + <edge from-layer="4449" from-port="0" to-layer="4450" to-port="0" /> + <edge from-layer="4450" from-port="1" to-layer="4451" to-port="1" /> + <edge from-layer="4451" from-port="2" to-layer="4453" to-port="0" /> + <edge from-layer="4452" from-port="0" to-layer="4453" to-port="1" /> + <edge from-layer="4453" from-port="2" to-layer="4454" to-port="0" /> + <edge from-layer="4454" from-port="1" to-layer="4455" to-port="1" /> + <edge from-layer="4455" from-port="2" to-layer="4461" to-port="0" /> + <edge from-layer="4456" from-port="1" to-layer="4459" to-port="0" /> + <edge from-layer="4457" from-port="0" to-layer="4459" to-port="1" /> + <edge from-layer="4458" from-port="0" to-layer="4459" to-port="2" /> + <edge from-layer="4459" from-port="3" to-layer="4460" to-port="0" /> + <edge from-layer="4459" from-port="3" to-layer="4518" to-port="0" /> + <edge from-layer="4459" from-port="3" to-layer="4624" to-port="0" /> + <edge from-layer="4460" from-port="3" to-layer="4461" to-port="1" /> + <edge from-layer="4461" from-port="2" to-layer="4463" to-port="0" /> + <edge from-layer="4462" from-port="0" to-layer="4463" to-port="1" /> + <edge from-layer="4463" from-port="2" to-layer="4476" to-port="0" /> + <edge from-layer="4463" from-port="2" to-layer="4473" to-port="0" /> + <edge from-layer="4463" from-port="2" to-layer="4487" to-port="0" /> + <edge from-layer="4463" from-port="2" to-layer="4493" to-port="0" /> + <edge from-layer="4464" from-port="0" to-layer="4465" to-port="0" /> + <edge from-layer="4465" from-port="1" to-layer="4466" to-port="0" /> + <edge from-layer="4466" from-port="2" to-layer="4467" to-port="0" /> + <edge from-layer="4467" from-port="2" to-layer="4469" to-port="0" /> + <edge from-layer="4468" from-port="0" to-layer="4469" to-port="1" /> + <edge from-layer="4469" from-port="2" to-layer="4470" to-port="1" /> + <edge from-layer="4469" from-port="2" to-layer="4470" to-port="0" /> + <edge from-layer="4470" from-port="2" to-layer="4471" to-port="0" /> + <edge from-layer="4470" from-port="2" to-layer="4495" to-port="0" /> + <edge from-layer="4471" from-port="1" to-layer="4472" to-port="0" /> + <edge from-layer="4472" from-port="2" to-layer="4473" to-port="1" /> + <edge from-layer="4472" from-port="2" to-layer="4522" to-port="1" /> + <edge from-layer="4473" from-port="2" to-layer="4498" to-port="0" /> + <edge from-layer="4474" from-port="0" to-layer="4484" to-port="0" /> + <edge from-layer="4475" from-port="0" to-layer="4484" to-port="1" /> + <edge from-layer="4476" from-port="1" to-layer="4478" to-port="0" /> + <edge from-layer="4477" from-port="0" to-layer="4478" to-port="2" /> + <edge from-layer="4478" from-port="3" to-layer="4479" to-port="0" /> + <edge from-layer="4479" from-port="2" to-layer="4480" to-port="0" /> + <edge from-layer="4480" from-port="1" to-layer="4482" to-port="0" /> + <edge from-layer="4481" from-port="0" to-layer="4482" to-port="1" /> + <edge from-layer="4482" from-port="2" to-layer="4484" to-port="2" /> + <edge from-layer="4483" from-port="0" to-layer="4484" to-port="3" /> + <edge from-layer="4484" from-port="4" to-layer="4487" to-port="1" /> + <edge from-layer="4484" from-port="4" to-layer="4493" to-port="2" /> + <edge from-layer="4485" from-port="0" to-layer="4487" to-port="2" /> + <edge from-layer="4486" from-port="0" to-layer="4487" to-port="3" /> + <edge from-layer="4487" from-port="4" to-layer="4490" to-port="0" /> + <edge from-layer="4488" from-port="0" to-layer="4489" to-port="0" /> + <edge from-layer="4489" from-port="1" to-layer="4490" to-port="1" /> + <edge from-layer="4490" from-port="2" to-layer="4494" to-port="0" /> + <edge from-layer="4491" from-port="0" to-layer="4493" to-port="1" /> + <edge from-layer="4492" from-port="0" to-layer="4493" to-port="3" /> + <edge from-layer="4493" from-port="4" to-layer="4494" to-port="1" /> + <edge from-layer="4494" from-port="2" to-layer="4497" to-port="0" /> + <edge from-layer="4495" from-port="1" to-layer="4496" to-port="0" /> + <edge from-layer="4496" from-port="2" to-layer="4544" to-port="1" /> + <edge from-layer="4496" from-port="2" to-layer="4497" to-port="1" /> + <edge from-layer="4497" from-port="2" to-layer="4498" to-port="1" /> + <edge from-layer="4498" from-port="2" to-layer="4621" to-port="0" /> + <edge from-layer="4499" from-port="0" to-layer="4500" to-port="0" /> + <edge from-layer="4500" from-port="1" to-layer="4505" to-port="0" /> + <edge from-layer="4501" from-port="0" to-layer="4504" to-port="1" /> + <edge from-layer="4502" from-port="0" to-layer="4504" to-port="2" /> + <edge from-layer="4503" from-port="0" to-layer="4504" to-port="3" /> + <edge from-layer="4504" from-port="4" to-layer="4505" to-port="1" /> + <edge from-layer="4505" from-port="2" to-layer="4506" to-port="0" /> + <edge from-layer="4506" from-port="1" to-layer="4508" to-port="0" /> + <edge from-layer="4507" from-port="0" to-layer="4508" to-port="2" /> + <edge from-layer="4508" from-port="3" to-layer="4546" to-port="0" /> + <edge from-layer="4509" from-port="0" to-layer="4510" to-port="0" /> + <edge from-layer="4510" from-port="1" to-layer="4513" to-port="0" /> + <edge from-layer="4511" from-port="0" to-layer="4512" to-port="0" /> + <edge from-layer="4512" from-port="1" to-layer="4513" to-port="1" /> + <edge from-layer="4513" from-port="2" to-layer="4515" to-port="0" /> + <edge from-layer="4514" from-port="0" to-layer="4515" to-port="1" /> + <edge from-layer="4515" from-port="2" to-layer="4516" to-port="0" /> + <edge from-layer="4516" from-port="1" to-layer="4517" to-port="1" /> + <edge from-layer="4517" from-port="2" to-layer="4519" to-port="0" /> + <edge from-layer="4518" from-port="3" to-layer="4519" to-port="1" /> + <edge from-layer="4518" from-port="3" to-layer="4586" to-port="1" /> + <edge from-layer="4519" from-port="2" to-layer="4521" to-port="0" /> + <edge from-layer="4520" from-port="0" to-layer="4521" to-port="1" /> + <edge from-layer="4521" from-port="2" to-layer="4536" to-port="0" /> + <edge from-layer="4521" from-port="2" to-layer="4542" to-port="0" /> + <edge from-layer="4521" from-port="2" to-layer="4525" to-port="0" /> + <edge from-layer="4521" from-port="2" to-layer="4522" to-port="0" /> + <edge from-layer="4522" from-port="2" to-layer="4545" to-port="0" /> + <edge from-layer="4523" from-port="0" to-layer="4533" to-port="0" /> + <edge from-layer="4524" from-port="0" to-layer="4533" to-port="1" /> + <edge from-layer="4525" from-port="1" to-layer="4527" to-port="0" /> + <edge from-layer="4526" from-port="0" to-layer="4527" to-port="2" /> + <edge from-layer="4527" from-port="3" to-layer="4528" to-port="0" /> + <edge from-layer="4528" from-port="2" to-layer="4529" to-port="0" /> + <edge from-layer="4529" from-port="1" to-layer="4531" to-port="0" /> + <edge from-layer="4530" from-port="0" to-layer="4531" to-port="1" /> + <edge from-layer="4531" from-port="2" to-layer="4533" to-port="2" /> + <edge from-layer="4532" from-port="0" to-layer="4533" to-port="3" /> + <edge from-layer="4533" from-port="4" to-layer="4542" to-port="2" /> + <edge from-layer="4533" from-port="4" to-layer="4536" to-port="1" /> + <edge from-layer="4534" from-port="0" to-layer="4536" to-port="2" /> + <edge from-layer="4535" from-port="0" to-layer="4536" to-port="3" /> + <edge from-layer="4536" from-port="4" to-layer="4539" to-port="0" /> + <edge from-layer="4537" from-port="0" to-layer="4538" to-port="0" /> + <edge from-layer="4538" from-port="1" to-layer="4539" to-port="1" /> + <edge from-layer="4539" from-port="2" to-layer="4543" to-port="0" /> + <edge from-layer="4540" from-port="0" to-layer="4542" to-port="1" /> + <edge from-layer="4541" from-port="0" to-layer="4542" to-port="3" /> + <edge from-layer="4542" from-port="4" to-layer="4543" to-port="1" /> + <edge from-layer="4543" from-port="2" to-layer="4544" to-port="0" /> + <edge from-layer="4544" from-port="2" to-layer="4545" to-port="1" /> + <edge from-layer="4545" from-port="2" to-layer="4546" to-port="1" /> + <edge from-layer="4546" from-port="2" to-layer="5536" to-port="0" /> + <edge from-layer="4546" from-port="2" to-layer="4548" to-port="0" /> + <edge from-layer="4546" from-port="2" to-layer="4547" to-port="0" /> + <edge from-layer="4547" from-port="2" to-layer="4561" to-port="0" /> + <edge from-layer="4548" from-port="1" to-layer="4558" to-port="0" /> + <edge from-layer="4548" from-port="1" to-layer="4553" to-port="0" /> + <edge from-layer="4548" from-port="1" to-layer="4551" to-port="0" /> + <edge from-layer="4549" from-port="0" to-layer="4551" to-port="1" /> + <edge from-layer="4550" from-port="0" to-layer="4551" to-port="2" /> + <edge from-layer="4551" from-port="3" to-layer="4559" to-port="0" /> + <edge from-layer="4551" from-port="3" to-layer="4565" to-port="0" /> + <edge from-layer="4552" from-port="0" to-layer="4553" to-port="2" /> + <edge from-layer="4553" from-port="3" to-layer="4555" to-port="0" /> + <edge from-layer="4553" from-port="3" to-layer="4562" to-port="0" /> + <edge from-layer="4554" from-port="0" to-layer="4555" to-port="1" /> + <edge from-layer="4555" from-port="2" to-layer="4559" to-port="1" /> + <edge from-layer="4556" from-port="0" to-layer="4558" to-port="1" /> + <edge from-layer="4557" from-port="0" to-layer="4558" to-port="2" /> + <edge from-layer="4558" from-port="3" to-layer="4559" to-port="3" /> + <edge from-layer="4558" from-port="3" to-layer="4565" to-port="2" /> + <edge from-layer="4559" from-port="4" to-layer="4560" to-port="0" /> + <edge from-layer="4560" from-port="1" to-layer="4561" to-port="1" /> + <edge from-layer="4561" from-port="2" to-layer="4566" to-port="0" /> + <edge from-layer="4562" from-port="2" to-layer="4564" to-port="0" /> + <edge from-layer="4563" from-port="0" to-layer="4564" to-port="1" /> + <edge from-layer="4564" from-port="2" to-layer="4565" to-port="1" /> + <edge from-layer="4565" from-port="3" to-layer="4566" to-port="1" /> + <edge from-layer="4566" from-port="2" to-layer="4613" to-port="0" /> + <edge from-layer="4566" from-port="2" to-layer="4621" to-port="1" /> + <edge from-layer="4567" from-port="0" to-layer="4568" to-port="0" /> + <edge from-layer="4568" from-port="1" to-layer="4573" to-port="0" /> + <edge from-layer="4569" from-port="0" to-layer="4572" to-port="1" /> + <edge from-layer="4570" from-port="0" to-layer="4572" to-port="2" /> + <edge from-layer="4571" from-port="0" to-layer="4572" to-port="3" /> + <edge from-layer="4572" from-port="4" to-layer="4573" to-port="1" /> + <edge from-layer="4573" from-port="2" to-layer="4574" to-port="0" /> + <edge from-layer="4574" from-port="1" to-layer="4576" to-port="0" /> + <edge from-layer="4575" from-port="0" to-layer="4576" to-port="2" /> + <edge from-layer="4576" from-port="3" to-layer="4589" to-port="0" /> + <edge from-layer="4577" from-port="0" to-layer="4578" to-port="0" /> + <edge from-layer="4578" from-port="1" to-layer="4581" to-port="0" /> + <edge from-layer="4579" from-port="0" to-layer="4580" to-port="0" /> + <edge from-layer="4580" from-port="1" to-layer="4581" to-port="1" /> + <edge from-layer="4581" from-port="2" to-layer="4583" to-port="0" /> + <edge from-layer="4582" from-port="0" to-layer="4583" to-port="1" /> + <edge from-layer="4583" from-port="2" to-layer="4584" to-port="0" /> + <edge from-layer="4584" from-port="1" to-layer="4585" to-port="1" /> + <edge from-layer="4585" from-port="2" to-layer="4586" to-port="0" /> + <edge from-layer="4586" from-port="2" to-layer="4588" to-port="0" /> + <edge from-layer="4587" from-port="0" to-layer="4588" to-port="1" /> + <edge from-layer="4588" from-port="2" to-layer="4589" to-port="1" /> + <edge from-layer="4589" from-port="2" to-layer="4590" to-port="0" /> + <edge from-layer="4589" from-port="2" to-layer="4591" to-port="0" /> + <edge from-layer="4589" from-port="2" to-layer="5535" to-port="0" /> + <edge from-layer="4590" from-port="2" to-layer="4604" to-port="0" /> + <edge from-layer="4591" from-port="1" to-layer="4601" to-port="0" /> + <edge from-layer="4591" from-port="1" to-layer="4596" to-port="0" /> + <edge from-layer="4591" from-port="1" to-layer="4594" to-port="0" /> + <edge from-layer="4592" from-port="0" to-layer="4594" to-port="1" /> + <edge from-layer="4593" from-port="0" to-layer="4594" to-port="2" /> + <edge from-layer="4594" from-port="3" to-layer="4602" to-port="0" /> + <edge from-layer="4594" from-port="3" to-layer="4608" to-port="0" /> + <edge from-layer="4595" from-port="0" to-layer="4596" to-port="2" /> + <edge from-layer="4596" from-port="3" to-layer="4605" to-port="0" /> + <edge from-layer="4596" from-port="3" to-layer="4598" to-port="0" /> + <edge from-layer="4597" from-port="0" to-layer="4598" to-port="1" /> + <edge from-layer="4598" from-port="2" to-layer="4602" to-port="1" /> + <edge from-layer="4599" from-port="0" to-layer="4601" to-port="1" /> + <edge from-layer="4600" from-port="0" to-layer="4601" to-port="2" /> + <edge from-layer="4601" from-port="3" to-layer="4608" to-port="2" /> + <edge from-layer="4601" from-port="3" to-layer="4602" to-port="3" /> + <edge from-layer="4602" from-port="4" to-layer="4603" to-port="0" /> + <edge from-layer="4603" from-port="1" to-layer="4604" to-port="1" /> + <edge from-layer="4604" from-port="2" to-layer="4609" to-port="0" /> + <edge from-layer="4605" from-port="2" to-layer="4607" to-port="0" /> + <edge from-layer="4606" from-port="0" to-layer="4607" to-port="1" /> + <edge from-layer="4607" from-port="2" to-layer="4608" to-port="1" /> + <edge from-layer="4608" from-port="3" to-layer="4609" to-port="1" /> + <edge from-layer="4609" from-port="2" to-layer="4621" to-port="2" /> + <edge from-layer="4610" from-port="0" to-layer="4620" to-port="1" /> + <edge from-layer="4611" from-port="0" to-layer="4618" to-port="0" /> + <edge from-layer="4612" from-port="0" to-layer="4618" to-port="1" /> + <edge from-layer="4613" from-port="1" to-layer="4616" to-port="0" /> + <edge from-layer="4614" from-port="0" to-layer="4616" to-port="1" /> + <edge from-layer="4615" from-port="0" to-layer="4616" to-port="2" /> + <edge from-layer="4616" from-port="3" to-layer="4618" to-port="2" /> + <edge from-layer="4617" from-port="0" to-layer="4618" to-port="3" /> + <edge from-layer="4618" from-port="4" to-layer="4620" to-port="2" /> + <edge from-layer="4619" from-port="0" to-layer="4620" to-port="3" /> + <edge from-layer="4620" from-port="4" to-layer="4621" to-port="3" /> + <edge from-layer="4621" from-port="4" to-layer="4623" to-port="0" /> + <edge from-layer="4622" from-port="0" to-layer="4623" to-port="1" /> + <edge from-layer="4623" from-port="2" to-layer="4625" to-port="0" /> + <edge from-layer="4624" from-port="2" to-layer="4625" to-port="1" /> + <edge from-layer="4625" from-port="2" to-layer="4634" to-port="0" /> + <edge from-layer="4626" from-port="0" to-layer="4627" to-port="0" /> + <edge from-layer="4627" from-port="1" to-layer="4630" to-port="0" /> + <edge from-layer="4628" from-port="0" to-layer="4629" to-port="0" /> + <edge from-layer="4629" from-port="1" to-layer="4630" to-port="1" /> + <edge from-layer="4630" from-port="2" to-layer="4632" to-port="0" /> + <edge from-layer="4631" from-port="0" to-layer="4632" to-port="1" /> + <edge from-layer="4632" from-port="2" to-layer="4633" to-port="0" /> + <edge from-layer="4633" from-port="1" to-layer="4634" to-port="1" /> + <edge from-layer="4634" from-port="2" to-layer="4635" to-port="1" /> + <edge from-layer="4635" from-port="2" to-layer="4650" to-port="0" /> + <edge from-layer="4635" from-port="2" to-layer="4681" to-port="0" /> + <edge from-layer="4635" from-port="2" to-layer="4642" to-port="0" /> + <edge from-layer="4636" from-port="0" to-layer="4637" to-port="0" /> + <edge from-layer="4637" from-port="1" to-layer="4651" to-port="0" /> + <edge from-layer="4638" from-port="0" to-layer="4639" to-port="0" /> + <edge from-layer="4639" from-port="1" to-layer="4649" to-port="0" /> + <edge from-layer="4640" from-port="0" to-layer="4641" to-port="0" /> + <edge from-layer="4641" from-port="1" to-layer="4642" to-port="1" /> + <edge from-layer="4642" from-port="2" to-layer="4644" to-port="0" /> + <edge from-layer="4643" from-port="0" to-layer="4644" to-port="1" /> + <edge from-layer="4644" from-port="2" to-layer="4647" to-port="0" /> + <edge from-layer="4645" from-port="0" to-layer="4646" to-port="0" /> + <edge from-layer="4646" from-port="1" to-layer="4647" to-port="1" /> + <edge from-layer="4647" from-port="2" to-layer="4648" to-port="0" /> + <edge from-layer="4648" from-port="1" to-layer="4649" to-port="1" /> + <edge from-layer="4649" from-port="2" to-layer="4650" to-port="1" /> + <edge from-layer="4650" from-port="2" to-layer="4651" to-port="1" /> + <edge from-layer="4651" from-port="2" to-layer="4670" to-port="0" /> + <edge from-layer="4651" from-port="2" to-layer="4660" to-port="0" /> + <edge from-layer="4652" from-port="0" to-layer="4653" to-port="0" /> + <edge from-layer="4653" from-port="1" to-layer="4656" to-port="0" /> + <edge from-layer="4654" from-port="0" to-layer="4655" to-port="0" /> + <edge from-layer="4655" from-port="1" to-layer="4656" to-port="1" /> + <edge from-layer="4656" from-port="2" to-layer="4658" to-port="0" /> + <edge from-layer="4657" from-port="0" to-layer="4658" to-port="1" /> + <edge from-layer="4658" from-port="2" to-layer="4659" to-port="0" /> + <edge from-layer="4659" from-port="1" to-layer="4660" to-port="1" /> + <edge from-layer="4660" from-port="2" to-layer="4661" to-port="0" /> + <edge from-layer="4661" from-port="1" to-layer="4671" to-port="0" /> + <edge from-layer="4662" from-port="0" to-layer="4663" to-port="0" /> + <edge from-layer="4663" from-port="1" to-layer="4666" to-port="0" /> + <edge from-layer="4664" from-port="0" to-layer="4665" to-port="0" /> + <edge from-layer="4665" from-port="1" to-layer="4666" to-port="1" /> + <edge from-layer="4666" from-port="2" to-layer="4668" to-port="0" /> + <edge from-layer="4667" from-port="0" to-layer="4668" to-port="1" /> + <edge from-layer="4668" from-port="2" to-layer="4669" to-port="0" /> + <edge from-layer="4669" from-port="1" to-layer="4670" to-port="1" /> + <edge from-layer="4670" from-port="2" to-layer="4671" to-port="1" /> + <edge from-layer="4671" from-port="2" to-layer="4680" to-port="0" /> + <edge from-layer="4672" from-port="0" to-layer="4673" to-port="0" /> + <edge from-layer="4673" from-port="1" to-layer="4676" to-port="0" /> + <edge from-layer="4674" from-port="0" to-layer="4675" to-port="0" /> + <edge from-layer="4675" from-port="1" to-layer="4676" to-port="1" /> + <edge from-layer="4676" from-port="2" to-layer="4678" to-port="0" /> + <edge from-layer="4677" from-port="0" to-layer="4678" to-port="1" /> + <edge from-layer="4678" from-port="2" to-layer="4679" to-port="0" /> + <edge from-layer="4679" from-port="1" to-layer="4680" to-port="1" /> + <edge from-layer="4680" from-port="2" to-layer="4681" to-port="1" /> + <edge from-layer="4681" from-port="2" to-layer="4688" to-port="0" /> + <edge from-layer="4681" from-port="2" to-layer="4696" to-port="0" /> + <edge from-layer="4681" from-port="2" to-layer="4886" to-port="0" /> + <edge from-layer="4682" from-port="0" to-layer="4683" to-port="0" /> + <edge from-layer="4683" from-port="1" to-layer="4697" to-port="0" /> + <edge from-layer="4684" from-port="0" to-layer="4685" to-port="0" /> + <edge from-layer="4685" from-port="1" to-layer="4695" to-port="0" /> + <edge from-layer="4686" from-port="0" to-layer="4687" to-port="0" /> + <edge from-layer="4687" from-port="1" to-layer="4688" to-port="1" /> + <edge from-layer="4688" from-port="2" to-layer="4690" to-port="0" /> + <edge from-layer="4689" from-port="0" to-layer="4690" to-port="1" /> + <edge from-layer="4690" from-port="2" to-layer="4693" to-port="0" /> + <edge from-layer="4691" from-port="0" to-layer="4692" to-port="0" /> + <edge from-layer="4692" from-port="1" to-layer="4693" to-port="1" /> + <edge from-layer="4693" from-port="2" to-layer="4694" to-port="0" /> + <edge from-layer="4694" from-port="1" to-layer="4695" to-port="1" /> + <edge from-layer="4695" from-port="2" to-layer="4696" to-port="1" /> + <edge from-layer="4696" from-port="2" to-layer="4697" to-port="1" /> + <edge from-layer="4697" from-port="2" to-layer="4836" to-port="0" /> + <edge from-layer="4697" from-port="2" to-layer="4707" to-port="0" /> + <edge from-layer="4697" from-port="2" to-layer="4706" to-port="0" /> + <edge from-layer="4697" from-port="2" to-layer="4768" to-port="0" /> + <edge from-layer="4698" from-port="0" to-layer="4699" to-port="0" /> + <edge from-layer="4699" from-port="1" to-layer="4702" to-port="0" /> + <edge from-layer="4700" from-port="0" to-layer="4701" to-port="0" /> + <edge from-layer="4701" from-port="1" to-layer="4702" to-port="1" /> + <edge from-layer="4702" from-port="2" to-layer="4704" to-port="0" /> + <edge from-layer="4703" from-port="0" to-layer="4704" to-port="1" /> + <edge from-layer="4704" from-port="2" to-layer="4705" to-port="0" /> + <edge from-layer="4705" from-port="1" to-layer="4706" to-port="1" /> + <edge from-layer="4706" from-port="2" to-layer="4712" to-port="0" /> + <edge from-layer="4707" from-port="1" to-layer="4710" to-port="0" /> + <edge from-layer="4708" from-port="0" to-layer="4710" to-port="1" /> + <edge from-layer="4709" from-port="0" to-layer="4710" to-port="2" /> + <edge from-layer="4710" from-port="3" to-layer="4875" to-port="0" /> + <edge from-layer="4710" from-port="3" to-layer="4769" to-port="0" /> + <edge from-layer="4710" from-port="3" to-layer="4711" to-port="0" /> + <edge from-layer="4711" from-port="3" to-layer="4712" to-port="1" /> + <edge from-layer="4712" from-port="2" to-layer="4714" to-port="0" /> + <edge from-layer="4713" from-port="0" to-layer="4714" to-port="1" /> + <edge from-layer="4714" from-port="2" to-layer="4738" to-port="0" /> + <edge from-layer="4714" from-port="2" to-layer="4724" to-port="0" /> + <edge from-layer="4714" from-port="2" to-layer="4727" to-port="0" /> + <edge from-layer="4714" from-port="2" to-layer="4744" to-port="0" /> + <edge from-layer="4715" from-port="0" to-layer="4716" to-port="0" /> + <edge from-layer="4716" from-port="1" to-layer="4717" to-port="0" /> + <edge from-layer="4717" from-port="2" to-layer="4718" to-port="0" /> + <edge from-layer="4718" from-port="2" to-layer="4720" to-port="0" /> + <edge from-layer="4719" from-port="0" to-layer="4720" to-port="1" /> + <edge from-layer="4720" from-port="2" to-layer="4721" to-port="0" /> + <edge from-layer="4720" from-port="2" to-layer="4721" to-port="1" /> + <edge from-layer="4721" from-port="2" to-layer="4746" to-port="0" /> + <edge from-layer="4721" from-port="2" to-layer="4722" to-port="0" /> + <edge from-layer="4722" from-port="1" to-layer="4723" to-port="0" /> + <edge from-layer="4723" from-port="2" to-layer="4773" to-port="1" /> + <edge from-layer="4723" from-port="2" to-layer="4724" to-port="1" /> + <edge from-layer="4724" from-port="2" to-layer="4749" to-port="0" /> + <edge from-layer="4725" from-port="0" to-layer="4735" to-port="0" /> + <edge from-layer="4726" from-port="0" to-layer="4735" to-port="1" /> + <edge from-layer="4727" from-port="1" to-layer="4729" to-port="0" /> + <edge from-layer="4728" from-port="0" to-layer="4729" to-port="2" /> + <edge from-layer="4729" from-port="3" to-layer="4730" to-port="0" /> + <edge from-layer="4730" from-port="2" to-layer="4731" to-port="0" /> + <edge from-layer="4731" from-port="1" to-layer="4733" to-port="0" /> + <edge from-layer="4732" from-port="0" to-layer="4733" to-port="1" /> + <edge from-layer="4733" from-port="2" to-layer="4735" to-port="2" /> + <edge from-layer="4734" from-port="0" to-layer="4735" to-port="3" /> + <edge from-layer="4735" from-port="4" to-layer="4744" to-port="2" /> + <edge from-layer="4735" from-port="4" to-layer="4738" to-port="1" /> + <edge from-layer="4736" from-port="0" to-layer="4738" to-port="2" /> + <edge from-layer="4737" from-port="0" to-layer="4738" to-port="3" /> + <edge from-layer="4738" from-port="4" to-layer="4741" to-port="0" /> + <edge from-layer="4739" from-port="0" to-layer="4740" to-port="0" /> + <edge from-layer="4740" from-port="1" to-layer="4741" to-port="1" /> + <edge from-layer="4741" from-port="2" to-layer="4745" to-port="0" /> + <edge from-layer="4742" from-port="0" to-layer="4744" to-port="1" /> + <edge from-layer="4743" from-port="0" to-layer="4744" to-port="3" /> + <edge from-layer="4744" from-port="4" to-layer="4745" to-port="1" /> + <edge from-layer="4745" from-port="2" to-layer="4748" to-port="0" /> + <edge from-layer="4746" from-port="1" to-layer="4747" to-port="0" /> + <edge from-layer="4747" from-port="2" to-layer="4748" to-port="1" /> + <edge from-layer="4747" from-port="2" to-layer="4795" to-port="1" /> + <edge from-layer="4748" from-port="2" to-layer="4749" to-port="1" /> + <edge from-layer="4749" from-port="2" to-layer="4872" to-port="0" /> + <edge from-layer="4750" from-port="0" to-layer="4751" to-port="0" /> + <edge from-layer="4751" from-port="1" to-layer="4756" to-port="0" /> + <edge from-layer="4752" from-port="0" to-layer="4755" to-port="1" /> + <edge from-layer="4753" from-port="0" to-layer="4755" to-port="2" /> + <edge from-layer="4754" from-port="0" to-layer="4755" to-port="3" /> + <edge from-layer="4755" from-port="4" to-layer="4756" to-port="1" /> + <edge from-layer="4756" from-port="2" to-layer="4757" to-port="0" /> + <edge from-layer="4757" from-port="1" to-layer="4759" to-port="0" /> + <edge from-layer="4758" from-port="0" to-layer="4759" to-port="2" /> + <edge from-layer="4759" from-port="3" to-layer="4797" to-port="0" /> + <edge from-layer="4760" from-port="0" to-layer="4761" to-port="0" /> + <edge from-layer="4761" from-port="1" to-layer="4764" to-port="0" /> + <edge from-layer="4762" from-port="0" to-layer="4763" to-port="0" /> + <edge from-layer="4763" from-port="1" to-layer="4764" to-port="1" /> + <edge from-layer="4764" from-port="2" to-layer="4766" to-port="0" /> + <edge from-layer="4765" from-port="0" to-layer="4766" to-port="1" /> + <edge from-layer="4766" from-port="2" to-layer="4767" to-port="0" /> + <edge from-layer="4767" from-port="1" to-layer="4768" to-port="1" /> + <edge from-layer="4768" from-port="2" to-layer="4770" to-port="0" /> + <edge from-layer="4769" from-port="3" to-layer="4770" to-port="1" /> + <edge from-layer="4769" from-port="3" to-layer="4837" to-port="1" /> + <edge from-layer="4770" from-port="2" to-layer="4772" to-port="0" /> + <edge from-layer="4771" from-port="0" to-layer="4772" to-port="1" /> + <edge from-layer="4772" from-port="2" to-layer="4776" to-port="0" /> + <edge from-layer="4772" from-port="2" to-layer="4773" to-port="0" /> + <edge from-layer="4772" from-port="2" to-layer="4787" to-port="0" /> + <edge from-layer="4772" from-port="2" to-layer="4793" to-port="0" /> + <edge from-layer="4773" from-port="2" to-layer="4796" to-port="0" /> + <edge from-layer="4774" from-port="0" to-layer="4784" to-port="0" /> + <edge from-layer="4775" from-port="0" to-layer="4784" to-port="1" /> + <edge from-layer="4776" from-port="1" to-layer="4778" to-port="0" /> + <edge from-layer="4777" from-port="0" to-layer="4778" to-port="2" /> + <edge from-layer="4778" from-port="3" to-layer="4779" to-port="0" /> + <edge from-layer="4779" from-port="2" to-layer="4780" to-port="0" /> + <edge from-layer="4780" from-port="1" to-layer="4782" to-port="0" /> + <edge from-layer="4781" from-port="0" to-layer="4782" to-port="1" /> + <edge from-layer="4782" from-port="2" to-layer="4784" to-port="2" /> + <edge from-layer="4783" from-port="0" to-layer="4784" to-port="3" /> + <edge from-layer="4784" from-port="4" to-layer="4793" to-port="2" /> + <edge from-layer="4784" from-port="4" to-layer="4787" to-port="1" /> + <edge from-layer="4785" from-port="0" to-layer="4787" to-port="2" /> + <edge from-layer="4786" from-port="0" to-layer="4787" to-port="3" /> + <edge from-layer="4787" from-port="4" to-layer="4790" to-port="0" /> + <edge from-layer="4788" from-port="0" to-layer="4789" to-port="0" /> + <edge from-layer="4789" from-port="1" to-layer="4790" to-port="1" /> + <edge from-layer="4790" from-port="2" to-layer="4794" to-port="0" /> + <edge from-layer="4791" from-port="0" to-layer="4793" to-port="1" /> + <edge from-layer="4792" from-port="0" to-layer="4793" to-port="3" /> + <edge from-layer="4793" from-port="4" to-layer="4794" to-port="1" /> + <edge from-layer="4794" from-port="2" to-layer="4795" to-port="0" /> + <edge from-layer="4795" from-port="2" to-layer="4796" to-port="1" /> + <edge from-layer="4796" from-port="2" to-layer="4797" to-port="1" /> + <edge from-layer="4797" from-port="2" to-layer="4798" to-port="0" /> + <edge from-layer="4797" from-port="2" to-layer="4799" to-port="0" /> + <edge from-layer="4797" from-port="2" to-layer="5534" to-port="0" /> + <edge from-layer="4798" from-port="2" to-layer="4812" to-port="0" /> + <edge from-layer="4799" from-port="1" to-layer="4802" to-port="0" /> + <edge from-layer="4799" from-port="1" to-layer="4804" to-port="0" /> + <edge from-layer="4799" from-port="1" to-layer="4809" to-port="0" /> + <edge from-layer="4800" from-port="0" to-layer="4802" to-port="1" /> + <edge from-layer="4801" from-port="0" to-layer="4802" to-port="2" /> + <edge from-layer="4802" from-port="3" to-layer="4816" to-port="0" /> + <edge from-layer="4802" from-port="3" to-layer="4810" to-port="0" /> + <edge from-layer="4803" from-port="0" to-layer="4804" to-port="2" /> + <edge from-layer="4804" from-port="3" to-layer="4806" to-port="0" /> + <edge from-layer="4804" from-port="3" to-layer="4813" to-port="0" /> + <edge from-layer="4805" from-port="0" to-layer="4806" to-port="1" /> + <edge from-layer="4806" from-port="2" to-layer="4810" to-port="1" /> + <edge from-layer="4807" from-port="0" to-layer="4809" to-port="1" /> + <edge from-layer="4808" from-port="0" to-layer="4809" to-port="2" /> + <edge from-layer="4809" from-port="3" to-layer="4810" to-port="3" /> + <edge from-layer="4809" from-port="3" to-layer="4816" to-port="2" /> + <edge from-layer="4810" from-port="4" to-layer="4811" to-port="0" /> + <edge from-layer="4811" from-port="1" to-layer="4812" to-port="1" /> + <edge from-layer="4812" from-port="2" to-layer="4817" to-port="0" /> + <edge from-layer="4813" from-port="2" to-layer="4815" to-port="0" /> + <edge from-layer="4814" from-port="0" to-layer="4815" to-port="1" /> + <edge from-layer="4815" from-port="2" to-layer="4816" to-port="1" /> + <edge from-layer="4816" from-port="3" to-layer="4817" to-port="1" /> + <edge from-layer="4817" from-port="2" to-layer="4872" to-port="1" /> + <edge from-layer="4817" from-port="2" to-layer="4864" to-port="0" /> + <edge from-layer="4818" from-port="0" to-layer="4819" to-port="0" /> + <edge from-layer="4819" from-port="1" to-layer="4824" to-port="0" /> + <edge from-layer="4820" from-port="0" to-layer="4823" to-port="1" /> + <edge from-layer="4821" from-port="0" to-layer="4823" to-port="2" /> + <edge from-layer="4822" from-port="0" to-layer="4823" to-port="3" /> + <edge from-layer="4823" from-port="4" to-layer="4824" to-port="1" /> + <edge from-layer="4824" from-port="2" to-layer="4825" to-port="0" /> + <edge from-layer="4825" from-port="1" to-layer="4827" to-port="0" /> + <edge from-layer="4826" from-port="0" to-layer="4827" to-port="2" /> + <edge from-layer="4827" from-port="3" to-layer="4840" to-port="0" /> + <edge from-layer="4828" from-port="0" to-layer="4829" to-port="0" /> + <edge from-layer="4829" from-port="1" to-layer="4832" to-port="0" /> + <edge from-layer="4830" from-port="0" to-layer="4831" to-port="0" /> + <edge from-layer="4831" from-port="1" to-layer="4832" to-port="1" /> + <edge from-layer="4832" from-port="2" to-layer="4834" to-port="0" /> + <edge from-layer="4833" from-port="0" to-layer="4834" to-port="1" /> + <edge from-layer="4834" from-port="2" to-layer="4835" to-port="0" /> + <edge from-layer="4835" from-port="1" to-layer="4836" to-port="1" /> + <edge from-layer="4836" from-port="2" to-layer="4837" to-port="0" /> + <edge from-layer="4837" from-port="2" to-layer="4839" to-port="0" /> + <edge from-layer="4838" from-port="0" to-layer="4839" to-port="1" /> + <edge from-layer="4839" from-port="2" to-layer="4840" to-port="1" /> + <edge from-layer="4840" from-port="2" to-layer="5533" to-port="0" /> + <edge from-layer="4840" from-port="2" to-layer="4842" to-port="0" /> + <edge from-layer="4840" from-port="2" to-layer="4841" to-port="0" /> + <edge from-layer="4841" from-port="2" to-layer="4855" to-port="0" /> + <edge from-layer="4842" from-port="1" to-layer="4845" to-port="0" /> + <edge from-layer="4842" from-port="1" to-layer="4847" to-port="0" /> + <edge from-layer="4842" from-port="1" to-layer="4852" to-port="0" /> + <edge from-layer="4843" from-port="0" to-layer="4845" to-port="1" /> + <edge from-layer="4844" from-port="0" to-layer="4845" to-port="2" /> + <edge from-layer="4845" from-port="3" to-layer="4859" to-port="0" /> + <edge from-layer="4845" from-port="3" to-layer="4853" to-port="0" /> + <edge from-layer="4846" from-port="0" to-layer="4847" to-port="2" /> + <edge from-layer="4847" from-port="3" to-layer="4856" to-port="0" /> + <edge from-layer="4847" from-port="3" to-layer="4849" to-port="0" /> + <edge from-layer="4848" from-port="0" to-layer="4849" to-port="1" /> + <edge from-layer="4849" from-port="2" to-layer="4853" to-port="1" /> + <edge from-layer="4850" from-port="0" to-layer="4852" to-port="1" /> + <edge from-layer="4851" from-port="0" to-layer="4852" to-port="2" /> + <edge from-layer="4852" from-port="3" to-layer="4859" to-port="2" /> + <edge from-layer="4852" from-port="3" to-layer="4853" to-port="3" /> + <edge from-layer="4853" from-port="4" to-layer="4854" to-port="0" /> + <edge from-layer="4854" from-port="1" to-layer="4855" to-port="1" /> + <edge from-layer="4855" from-port="2" to-layer="4860" to-port="0" /> + <edge from-layer="4856" from-port="2" to-layer="4858" to-port="0" /> + <edge from-layer="4857" from-port="0" to-layer="4858" to-port="1" /> + <edge from-layer="4858" from-port="2" to-layer="4859" to-port="1" /> + <edge from-layer="4859" from-port="3" to-layer="4860" to-port="1" /> + <edge from-layer="4860" from-port="2" to-layer="4872" to-port="2" /> + <edge from-layer="4861" from-port="0" to-layer="4871" to-port="1" /> + <edge from-layer="4862" from-port="0" to-layer="4869" to-port="0" /> + <edge from-layer="4863" from-port="0" to-layer="4869" to-port="1" /> + <edge from-layer="4864" from-port="1" to-layer="4867" to-port="0" /> + <edge from-layer="4865" from-port="0" to-layer="4867" to-port="1" /> + <edge from-layer="4866" from-port="0" to-layer="4867" to-port="2" /> + <edge from-layer="4867" from-port="3" to-layer="4869" to-port="2" /> + <edge from-layer="4868" from-port="0" to-layer="4869" to-port="3" /> + <edge from-layer="4869" from-port="4" to-layer="4871" to-port="2" /> + <edge from-layer="4870" from-port="0" to-layer="4871" to-port="3" /> + <edge from-layer="4871" from-port="4" to-layer="4872" to-port="3" /> + <edge from-layer="4872" from-port="4" to-layer="4874" to-port="0" /> + <edge from-layer="4873" from-port="0" to-layer="4874" to-port="1" /> + <edge from-layer="4874" from-port="2" to-layer="4876" to-port="0" /> + <edge from-layer="4875" from-port="2" to-layer="4876" to-port="1" /> + <edge from-layer="4876" from-port="2" to-layer="4885" to-port="0" /> + <edge from-layer="4877" from-port="0" to-layer="4878" to-port="0" /> + <edge from-layer="4878" from-port="1" to-layer="4881" to-port="0" /> + <edge from-layer="4879" from-port="0" to-layer="4880" to-port="0" /> + <edge from-layer="4880" from-port="1" to-layer="4881" to-port="1" /> + <edge from-layer="4881" from-port="2" to-layer="4883" to-port="0" /> + <edge from-layer="4882" from-port="0" to-layer="4883" to-port="1" /> + <edge from-layer="4883" from-port="2" to-layer="4884" to-port="0" /> + <edge from-layer="4884" from-port="1" to-layer="4885" to-port="1" /> + <edge from-layer="4885" from-port="2" to-layer="4886" to-port="1" /> + <edge from-layer="4886" from-port="2" to-layer="4893" to-port="0" /> + <edge from-layer="4886" from-port="2" to-layer="4901" to-port="0" /> + <edge from-layer="4886" from-port="2" to-layer="4932" to-port="0" /> + <edge from-layer="4887" from-port="0" to-layer="4888" to-port="0" /> + <edge from-layer="4888" from-port="1" to-layer="4902" to-port="0" /> + <edge from-layer="4889" from-port="0" to-layer="4890" to-port="0" /> + <edge from-layer="4890" from-port="1" to-layer="4900" to-port="0" /> + <edge from-layer="4891" from-port="0" to-layer="4892" to-port="0" /> + <edge from-layer="4892" from-port="1" to-layer="4893" to-port="1" /> + <edge from-layer="4893" from-port="2" to-layer="4895" to-port="0" /> + <edge from-layer="4894" from-port="0" to-layer="4895" to-port="1" /> + <edge from-layer="4895" from-port="2" to-layer="4898" to-port="0" /> + <edge from-layer="4896" from-port="0" to-layer="4897" to-port="0" /> + <edge from-layer="4897" from-port="1" to-layer="4898" to-port="1" /> + <edge from-layer="4898" from-port="2" to-layer="4899" to-port="0" /> + <edge from-layer="4899" from-port="1" to-layer="4900" to-port="1" /> + <edge from-layer="4900" from-port="2" to-layer="4901" to-port="1" /> + <edge from-layer="4901" from-port="2" to-layer="4902" to-port="1" /> + <edge from-layer="4902" from-port="2" to-layer="4921" to-port="0" /> + <edge from-layer="4902" from-port="2" to-layer="4911" to-port="0" /> + <edge from-layer="4903" from-port="0" to-layer="4904" to-port="0" /> + <edge from-layer="4904" from-port="1" to-layer="4907" to-port="0" /> + <edge from-layer="4905" from-port="0" to-layer="4906" to-port="0" /> + <edge from-layer="4906" from-port="1" to-layer="4907" to-port="1" /> + <edge from-layer="4907" from-port="2" to-layer="4909" to-port="0" /> + <edge from-layer="4908" from-port="0" to-layer="4909" to-port="1" /> + <edge from-layer="4909" from-port="2" to-layer="4910" to-port="0" /> + <edge from-layer="4910" from-port="1" to-layer="4911" to-port="1" /> + <edge from-layer="4911" from-port="2" to-layer="4912" to-port="0" /> + <edge from-layer="4912" from-port="1" to-layer="4922" to-port="0" /> + <edge from-layer="4913" from-port="0" to-layer="4914" to-port="0" /> + <edge from-layer="4914" from-port="1" to-layer="4917" to-port="0" /> + <edge from-layer="4915" from-port="0" to-layer="4916" to-port="0" /> + <edge from-layer="4916" from-port="1" to-layer="4917" to-port="1" /> + <edge from-layer="4917" from-port="2" to-layer="4919" to-port="0" /> + <edge from-layer="4918" from-port="0" to-layer="4919" to-port="1" /> + <edge from-layer="4919" from-port="2" to-layer="4920" to-port="0" /> + <edge from-layer="4920" from-port="1" to-layer="4921" to-port="1" /> + <edge from-layer="4921" from-port="2" to-layer="4922" to-port="1" /> + <edge from-layer="4922" from-port="2" to-layer="4931" to-port="0" /> + <edge from-layer="4923" from-port="0" to-layer="4924" to-port="0" /> + <edge from-layer="4924" from-port="1" to-layer="4927" to-port="0" /> + <edge from-layer="4925" from-port="0" to-layer="4926" to-port="0" /> + <edge from-layer="4926" from-port="1" to-layer="4927" to-port="1" /> + <edge from-layer="4927" from-port="2" to-layer="4929" to-port="0" /> + <edge from-layer="4928" from-port="0" to-layer="4929" to-port="1" /> + <edge from-layer="4929" from-port="2" to-layer="4930" to-port="0" /> + <edge from-layer="4930" from-port="1" to-layer="4931" to-port="1" /> + <edge from-layer="4931" from-port="2" to-layer="4932" to-port="1" /> + <edge from-layer="4932" from-port="2" to-layer="4947" to-port="0" /> + <edge from-layer="4932" from-port="2" to-layer="5137" to-port="0" /> + <edge from-layer="4932" from-port="2" to-layer="4939" to-port="0" /> + <edge from-layer="4933" from-port="0" to-layer="4934" to-port="0" /> + <edge from-layer="4934" from-port="1" to-layer="4948" to-port="0" /> + <edge from-layer="4935" from-port="0" to-layer="4936" to-port="0" /> + <edge from-layer="4936" from-port="1" to-layer="4946" to-port="0" /> + <edge from-layer="4937" from-port="0" to-layer="4938" to-port="0" /> + <edge from-layer="4938" from-port="1" to-layer="4939" to-port="1" /> + <edge from-layer="4939" from-port="2" to-layer="4941" to-port="0" /> + <edge from-layer="4940" from-port="0" to-layer="4941" to-port="1" /> + <edge from-layer="4941" from-port="2" to-layer="4944" to-port="0" /> + <edge from-layer="4942" from-port="0" to-layer="4943" to-port="0" /> + <edge from-layer="4943" from-port="1" to-layer="4944" to-port="1" /> + <edge from-layer="4944" from-port="2" to-layer="4945" to-port="0" /> + <edge from-layer="4945" from-port="1" to-layer="4946" to-port="1" /> + <edge from-layer="4946" from-port="2" to-layer="4947" to-port="1" /> + <edge from-layer="4947" from-port="2" to-layer="4948" to-port="1" /> + <edge from-layer="4948" from-port="2" to-layer="4958" to-port="0" /> + <edge from-layer="4948" from-port="2" to-layer="5019" to-port="0" /> + <edge from-layer="4948" from-port="2" to-layer="5087" to-port="0" /> + <edge from-layer="4948" from-port="2" to-layer="4957" to-port="0" /> + <edge from-layer="4949" from-port="0" to-layer="4950" to-port="0" /> + <edge from-layer="4950" from-port="1" to-layer="4953" to-port="0" /> + <edge from-layer="4951" from-port="0" to-layer="4952" to-port="0" /> + <edge from-layer="4952" from-port="1" to-layer="4953" to-port="1" /> + <edge from-layer="4953" from-port="2" to-layer="4955" to-port="0" /> + <edge from-layer="4954" from-port="0" to-layer="4955" to-port="1" /> + <edge from-layer="4955" from-port="2" to-layer="4956" to-port="0" /> + <edge from-layer="4956" from-port="1" to-layer="4957" to-port="1" /> + <edge from-layer="4957" from-port="2" to-layer="4963" to-port="0" /> + <edge from-layer="4958" from-port="1" to-layer="4961" to-port="0" /> + <edge from-layer="4959" from-port="0" to-layer="4961" to-port="1" /> + <edge from-layer="4960" from-port="0" to-layer="4961" to-port="2" /> + <edge from-layer="4961" from-port="3" to-layer="5020" to-port="0" /> + <edge from-layer="4961" from-port="3" to-layer="4962" to-port="0" /> + <edge from-layer="4961" from-port="3" to-layer="5126" to-port="0" /> + <edge from-layer="4962" from-port="3" to-layer="4963" to-port="1" /> + <edge from-layer="4963" from-port="2" to-layer="4965" to-port="0" /> + <edge from-layer="4964" from-port="0" to-layer="4965" to-port="1" /> + <edge from-layer="4965" from-port="2" to-layer="4978" to-port="0" /> + <edge from-layer="4965" from-port="2" to-layer="4975" to-port="0" /> + <edge from-layer="4965" from-port="2" to-layer="4989" to-port="0" /> + <edge from-layer="4965" from-port="2" to-layer="4995" to-port="0" /> + <edge from-layer="4966" from-port="0" to-layer="4967" to-port="0" /> + <edge from-layer="4967" from-port="1" to-layer="4968" to-port="0" /> + <edge from-layer="4968" from-port="2" to-layer="4969" to-port="0" /> + <edge from-layer="4969" from-port="2" to-layer="4971" to-port="0" /> + <edge from-layer="4970" from-port="0" to-layer="4971" to-port="1" /> + <edge from-layer="4971" from-port="2" to-layer="4972" to-port="0" /> + <edge from-layer="4971" from-port="2" to-layer="4972" to-port="1" /> + <edge from-layer="4972" from-port="2" to-layer="4997" to-port="0" /> + <edge from-layer="4972" from-port="2" to-layer="4973" to-port="0" /> + <edge from-layer="4973" from-port="1" to-layer="4974" to-port="0" /> + <edge from-layer="4974" from-port="2" to-layer="4975" to-port="1" /> + <edge from-layer="4974" from-port="2" to-layer="5024" to-port="1" /> + <edge from-layer="4975" from-port="2" to-layer="5000" to-port="0" /> + <edge from-layer="4976" from-port="0" to-layer="4986" to-port="0" /> + <edge from-layer="4977" from-port="0" to-layer="4986" to-port="1" /> + <edge from-layer="4978" from-port="1" to-layer="4980" to-port="0" /> + <edge from-layer="4979" from-port="0" to-layer="4980" to-port="2" /> + <edge from-layer="4980" from-port="3" to-layer="4981" to-port="0" /> + <edge from-layer="4981" from-port="2" to-layer="4982" to-port="0" /> + <edge from-layer="4982" from-port="1" to-layer="4984" to-port="0" /> + <edge from-layer="4983" from-port="0" to-layer="4984" to-port="1" /> + <edge from-layer="4984" from-port="2" to-layer="4986" to-port="2" /> + <edge from-layer="4985" from-port="0" to-layer="4986" to-port="3" /> + <edge from-layer="4986" from-port="4" to-layer="4989" to-port="1" /> + <edge from-layer="4986" from-port="4" to-layer="4995" to-port="2" /> + <edge from-layer="4987" from-port="0" to-layer="4989" to-port="2" /> + <edge from-layer="4988" from-port="0" to-layer="4989" to-port="3" /> + <edge from-layer="4989" from-port="4" to-layer="4992" to-port="0" /> + <edge from-layer="4990" from-port="0" to-layer="4991" to-port="0" /> + <edge from-layer="4991" from-port="1" to-layer="4992" to-port="1" /> + <edge from-layer="4992" from-port="2" to-layer="4996" to-port="0" /> + <edge from-layer="4993" from-port="0" to-layer="4995" to-port="1" /> + <edge from-layer="4994" from-port="0" to-layer="4995" to-port="3" /> + <edge from-layer="4995" from-port="4" to-layer="4996" to-port="1" /> + <edge from-layer="4996" from-port="2" to-layer="4999" to-port="0" /> + <edge from-layer="4997" from-port="1" to-layer="4998" to-port="0" /> + <edge from-layer="4998" from-port="2" to-layer="4999" to-port="1" /> + <edge from-layer="4998" from-port="2" to-layer="5046" to-port="1" /> + <edge from-layer="4999" from-port="2" to-layer="5000" to-port="1" /> + <edge from-layer="5000" from-port="2" to-layer="5123" to-port="0" /> + <edge from-layer="5001" from-port="0" to-layer="5002" to-port="0" /> + <edge from-layer="5002" from-port="1" to-layer="5007" to-port="0" /> + <edge from-layer="5003" from-port="0" to-layer="5006" to-port="1" /> + <edge from-layer="5004" from-port="0" to-layer="5006" to-port="2" /> + <edge from-layer="5005" from-port="0" to-layer="5006" to-port="3" /> + <edge from-layer="5006" from-port="4" to-layer="5007" to-port="1" /> + <edge from-layer="5007" from-port="2" to-layer="5008" to-port="0" /> + <edge from-layer="5008" from-port="1" to-layer="5010" to-port="0" /> + <edge from-layer="5009" from-port="0" to-layer="5010" to-port="2" /> + <edge from-layer="5010" from-port="3" to-layer="5048" to-port="0" /> + <edge from-layer="5011" from-port="0" to-layer="5012" to-port="0" /> + <edge from-layer="5012" from-port="1" to-layer="5015" to-port="0" /> + <edge from-layer="5013" from-port="0" to-layer="5014" to-port="0" /> + <edge from-layer="5014" from-port="1" to-layer="5015" to-port="1" /> + <edge from-layer="5015" from-port="2" to-layer="5017" to-port="0" /> + <edge from-layer="5016" from-port="0" to-layer="5017" to-port="1" /> + <edge from-layer="5017" from-port="2" to-layer="5018" to-port="0" /> + <edge from-layer="5018" from-port="1" to-layer="5019" to-port="1" /> + <edge from-layer="5019" from-port="2" to-layer="5021" to-port="0" /> + <edge from-layer="5020" from-port="3" to-layer="5088" to-port="1" /> + <edge from-layer="5020" from-port="3" to-layer="5021" to-port="1" /> + <edge from-layer="5021" from-port="2" to-layer="5023" to-port="0" /> + <edge from-layer="5022" from-port="0" to-layer="5023" to-port="1" /> + <edge from-layer="5023" from-port="2" to-layer="5044" to-port="0" /> + <edge from-layer="5023" from-port="2" to-layer="5027" to-port="0" /> + <edge from-layer="5023" from-port="2" to-layer="5024" to-port="0" /> + <edge from-layer="5023" from-port="2" to-layer="5038" to-port="0" /> + <edge from-layer="5024" from-port="2" to-layer="5047" to-port="0" /> + <edge from-layer="5025" from-port="0" to-layer="5035" to-port="0" /> + <edge from-layer="5026" from-port="0" to-layer="5035" to-port="1" /> + <edge from-layer="5027" from-port="1" to-layer="5029" to-port="0" /> + <edge from-layer="5028" from-port="0" to-layer="5029" to-port="2" /> + <edge from-layer="5029" from-port="3" to-layer="5030" to-port="0" /> + <edge from-layer="5030" from-port="2" to-layer="5031" to-port="0" /> + <edge from-layer="5031" from-port="1" to-layer="5033" to-port="0" /> + <edge from-layer="5032" from-port="0" to-layer="5033" to-port="1" /> + <edge from-layer="5033" from-port="2" to-layer="5035" to-port="2" /> + <edge from-layer="5034" from-port="0" to-layer="5035" to-port="3" /> + <edge from-layer="5035" from-port="4" to-layer="5044" to-port="2" /> + <edge from-layer="5035" from-port="4" to-layer="5038" to-port="1" /> + <edge from-layer="5036" from-port="0" to-layer="5038" to-port="2" /> + <edge from-layer="5037" from-port="0" to-layer="5038" to-port="3" /> + <edge from-layer="5038" from-port="4" to-layer="5041" to-port="0" /> + <edge from-layer="5039" from-port="0" to-layer="5040" to-port="0" /> + <edge from-layer="5040" from-port="1" to-layer="5041" to-port="1" /> + <edge from-layer="5041" from-port="2" to-layer="5045" to-port="0" /> + <edge from-layer="5042" from-port="0" to-layer="5044" to-port="1" /> + <edge from-layer="5043" from-port="0" to-layer="5044" to-port="3" /> + <edge from-layer="5044" from-port="4" to-layer="5045" to-port="1" /> + <edge from-layer="5045" from-port="2" to-layer="5046" to-port="0" /> + <edge from-layer="5046" from-port="2" to-layer="5047" to-port="1" /> + <edge from-layer="5047" from-port="2" to-layer="5048" to-port="1" /> + <edge from-layer="5048" from-port="2" to-layer="5049" to-port="0" /> + <edge from-layer="5048" from-port="2" to-layer="5050" to-port="0" /> + <edge from-layer="5048" from-port="2" to-layer="5532" to-port="0" /> + <edge from-layer="5049" from-port="2" to-layer="5063" to-port="0" /> + <edge from-layer="5050" from-port="1" to-layer="5055" to-port="0" /> + <edge from-layer="5050" from-port="1" to-layer="5053" to-port="0" /> + <edge from-layer="5050" from-port="1" to-layer="5060" to-port="0" /> + <edge from-layer="5051" from-port="0" to-layer="5053" to-port="1" /> + <edge from-layer="5052" from-port="0" to-layer="5053" to-port="2" /> + <edge from-layer="5053" from-port="3" to-layer="5067" to-port="0" /> + <edge from-layer="5053" from-port="3" to-layer="5061" to-port="0" /> + <edge from-layer="5054" from-port="0" to-layer="5055" to-port="2" /> + <edge from-layer="5055" from-port="3" to-layer="5057" to-port="0" /> + <edge from-layer="5055" from-port="3" to-layer="5064" to-port="0" /> + <edge from-layer="5056" from-port="0" to-layer="5057" to-port="1" /> + <edge from-layer="5057" from-port="2" to-layer="5061" to-port="1" /> + <edge from-layer="5058" from-port="0" to-layer="5060" to-port="1" /> + <edge from-layer="5059" from-port="0" to-layer="5060" to-port="2" /> + <edge from-layer="5060" from-port="3" to-layer="5067" to-port="2" /> + <edge from-layer="5060" from-port="3" to-layer="5061" to-port="3" /> + <edge from-layer="5061" from-port="4" to-layer="5062" to-port="0" /> + <edge from-layer="5062" from-port="1" to-layer="5063" to-port="1" /> + <edge from-layer="5063" from-port="2" to-layer="5068" to-port="0" /> + <edge from-layer="5064" from-port="2" to-layer="5066" to-port="0" /> + <edge from-layer="5065" from-port="0" to-layer="5066" to-port="1" /> + <edge from-layer="5066" from-port="2" to-layer="5067" to-port="1" /> + <edge from-layer="5067" from-port="3" to-layer="5068" to-port="1" /> + <edge from-layer="5068" from-port="2" to-layer="5123" to-port="1" /> + <edge from-layer="5068" from-port="2" to-layer="5115" to-port="0" /> + <edge from-layer="5069" from-port="0" to-layer="5070" to-port="0" /> + <edge from-layer="5070" from-port="1" to-layer="5075" to-port="0" /> + <edge from-layer="5071" from-port="0" to-layer="5074" to-port="1" /> + <edge from-layer="5072" from-port="0" to-layer="5074" to-port="2" /> + <edge from-layer="5073" from-port="0" to-layer="5074" to-port="3" /> + <edge from-layer="5074" from-port="4" to-layer="5075" to-port="1" /> + <edge from-layer="5075" from-port="2" to-layer="5076" to-port="0" /> + <edge from-layer="5076" from-port="1" to-layer="5078" to-port="0" /> + <edge from-layer="5077" from-port="0" to-layer="5078" to-port="2" /> + <edge from-layer="5078" from-port="3" to-layer="5091" to-port="0" /> + <edge from-layer="5079" from-port="0" to-layer="5080" to-port="0" /> + <edge from-layer="5080" from-port="1" to-layer="5083" to-port="0" /> + <edge from-layer="5081" from-port="0" to-layer="5082" to-port="0" /> + <edge from-layer="5082" from-port="1" to-layer="5083" to-port="1" /> + <edge from-layer="5083" from-port="2" to-layer="5085" to-port="0" /> + <edge from-layer="5084" from-port="0" to-layer="5085" to-port="1" /> + <edge from-layer="5085" from-port="2" to-layer="5086" to-port="0" /> + <edge from-layer="5086" from-port="1" to-layer="5087" to-port="1" /> + <edge from-layer="5087" from-port="2" to-layer="5088" to-port="0" /> + <edge from-layer="5088" from-port="2" to-layer="5090" to-port="0" /> + <edge from-layer="5089" from-port="0" to-layer="5090" to-port="1" /> + <edge from-layer="5090" from-port="2" to-layer="5091" to-port="1" /> + <edge from-layer="5091" from-port="2" to-layer="5531" to-port="0" /> + <edge from-layer="5091" from-port="2" to-layer="5093" to-port="0" /> + <edge from-layer="5091" from-port="2" to-layer="5092" to-port="0" /> + <edge from-layer="5092" from-port="2" to-layer="5106" to-port="0" /> + <edge from-layer="5093" from-port="1" to-layer="5103" to-port="0" /> + <edge from-layer="5093" from-port="1" to-layer="5098" to-port="0" /> + <edge from-layer="5093" from-port="1" to-layer="5096" to-port="0" /> + <edge from-layer="5094" from-port="0" to-layer="5096" to-port="1" /> + <edge from-layer="5095" from-port="0" to-layer="5096" to-port="2" /> + <edge from-layer="5096" from-port="3" to-layer="5104" to-port="0" /> + <edge from-layer="5096" from-port="3" to-layer="5110" to-port="0" /> + <edge from-layer="5097" from-port="0" to-layer="5098" to-port="2" /> + <edge from-layer="5098" from-port="3" to-layer="5100" to-port="0" /> + <edge from-layer="5098" from-port="3" to-layer="5107" to-port="0" /> + <edge from-layer="5099" from-port="0" to-layer="5100" to-port="1" /> + <edge from-layer="5100" from-port="2" to-layer="5104" to-port="1" /> + <edge from-layer="5101" from-port="0" to-layer="5103" to-port="1" /> + <edge from-layer="5102" from-port="0" to-layer="5103" to-port="2" /> + <edge from-layer="5103" from-port="3" to-layer="5110" to-port="2" /> + <edge from-layer="5103" from-port="3" to-layer="5104" to-port="3" /> + <edge from-layer="5104" from-port="4" to-layer="5105" to-port="0" /> + <edge from-layer="5105" from-port="1" to-layer="5106" to-port="1" /> + <edge from-layer="5106" from-port="2" to-layer="5111" to-port="0" /> + <edge from-layer="5107" from-port="2" to-layer="5109" to-port="0" /> + <edge from-layer="5108" from-port="0" to-layer="5109" to-port="1" /> + <edge from-layer="5109" from-port="2" to-layer="5110" to-port="1" /> + <edge from-layer="5110" from-port="3" to-layer="5111" to-port="1" /> + <edge from-layer="5111" from-port="2" to-layer="5123" to-port="2" /> + <edge from-layer="5112" from-port="0" to-layer="5122" to-port="1" /> + <edge from-layer="5113" from-port="0" to-layer="5120" to-port="0" /> + <edge from-layer="5114" from-port="0" to-layer="5120" to-port="1" /> + <edge from-layer="5115" from-port="1" to-layer="5118" to-port="0" /> + <edge from-layer="5116" from-port="0" to-layer="5118" to-port="1" /> + <edge from-layer="5117" from-port="0" to-layer="5118" to-port="2" /> + <edge from-layer="5118" from-port="3" to-layer="5120" to-port="2" /> + <edge from-layer="5119" from-port="0" to-layer="5120" to-port="3" /> + <edge from-layer="5120" from-port="4" to-layer="5122" to-port="2" /> + <edge from-layer="5121" from-port="0" to-layer="5122" to-port="3" /> + <edge from-layer="5122" from-port="4" to-layer="5123" to-port="3" /> + <edge from-layer="5123" from-port="4" to-layer="5125" to-port="0" /> + <edge from-layer="5124" from-port="0" to-layer="5125" to-port="1" /> + <edge from-layer="5125" from-port="2" to-layer="5127" to-port="0" /> + <edge from-layer="5126" from-port="2" to-layer="5127" to-port="1" /> + <edge from-layer="5127" from-port="2" to-layer="5136" to-port="0" /> + <edge from-layer="5128" from-port="0" to-layer="5129" to-port="0" /> + <edge from-layer="5129" from-port="1" to-layer="5132" to-port="0" /> + <edge from-layer="5130" from-port="0" to-layer="5131" to-port="0" /> + <edge from-layer="5131" from-port="1" to-layer="5132" to-port="1" /> + <edge from-layer="5132" from-port="2" to-layer="5134" to-port="0" /> + <edge from-layer="5133" from-port="0" to-layer="5134" to-port="1" /> + <edge from-layer="5134" from-port="2" to-layer="5135" to-port="0" /> + <edge from-layer="5135" from-port="1" to-layer="5136" to-port="1" /> + <edge from-layer="5136" from-port="2" to-layer="5137" to-port="1" /> + <edge from-layer="5137" from-port="2" to-layer="5183" to-port="0" /> + <edge from-layer="5137" from-port="2" to-layer="5144" to-port="0" /> + <edge from-layer="5137" from-port="2" to-layer="5152" to-port="0" /> + <edge from-layer="5138" from-port="0" to-layer="5139" to-port="0" /> + <edge from-layer="5139" from-port="1" to-layer="5153" to-port="0" /> + <edge from-layer="5140" from-port="0" to-layer="5141" to-port="0" /> + <edge from-layer="5141" from-port="1" to-layer="5151" to-port="0" /> + <edge from-layer="5142" from-port="0" to-layer="5143" to-port="0" /> + <edge from-layer="5143" from-port="1" to-layer="5144" to-port="1" /> + <edge from-layer="5144" from-port="2" to-layer="5146" to-port="0" /> + <edge from-layer="5145" from-port="0" to-layer="5146" to-port="1" /> + <edge from-layer="5146" from-port="2" to-layer="5149" to-port="0" /> + <edge from-layer="5147" from-port="0" to-layer="5148" to-port="0" /> + <edge from-layer="5148" from-port="1" to-layer="5149" to-port="1" /> + <edge from-layer="5149" from-port="2" to-layer="5150" to-port="0" /> + <edge from-layer="5150" from-port="1" to-layer="5151" to-port="1" /> + <edge from-layer="5151" from-port="2" to-layer="5152" to-port="1" /> + <edge from-layer="5152" from-port="2" to-layer="5153" to-port="1" /> + <edge from-layer="5153" from-port="2" to-layer="5172" to-port="0" /> + <edge from-layer="5153" from-port="2" to-layer="5162" to-port="0" /> + <edge from-layer="5154" from-port="0" to-layer="5155" to-port="0" /> + <edge from-layer="5155" from-port="1" to-layer="5158" to-port="0" /> + <edge from-layer="5156" from-port="0" to-layer="5157" to-port="0" /> + <edge from-layer="5157" from-port="1" to-layer="5158" to-port="1" /> + <edge from-layer="5158" from-port="2" to-layer="5160" to-port="0" /> + <edge from-layer="5159" from-port="0" to-layer="5160" to-port="1" /> + <edge from-layer="5160" from-port="2" to-layer="5161" to-port="0" /> + <edge from-layer="5161" from-port="1" to-layer="5162" to-port="1" /> + <edge from-layer="5162" from-port="2" to-layer="5163" to-port="0" /> + <edge from-layer="5163" from-port="1" to-layer="5173" to-port="0" /> + <edge from-layer="5164" from-port="0" to-layer="5165" to-port="0" /> + <edge from-layer="5165" from-port="1" to-layer="5168" to-port="0" /> + <edge from-layer="5166" from-port="0" to-layer="5167" to-port="0" /> + <edge from-layer="5167" from-port="1" to-layer="5168" to-port="1" /> + <edge from-layer="5168" from-port="2" to-layer="5170" to-port="0" /> + <edge from-layer="5169" from-port="0" to-layer="5170" to-port="1" /> + <edge from-layer="5170" from-port="2" to-layer="5171" to-port="0" /> + <edge from-layer="5171" from-port="1" to-layer="5172" to-port="1" /> + <edge from-layer="5172" from-port="2" to-layer="5173" to-port="1" /> + <edge from-layer="5173" from-port="2" to-layer="5182" to-port="0" /> + <edge from-layer="5174" from-port="0" to-layer="5175" to-port="0" /> + <edge from-layer="5175" from-port="1" to-layer="5178" to-port="0" /> + <edge from-layer="5176" from-port="0" to-layer="5177" to-port="0" /> + <edge from-layer="5177" from-port="1" to-layer="5178" to-port="1" /> + <edge from-layer="5178" from-port="2" to-layer="5180" to-port="0" /> + <edge from-layer="5179" from-port="0" to-layer="5180" to-port="1" /> + <edge from-layer="5180" from-port="2" to-layer="5181" to-port="0" /> + <edge from-layer="5181" from-port="1" to-layer="5182" to-port="1" /> + <edge from-layer="5182" from-port="2" to-layer="5183" to-port="1" /> + <edge from-layer="5183" from-port="2" to-layer="5388" to-port="0" /> + <edge from-layer="5183" from-port="2" to-layer="5190" to-port="0" /> + <edge from-layer="5183" from-port="2" to-layer="5198" to-port="0" /> + <edge from-layer="5184" from-port="0" to-layer="5185" to-port="0" /> + <edge from-layer="5185" from-port="1" to-layer="5199" to-port="0" /> + <edge from-layer="5186" from-port="0" to-layer="5187" to-port="0" /> + <edge from-layer="5187" from-port="1" to-layer="5197" to-port="0" /> + <edge from-layer="5188" from-port="0" to-layer="5189" to-port="0" /> + <edge from-layer="5189" from-port="1" to-layer="5190" to-port="1" /> + <edge from-layer="5190" from-port="2" to-layer="5192" to-port="0" /> + <edge from-layer="5191" from-port="0" to-layer="5192" to-port="1" /> + <edge from-layer="5192" from-port="2" to-layer="5195" to-port="0" /> + <edge from-layer="5193" from-port="0" to-layer="5194" to-port="0" /> + <edge from-layer="5194" from-port="1" to-layer="5195" to-port="1" /> + <edge from-layer="5195" from-port="2" to-layer="5196" to-port="0" /> + <edge from-layer="5196" from-port="1" to-layer="5197" to-port="1" /> + <edge from-layer="5197" from-port="2" to-layer="5198" to-port="1" /> + <edge from-layer="5198" from-port="2" to-layer="5199" to-port="1" /> + <edge from-layer="5199" from-port="2" to-layer="5338" to-port="0" /> + <edge from-layer="5199" from-port="2" to-layer="5209" to-port="0" /> + <edge from-layer="5199" from-port="2" to-layer="5208" to-port="0" /> + <edge from-layer="5199" from-port="2" to-layer="5270" to-port="0" /> + <edge from-layer="5200" from-port="0" to-layer="5201" to-port="0" /> + <edge from-layer="5201" from-port="1" to-layer="5204" to-port="0" /> + <edge from-layer="5202" from-port="0" to-layer="5203" to-port="0" /> + <edge from-layer="5203" from-port="1" to-layer="5204" to-port="1" /> + <edge from-layer="5204" from-port="2" to-layer="5206" to-port="0" /> + <edge from-layer="5205" from-port="0" to-layer="5206" to-port="1" /> + <edge from-layer="5206" from-port="2" to-layer="5207" to-port="0" /> + <edge from-layer="5207" from-port="1" to-layer="5208" to-port="1" /> + <edge from-layer="5208" from-port="2" to-layer="5214" to-port="0" /> + <edge from-layer="5209" from-port="1" to-layer="5212" to-port="0" /> + <edge from-layer="5210" from-port="0" to-layer="5212" to-port="1" /> + <edge from-layer="5211" from-port="0" to-layer="5212" to-port="2" /> + <edge from-layer="5212" from-port="3" to-layer="5213" to-port="0" /> + <edge from-layer="5212" from-port="3" to-layer="5271" to-port="0" /> + <edge from-layer="5212" from-port="3" to-layer="5377" to-port="0" /> + <edge from-layer="5213" from-port="3" to-layer="5214" to-port="1" /> + <edge from-layer="5214" from-port="2" to-layer="5216" to-port="0" /> + <edge from-layer="5215" from-port="0" to-layer="5216" to-port="1" /> + <edge from-layer="5216" from-port="2" to-layer="5226" to-port="0" /> + <edge from-layer="5216" from-port="2" to-layer="5240" to-port="0" /> + <edge from-layer="5216" from-port="2" to-layer="5229" to-port="0" /> + <edge from-layer="5216" from-port="2" to-layer="5246" to-port="0" /> + <edge from-layer="5217" from-port="0" to-layer="5218" to-port="0" /> + <edge from-layer="5218" from-port="1" to-layer="5219" to-port="0" /> + <edge from-layer="5219" from-port="2" to-layer="5220" to-port="0" /> + <edge from-layer="5220" from-port="2" to-layer="5222" to-port="0" /> + <edge from-layer="5221" from-port="0" to-layer="5222" to-port="1" /> + <edge from-layer="5222" from-port="2" to-layer="5223" to-port="0" /> + <edge from-layer="5222" from-port="2" to-layer="5223" to-port="1" /> + <edge from-layer="5223" from-port="2" to-layer="5248" to-port="0" /> + <edge from-layer="5223" from-port="2" to-layer="5224" to-port="0" /> + <edge from-layer="5224" from-port="1" to-layer="5225" to-port="0" /> + <edge from-layer="5225" from-port="2" to-layer="5226" to-port="1" /> + <edge from-layer="5225" from-port="2" to-layer="5275" to-port="1" /> + <edge from-layer="5226" from-port="2" to-layer="5251" to-port="0" /> + <edge from-layer="5227" from-port="0" to-layer="5237" to-port="0" /> + <edge from-layer="5228" from-port="0" to-layer="5237" to-port="1" /> + <edge from-layer="5229" from-port="1" to-layer="5231" to-port="0" /> + <edge from-layer="5230" from-port="0" to-layer="5231" to-port="2" /> + <edge from-layer="5231" from-port="3" to-layer="5232" to-port="0" /> + <edge from-layer="5232" from-port="2" to-layer="5233" to-port="0" /> + <edge from-layer="5233" from-port="1" to-layer="5235" to-port="0" /> + <edge from-layer="5234" from-port="0" to-layer="5235" to-port="1" /> + <edge from-layer="5235" from-port="2" to-layer="5237" to-port="2" /> + <edge from-layer="5236" from-port="0" to-layer="5237" to-port="3" /> + <edge from-layer="5237" from-port="4" to-layer="5246" to-port="2" /> + <edge from-layer="5237" from-port="4" to-layer="5240" to-port="1" /> + <edge from-layer="5238" from-port="0" to-layer="5240" to-port="2" /> + <edge from-layer="5239" from-port="0" to-layer="5240" to-port="3" /> + <edge from-layer="5240" from-port="4" to-layer="5243" to-port="0" /> + <edge from-layer="5241" from-port="0" to-layer="5242" to-port="0" /> + <edge from-layer="5242" from-port="1" to-layer="5243" to-port="1" /> + <edge from-layer="5243" from-port="2" to-layer="5247" to-port="0" /> + <edge from-layer="5244" from-port="0" to-layer="5246" to-port="1" /> + <edge from-layer="5245" from-port="0" to-layer="5246" to-port="3" /> + <edge from-layer="5246" from-port="4" to-layer="5247" to-port="1" /> + <edge from-layer="5247" from-port="2" to-layer="5250" to-port="0" /> + <edge from-layer="5248" from-port="1" to-layer="5249" to-port="0" /> + <edge from-layer="5249" from-port="2" to-layer="5297" to-port="1" /> + <edge from-layer="5249" from-port="2" to-layer="5250" to-port="1" /> + <edge from-layer="5250" from-port="2" to-layer="5251" to-port="1" /> + <edge from-layer="5251" from-port="2" to-layer="5374" to-port="0" /> + <edge from-layer="5252" from-port="0" to-layer="5253" to-port="0" /> + <edge from-layer="5253" from-port="1" to-layer="5258" to-port="0" /> + <edge from-layer="5254" from-port="0" to-layer="5257" to-port="1" /> + <edge from-layer="5255" from-port="0" to-layer="5257" to-port="2" /> + <edge from-layer="5256" from-port="0" to-layer="5257" to-port="3" /> + <edge from-layer="5257" from-port="4" to-layer="5258" to-port="1" /> + <edge from-layer="5258" from-port="2" to-layer="5259" to-port="0" /> + <edge from-layer="5259" from-port="1" to-layer="5261" to-port="0" /> + <edge from-layer="5260" from-port="0" to-layer="5261" to-port="2" /> + <edge from-layer="5261" from-port="3" to-layer="5299" to-port="0" /> + <edge from-layer="5262" from-port="0" to-layer="5263" to-port="0" /> + <edge from-layer="5263" from-port="1" to-layer="5266" to-port="0" /> + <edge from-layer="5264" from-port="0" to-layer="5265" to-port="0" /> + <edge from-layer="5265" from-port="1" to-layer="5266" to-port="1" /> + <edge from-layer="5266" from-port="2" to-layer="5268" to-port="0" /> + <edge from-layer="5267" from-port="0" to-layer="5268" to-port="1" /> + <edge from-layer="5268" from-port="2" to-layer="5269" to-port="0" /> + <edge from-layer="5269" from-port="1" to-layer="5270" to-port="1" /> + <edge from-layer="5270" from-port="2" to-layer="5272" to-port="0" /> + <edge from-layer="5271" from-port="3" to-layer="5339" to-port="1" /> + <edge from-layer="5271" from-port="3" to-layer="5272" to-port="1" /> + <edge from-layer="5272" from-port="2" to-layer="5274" to-port="0" /> + <edge from-layer="5273" from-port="0" to-layer="5274" to-port="1" /> + <edge from-layer="5274" from-port="2" to-layer="5295" to-port="0" /> + <edge from-layer="5274" from-port="2" to-layer="5278" to-port="0" /> + <edge from-layer="5274" from-port="2" to-layer="5275" to-port="0" /> + <edge from-layer="5274" from-port="2" to-layer="5289" to-port="0" /> + <edge from-layer="5275" from-port="2" to-layer="5298" to-port="0" /> + <edge from-layer="5276" from-port="0" to-layer="5286" to-port="0" /> + <edge from-layer="5277" from-port="0" to-layer="5286" to-port="1" /> + <edge from-layer="5278" from-port="1" to-layer="5280" to-port="0" /> + <edge from-layer="5279" from-port="0" to-layer="5280" to-port="2" /> + <edge from-layer="5280" from-port="3" to-layer="5281" to-port="0" /> + <edge from-layer="5281" from-port="2" to-layer="5282" to-port="0" /> + <edge from-layer="5282" from-port="1" to-layer="5284" to-port="0" /> + <edge from-layer="5283" from-port="0" to-layer="5284" to-port="1" /> + <edge from-layer="5284" from-port="2" to-layer="5286" to-port="2" /> + <edge from-layer="5285" from-port="0" to-layer="5286" to-port="3" /> + <edge from-layer="5286" from-port="4" to-layer="5289" to-port="1" /> + <edge from-layer="5286" from-port="4" to-layer="5295" to-port="2" /> + <edge from-layer="5287" from-port="0" to-layer="5289" to-port="2" /> + <edge from-layer="5288" from-port="0" to-layer="5289" to-port="3" /> + <edge from-layer="5289" from-port="4" to-layer="5292" to-port="0" /> + <edge from-layer="5290" from-port="0" to-layer="5291" to-port="0" /> + <edge from-layer="5291" from-port="1" to-layer="5292" to-port="1" /> + <edge from-layer="5292" from-port="2" to-layer="5296" to-port="0" /> + <edge from-layer="5293" from-port="0" to-layer="5295" to-port="1" /> + <edge from-layer="5294" from-port="0" to-layer="5295" to-port="3" /> + <edge from-layer="5295" from-port="4" to-layer="5296" to-port="1" /> + <edge from-layer="5296" from-port="2" to-layer="5297" to-port="0" /> + <edge from-layer="5297" from-port="2" to-layer="5298" to-port="1" /> + <edge from-layer="5298" from-port="2" to-layer="5299" to-port="1" /> + <edge from-layer="5299" from-port="2" to-layer="5528" to-port="0" /> + <edge from-layer="5299" from-port="2" to-layer="5300" to-port="0" /> + <edge from-layer="5299" from-port="2" to-layer="5301" to-port="0" /> + <edge from-layer="5300" from-port="2" to-layer="5314" to-port="0" /> + <edge from-layer="5301" from-port="1" to-layer="5311" to-port="0" /> + <edge from-layer="5301" from-port="1" to-layer="5304" to-port="0" /> + <edge from-layer="5301" from-port="1" to-layer="5306" to-port="0" /> + <edge from-layer="5302" from-port="0" to-layer="5304" to-port="1" /> + <edge from-layer="5303" from-port="0" to-layer="5304" to-port="2" /> + <edge from-layer="5304" from-port="3" to-layer="5312" to-port="0" /> + <edge from-layer="5304" from-port="3" to-layer="5318" to-port="0" /> + <edge from-layer="5305" from-port="0" to-layer="5306" to-port="2" /> + <edge from-layer="5306" from-port="3" to-layer="5308" to-port="0" /> + <edge from-layer="5306" from-port="3" to-layer="5315" to-port="0" /> + <edge from-layer="5307" from-port="0" to-layer="5308" to-port="1" /> + <edge from-layer="5308" from-port="2" to-layer="5312" to-port="1" /> + <edge from-layer="5309" from-port="0" to-layer="5311" to-port="1" /> + <edge from-layer="5310" from-port="0" to-layer="5311" to-port="2" /> + <edge from-layer="5311" from-port="3" to-layer="5312" to-port="3" /> + <edge from-layer="5311" from-port="3" to-layer="5318" to-port="2" /> + <edge from-layer="5312" from-port="4" to-layer="5313" to-port="0" /> + <edge from-layer="5313" from-port="1" to-layer="5314" to-port="1" /> + <edge from-layer="5314" from-port="2" to-layer="5319" to-port="0" /> + <edge from-layer="5315" from-port="2" to-layer="5317" to-port="0" /> + <edge from-layer="5316" from-port="0" to-layer="5317" to-port="1" /> + <edge from-layer="5317" from-port="2" to-layer="5318" to-port="1" /> + <edge from-layer="5318" from-port="3" to-layer="5319" to-port="1" /> + <edge from-layer="5319" from-port="2" to-layer="5366" to-port="0" /> + <edge from-layer="5319" from-port="2" to-layer="5374" to-port="1" /> + <edge from-layer="5320" from-port="0" to-layer="5321" to-port="0" /> + <edge from-layer="5321" from-port="1" to-layer="5326" to-port="0" /> + <edge from-layer="5322" from-port="0" to-layer="5325" to-port="1" /> + <edge from-layer="5323" from-port="0" to-layer="5325" to-port="2" /> + <edge from-layer="5324" from-port="0" to-layer="5325" to-port="3" /> + <edge from-layer="5325" from-port="4" to-layer="5326" to-port="1" /> + <edge from-layer="5326" from-port="2" to-layer="5327" to-port="0" /> + <edge from-layer="5327" from-port="1" to-layer="5329" to-port="0" /> + <edge from-layer="5328" from-port="0" to-layer="5329" to-port="2" /> + <edge from-layer="5329" from-port="3" to-layer="5342" to-port="0" /> + <edge from-layer="5330" from-port="0" to-layer="5331" to-port="0" /> + <edge from-layer="5331" from-port="1" to-layer="5334" to-port="0" /> + <edge from-layer="5332" from-port="0" to-layer="5333" to-port="0" /> + <edge from-layer="5333" from-port="1" to-layer="5334" to-port="1" /> + <edge from-layer="5334" from-port="2" to-layer="5336" to-port="0" /> + <edge from-layer="5335" from-port="0" to-layer="5336" to-port="1" /> + <edge from-layer="5336" from-port="2" to-layer="5337" to-port="0" /> + <edge from-layer="5337" from-port="1" to-layer="5338" to-port="1" /> + <edge from-layer="5338" from-port="2" to-layer="5339" to-port="0" /> + <edge from-layer="5339" from-port="2" to-layer="5341" to-port="0" /> + <edge from-layer="5340" from-port="0" to-layer="5341" to-port="1" /> + <edge from-layer="5341" from-port="2" to-layer="5342" to-port="1" /> + <edge from-layer="5342" from-port="2" to-layer="5344" to-port="0" /> + <edge from-layer="5342" from-port="2" to-layer="5527" to-port="0" /> + <edge from-layer="5342" from-port="2" to-layer="5343" to-port="0" /> + <edge from-layer="5343" from-port="2" to-layer="5357" to-port="0" /> + <edge from-layer="5344" from-port="1" to-layer="5354" to-port="0" /> + <edge from-layer="5344" from-port="1" to-layer="5347" to-port="0" /> + <edge from-layer="5344" from-port="1" to-layer="5349" to-port="0" /> + <edge from-layer="5345" from-port="0" to-layer="5347" to-port="1" /> + <edge from-layer="5346" from-port="0" to-layer="5347" to-port="2" /> + <edge from-layer="5347" from-port="3" to-layer="5355" to-port="0" /> + <edge from-layer="5347" from-port="3" to-layer="5361" to-port="0" /> + <edge from-layer="5348" from-port="0" to-layer="5349" to-port="2" /> + <edge from-layer="5349" from-port="3" to-layer="5351" to-port="0" /> + <edge from-layer="5349" from-port="3" to-layer="5358" to-port="0" /> + <edge from-layer="5350" from-port="0" to-layer="5351" to-port="1" /> + <edge from-layer="5351" from-port="2" to-layer="5355" to-port="1" /> + <edge from-layer="5352" from-port="0" to-layer="5354" to-port="1" /> + <edge from-layer="5353" from-port="0" to-layer="5354" to-port="2" /> + <edge from-layer="5354" from-port="3" to-layer="5361" to-port="2" /> + <edge from-layer="5354" from-port="3" to-layer="5355" to-port="3" /> + <edge from-layer="5355" from-port="4" to-layer="5356" to-port="0" /> + <edge from-layer="5356" from-port="1" to-layer="5357" to-port="1" /> + <edge from-layer="5357" from-port="2" to-layer="5362" to-port="0" /> + <edge from-layer="5358" from-port="2" to-layer="5360" to-port="0" /> + <edge from-layer="5359" from-port="0" to-layer="5360" to-port="1" /> + <edge from-layer="5360" from-port="2" to-layer="5361" to-port="1" /> + <edge from-layer="5361" from-port="3" to-layer="5362" to-port="1" /> + <edge from-layer="5362" from-port="2" to-layer="5374" to-port="2" /> + <edge from-layer="5363" from-port="0" to-layer="5373" to-port="1" /> + <edge from-layer="5364" from-port="0" to-layer="5371" to-port="0" /> + <edge from-layer="5365" from-port="0" to-layer="5371" to-port="1" /> + <edge from-layer="5366" from-port="1" to-layer="5369" to-port="0" /> + <edge from-layer="5367" from-port="0" to-layer="5369" to-port="1" /> + <edge from-layer="5368" from-port="0" to-layer="5369" to-port="2" /> + <edge from-layer="5369" from-port="3" to-layer="5371" to-port="2" /> + <edge from-layer="5370" from-port="0" to-layer="5371" to-port="3" /> + <edge from-layer="5371" from-port="4" to-layer="5373" to-port="2" /> + <edge from-layer="5372" from-port="0" to-layer="5373" to-port="3" /> + <edge from-layer="5373" from-port="4" to-layer="5374" to-port="3" /> + <edge from-layer="5374" from-port="4" to-layer="5376" to-port="0" /> + <edge from-layer="5375" from-port="0" to-layer="5376" to-port="1" /> + <edge from-layer="5376" from-port="2" to-layer="5378" to-port="0" /> + <edge from-layer="5377" from-port="2" to-layer="5378" to-port="1" /> + <edge from-layer="5378" from-port="2" to-layer="5387" to-port="0" /> + <edge from-layer="5379" from-port="0" to-layer="5380" to-port="0" /> + <edge from-layer="5380" from-port="1" to-layer="5383" to-port="0" /> + <edge from-layer="5381" from-port="0" to-layer="5382" to-port="0" /> + <edge from-layer="5382" from-port="1" to-layer="5383" to-port="1" /> + <edge from-layer="5383" from-port="2" to-layer="5385" to-port="0" /> + <edge from-layer="5384" from-port="0" to-layer="5385" to-port="1" /> + <edge from-layer="5385" from-port="2" to-layer="5386" to-port="0" /> + <edge from-layer="5386" from-port="1" to-layer="5387" to-port="1" /> + <edge from-layer="5387" from-port="2" to-layer="5388" to-port="1" /> + <edge from-layer="5388" from-port="2" to-layer="5395" to-port="0" /> + <edge from-layer="5388" from-port="2" to-layer="5403" to-port="0" /> + <edge from-layer="5388" from-port="2" to-layer="5434" to-port="0" /> + <edge from-layer="5389" from-port="0" to-layer="5390" to-port="0" /> + <edge from-layer="5390" from-port="1" to-layer="5404" to-port="0" /> + <edge from-layer="5391" from-port="0" to-layer="5392" to-port="0" /> + <edge from-layer="5392" from-port="1" to-layer="5402" to-port="0" /> + <edge from-layer="5393" from-port="0" to-layer="5394" to-port="0" /> + <edge from-layer="5394" from-port="1" to-layer="5395" to-port="1" /> + <edge from-layer="5395" from-port="2" to-layer="5397" to-port="0" /> + <edge from-layer="5396" from-port="0" to-layer="5397" to-port="1" /> + <edge from-layer="5397" from-port="2" to-layer="5400" to-port="0" /> + <edge from-layer="5398" from-port="0" to-layer="5399" to-port="0" /> + <edge from-layer="5399" from-port="1" to-layer="5400" to-port="1" /> + <edge from-layer="5400" from-port="2" to-layer="5401" to-port="0" /> + <edge from-layer="5401" from-port="1" to-layer="5402" to-port="1" /> + <edge from-layer="5402" from-port="2" to-layer="5403" to-port="1" /> + <edge from-layer="5403" from-port="2" to-layer="5404" to-port="1" /> + <edge from-layer="5404" from-port="2" to-layer="5423" to-port="0" /> + <edge from-layer="5404" from-port="2" to-layer="5413" to-port="0" /> + <edge from-layer="5405" from-port="0" to-layer="5406" to-port="0" /> + <edge from-layer="5406" from-port="1" to-layer="5409" to-port="0" /> + <edge from-layer="5407" from-port="0" to-layer="5408" to-port="0" /> + <edge from-layer="5408" from-port="1" to-layer="5409" to-port="1" /> + <edge from-layer="5409" from-port="2" to-layer="5411" to-port="0" /> + <edge from-layer="5410" from-port="0" to-layer="5411" to-port="1" /> + <edge from-layer="5411" from-port="2" to-layer="5412" to-port="0" /> + <edge from-layer="5412" from-port="1" to-layer="5413" to-port="1" /> + <edge from-layer="5413" from-port="2" to-layer="5414" to-port="0" /> + <edge from-layer="5414" from-port="1" to-layer="5424" to-port="0" /> + <edge from-layer="5415" from-port="0" to-layer="5416" to-port="0" /> + <edge from-layer="5416" from-port="1" to-layer="5419" to-port="0" /> + <edge from-layer="5417" from-port="0" to-layer="5418" to-port="0" /> + <edge from-layer="5418" from-port="1" to-layer="5419" to-port="1" /> + <edge from-layer="5419" from-port="2" to-layer="5421" to-port="0" /> + <edge from-layer="5420" from-port="0" to-layer="5421" to-port="1" /> + <edge from-layer="5421" from-port="2" to-layer="5422" to-port="0" /> + <edge from-layer="5422" from-port="1" to-layer="5423" to-port="1" /> + <edge from-layer="5423" from-port="2" to-layer="5424" to-port="1" /> + <edge from-layer="5424" from-port="2" to-layer="5433" to-port="0" /> + <edge from-layer="5425" from-port="0" to-layer="5426" to-port="0" /> + <edge from-layer="5426" from-port="1" to-layer="5429" to-port="0" /> + <edge from-layer="5427" from-port="0" to-layer="5428" to-port="0" /> + <edge from-layer="5428" from-port="1" to-layer="5429" to-port="1" /> + <edge from-layer="5429" from-port="2" to-layer="5431" to-port="0" /> + <edge from-layer="5430" from-port="0" to-layer="5431" to-port="1" /> + <edge from-layer="5431" from-port="2" to-layer="5432" to-port="0" /> + <edge from-layer="5432" from-port="1" to-layer="5433" to-port="1" /> + <edge from-layer="5433" from-port="2" to-layer="5434" to-port="1" /> + <edge from-layer="5434" from-port="2" to-layer="5659" to-port="0" /> + <edge from-layer="5434" from-port="2" to-layer="5439" to-port="0" /> + <edge from-layer="5434" from-port="2" to-layer="5447" to-port="0" /> + <edge from-layer="5435" from-port="0" to-layer="5436" to-port="0" /> + <edge from-layer="5436" from-port="1" to-layer="5446" to-port="0" /> + <edge from-layer="5437" from-port="0" to-layer="5438" to-port="0" /> + <edge from-layer="5438" from-port="1" to-layer="5439" to-port="1" /> + <edge from-layer="5439" from-port="2" to-layer="5441" to-port="0" /> + <edge from-layer="5440" from-port="0" to-layer="5441" to-port="1" /> + <edge from-layer="5441" from-port="2" to-layer="5444" to-port="0" /> + <edge from-layer="5442" from-port="0" to-layer="5443" to-port="0" /> + <edge from-layer="5443" from-port="1" to-layer="5444" to-port="1" /> + <edge from-layer="5444" from-port="2" to-layer="5445" to-port="0" /> + <edge from-layer="5445" from-port="1" to-layer="5446" to-port="1" /> + <edge from-layer="5446" from-port="2" to-layer="5447" to-port="1" /> + <edge from-layer="5447" from-port="2" to-layer="5448" to-port="1" /> + <edge from-layer="5448" from-port="2" to-layer="5565" to-port="0" /> + <edge from-layer="5448" from-port="2" to-layer="5457" to-port="0" /> + <edge from-layer="5448" from-port="2" to-layer="5486" to-port="0" /> + <edge from-layer="5448" from-port="2" to-layer="5458" to-port="0" /> + <edge from-layer="5449" from-port="0" to-layer="5450" to-port="0" /> + <edge from-layer="5450" from-port="1" to-layer="5453" to-port="0" /> + <edge from-layer="5451" from-port="0" to-layer="5452" to-port="0" /> + <edge from-layer="5452" from-port="1" to-layer="5453" to-port="1" /> + <edge from-layer="5453" from-port="2" to-layer="5455" to-port="0" /> + <edge from-layer="5454" from-port="0" to-layer="5455" to-port="1" /> + <edge from-layer="5455" from-port="2" to-layer="5456" to-port="0" /> + <edge from-layer="5456" from-port="1" to-layer="5457" to-port="1" /> + <edge from-layer="5457" from-port="2" to-layer="5463" to-port="0" /> + <edge from-layer="5458" from-port="1" to-layer="5461" to-port="0" /> + <edge from-layer="5459" from-port="0" to-layer="5461" to-port="1" /> + <edge from-layer="5460" from-port="0" to-layer="5461" to-port="2" /> + <edge from-layer="5461" from-port="3" to-layer="5566" to-port="0" /> + <edge from-layer="5461" from-port="3" to-layer="5462" to-port="0" /> + <edge from-layer="5461" from-port="3" to-layer="5648" to-port="0" /> + <edge from-layer="5462" from-port="3" to-layer="5463" to-port="1" /> + <edge from-layer="5462" from-port="3" to-layer="5487" to-port="1" /> + <edge from-layer="5463" from-port="2" to-layer="5465" to-port="0" /> + <edge from-layer="5464" from-port="0" to-layer="5465" to-port="1" /> + <edge from-layer="5465" from-port="2" to-layer="5466" to-port="1" /> + <edge from-layer="5466" from-port="2" to-layer="5614" to-port="0" /> + <edge from-layer="5466" from-port="2" to-layer="5615" to-port="0" /> + <edge from-layer="5466" from-port="2" to-layer="5467" to-port="0" /> + <edge from-layer="5468" from-port="0" to-layer="5469" to-port="0" /> + <edge from-layer="5469" from-port="1" to-layer="5474" to-port="0" /> + <edge from-layer="5470" from-port="0" to-layer="5473" to-port="1" /> + <edge from-layer="5471" from-port="0" to-layer="5473" to-port="2" /> + <edge from-layer="5472" from-port="0" to-layer="5473" to-port="3" /> + <edge from-layer="5473" from-port="4" to-layer="5474" to-port="1" /> + <edge from-layer="5474" from-port="2" to-layer="5475" to-port="0" /> + <edge from-layer="5475" from-port="1" to-layer="5477" to-port="0" /> + <edge from-layer="5476" from-port="0" to-layer="5477" to-port="2" /> + <edge from-layer="5477" from-port="3" to-layer="5525" to-port="0" /> + <edge from-layer="5478" from-port="0" to-layer="5479" to-port="0" /> + <edge from-layer="5479" from-port="1" to-layer="5482" to-port="0" /> + <edge from-layer="5480" from-port="0" to-layer="5481" to-port="0" /> + <edge from-layer="5481" from-port="1" to-layer="5482" to-port="1" /> + <edge from-layer="5482" from-port="2" to-layer="5484" to-port="0" /> + <edge from-layer="5483" from-port="0" to-layer="5484" to-port="1" /> + <edge from-layer="5484" from-port="2" to-layer="5485" to-port="0" /> + <edge from-layer="5485" from-port="1" to-layer="5486" to-port="1" /> + <edge from-layer="5486" from-port="2" to-layer="5487" to-port="0" /> + <edge from-layer="5487" from-port="2" to-layer="5489" to-port="0" /> + <edge from-layer="5488" from-port="0" to-layer="5489" to-port="1" /> + <edge from-layer="5489" from-port="2" to-layer="5513" to-port="0" /> + <edge from-layer="5489" from-port="2" to-layer="5502" to-port="0" /> + <edge from-layer="5489" from-port="2" to-layer="5499" to-port="0" /> + <edge from-layer="5489" from-port="2" to-layer="5519" to-port="0" /> + <edge from-layer="5490" from-port="0" to-layer="5491" to-port="0" /> + <edge from-layer="5491" from-port="1" to-layer="5492" to-port="0" /> + <edge from-layer="5492" from-port="2" to-layer="5493" to-port="0" /> + <edge from-layer="5493" from-port="2" to-layer="5495" to-port="0" /> + <edge from-layer="5494" from-port="0" to-layer="5495" to-port="1" /> + <edge from-layer="5495" from-port="2" to-layer="5496" to-port="0" /> + <edge from-layer="5495" from-port="2" to-layer="5496" to-port="1" /> + <edge from-layer="5496" from-port="2" to-layer="5497" to-port="0" /> + <edge from-layer="5496" from-port="2" to-layer="5521" to-port="0" /> + <edge from-layer="5497" from-port="1" to-layer="5498" to-port="0" /> + <edge from-layer="5498" from-port="2" to-layer="5570" to-port="1" /> + <edge from-layer="5498" from-port="2" to-layer="5499" to-port="1" /> + <edge from-layer="5499" from-port="2" to-layer="5524" to-port="0" /> + <edge from-layer="5500" from-port="0" to-layer="5510" to-port="0" /> + <edge from-layer="5501" from-port="0" to-layer="5510" to-port="1" /> + <edge from-layer="5502" from-port="1" to-layer="5504" to-port="0" /> + <edge from-layer="5503" from-port="0" to-layer="5504" to-port="2" /> + <edge from-layer="5504" from-port="3" to-layer="5505" to-port="0" /> + <edge from-layer="5505" from-port="2" to-layer="5506" to-port="0" /> + <edge from-layer="5506" from-port="1" to-layer="5508" to-port="0" /> + <edge from-layer="5507" from-port="0" to-layer="5508" to-port="1" /> + <edge from-layer="5508" from-port="2" to-layer="5510" to-port="2" /> + <edge from-layer="5509" from-port="0" to-layer="5510" to-port="3" /> + <edge from-layer="5510" from-port="4" to-layer="5519" to-port="2" /> + <edge from-layer="5510" from-port="4" to-layer="5513" to-port="1" /> + <edge from-layer="5511" from-port="0" to-layer="5513" to-port="2" /> + <edge from-layer="5512" from-port="0" to-layer="5513" to-port="3" /> + <edge from-layer="5513" from-port="4" to-layer="5516" to-port="0" /> + <edge from-layer="5514" from-port="0" to-layer="5515" to-port="0" /> + <edge from-layer="5515" from-port="1" to-layer="5516" to-port="1" /> + <edge from-layer="5516" from-port="2" to-layer="5520" to-port="0" /> + <edge from-layer="5517" from-port="0" to-layer="5519" to-port="1" /> + <edge from-layer="5518" from-port="0" to-layer="5519" to-port="3" /> + <edge from-layer="5519" from-port="4" to-layer="5520" to-port="1" /> + <edge from-layer="5520" from-port="2" to-layer="5523" to-port="0" /> + <edge from-layer="5521" from-port="1" to-layer="5522" to-port="0" /> + <edge from-layer="5522" from-port="2" to-layer="5523" to-port="1" /> + <edge from-layer="5522" from-port="2" to-layer="5592" to-port="1" /> + <edge from-layer="5523" from-port="2" to-layer="5524" to-port="1" /> + <edge from-layer="5524" from-port="2" to-layer="5525" to-port="1" /> + <edge from-layer="5525" from-port="2" to-layer="5594" to-port="0" /> + <edge from-layer="5525" from-port="2" to-layer="5595" to-port="0" /> + <edge from-layer="5525" from-port="2" to-layer="5526" to-port="0" /> + <edge from-layer="5555" from-port="0" to-layer="5556" to-port="0" /> + <edge from-layer="5556" from-port="1" to-layer="5719" to-port="0" /> + <edge from-layer="5557" from-port="0" to-layer="5558" to-port="0" /> + <edge from-layer="5558" from-port="1" to-layer="5561" to-port="0" /> + <edge from-layer="5559" from-port="0" to-layer="5560" to-port="0" /> + <edge from-layer="5560" from-port="1" to-layer="5561" to-port="1" /> + <edge from-layer="5561" from-port="2" to-layer="5563" to-port="0" /> + <edge from-layer="5562" from-port="0" to-layer="5563" to-port="1" /> + <edge from-layer="5563" from-port="2" to-layer="5564" to-port="0" /> + <edge from-layer="5564" from-port="1" to-layer="5565" to-port="1" /> + <edge from-layer="5565" from-port="2" to-layer="5567" to-port="0" /> + <edge from-layer="5566" from-port="3" to-layer="5567" to-port="1" /> + <edge from-layer="5567" from-port="2" to-layer="5569" to-port="0" /> + <edge from-layer="5568" from-port="0" to-layer="5569" to-port="1" /> + <edge from-layer="5569" from-port="2" to-layer="5590" to-port="0" /> + <edge from-layer="5569" from-port="2" to-layer="5573" to-port="0" /> + <edge from-layer="5569" from-port="2" to-layer="5570" to-port="0" /> + <edge from-layer="5569" from-port="2" to-layer="5584" to-port="0" /> + <edge from-layer="5570" from-port="2" to-layer="5593" to-port="0" /> + <edge from-layer="5571" from-port="0" to-layer="5581" to-port="0" /> + <edge from-layer="5572" from-port="0" to-layer="5581" to-port="1" /> + <edge from-layer="5573" from-port="1" to-layer="5575" to-port="0" /> + <edge from-layer="5574" from-port="0" to-layer="5575" to-port="2" /> + <edge from-layer="5575" from-port="3" to-layer="5576" to-port="0" /> + <edge from-layer="5576" from-port="2" to-layer="5577" to-port="0" /> + <edge from-layer="5577" from-port="1" to-layer="5579" to-port="0" /> + <edge from-layer="5578" from-port="0" to-layer="5579" to-port="1" /> + <edge from-layer="5579" from-port="2" to-layer="5581" to-port="2" /> + <edge from-layer="5580" from-port="0" to-layer="5581" to-port="3" /> + <edge from-layer="5581" from-port="4" to-layer="5584" to-port="1" /> + <edge from-layer="5581" from-port="4" to-layer="5590" to-port="2" /> + <edge from-layer="5582" from-port="0" to-layer="5584" to-port="2" /> + <edge from-layer="5583" from-port="0" to-layer="5584" to-port="3" /> + <edge from-layer="5584" from-port="4" to-layer="5587" to-port="0" /> + <edge from-layer="5585" from-port="0" to-layer="5586" to-port="0" /> + <edge from-layer="5586" from-port="1" to-layer="5587" to-port="1" /> + <edge from-layer="5587" from-port="2" to-layer="5591" to-port="0" /> + <edge from-layer="5588" from-port="0" to-layer="5590" to-port="1" /> + <edge from-layer="5589" from-port="0" to-layer="5590" to-port="3" /> + <edge from-layer="5590" from-port="4" to-layer="5591" to-port="1" /> + <edge from-layer="5591" from-port="2" to-layer="5592" to-port="0" /> + <edge from-layer="5592" from-port="2" to-layer="5593" to-port="1" /> + <edge from-layer="5593" from-port="2" to-layer="5645" to-port="0" /> + <edge from-layer="5594" from-port="2" to-layer="5608" to-port="0" /> + <edge from-layer="5595" from-port="1" to-layer="5600" to-port="0" /> + <edge from-layer="5595" from-port="1" to-layer="5605" to-port="0" /> + <edge from-layer="5595" from-port="1" to-layer="5598" to-port="0" /> + <edge from-layer="5596" from-port="0" to-layer="5598" to-port="1" /> + <edge from-layer="5597" from-port="0" to-layer="5598" to-port="2" /> + <edge from-layer="5598" from-port="3" to-layer="5606" to-port="0" /> + <edge from-layer="5598" from-port="3" to-layer="5612" to-port="0" /> + <edge from-layer="5599" from-port="0" to-layer="5600" to-port="2" /> + <edge from-layer="5600" from-port="3" to-layer="5609" to-port="0" /> + <edge from-layer="5600" from-port="3" to-layer="5602" to-port="0" /> + <edge from-layer="5601" from-port="0" to-layer="5602" to-port="1" /> + <edge from-layer="5602" from-port="2" to-layer="5606" to-port="1" /> + <edge from-layer="5603" from-port="0" to-layer="5605" to-port="1" /> + <edge from-layer="5604" from-port="0" to-layer="5605" to-port="2" /> + <edge from-layer="5605" from-port="3" to-layer="5612" to-port="2" /> + <edge from-layer="5605" from-port="3" to-layer="5606" to-port="3" /> + <edge from-layer="5606" from-port="4" to-layer="5607" to-port="0" /> + <edge from-layer="5607" from-port="1" to-layer="5608" to-port="1" /> + <edge from-layer="5608" from-port="2" to-layer="5613" to-port="0" /> + <edge from-layer="5609" from-port="2" to-layer="5611" to-port="0" /> + <edge from-layer="5610" from-port="0" to-layer="5611" to-port="1" /> + <edge from-layer="5611" from-port="2" to-layer="5612" to-port="1" /> + <edge from-layer="5612" from-port="3" to-layer="5613" to-port="1" /> + <edge from-layer="5613" from-port="2" to-layer="5645" to-port="1" /> + <edge from-layer="5613" from-port="2" to-layer="5637" to-port="0" /> + <edge from-layer="5614" from-port="2" to-layer="5628" to-port="0" /> + <edge from-layer="5615" from-port="1" to-layer="5625" to-port="0" /> + <edge from-layer="5615" from-port="1" to-layer="5620" to-port="0" /> + <edge from-layer="5615" from-port="1" to-layer="5618" to-port="0" /> + <edge from-layer="5616" from-port="0" to-layer="5618" to-port="1" /> + <edge from-layer="5617" from-port="0" to-layer="5618" to-port="2" /> + <edge from-layer="5618" from-port="3" to-layer="5626" to-port="0" /> + <edge from-layer="5618" from-port="3" to-layer="5632" to-port="0" /> + <edge from-layer="5619" from-port="0" to-layer="5620" to-port="2" /> + <edge from-layer="5620" from-port="3" to-layer="5622" to-port="0" /> + <edge from-layer="5620" from-port="3" to-layer="5629" to-port="0" /> + <edge from-layer="5621" from-port="0" to-layer="5622" to-port="1" /> + <edge from-layer="5622" from-port="2" to-layer="5626" to-port="1" /> + <edge from-layer="5623" from-port="0" to-layer="5625" to-port="1" /> + <edge from-layer="5624" from-port="0" to-layer="5625" to-port="2" /> + <edge from-layer="5625" from-port="3" to-layer="5626" to-port="3" /> + <edge from-layer="5625" from-port="3" to-layer="5632" to-port="2" /> + <edge from-layer="5626" from-port="4" to-layer="5627" to-port="0" /> + <edge from-layer="5627" from-port="1" to-layer="5628" to-port="1" /> + <edge from-layer="5628" from-port="2" to-layer="5633" to-port="0" /> + <edge from-layer="5629" from-port="2" to-layer="5631" to-port="0" /> + <edge from-layer="5630" from-port="0" to-layer="5631" to-port="1" /> + <edge from-layer="5631" from-port="2" to-layer="5632" to-port="1" /> + <edge from-layer="5632" from-port="3" to-layer="5633" to-port="1" /> + <edge from-layer="5633" from-port="2" to-layer="5645" to-port="2" /> + <edge from-layer="5634" from-port="0" to-layer="5644" to-port="1" /> + <edge from-layer="5635" from-port="0" to-layer="5642" to-port="0" /> + <edge from-layer="5636" from-port="0" to-layer="5642" to-port="1" /> + <edge from-layer="5637" from-port="1" to-layer="5640" to-port="0" /> + <edge from-layer="5638" from-port="0" to-layer="5640" to-port="1" /> + <edge from-layer="5639" from-port="0" to-layer="5640" to-port="2" /> + <edge from-layer="5640" from-port="3" to-layer="5642" to-port="2" /> + <edge from-layer="5641" from-port="0" to-layer="5642" to-port="3" /> + <edge from-layer="5642" from-port="4" to-layer="5644" to-port="2" /> + <edge from-layer="5643" from-port="0" to-layer="5644" to-port="3" /> + <edge from-layer="5644" from-port="4" to-layer="5645" to-port="3" /> + <edge from-layer="5645" from-port="4" to-layer="5647" to-port="0" /> + <edge from-layer="5646" from-port="0" to-layer="5647" to-port="1" /> + <edge from-layer="5647" from-port="2" to-layer="5649" to-port="0" /> + <edge from-layer="5648" from-port="2" to-layer="5649" to-port="1" /> + <edge from-layer="5649" from-port="2" to-layer="5658" to-port="0" /> + <edge from-layer="5650" from-port="0" to-layer="5651" to-port="0" /> + <edge from-layer="5651" from-port="1" to-layer="5654" to-port="0" /> + <edge from-layer="5652" from-port="0" to-layer="5653" to-port="0" /> + <edge from-layer="5653" from-port="1" to-layer="5654" to-port="1" /> + <edge from-layer="5654" from-port="2" to-layer="5656" to-port="0" /> + <edge from-layer="5655" from-port="0" to-layer="5656" to-port="1" /> + <edge from-layer="5656" from-port="2" to-layer="5657" to-port="0" /> + <edge from-layer="5657" from-port="1" to-layer="5658" to-port="1" /> + <edge from-layer="5658" from-port="2" to-layer="5659" to-port="1" /> + <edge from-layer="5659" from-port="2" to-layer="5705" to-port="0" /> + <edge from-layer="5659" from-port="2" to-layer="5674" to-port="0" /> + <edge from-layer="5659" from-port="2" to-layer="5666" to-port="0" /> + <edge from-layer="5660" from-port="0" to-layer="5661" to-port="0" /> + <edge from-layer="5661" from-port="1" to-layer="5675" to-port="0" /> + <edge from-layer="5662" from-port="0" to-layer="5663" to-port="0" /> + <edge from-layer="5663" from-port="1" to-layer="5673" to-port="0" /> + <edge from-layer="5664" from-port="0" to-layer="5665" to-port="0" /> + <edge from-layer="5665" from-port="1" to-layer="5666" to-port="1" /> + <edge from-layer="5666" from-port="2" to-layer="5668" to-port="0" /> + <edge from-layer="5667" from-port="0" to-layer="5668" to-port="1" /> + <edge from-layer="5668" from-port="2" to-layer="5671" to-port="0" /> + <edge from-layer="5669" from-port="0" to-layer="5670" to-port="0" /> + <edge from-layer="5670" from-port="1" to-layer="5671" to-port="1" /> + <edge from-layer="5671" from-port="2" to-layer="5672" to-port="0" /> + <edge from-layer="5672" from-port="1" to-layer="5673" to-port="1" /> + <edge from-layer="5673" from-port="2" to-layer="5674" to-port="1" /> + <edge from-layer="5674" from-port="2" to-layer="5675" to-port="1" /> + <edge from-layer="5675" from-port="2" to-layer="5694" to-port="0" /> + <edge from-layer="5675" from-port="2" to-layer="5684" to-port="0" /> + <edge from-layer="5676" from-port="0" to-layer="5677" to-port="0" /> + <edge from-layer="5677" from-port="1" to-layer="5680" to-port="0" /> + <edge from-layer="5678" from-port="0" to-layer="5679" to-port="0" /> + <edge from-layer="5679" from-port="1" to-layer="5680" to-port="1" /> + <edge from-layer="5680" from-port="2" to-layer="5682" to-port="0" /> + <edge from-layer="5681" from-port="0" to-layer="5682" to-port="1" /> + <edge from-layer="5682" from-port="2" to-layer="5683" to-port="0" /> + <edge from-layer="5683" from-port="1" to-layer="5684" to-port="1" /> + <edge from-layer="5684" from-port="2" to-layer="5685" to-port="0" /> + <edge from-layer="5685" from-port="1" to-layer="5695" to-port="0" /> + <edge from-layer="5686" from-port="0" to-layer="5687" to-port="0" /> + <edge from-layer="5687" from-port="1" to-layer="5690" to-port="0" /> + <edge from-layer="5688" from-port="0" to-layer="5689" to-port="0" /> + <edge from-layer="5689" from-port="1" to-layer="5690" to-port="1" /> + <edge from-layer="5690" from-port="2" to-layer="5692" to-port="0" /> + <edge from-layer="5691" from-port="0" to-layer="5692" to-port="1" /> + <edge from-layer="5692" from-port="2" to-layer="5693" to-port="0" /> + <edge from-layer="5693" from-port="1" to-layer="5694" to-port="1" /> + <edge from-layer="5694" from-port="2" to-layer="5695" to-port="1" /> + <edge from-layer="5695" from-port="2" to-layer="5704" to-port="0" /> + <edge from-layer="5696" from-port="0" to-layer="5697" to-port="0" /> + <edge from-layer="5697" from-port="1" to-layer="5700" to-port="0" /> + <edge from-layer="5698" from-port="0" to-layer="5699" to-port="0" /> + <edge from-layer="5699" from-port="1" to-layer="5700" to-port="1" /> + <edge from-layer="5700" from-port="2" to-layer="5702" to-port="0" /> + <edge from-layer="5701" from-port="0" to-layer="5702" to-port="1" /> + <edge from-layer="5702" from-port="2" to-layer="5703" to-port="0" /> + <edge from-layer="5703" from-port="1" to-layer="5704" to-port="1" /> + <edge from-layer="5704" from-port="2" to-layer="5705" to-port="1" /> + <edge from-layer="5705" from-port="2" to-layer="5710" to-port="0" /> + <edge from-layer="5705" from-port="2" to-layer="5718" to-port="0" /> + <edge from-layer="5706" from-port="0" to-layer="5707" to-port="0" /> + <edge from-layer="5707" from-port="1" to-layer="5717" to-port="0" /> + <edge from-layer="5708" from-port="0" to-layer="5709" to-port="0" /> + <edge from-layer="5709" from-port="1" to-layer="5710" to-port="1" /> + <edge from-layer="5710" from-port="2" to-layer="5712" to-port="0" /> + <edge from-layer="5711" from-port="0" to-layer="5712" to-port="1" /> + <edge from-layer="5712" from-port="2" to-layer="5715" to-port="0" /> + <edge from-layer="5713" from-port="0" to-layer="5714" to-port="0" /> + <edge from-layer="5714" from-port="1" to-layer="5715" to-port="1" /> + <edge from-layer="5715" from-port="2" to-layer="5716" to-port="0" /> + <edge from-layer="5716" from-port="1" to-layer="5717" to-port="1" /> + <edge from-layer="5717" from-port="2" to-layer="5718" to-port="1" /> + <edge from-layer="5718" from-port="2" to-layer="5719" to-port="1" /> + <edge from-layer="5719" from-port="2" to-layer="5728" to-port="0" /> + <edge from-layer="5720" from-port="0" to-layer="5721" to-port="0" /> + <edge from-layer="5721" from-port="1" to-layer="5724" to-port="0" /> + <edge from-layer="5722" from-port="0" to-layer="5723" to-port="0" /> + <edge from-layer="5723" from-port="1" to-layer="5724" to-port="1" /> + <edge from-layer="5724" from-port="2" to-layer="5726" to-port="0" /> + <edge from-layer="5725" from-port="0" to-layer="5726" to-port="1" /> + <edge from-layer="5726" from-port="2" to-layer="5727" to-port="0" /> + <edge from-layer="5727" from-port="1" to-layer="5728" to-port="1" /> + <edge from-layer="5728" from-port="2" to-layer="5729" to-port="0" /> + </edges> + <rt_info> + <Runtime_version value="2024.2.0-15229-1b5a6c49e9b" /> + <conversion_parameters> + <framework value="pytorch" /> + <is_python_object value="True" /> + </conversion_parameters> + <nncf> + <friendly_names_were_updated value="True" /> + <weight_compression> + <all_layers value="False" /> + <awq value="False" /> + <group_size value="-1" /> + <ignored_scope> + <names value="[]" /> + <patterns value="[]" /> + <subgraphs value="[]" /> + <types value="[]" /> + <validate value="True" /> + </ignored_scope> + <mode value="int8_asym" /> + <ratio value="1" /> + <sensitivity_metric value="weight_quantization_error" /> + </weight_compression> + </nncf> + <optimum> + <optimum_intel_version value="1.17.0.dev0+e1b6a59" /> + <optimum_version value="1.19.1" /> + <pytorch_version value="2.3.0+cpu" /> + <transformers_version value="4.40.1" /> + </optimum> + </rt_info> +</net>