diff --git "a/ndarray-cache-b16.json" "b/ndarray-cache-b16.json" --- "a/ndarray-cache-b16.json" +++ "b/ndarray-cache-b16.json" @@ -1,8 +1,8 @@ { "metadata": { - "ParamSize": 869, - "ParamBytes": 2595488384.0, - "BitsPerParam": 5.12927057608146 + "ParamSize": 724, + "ParamBytes": 3616698368.0, + "BitsPerParam": 7.147411884369364 }, "records": [ { @@ -4122,7 +4122,7 @@ { "dataPath": "params_shard_81.bin", "format": "raw-shard", - "nbytes": 32262720, + "nbytes": 32036864, "records": [ { "name": "model.h.9.mixer.out_proj.q_weight", @@ -4204,26 +4204,15 @@ "byteOffset": 21243904 }, { - "name": "vision_embed_tokens.img_processor.vision_model.embeddings.position_embedding.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.embeddings.position_embedding.weight", "shape": [ 577, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 295424, - "byteOffset": 22448128 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.embeddings.position_embedding.q_scale", - "shape": [ - 577, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 36928, - "byteOffset": 22743552 + "nbytes": 1181696, + "byteOffset": 22448128 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.layer_norm1.bias", @@ -4233,7 +4222,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 22780480 + "byteOffset": 23629824 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.layer_norm1.weight", @@ -4243,7 +4232,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 22782528 + "byteOffset": 23631872 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.layer_norm2.bias", @@ -4253,7 +4242,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 22784576 + "byteOffset": 23633920 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.layer_norm2.weight", @@ -4263,7 +4252,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 22786624 + "byteOffset": 23635968 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc1.bias", @@ -4273,29 +4262,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 22788672 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 22796864 + "byteOffset": 23638016 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 24894016 + "nbytes": 8388608, + "byteOffset": 23646208 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc2.bias", @@ -4305,29 +4283,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 25156160 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 25158208 - }, + "byteOffset": 32034816 + } + ], + "md5sum": "a7d9ae4deaea6c486bff82f9bde35623" + }, + { + "dataPath": "params_shard_82.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 27255360 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.k_proj.bias", @@ -4337,29 +4312,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 27517504 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 27519552 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 28043840 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.out_proj.bias", @@ -4369,29 +4333,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28109376 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 28111424 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 28635712 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.q_proj.bias", @@ -4401,29 +4354,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28701248 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 28703296 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 29227584 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.v_proj.bias", @@ -4433,29 +4375,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 29293120 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 29295168 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.0.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 29819456 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.layer_norm1.bias", @@ -4465,7 +4396,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 29884992 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.layer_norm1.weight", @@ -4475,7 +4406,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 29887040 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.layer_norm2.bias", @@ -4485,7 +4416,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 29889088 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.layer_norm2.weight", @@ -4495,7 +4426,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 29891136 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc1.bias", @@ -4505,29 +4436,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 29893184 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 29901376 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 31998528 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc2.bias", @@ -4537,37 +4457,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 32260672 + "byteOffset": 25190400 } ], - "md5sum": "71eb5c90f02b519dc17163d065af1427" + "md5sum": "c8b538e60f975970e8d0114577dee382" }, { - "dataPath": "params_shard_82.bin", + "dataPath": "params_shard_83.bin", "format": "raw-shard", - "nbytes": 33161216, + "nbytes": 25192448, "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 0 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 2097152 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.k_proj.bias", @@ -4577,29 +4486,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2359296 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 2361344 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 2885632 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.out_proj.bias", @@ -4609,29 +4507,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2951168 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 2953216 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 3477504 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.q_proj.bias", @@ -4641,29 +4528,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 3543040 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 3545088 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 4069376 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.v_proj.bias", @@ -4673,29 +4549,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4134912 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 4136960 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.1.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 4661248 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.layer_norm1.bias", @@ -4705,7 +4570,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4726784 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.layer_norm1.weight", @@ -4715,7 +4580,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4728832 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.layer_norm2.bias", @@ -4725,7 +4590,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4730880 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.layer_norm2.weight", @@ -4735,7 +4600,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4732928 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc1.bias", @@ -4745,29 +4610,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 4734976 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 4743168 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 6840320 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc2.bias", @@ -4777,29 +4631,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7102464 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 7104512 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "683e2f92ba10f7a9705f52c6cb51368e" + }, + { + "dataPath": "params_shard_84.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 9201664 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.k_proj.bias", @@ -4809,29 +4660,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 9463808 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.k_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.k_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 9465856 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.k_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 9990144 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.out_proj.bias", @@ -4841,29 +4681,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 10055680 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 10057728 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 10582016 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.q_proj.bias", @@ -4873,29 +4702,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 10647552 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 10649600 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 11173888 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.v_proj.bias", @@ -4905,29 +4723,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11239424 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 11241472 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.10.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 11765760 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.layer_norm1.bias", @@ -4937,7 +4744,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11831296 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.layer_norm1.weight", @@ -4947,7 +4754,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11833344 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.layer_norm2.bias", @@ -4957,7 +4764,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11835392 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.layer_norm2.weight", @@ -4967,7 +4774,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11837440 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc1.bias", @@ -4977,29 +4784,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 11839488 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 11847680 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 13944832 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc2.bias", @@ -5009,29 +4805,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14206976 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 14209024 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "adfafbe7d6959d9569c06ceadc8d7c34" + }, + { + "dataPath": "params_shard_85.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 16306176 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.k_proj.bias", @@ -5041,29 +4834,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 16568320 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 16570368 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 17094656 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.out_proj.bias", @@ -5073,29 +4855,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 17160192 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.out_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.out_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 17162240 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.out_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 17686528 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.q_proj.bias", @@ -5105,29 +4876,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 17752064 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 17754112 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 18278400 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.v_proj.bias", @@ -5137,29 +4897,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18343936 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 18345984 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.11.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 18870272 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.layer_norm1.bias", @@ -5169,7 +4918,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18935808 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.layer_norm1.weight", @@ -5179,7 +4928,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18937856 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.layer_norm2.bias", @@ -5189,7 +4938,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18939904 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.layer_norm2.weight", @@ -5199,7 +4948,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18941952 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc1.bias", @@ -5209,29 +4958,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 18944000 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 18952192 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 21049344 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc2.bias", @@ -5241,29 +4979,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21311488 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 21313536 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "24d41ecf557c5f984d40d215effcb6c9" + }, + { + "dataPath": "params_shard_86.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 23410688 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.k_proj.bias", @@ -5273,29 +5008,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23672832 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 23674880 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 24199168 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.out_proj.bias", @@ -5305,29 +5029,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 24264704 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 24266752 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 24791040 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.q_proj.bias", @@ -5337,29 +5050,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 24856576 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.q_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.q_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 24858624 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.q_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 25382912 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.v_proj.bias", @@ -5369,29 +5071,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 25448448 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 25450496 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.12.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 25974784 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.layer_norm1.bias", @@ -5401,7 +5092,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26040320 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.layer_norm1.weight", @@ -5411,7 +5102,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26042368 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.layer_norm2.bias", @@ -5421,7 +5112,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26044416 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.layer_norm2.weight", @@ -5431,7 +5122,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26046464 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc1.bias", @@ -5441,29 +5132,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 26048512 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 26056704 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 28153856 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc2.bias", @@ -5473,29 +5153,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28416000 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 28418048 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "73750663aec89f8ee4fa3d34c3795150" + }, + { + "dataPath": "params_shard_87.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 30515200 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.k_proj.bias", @@ -5505,29 +5182,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 30777344 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 30779392 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 31303680 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.out_proj.bias", @@ -5537,29 +5203,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 31369216 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 31371264 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 31895552 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.q_proj.bias", @@ -5569,29 +5224,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 31961088 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 31963136 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 32487424 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.v_proj.bias", @@ -5601,29 +5245,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 32552960 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.v_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.v_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 32555008 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.13.self_attn.v_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 33079296 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.layer_norm1.bias", @@ -5633,7 +5266,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33144832 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.layer_norm1.weight", @@ -5643,7 +5276,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33146880 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.layer_norm2.bias", @@ -5653,7 +5286,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33148928 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.layer_norm2.weight", @@ -5663,7 +5296,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33150976 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc1.bias", @@ -5673,37 +5306,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 33153024 - } - ], - "md5sum": "479e69b8802cd150602526108ff0ec95" - }, - { - "dataPath": "params_shard_83.bin", - "format": "raw-shard", - "nbytes": 33140736, - "records": [ - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 0 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 2097152 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc2.bias", @@ -5713,29 +5327,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2359296 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 2361344 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "a8fdf964a0bf7c0f2fd2e43b58cf8882" + }, + { + "dataPath": "params_shard_88.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 4458496 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.k_proj.bias", @@ -5745,29 +5356,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4720640 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 4722688 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 5246976 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.out_proj.bias", @@ -5777,29 +5377,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 5312512 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 5314560 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 5838848 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.q_proj.bias", @@ -5809,29 +5398,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 5904384 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 5906432 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 6430720 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.v_proj.bias", @@ -5841,29 +5419,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 6496256 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 6498304 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.14.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 7022592 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.layer_norm1.bias", @@ -5873,7 +5440,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7088128 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.layer_norm1.weight", @@ -5883,7 +5450,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7090176 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.layer_norm2.bias", @@ -5893,7 +5460,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7092224 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.layer_norm2.weight", @@ -5903,7 +5470,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7094272 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc1.bias", @@ -5913,29 +5480,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 7096320 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 7104512 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 9201664 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc2.bias", @@ -5945,29 +5501,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 9463808 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 9465856 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "6bedb2090ba553bf7ce53c5b11227500" + }, + { + "dataPath": "params_shard_89.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 11563008 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.k_proj.bias", @@ -5977,29 +5530,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11825152 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 11827200 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 12351488 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.out_proj.bias", @@ -6009,29 +5551,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 12417024 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 12419072 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 12943360 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.q_proj.bias", @@ -6041,29 +5572,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 13008896 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.q_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.q_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 13010944 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.q_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 13535232 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.v_proj.bias", @@ -6073,29 +5593,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 13600768 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.v_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.v_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 13602816 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.15.self_attn.v_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 14127104 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.layer_norm1.bias", @@ -6105,7 +5614,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14192640 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.layer_norm1.weight", @@ -6115,7 +5624,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14194688 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.layer_norm2.bias", @@ -6125,7 +5634,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14196736 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.layer_norm2.weight", @@ -6135,7 +5644,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14198784 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc1.bias", @@ -6145,29 +5654,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 14200832 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 14209024 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 16306176 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc2.bias", @@ -6177,29 +5675,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 16568320 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 16570368 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "7ba91f13cad2b1eb14419a63379f185d" + }, + { + "dataPath": "params_shard_90.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 18667520 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.k_proj.bias", @@ -6209,29 +5704,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18929664 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 18931712 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 19456000 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.out_proj.bias", @@ -6241,29 +5725,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 19521536 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.out_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.out_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 19523584 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.out_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 20047872 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.q_proj.bias", @@ -6273,29 +5746,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 20113408 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 20115456 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 20639744 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.v_proj.bias", @@ -6305,29 +5767,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 20705280 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.v_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.v_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 20707328 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.16.self_attn.v_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 21231616 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.layer_norm1.bias", @@ -6337,7 +5788,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21297152 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.layer_norm1.weight", @@ -6347,7 +5798,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21299200 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.layer_norm2.bias", @@ -6357,7 +5808,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21301248 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.layer_norm2.weight", @@ -6367,7 +5818,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21303296 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc1.bias", @@ -6377,29 +5828,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 21305344 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc1.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc1.weight", "shape": [ 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 21313536 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc1.q_scale", - "shape": [ - 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 23410688 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc2.bias", @@ -6409,29 +5849,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23672832 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 23674880 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "b00d568dfa33f1e269e838b50ae68cec" + }, + { + "dataPath": "params_shard_91.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 25772032 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.k_proj.bias", @@ -6441,29 +5878,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26034176 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 26036224 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 26560512 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.out_proj.bias", @@ -6473,29 +5899,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26626048 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.out_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.out_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 26628096 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.out_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 27152384 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.q_proj.bias", @@ -6505,29 +5920,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 27217920 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.q_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.q_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 27219968 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.q_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 27744256 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.v_proj.bias", @@ -6537,29 +5941,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 27809792 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.v_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.v_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 27811840 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.17.self_attn.v_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 28336128 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.layer_norm1.bias", @@ -6569,7 +5962,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28401664 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.layer_norm1.weight", @@ -6579,7 +5972,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28403712 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.layer_norm2.bias", @@ -6589,7 +5982,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28405760 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.layer_norm2.weight", @@ -6599,7 +5992,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28407808 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc1.bias", @@ -6609,29 +6002,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 28409856 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 28418048 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 30515200 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc2.bias", @@ -6641,29 +6023,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 30777344 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 30779392 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "9f6fa0f980598d760dc7579e92031891" + }, + { + "dataPath": "params_shard_92.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 32876544 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.k_proj.bias", @@ -6673,37 +6052,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33138688 - } - ], - "md5sum": "0d5f74cda6a152ad76b96fff892004bf" - }, - { - "dataPath": "params_shard_84.bin", - "format": "raw-shard", - "nbytes": 33161216, - "records": [ - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 0 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 524288 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.out_proj.bias", @@ -6713,29 +6073,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 589824 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.out_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.out_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 591872 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.out_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 1116160 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.q_proj.bias", @@ -6745,29 +6094,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 1181696 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.q_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.q_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 1183744 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.q_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 1708032 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.v_proj.bias", @@ -6777,29 +6115,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 1773568 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 1775616 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.18.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 2299904 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.layer_norm1.bias", @@ -6809,7 +6136,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2365440 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.layer_norm1.weight", @@ -6819,7 +6146,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2367488 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.layer_norm2.bias", @@ -6829,7 +6156,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2369536 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.layer_norm2.weight", @@ -6839,7 +6166,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2371584 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc1.bias", @@ -6849,29 +6176,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 2373632 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc1.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc1.weight", "shape": [ 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 2381824 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc1.q_scale", - "shape": [ - 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 4478976 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc2.bias", @@ -6881,29 +6197,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4741120 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 4743168 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "7c807048ef28f5cb8ec6ba0ffff9a5e6" + }, + { + "dataPath": "params_shard_93.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 6840320 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.k_proj.bias", @@ -6913,29 +6226,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7102464 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 7104512 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 7628800 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.out_proj.bias", @@ -6945,29 +6247,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7694336 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 7696384 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 8220672 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.q_proj.bias", @@ -6977,29 +6268,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 8286208 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 8288256 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 8812544 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.v_proj.bias", @@ -7009,29 +6289,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 8878080 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.v_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.v_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 8880128 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.19.self_attn.v_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 9404416 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.layer_norm1.bias", @@ -7041,7 +6310,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 9469952 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.layer_norm1.weight", @@ -7051,7 +6320,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 9472000 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.layer_norm2.bias", @@ -7061,7 +6330,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 9474048 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.layer_norm2.weight", @@ -7071,7 +6340,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 9476096 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc1.bias", @@ -7081,29 +6350,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 9478144 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc1.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc1.weight", "shape": [ 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 9486336 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc1.q_scale", - "shape": [ - 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 11583488 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc2.bias", @@ -7113,29 +6371,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11845632 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 11847680 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "0d349d70e68ac4565a701add22a3f848" + }, + { + "dataPath": "params_shard_94.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 13944832 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.k_proj.bias", @@ -7145,29 +6400,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14206976 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 14209024 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 14733312 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.out_proj.bias", @@ -7177,29 +6421,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14798848 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.out_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.out_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 14800896 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.out_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 15325184 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.q_proj.bias", @@ -7209,29 +6442,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 15390720 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 15392768 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 15917056 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.v_proj.bias", @@ -7241,29 +6463,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 15982592 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.v_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.v_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 15984640 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.2.self_attn.v_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 16508928 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.layer_norm1.bias", @@ -7273,7 +6484,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 16574464 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.layer_norm1.weight", @@ -7283,7 +6494,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 16576512 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.layer_norm2.bias", @@ -7293,7 +6504,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 16578560 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.layer_norm2.weight", @@ -7303,39 +6514,28 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 16580608 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc1.bias", "shape": [ - 4096 - ], - "dtype": "bfloat16", - "format": "raw", - "nbytes": 8192, - "byteOffset": 16582656 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 16590848 + 4096 + ], + "dtype": "bfloat16", + "format": "raw", + "nbytes": 8192, + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 18688000 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc2.bias", @@ -7345,29 +6545,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18950144 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 18952192 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "b035699cefdbe31b0aec42eb8e1b8031" + }, + { + "dataPath": "params_shard_95.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 21049344 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.k_proj.bias", @@ -7377,29 +6574,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21311488 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 21313536 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 21837824 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.out_proj.bias", @@ -7409,29 +6595,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21903360 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 21905408 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 22429696 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.q_proj.bias", @@ -7441,29 +6616,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 22495232 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 22497280 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 23021568 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.v_proj.bias", @@ -7473,29 +6637,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23087104 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 23089152 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.20.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 23613440 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.layer_norm1.bias", @@ -7505,7 +6658,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23678976 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.layer_norm1.weight", @@ -7515,7 +6668,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23681024 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.layer_norm2.bias", @@ -7525,7 +6678,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23683072 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.layer_norm2.weight", @@ -7535,7 +6688,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23685120 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc1.bias", @@ -7545,29 +6698,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 23687168 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 23695360 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 25792512 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc2.bias", @@ -7577,29 +6719,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26054656 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 26056704 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "55ca8afceb79e197e4f486c30fa99d2f" + }, + { + "dataPath": "params_shard_96.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 28153856 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.k_proj.bias", @@ -7609,29 +6748,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28416000 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 28418048 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 28942336 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.out_proj.bias", @@ -7641,29 +6769,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 29007872 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 29009920 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 29534208 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.q_proj.bias", @@ -7673,29 +6790,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 29599744 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 29601792 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 30126080 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.v_proj.bias", @@ -7705,29 +6811,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 30191616 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 30193664 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.21.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 30717952 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.layer_norm1.bias", @@ -7737,7 +6832,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 30783488 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.layer_norm1.weight", @@ -7747,7 +6842,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 30785536 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.layer_norm2.bias", @@ -7757,7 +6852,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 30787584 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.layer_norm2.weight", @@ -7767,7 +6862,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 30789632 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc1.bias", @@ -7777,29 +6872,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 30791680 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 30799872 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 32897024 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc2.bias", @@ -7809,37 +6893,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33159168 + "byteOffset": 25190400 } ], - "md5sum": "f83684667515df399fab68f06af18213" + "md5sum": "5a06a1571b9cacbc8316c5b4cd99abf1" }, { - "dataPath": "params_shard_85.bin", + "dataPath": "params_shard_97.bin", "format": "raw-shard", - "nbytes": 33161216, + "nbytes": 25192448, "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 0 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 2097152 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.k_proj.bias", @@ -7849,29 +6922,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2359296 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 2361344 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 2885632 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.out_proj.bias", @@ -7881,29 +6943,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2951168 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 2953216 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 3477504 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.q_proj.bias", @@ -7913,29 +6964,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 3543040 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 3545088 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 4069376 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.v_proj.bias", @@ -7945,29 +6985,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4134912 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 4136960 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.22.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 4661248 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.layer_norm1.bias", @@ -7977,7 +7006,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4726784 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.layer_norm1.weight", @@ -7987,7 +7016,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4728832 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.layer_norm2.bias", @@ -7997,7 +7026,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4730880 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.layer_norm2.weight", @@ -8007,7 +7036,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4732928 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc1.bias", @@ -8017,29 +7046,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 4734976 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 4743168 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 6840320 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc2.bias", @@ -8049,29 +7067,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7102464 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 7104512 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "344a5b0799be7fedcb55796de58bff4e" + }, + { + "dataPath": "params_shard_98.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 9201664 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.k_proj.bias", @@ -8081,29 +7096,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 9463808 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 9465856 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 9990144 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.out_proj.bias", @@ -8113,29 +7117,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 10055680 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 10057728 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 10582016 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.q_proj.bias", @@ -8144,30 +7137,19 @@ ], "dtype": "bfloat16", "format": "raw", - "nbytes": 2048, - "byteOffset": 10647552 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 10649600 + "nbytes": 2048, + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 11173888 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.v_proj.bias", @@ -8177,29 +7159,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11239424 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 11241472 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.23.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 11765760 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.layer_norm1.bias", @@ -8209,7 +7180,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11831296 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.layer_norm1.weight", @@ -8219,7 +7190,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11833344 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.layer_norm2.bias", @@ -8229,7 +7200,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11835392 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.layer_norm2.weight", @@ -8239,7 +7210,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11837440 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc1.bias", @@ -8249,29 +7220,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 11839488 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 11847680 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 13944832 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc2.bias", @@ -8281,29 +7241,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14206976 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 14209024 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "d98c360eec434f7dcda0392eaae948d5" + }, + { + "dataPath": "params_shard_99.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 16306176 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.k_proj.bias", @@ -8313,29 +7270,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 16568320 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 16570368 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 17094656 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.out_proj.bias", @@ -8345,29 +7291,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 17160192 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 17162240 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 17686528 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.q_proj.bias", @@ -8377,29 +7312,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 17752064 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 17754112 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 18278400 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.v_proj.bias", @@ -8409,29 +7333,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18343936 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 18345984 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.3.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 18870272 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.layer_norm1.bias", @@ -8441,7 +7354,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18935808 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.layer_norm1.weight", @@ -8451,7 +7364,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18937856 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.layer_norm2.bias", @@ -8461,7 +7374,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18939904 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.layer_norm2.weight", @@ -8471,7 +7384,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18941952 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc1.bias", @@ -8481,29 +7394,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 18944000 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 18952192 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 21049344 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc2.bias", @@ -8513,29 +7415,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21311488 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 21313536 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "bffa470a26c1841e2812df3f7309e1e8" + }, + { + "dataPath": "params_shard_100.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 23410688 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.k_proj.bias", @@ -8545,29 +7444,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23672832 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 23674880 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 24199168 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.out_proj.bias", @@ -8577,29 +7465,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 24264704 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 24266752 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 24791040 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.q_proj.bias", @@ -8609,29 +7486,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 24856576 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 24858624 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 25382912 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.v_proj.bias", @@ -8641,29 +7507,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 25448448 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 25450496 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.4.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 25974784 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.layer_norm1.bias", @@ -8673,7 +7528,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26040320 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.layer_norm1.weight", @@ -8683,7 +7538,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26042368 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.layer_norm2.bias", @@ -8693,7 +7548,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26044416 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.layer_norm2.weight", @@ -8703,7 +7558,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26046464 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc1.bias", @@ -8713,29 +7568,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 26048512 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 26056704 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 28153856 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc2.bias", @@ -8745,29 +7589,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28416000 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 28418048 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "83b41f7765f019cc8a8166b401989905" + }, + { + "dataPath": "params_shard_101.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 30515200 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.k_proj.bias", @@ -8777,29 +7618,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 30777344 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 30779392 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 31303680 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.out_proj.bias", @@ -8809,29 +7639,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 31369216 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 31371264 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 31895552 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.q_proj.bias", @@ -8841,29 +7660,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 31961088 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 31963136 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 32487424 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.v_proj.bias", @@ -8873,29 +7681,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 32552960 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 32555008 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.5.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 33079296 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.layer_norm1.bias", @@ -8905,7 +7702,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33144832 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.layer_norm1.weight", @@ -8915,7 +7712,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33146880 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.layer_norm2.bias", @@ -8925,7 +7722,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33148928 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.layer_norm2.weight", @@ -8935,7 +7732,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 33150976 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc1.bias", @@ -8945,37 +7742,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 33153024 - } - ], - "md5sum": "e959b0c27c59fabcc0443c6b269f21c5" - }, - { - "dataPath": "params_shard_86.bin", - "format": "raw-shard", - "nbytes": 28416000, - "records": [ - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 0 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 2097152 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc2.bias", @@ -8985,29 +7763,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 2359296 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 2361344 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "49f40ce6f5f45f4d3717724aab652667" + }, + { + "dataPath": "params_shard_102.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 4458496 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.k_proj.bias", @@ -9017,29 +7792,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 4720640 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.k_proj.q_weight", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.k_proj.weight", "shape": [ 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 4722688 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.k_proj.q_scale", - "shape": [ - 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 5246976 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.out_proj.bias", @@ -9049,29 +7813,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 5312512 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 5314560 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 5838848 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.q_proj.bias", @@ -9081,29 +7834,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 5904384 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 5906432 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 6430720 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.v_proj.bias", @@ -9113,29 +7855,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 6496256 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 6498304 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.6.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 7022592 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.layer_norm1.bias", @@ -9145,7 +7876,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7088128 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.layer_norm1.weight", @@ -9155,7 +7886,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7090176 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.layer_norm2.bias", @@ -9165,7 +7896,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7092224 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.layer_norm2.weight", @@ -9175,7 +7906,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 7094272 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc1.bias", @@ -9185,29 +7916,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 7096320 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 7104512 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 9201664 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc2.bias", @@ -9217,29 +7937,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 9463808 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 9465856 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "5cd95026a2d7f8360f1d3bf2e2f2ff7f" + }, + { + "dataPath": "params_shard_103.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 11563008 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.k_proj.bias", @@ -9249,29 +7966,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 11825152 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 11827200 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 12351488 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.out_proj.bias", @@ -9281,29 +7987,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 12417024 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 12419072 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 12943360 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.q_proj.bias", @@ -9313,29 +8008,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 13008896 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 13010944 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 13535232 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.v_proj.bias", @@ -9345,29 +8029,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 13600768 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 13602816 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.7.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 14127104 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.layer_norm1.bias", @@ -9377,7 +8050,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14192640 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.layer_norm1.weight", @@ -9387,7 +8060,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14194688 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.layer_norm2.bias", @@ -9397,7 +8070,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14196736 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.layer_norm2.weight", @@ -9407,7 +8080,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 14198784 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc1.bias", @@ -9417,29 +8090,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 14200832 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 14209024 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 16306176 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc2.bias", @@ -9449,29 +8111,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 16568320 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 16570368 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "04877e8c02cbd383419f46beaebe7d6a" + }, + { + "dataPath": "params_shard_104.bin", + "format": "raw-shard", + "nbytes": 25192448, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 18667520 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.k_proj.bias", @@ -9481,29 +8140,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 18929664 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 18931712 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 19456000 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.out_proj.bias", @@ -9513,29 +8161,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 19521536 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 19523584 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 20047872 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.q_proj.bias", @@ -9545,29 +8182,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 20113408 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 20115456 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 20639744 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.v_proj.bias", @@ -9577,29 +8203,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 20705280 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 20707328 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.8.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 21231616 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.layer_norm1.bias", @@ -9609,7 +8224,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21297152 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.layer_norm1.weight", @@ -9619,7 +8234,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21299200 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.layer_norm2.bias", @@ -9629,7 +8244,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21301248 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.layer_norm2.weight", @@ -9639,7 +8254,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 21303296 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc1.bias", @@ -9649,29 +8264,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 21305344 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc1.q_weight", - "shape": [ - 4096, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 21313536 + "byteOffset": 16793600 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc1.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc1.weight", "shape": [ 4096, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 23410688 + "nbytes": 8388608, + "byteOffset": 16801792 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc2.bias", @@ -9681,29 +8285,26 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 23672832 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc2.q_weight", - "shape": [ - 1024, - 512 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 2097152, - "byteOffset": 23674880 - }, + "byteOffset": 25190400 + } + ], + "md5sum": "7cf1ead00f736b6144ae39774c21f3f2" + }, + { + "dataPath": "params_shard_105.bin", + "format": "raw-shard", + "nbytes": 29200384, + "records": [ { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc2.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.mlp.fc2.weight", "shape": [ 1024, - 128 + 4096 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 262144, - "byteOffset": 25772032 + "nbytes": 8388608, + "byteOffset": 0 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.k_proj.bias", @@ -9713,29 +8314,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26034176 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.k_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 26036224 + "byteOffset": 8388608 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.k_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.k_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 26560512 + "nbytes": 2097152, + "byteOffset": 8390656 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.out_proj.bias", @@ -9745,29 +8335,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 26626048 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.out_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 26628096 + "byteOffset": 10487808 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.out_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.out_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 27152384 + "nbytes": 2097152, + "byteOffset": 10489856 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.q_proj.bias", @@ -9777,29 +8356,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 27217920 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.q_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 27219968 + "byteOffset": 12587008 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.q_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.q_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 27744256 + "nbytes": 2097152, + "byteOffset": 12589056 }, { "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.v_proj.bias", @@ -9809,29 +8377,18 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 27809792 - }, - { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.v_proj.q_weight", - "shape": [ - 1024, - 128 - ], - "dtype": "uint32", - "format": "f32-to-bf16", - "nbytes": 524288, - "byteOffset": 27811840 + "byteOffset": 14686208 }, { - "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.v_proj.q_scale", + "name": "vision_embed_tokens.img_processor.vision_model.encoder.layers.9.self_attn.v_proj.weight", "shape": [ 1024, - 32 + 1024 ], "dtype": "bfloat16", "format": "raw", - "nbytes": 65536, - "byteOffset": 28336128 + "nbytes": 2097152, + "byteOffset": 14688256 }, { "name": "vision_embed_tokens.img_processor.vision_model.post_layernorm.bias", @@ -9841,7 +8398,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28401664 + "byteOffset": 16785408 }, { "name": "vision_embed_tokens.img_processor.vision_model.post_layernorm.weight", @@ -9851,7 +8408,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28403712 + "byteOffset": 16787456 }, { "name": "vision_embed_tokens.img_processor.vision_model.pre_layrnorm.bias", @@ -9861,7 +8418,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28405760 + "byteOffset": 16789504 }, { "name": "vision_embed_tokens.img_processor.vision_model.pre_layrnorm.weight", @@ -9871,7 +8428,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 2048, - "byteOffset": 28407808 + "byteOffset": 16791552 }, { "name": "vision_embed_tokens.img_projection.linear_1.bias", @@ -9881,16 +8438,8 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 6144, - "byteOffset": 28409856 - } - ], - "md5sum": "19070c9325e7d2efe559224ba001ae56" - }, - { - "dataPath": "params_shard_87.bin", - "format": "raw-shard", - "nbytes": 12400640, - "records": [ + "byteOffset": 16793600 + }, { "name": "vision_embed_tokens.img_projection.linear_1.q_weight", "shape": [ @@ -9900,7 +8449,7 @@ "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 6291456, - "byteOffset": 0 + "byteOffset": 16799744 }, { "name": "vision_embed_tokens.img_projection.linear_1.q_scale", @@ -9911,7 +8460,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 786432, - "byteOffset": 6291456 + "byteOffset": 23091200 }, { "name": "vision_embed_tokens.img_projection.linear_2.bias", @@ -9921,7 +8470,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 6144, - "byteOffset": 7077888 + "byteOffset": 23877632 }, { "name": "vision_embed_tokens.img_projection.linear_2.q_weight", @@ -9932,7 +8481,7 @@ "dtype": "uint32", "format": "f32-to-bf16", "nbytes": 4718592, - "byteOffset": 7084032 + "byteOffset": 23883776 }, { "name": "vision_embed_tokens.img_projection.linear_2.q_scale", @@ -9943,7 +8492,7 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 589824, - "byteOffset": 11802624 + "byteOffset": 28602368 }, { "name": "vision_embed_tokens.sub_GN", @@ -9956,10 +8505,10 @@ "dtype": "bfloat16", "format": "raw", "nbytes": 8192, - "byteOffset": 12392448 + "byteOffset": 29192192 } ], - "md5sum": "0e6cede73fde964dc6681f3ad5f7c16b" + "md5sum": "205528c07e65479e4ca999084a25016b" } ] } \ No newline at end of file