Hugging Face's logo

onnx-community
/
lite-whisper-large-v3-turbo-ONNX

Automatic Speech Recognition
Transformers.js
ONNX
lite-whisper
text2text-generation
Model card Files Community
Xenova HF staff commited on
Commit
b07bfcc
·
verified ·
1 Parent(s): 776c726

Upload folder using huggingface_hub

Browse files
Files changed (34) hide show
  1. config.json +283 -0
  2. generation_config.json +12 -0
  3. onnx/decoder_model.onnx +3 -0
  4. onnx/decoder_model_bnb4.onnx +3 -0
  5. onnx/decoder_model_fp16.onnx +3 -0
  6. onnx/decoder_model_int8.onnx +3 -0
  7. onnx/decoder_model_merged.onnx +3 -0
  8. onnx/decoder_model_merged_bnb4.onnx +3 -0
  9. onnx/decoder_model_merged_fp16.onnx +3 -0
  10. onnx/decoder_model_merged_int8.onnx +3 -0
  11. onnx/decoder_model_merged_q4.onnx +3 -0
  12. onnx/decoder_model_merged_q4f16.onnx +3 -0
  13. onnx/decoder_model_merged_quantized.onnx +3 -0
  14. onnx/decoder_model_merged_uint8.onnx +3 -0
  15. onnx/decoder_model_q4.onnx +3 -0
  16. onnx/decoder_model_q4f16.onnx +3 -0
  17. onnx/decoder_model_quantized.onnx +3 -0
  18. onnx/decoder_model_uint8.onnx +3 -0
  19. onnx/decoder_with_past_model.onnx +3 -0
  20. onnx/decoder_with_past_model_bnb4.onnx +3 -0
  21. onnx/decoder_with_past_model_fp16.onnx +3 -0
  22. onnx/decoder_with_past_model_int8.onnx +3 -0
  23. onnx/decoder_with_past_model_q4.onnx +3 -0
  24. onnx/decoder_with_past_model_q4f16.onnx +3 -0
  25. onnx/decoder_with_past_model_quantized.onnx +3 -0
  26. onnx/decoder_with_past_model_uint8.onnx +3 -0
  27. onnx/encoder_model.onnx +3 -0
  28. onnx/encoder_model_bnb4.onnx +3 -0
  29. onnx/encoder_model_fp16.onnx +3 -0
  30. onnx/encoder_model_int8.onnx +3 -0
  31. onnx/encoder_model_q4.onnx +3 -0
  32. onnx/encoder_model_q4f16.onnx +3 -0
  33. onnx/encoder_model_quantized.onnx +3 -0
  34. onnx/encoder_model_uint8.onnx +3 -0
config.json ADDED
@@ -0,0 +1,283 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_attn_implementation_autoset": true,
3
+ "_name_or_path": "efficient-speech/lite-whisper-large-v3-turbo",
4
+ "activation_dropout": 0.0,
5
+ "activation_function": "gelu",
6
+ "apply_spec_augment": false,
7
+ "architectures": [
8
+ "LiteWhisperForConditionalGeneration"
9
+ ],
10
+ "attention_dropout": 0.0,
11
+ "auto_map": {
12
+ "AutoConfig": "efficient-speech/lite-whisper-large-v3-turbo--configuration_lite_whisper.LiteWhisperConfig",
13
+ "AutoModel": "efficient-speech/lite-whisper-large-v3-turbo--modeling_lite_whisper.LiteWhisperForConditionalGeneration"
14
+ },
15
+ "begin_suppress_tokens": [
16
+ 220,
17
+ 50256
18
+ ],
19
+ "bos_token_id": 50257,
20
+ "classifier_proj_size": 256,
21
+ "d_model": 1280,
22
+ "decoder_attention_heads": 20,
23
+ "decoder_ffn_dim": 5120,
24
+ "decoder_layerdrop": 0.0,
25
+ "decoder_layers": 4,
26
+ "decoder_start_token_id": 50258,
27
+ "dropout": 0.0,
28
+ "encoder_attention_heads": 20,
29
+ "encoder_ffn_dim": 5120,
30
+ "encoder_layerdrop": 0.0,
31
+ "encoder_layers": 32,
32
+ "eos_token_id": 50257,
33
+ "init_std": 0.02,
34
+ "is_encoder_decoder": true,
35
+ "low_rank_config": [
36
+ {
37
+ "fc1": 352,
38
+ "fc2": 304,
39
+ "k_proj": 32,
40
+ "out_proj": 32,
41
+ "q_proj": 32,
42
+ "v_proj": 48
43
+ },
44
+ {
45
+ "fc1": 240,
46
+ "fc2": 256,
47
+ "k_proj": 48,
48
+ "out_proj": 80,
49
+ "q_proj": 48,
50
+ "v_proj": 112
51
+ },
52
+ {
53
+ "fc1": 192,
54
+ "fc2": 144,
55
+ "k_proj": 32,
56
+ "out_proj": 112,
57
+ "q_proj": 32,
58
+ "v_proj": 112
59
+ },
60
+ {
61
+ "fc1": 256,
62
+ "fc2": 128,
63
+ "k_proj": 48,
64
+ "out_proj": 160,
65
+ "q_proj": 48,
66
+ "v_proj": 176
67
+ },
68
+ {
69
+ "fc1": 272,
70
+ "fc2": 192,
71
+ "k_proj": 48,
72
+ "out_proj": 112,
73
+ "q_proj": 48,
74
+ "v_proj": 176
75
+ },
76
+ {
77
+ "fc1": 272,
78
+ "fc2": 304,
79
+ "k_proj": 32,
80
+ "out_proj": 112,
81
+ "q_proj": 32,
82
+ "v_proj": 160
83
+ },
84
+ {
85
+ "fc1": 352,
86
+ "fc2": 320,
87
+ "k_proj": 64,
88
+ "out_proj": 192,
89
+ "q_proj": 64,
90
+ "v_proj": 304
91
+ },
92
+ {
93
+ "fc1": 368,
94
+ "fc2": 560,
95
+ "k_proj": 64,
96
+ "out_proj": 192,
97
+ "q_proj": 80,
98
+ "v_proj": 272
99
+ },
100
+ {
101
+ "fc1": 464,
102
+ "fc2": 672,
103
+ "k_proj": 64,
104
+ "out_proj": 224,
105
+ "q_proj": 80,
106
+ "v_proj": 272
107
+ },
108
+ {
109
+ "fc1": 464,
110
+ "fc2": 688,
111
+ "k_proj": 80,
112
+ "out_proj": 224,
113
+ "q_proj": 96,
114
+ "v_proj": 304
115
+ },
116
+ {
117
+ "fc1": 528,
118
+ "fc2": 720,
119
+ "k_proj": 48,
120
+ "out_proj": 240,
121
+ "q_proj": 48,
122
+ "v_proj": 288
123
+ },
124
+ {
125
+ "fc1": 512,
126
+ "fc2": 720,
127
+ "k_proj": 64,
128
+ "out_proj": 224,
129
+ "q_proj": 64,
130
+ "v_proj": 320
131
+ },
132
+ {
133
+ "fc1": 544,
134
+ "fc2": 800,
135
+ "k_proj": 48,
136
+ "out_proj": 224,
137
+ "q_proj": 64,
138
+ "v_proj": 368
139
+ },
140
+ {
141
+ "fc1": 624,
142
+ "fc2": 928,
143
+ "k_proj": 48,
144
+ "out_proj": 272,
145
+ "q_proj": 48,
146
+ "v_proj": 304
147
+ },
148
+ {
149
+ "fc1": 704,
150
+ "fc2": 880,
151
+ "k_proj": 64,
152
+ "out_proj": 288,
153
+ "q_proj": 64,
154
+ "v_proj": 464
155
+ },
156
+ {
157
+ "fc1": 592,
158
+ "fc2": 992,
159
+ "k_proj": 32,
160
+ "out_proj": 352,
161
+ "q_proj": 48,
162
+ "v_proj": 400
163
+ },
164
+ {
165
+ "fc1": 640,
166
+ "fc2": 1024,
167
+ "k_proj": 48,
168
+ "out_proj": 224,
169
+ "q_proj": 48,
170
+ "v_proj": 288
171
+ },
172
+ {
173
+ "fc1": 736,
174
+ "k_proj": 80,
175
+ "out_proj": 304,
176
+ "q_proj": 96,
177
+ "v_proj": 384
178
+ },
179
+ {
180
+ "fc1": 800,
181
+ "k_proj": 64,
182
+ "out_proj": 272,
183
+ "q_proj": 64,
184
+ "v_proj": 320
185
+ },
186
+ {
187
+ "fc1": 848,
188
+ "k_proj": 64,
189
+ "out_proj": 272,
190
+ "q_proj": 64,
191
+ "v_proj": 416
192
+ },
193
+ {
194
+ "fc1": 832,
195
+ "fc2": 16,
196
+ "k_proj": 64,
197
+ "out_proj": 352,
198
+ "q_proj": 64,
199
+ "v_proj": 432
200
+ },
201
+ {
202
+ "fc1": 976,
203
+ "k_proj": 96,
204
+ "out_proj": 368,
205
+ "q_proj": 112,
206
+ "v_proj": 512
207
+ },
208
+ {
209
+ "fc1": 960,
210
+ "k_proj": 112,
211
+ "out_proj": 496,
212
+ "q_proj": 112,
213
+ "v_proj": 576
214
+ },
215
+ {
216
+ "k_proj": 96,
217
+ "out_proj": 496,
218
+ "q_proj": 112,
219
+ "v_proj": 528
220
+ },
221
+ {
222
+ "k_proj": 144,
223
+ "out_proj": 512,
224
+ "q_proj": 160,
225
+ "v_proj": 624
226
+ },
227
+ {
228
+ "k_proj": 128,
229
+ "out_proj": 464,
230
+ "q_proj": 144
231
+ },
232
+ {
233
+ "k_proj": 160,
234
+ "out_proj": 576,
235
+ "q_proj": 176
236
+ },
237
+ {
238
+ "k_proj": 160,
239
+ "out_proj": 560,
240
+ "q_proj": 192
241
+ },
242
+ {
243
+ "k_proj": 176,
244
+ "out_proj": 576,
245
+ "q_proj": 192
246
+ },
247
+ {
248
+ "k_proj": 240,
249
+ "out_proj": 624,
250
+ "q_proj": 272,
251
+ "v_proj": 608
252
+ },
253
+ {
254
+ "k_proj": 272,
255
+ "out_proj": 640,
256
+ "q_proj": 320,
257
+ "v_proj": 640
258
+ },
259
+ {
260
+ "k_proj": 272,
261
+ "q_proj": 320,
262
+ "v_proj": 592
263
+ }
264
+ ],
265
+ "mask_feature_length": 10,
266
+ "mask_feature_min_masks": 0,
267
+ "mask_feature_prob": 0.0,
268
+ "mask_time_length": 10,
269
+ "mask_time_min_masks": 2,
270
+ "mask_time_prob": 0.05,
271
+ "max_source_positions": 1500,
272
+ "max_target_positions": 448,
273
+ "median_filter_width": 7,
274
+ "model_type": "lite-whisper",
275
+ "num_hidden_layers": 32,
276
+ "num_mel_bins": 128,
277
+ "pad_token_id": 50257,
278
+ "scale_embedding": false,
279
+ "transformers_version": "4.48.3",
280
+ "use_cache": true,
281
+ "use_weighted_layer_sum": false,
282
+ "vocab_size": 51866
283
+ }
generation_config.json ADDED
@@ -0,0 +1,12 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "begin_suppress_tokens": [
4
+ 220,
5
+ 50256
6
+ ],
7
+ "bos_token_id": 50257,
8
+ "decoder_start_token_id": 50258,
9
+ "eos_token_id": 50257,
10
+ "pad_token_id": 50257,
11
+ "transformers_version": "4.48.3"
12
+ }
onnx/decoder_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:18b38a8941b60f29099d7f9af24d3fcf6c259492b1e181d3b97e38b243d0a9f6
3
+ size 687782706
onnx/decoder_model_bnb4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:50e018ca780b0397e7b267741c194480484823deb8d02ab00b324a359f186d05
3
+ size 327340989
onnx/decoder_model_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2faafc57a07b502db0e3daa92eace0e7ce83bf46861cf9bdb8162df97ea28be5
3
+ size 343968493
onnx/decoder_model_int8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:933f05ea4171d5f17402e5b71cebc22f3fd3a4a8f99486119ba36ec51716d443
3
+ size 437927549
onnx/decoder_model_merged.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:16908afdadaaa0376e2fc27e9bf4f741013e49995249dfac38bf43dc82ab6b21
3
+ size 687963426
onnx/decoder_model_merged_bnb4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be5071adb576711c37bfd8a607d452c3377b2cb75bde6fbe5d647bc5ce9d626b
3
+ size 327526397
onnx/decoder_model_merged_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bd3150fb834ac0a85318c1fcda7c1dad7dfa8c3f9ed4f653d68f9a9bb68222b
3
+ size 661764234
onnx/decoder_model_merged_int8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89ffbd76b5ad8fb232c08f7bc1d7fd23cbc7ef7d274a14a6b9c9ae3ff6b8aa2f
3
+ size 438149217
onnx/decoder_model_merged_q4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d3e8d0af4e7feca8d7cc71dd5519d1c27917fc93b72a156f8d0514f116c39648
3
+ size 334079385
onnx/decoder_model_merged_q4f16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:083607a212890a22cd081f17b8258c1df76b18e2952e0654c07689908882d959
3
+ size 379152361
onnx/decoder_model_merged_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:89ffbd76b5ad8fb232c08f7bc1d7fd23cbc7ef7d274a14a6b9c9ae3ff6b8aa2f
3
+ size 438149217
onnx/decoder_model_merged_uint8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:34d1f9ea7d9e08f7b799cb7ef55783c8051108ef70c65a048c84aa65c5d06cf3
3
+ size 438149238
onnx/decoder_model_q4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd93d4438c480ae152316106d9766d219429007a19befddcb01856d12205c860
3
+ size 333894265
onnx/decoder_model_q4f16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ac893fd97634833b8bfa35402ee1e3f98c940d2b08ad07af3dabd590a68e6a5f
3
+ size 193241747
onnx/decoder_model_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:933f05ea4171d5f17402e5b71cebc22f3fd3a4a8f99486119ba36ec51716d443
3
+ size 437927549
onnx/decoder_model_uint8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8e1cc209f553fb8170169dcafe8400f7b6a633b74f002241801b1dcd75c30b88
3
+ size 437927570
onnx/decoder_with_past_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5a7bef3b1db03a996beb617d2d814cbb045f9b02c42f9753821b6b24707e4d05
3
+ size 635300182
onnx/decoder_with_past_model_bnb4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8bf6860b06a7321d8cf7cf2de400e5f2151c6310f7d85f76def1f98c15391deb
3
+ size 319913217
onnx/decoder_with_past_model_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0f4beb01c3fbe585b49616fa7a5fa499c0d71b07dd72da1ae8cf1802c3bd6d2
3
+ size 317714162
onnx/decoder_with_past_model_int8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7d2222b05959c7f9ab3b0231e7462e0ca13a887919b9fc0e5c70f04956c32bc
3
+ size 424758658
onnx/decoder_with_past_model_q4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:463c49d86d8fc2fe78f7dc3033766717bc4269763147317fd3e337b11f771605
3
+ size 325647357
onnx/decoder_with_past_model_q4f16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:48a7f0f8392a3086dc0cc4c8c04ee9a3f0cf192d14a1263f8ecb5aeceea61793
3
+ size 185827736
onnx/decoder_with_past_model_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7d2222b05959c7f9ab3b0231e7462e0ca13a887919b9fc0e5c70f04956c32bc
3
+ size 424758658
onnx/decoder_with_past_model_uint8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e60240d04cd89aae85df90663c4738664d26dc9e4ff3ef7dc0e708370b3ffa66
3
+ size 424758674
onnx/encoder_model.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:004bdc0d8847302615df5e35d94bf23c7ab3d7f8b2f6c71d655daab939338ea8
3
+ size 1502694019
onnx/encoder_model_bnb4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:557ff22f1694432156f9294c2c44425c59191d6131cb5bdaba0fe57ab57d1a30
3
+ size 238797049
onnx/encoder_model_fp16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2b6b5700f095cb3cbc24a4c021518cbb630a73d4ae446e9a97ef4c29f0115d06
3
+ size 751569592
onnx/encoder_model_int8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ae0c159ff825a7c6ae063e1f587fc5b71f897c33fc57fd9ea84f9c98d084d13
3
+ size 383783309
onnx/encoder_model_q4.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a293f1bc52367fba0b58540cb84fcbefce69dfbf9437c1dc5d38f4acf8882a7
3
+ size 263247436
onnx/encoder_model_q4f16.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32fa7fa91da715c323379dc4da1e5853d52fd3d0b0d6c608f68d0c12e564e49f
3
+ size 224390539
onnx/encoder_model_quantized.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54889546b4114e0e62d8d6a4d2f2b4e86830aaf403b7bac82f1a47821a372cb2
3
+ size 383783467
onnx/encoder_model_uint8.onnx ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:54889546b4114e0e62d8d6a4d2f2b4e86830aaf403b7bac82f1a47821a372cb2
3
+ size 383783467