Hugging Face's logo

Spaces:
intelli-zen
/
asr
Sleeping

App Files Community
HoneyTian commited on
Commit
f1ed46a
•
1 Parent(s): e7f863b

update

Browse files
Files changed (2) hide show
  1. examples.py +1 -382
  2. main.py +13 -0
examples.py CHANGED
@@ -3,393 +3,12 @@
3
 
4
 
5
  examples = [
6
- [
7
- "Chinese+English",
8
- "csukuangfj/sherpa-onnx-streaming-zipformer-bilingual-zh-en-2023-02-20",
9
- "greedy_search",
10
- 4,
11
- "Yes",
12
- "./data/test_wavs/tal_csasr/0.wav",
13
- ],
14
- [
15
- "Chinese+English+Cantonese",
16
- "csukuangfj/sherpa-onnx-paraformer-trilingual-zh-cantonese-en",
17
- "greedy_search",
18
- 4,
19
- "Yes",
20
- "./data/test_wavs/cantonese/2.wav",
21
- ],
22
- [
23
- "Cantonese",
24
- "zrjin/icefall-asr-mdcc-zipformer-2024-03-11",
25
- "greedy_search",
26
- 4,
27
- "Yes",
28
- "./data/test_wavs/cantonese/1.wav",
29
- ],
30
- [
31
- "English",
32
- "whisper-base.en",
33
- "greedy_search",
34
- 4,
35
- "Yes",
36
- "./data/test_wavs/librispeech/1089-134686-0001.wav",
37
- ],
38
- [
39
- "Chinese",
40
- "csukuangfj/sherpa-onnx-paraformer-zh-2024-03-09",
41
- "greedy_search",
42
- 4,
43
- "Yes",
44
- "./data/test_wavs/paraformer-zh/四川话.wav",
45
- ],
46
- [
47
- "Russian",
48
- "alphacep/vosk-model-ru",
49
- "greedy_search",
50
- 4,
51
- "No",
52
- "./data/test_wavs/russian/russian-i-love-you.wav",
53
- ],
54
- [
55
- "Russian",
56
- "alphacep/vosk-model-ru",
57
- "greedy_search",
58
- 4,
59
- "No",
60
- "./data/test_wavs/russian/test.wav",
61
- ],
62
- [
63
- "German",
64
- "csukuangfj/wav2vec2.0-torchaudio",
65
- "greedy_search",
66
- 4,
67
- "No",
68
- "./data/test_wavs/german/20170517-0900-PLENARY-16-de_20170517.wav",
69
- ],
70
- [
71
- "Arabic",
72
- "AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06",
73
- "greedy_search",
74
- 4,
75
- "No",
76
- "./data/test_wavs/arabic/a.wav",
77
- ],
78
- [
79
- "Tibetan",
80
- "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless7-2022-12-02",
81
- "greedy_search",
82
- 4,
83
- "No",
84
- "./data/test_wavs/tibetan/a_0_cacm-A70_31117.wav",
85
- ],
86
- [
87
- "French",
88
- "shaojieli/sherpa-onnx-streaming-zipformer-fr-2023-04-14",
89
- "greedy_search",
90
- 4,
91
- "No",
92
- "./data/test_wavs/french/common_voice_fr_19364697.wav",
93
- ],
94
- [
95
- "Chinese",
96
- "desh2608/icefall-asr-alimeeting-pruned-transducer-stateless7",
97
- "greedy_search",
98
- 4,
99
- "Yes",
100
- "./data/test_wavs/alimeeting/R8003_M8001-8004-165.wav",
101
- ],
102
- [
103
- "Chinese",
104
- "csukuangfj/sherpa-onnx-paraformer-zh-2024-03-09",
105
- "greedy_search",
106
- 4,
107
- "Yes",
108
- "./data/test_wavs/paraformer-zh/天津话.wav",
109
- ],
110
  [
111
  "Chinese",
112
  "csukuangfj/sherpa-onnx-paraformer-zh-2024-03-09",
113
  "greedy_search",
114
  4,
115
  "Yes",
116
- "./data/test_wavs/paraformer-zh/郑州话.wav",
117
- ],
118
- [
119
- "Chinese",
120
- "desh2608/icefall-asr-alimeeting-pruned-transducer-stateless7",
121
- "greedy_search",
122
- 4,
123
- "Yes",
124
- "./data/test_wavs/alimeeting/R8008_M8013-8049-74.wav",
125
- ],
126
- [
127
- "Chinese",
128
- "desh2608/icefall-asr-alimeeting-pruned-transducer-stateless7",
129
- "greedy_search",
130
- 4,
131
- "Yes",
132
- "./data/test_wavs/alimeeting/R8009_M8020_N_SPK8026-8026-209.wav",
133
- ],
134
- [
135
- "English",
136
- "videodanchik/icefall-asr-tedlium3-conformer-ctc2",
137
- "greedy_search",
138
- 4,
139
- "Yes",
140
- "./data/test_wavs/tedlium3/DanBarber_2010-219.wav",
141
- ],
142
- [
143
- "English",
144
- "whisper-base.en",
145
- "greedy_search",
146
- 4,
147
- "Yes",
148
- "./data/test_wavs/tedlium3/DanielKahneman_2010-157.wav",
149
- ],
150
- [
151
- "English",
152
- "videodanchik/icefall-asr-tedlium3-conformer-ctc2",
153
- "greedy_search",
154
- 4,
155
- "Yes",
156
- "./data/test_wavs/tedlium3/RobertGupta_2010U-15.wav",
157
- ],
158
- # librispeech
159
- # https://huggingface.co/csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless5-2022-05-13/tree/main/data/test_wavs
160
- [
161
- "English",
162
- "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13",
163
- "greedy_search",
164
- 4,
165
- "Yes",
166
- "./data/test_wavs/librispeech/1089-134686-0001.wav",
167
- ],
168
- [
169
- "English",
170
- "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13",
171
- "greedy_search",
172
- 4,
173
- "Yes",
174
- "./data/test_wavs/librispeech/1221-135766-0001.wav",
175
- ],
176
- [
177
- "English",
178
- "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13",
179
- "greedy_search",
180
- 4,
181
- "Yes",
182
- "./data/test_wavs/librispeech/1221-135766-0002.wav",
183
- ],
184
- # gigaspeech
185
- [
186
- "English",
187
- "wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2",
188
- "greedy_search",
189
- 4,
190
- "Yes",
191
- "./data/test_wavs/gigaspeech/1-minute-audiobook.opus",
192
- ],
193
- [
194
- "English",
195
- "wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2",
196
- "greedy_search",
197
- 4,
198
- "Yes",
199
- "./data/test_wavs/gigaspeech/100-seconds-podcast.opus",
200
- ],
201
- [
202
- "English",
203
- "wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2",
204
- "greedy_search",
205
- 4,
206
- "Yes",
207
- "./data/test_wavs/gigaspeech/100-seconds-youtube.opus",
208
- ],
209
- # wenetspeech
210
- # https://huggingface.co/luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2/tree/main/data/test_wavs
211
- [
212
- "Chinese",
213
- "luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2",
214
- "greedy_search",
215
- 4,
216
- "Yes",
217
- "./data/test_wavs/wenetspeech/DEV_T0000000000.opus",
218
- ],
219
- [
220
- "Chinese",
221
- "luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2",
222
- "greedy_search",
223
- 4,
224
- "Yes",
225
- "./data/test_wavs/wenetspeech/DEV_T0000000001.opus",
226
- ],
227
- [
228
- "Chinese",
229
- "luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2",
230
- "greedy_search",
231
- 4,
232
- "Yes",
233
- "./data/test_wavs/wenetspeech/DEV_T0000000002.opus",
234
- ],
235
- # aishell2-A
236
- # https://huggingface.co/yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12/tree/main/data/test_wavs
237
- [
238
- "Chinese",
239
- "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12",
240
- "greedy_search",
241
- 4,
242
- "Yes",
243
- "./data/test_wavs/aishell2/ID0012W0030.wav",
244
- ],
245
- [
246
- "Chinese",
247
- "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12",
248
- "greedy_search",
249
- 4,
250
- "Yes",
251
- "./data/test_wavs/aishell2/ID0012W0162.wav",
252
- ],
253
- [
254
- "Chinese",
255
- "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12",
256
- "greedy_search",
257
- 4,
258
- "Yes",
259
- "./data/test_wavs/aishell2/ID0012W0215.wav",
260
- ],
261
- # aishell2-B
262
- # https://huggingface.co/yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12/tree/main/data/test_wavs
263
- [
264
- "Chinese",
265
- "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-B-2022-07-12",
266
- "greedy_search",
267
- 4,
268
- "Yes",
269
- "./data/test_wavs/aishell2/ID0012W0030.wav",
270
- ],
271
- [
272
- "Chinese",
273
- "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-B-2022-07-12",
274
- "greedy_search",
275
- 4,
276
- "Yes",
277
- "./data/test_wavs/aishell2/ID0012W0162.wav",
278
- ],
279
- [
280
- "Chinese",
281
- "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-B-2022-07-12",
282
- "greedy_search",
283
- 4,
284
- "Yes",
285
- "./data/test_wavs/aishell2/ID0012W0215.wav",
286
- ],
287
- # aishell2-B
288
- # https://huggingface.co/luomingshuang/icefall_asr_aidatatang-200zh_pruned_transducer_stateless2/tree/main/data/test_wavs
289
- [
290
- "Chinese",
291
- "luomingshuang/icefall_asr_aidatatang-200zh_pruned_transducer_stateless2",
292
- "greedy_search",
293
- 4,
294
- "Yes",
295
- "./data/test_wavs/aidatatang_200zh/T0055G0036S0002.wav",
296
- ],
297
- [
298
- "Chinese",
299
- "luomingshuang/icefall_asr_aidatatang-200zh_pruned_transducer_stateless2",
300
- "greedy_search",
301
- 4,
302
- "Yes",
303
- "./data/test_wavs/aidatatang_200zh/T0055G0036S0003.wav",
304
- ],
305
- [
306
- "Chinese",
307
- "luomingshuang/icefall_asr_aidatatang-200zh_pruned_transducer_stateless2",
308
- "greedy_search",
309
- 4,
310
- "Yes",
311
- "./data/test_wavs/aidatatang_200zh/T0055G0036S0004.wav",
312
- ],
313
- # tal_csasr
314
- [
315
- "Chinese+English",
316
- "ptrnull/icefall-asr-conv-emformer-transducer-stateless2-zh",
317
- "greedy_search",
318
- 4,
319
- "Yes",
320
- "./data/test_wavs/tal_csasr/210_36476_210_8341_1_1533271973_7057520_132.wav",
321
- ],
322
- [
323
- "Chinese+English",
324
- "ptrnull/icefall-asr-conv-emformer-transducer-stateless2-zh",
325
- "greedy_search",
326
- 4,
327
- "Yes",
328
- "./data/test_wavs/tal_csasr/210_36476_210_8341_1_1533271973_7057520_138.wav",
329
- ],
330
- [
331
- "Chinese+English",
332
- "ptrnull/icefall-asr-conv-emformer-transducer-stateless2-zh",
333
- "greedy_search",
334
- 4,
335
- "Yes",
336
- "./data/test_wavs/tal_csasr/210_36476_210_8341_1_1533271973_7057520_145.wav",
337
- ],
338
- [
339
- "Tibetan",
340
- "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless7-2022-12-02",
341
- "greedy_search",
342
- 4,
343
- "No",
344
- "./data/test_wavs/tibetan/a_0_cacm-A70_31116.wav",
345
- ],
346
- [
347
- "Tibetan",
348
- "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless7-2022-12-02",
349
- "greedy_search",
350
- 4,
351
- "No",
352
- "./data/test_wavs/tibetan/a_0_cacm-A70_31118.wav",
353
- ],
354
- # arabic
355
- [
356
- "Arabic",
357
- "AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06",
358
- "greedy_search",
359
- 4,
360
- "No",
361
- "./data/test_wavs/arabic/b.wav",
362
- ],
363
- [
364
- "Arabic",
365
- "AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06",
366
- "greedy_search",
367
- 4,
368
- "No",
369
- "./data/test_wavs/arabic/c.wav",
370
- ],
371
- [
372
- "German",
373
- "csukuangfj/wav2vec2.0-torchaudio",
374
- "greedy_search",
375
- 4,
376
- "No",
377
- "./data/test_wavs/german/20120315-0900-PLENARY-14-de_20120315.wav",
378
- ],
379
- [
380
- "French",
381
- "shaojieli/sherpa-onnx-streaming-zipformer-fr-2023-04-14",
382
- "greedy_search",
383
- 4,
384
- "No",
385
- "./data/test_wavs/french/common_voice_fr_19738183.wav",
386
- ],
387
- [
388
- "French",
389
- "shaojieli/sherpa-onnx-streaming-zipformer-fr-2023-04-14",
390
- "greedy_search",
391
- 4,
392
- "No",
393
- "./data/test_wavs/french/common_voice_fr_27024649.wav",
394
  ],
395
  ]
 
3
 
4
 
5
  examples = [
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6
  [
7
  "Chinese",
8
  "csukuangfj/sherpa-onnx-paraformer-zh-2024-03-09",
9
  "greedy_search",
10
  4,
11
  "Yes",
12
+ "./data/test_wavs/paraformer-zh/si_chuan_hua.wav",
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
13
  ],
14
  ]
main.py CHANGED
@@ -134,6 +134,19 @@ def main():
134
  fn=process_uploaded_file,
135
  )
136
 
 
 
 
 
 
 
 
 
 
 
 
 
 
137
  language_radio.change(
138
  update_model_dropdown,
139
  inputs=language_radio,
 
134
  fn=process_uploaded_file,
135
  )
136
 
137
+ upload_button.click(
138
+ process_uploaded_file,
139
+ inputs=[
140
+ language_radio,
141
+ model_dropdown,
142
+ decoding_method_radio,
143
+ num_active_paths_slider,
144
+ punct_radio,
145
+ uploaded_file,
146
+ ],
147
+ outputs=[uploaded_output, uploaded_html_info],
148
+ )
149
+
150
  language_radio.change(
151
  update_model_dropdown,
152
  inputs=language_radio,