Upload improved quantized ONNX weights (#2)
Browse files- Upload improved quantized ONNX weights (f20b810422e70b61f57a47ae23665c2ed2f2a44f)
- Update config.json (9a85d1c543b135e051aac844b88285849bfd6fb0)
- config.json +12 -17
- onnx/decoder_model_merged_quantized.onnx +2 -2
- onnx/encoder_model_quantized.onnx +2 -2
config.json
CHANGED
@@ -1,7 +1,4 @@
|
|
1 |
{
|
2 |
-
"_attn_implementation_autoset": true,
|
3 |
-
"_name_or_path": "Xenova/moonshine-base-hf",
|
4 |
-
"apply_spec_augment": false,
|
5 |
"architectures": [
|
6 |
"MoonshineForConditionalGeneration"
|
7 |
],
|
@@ -9,31 +6,29 @@
|
|
9 |
"attention_dropout": 0.0,
|
10 |
"bos_token_id": 1,
|
11 |
"decoder_hidden_act": "silu",
|
|
|
|
|
|
|
12 |
"decoder_start_token_id": 1,
|
13 |
"encoder_hidden_act": "gelu",
|
|
|
|
|
|
|
14 |
"eos_token_id": 2,
|
15 |
-
"ff_mult": 4,
|
16 |
"hidden_size": 416,
|
17 |
"initializer_range": 0.02,
|
18 |
-
"intermediate_size":
|
19 |
"is_encoder_decoder": true,
|
20 |
-
"
|
21 |
-
"mask_feature_length": 10,
|
22 |
-
"mask_feature_min_masks": 0,
|
23 |
-
"mask_feature_prob": 0.0,
|
24 |
-
"mask_time_length": 10,
|
25 |
-
"mask_time_min_masks": 2,
|
26 |
-
"mask_time_prob": 0.05,
|
27 |
-
"max_position_embeddings": 2048,
|
28 |
"model_type": "moonshine",
|
29 |
"num_attention_heads": 8,
|
30 |
"num_hidden_layers": 8,
|
31 |
"num_key_value_heads": 8,
|
32 |
-
"partial_rotary_factor": 0.
|
33 |
-
"qk_layernorm": false,
|
34 |
"rope_scaling": null,
|
35 |
"rope_theta": 10000.0,
|
36 |
-
"
|
|
|
37 |
"use_cache": true,
|
38 |
"vocab_size": 32768
|
39 |
-
}
|
|
|
1 |
{
|
|
|
|
|
|
|
2 |
"architectures": [
|
3 |
"MoonshineForConditionalGeneration"
|
4 |
],
|
|
|
6 |
"attention_dropout": 0.0,
|
7 |
"bos_token_id": 1,
|
8 |
"decoder_hidden_act": "silu",
|
9 |
+
"decoder_num_attention_heads": 8,
|
10 |
+
"decoder_num_hidden_layers": 8,
|
11 |
+
"decoder_num_key_value_heads": 8,
|
12 |
"decoder_start_token_id": 1,
|
13 |
"encoder_hidden_act": "gelu",
|
14 |
+
"encoder_num_attention_heads": 8,
|
15 |
+
"encoder_num_hidden_layers": 8,
|
16 |
+
"encoder_num_key_value_heads": 8,
|
17 |
"eos_token_id": 2,
|
|
|
18 |
"hidden_size": 416,
|
19 |
"initializer_range": 0.02,
|
20 |
+
"intermediate_size": 1664,
|
21 |
"is_encoder_decoder": true,
|
22 |
+
"max_position_embeddings": 512,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
23 |
"model_type": "moonshine",
|
24 |
"num_attention_heads": 8,
|
25 |
"num_hidden_layers": 8,
|
26 |
"num_key_value_heads": 8,
|
27 |
+
"partial_rotary_factor": 0.62,
|
|
|
28 |
"rope_scaling": null,
|
29 |
"rope_theta": 10000.0,
|
30 |
+
"torch_dtype": "float32",
|
31 |
+
"transformers_version": "4.48.0.dev0",
|
32 |
"use_cache": true,
|
33 |
"vocab_size": 32768
|
34 |
+
}
|
onnx/decoder_model_merged_quantized.onnx
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:cc9f3cd6698a369c6008b41aa60aa3fb3322e7f03c9bdf19d8e6b7200afca4f3
|
3 |
+
size 42498870
|
onnx/encoder_model_quantized.onnx
CHANGED
@@ -1,3 +1,3 @@
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
-
oid sha256:
|
3 |
-
size
|
|
|
1 |
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1dd9ab0a7f987113d30affcba5a068d11c8f90fa0223caa3e491ade431ad9751
|
3 |
+
size 20513063
|