File size: 16,953 Bytes
071812d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1d48bfe
071812d
1d48bfe
 
 
 
 
 
071812d
 
 
 
 
a97e72d
1d48bfe
 
 
 
 
a97e72d
1d48bfe
 
 
a97e72d
1d48bfe
 
 
a97e72d
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
a97e72d
 
 
 
 
 
 
 
 
071812d
a97e72d
 
 
071812d
a97e72d
 
071812d
a97e72d
 
 
 
 
071812d
a97e72d
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1d48bfe
 
 
 
 
a97e72d
 
 
 
 
c916351
a97e72d
1d48bfe
a97e72d
 
 
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
a97e72d
 
1d48bfe
 
 
 
a97e72d
 
1d48bfe
 
 
 
 
a97e72d
 
c916351
a97e72d
1d48bfe
c916351
 
071812d
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
071812d
 
1d48bfe
 
 
 
071812d
a97e72d
1d48bfe
 
 
 
 
a97e72d
1d48bfe
a97e72d
1d48bfe
 
c916351
a97e72d
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
a97e72d
1d48bfe
a97e72d
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
a97e72d
 
1d48bfe
 
 
 
a97e72d
 
1d48bfe
 
 
 
 
a97e72d
 
c916351
a97e72d
1d48bfe
a97e72d
071812d
 
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
a97e72d
 
1d48bfe
 
 
 
a97e72d
 
1d48bfe
 
 
 
 
a97e72d
 
c916351
a97e72d
1d48bfe
a97e72d
 
071812d
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
071812d
a97e72d
1d48bfe
 
 
 
a97e72d
c4a1766
1d48bfe
 
 
 
 
c4a1766
 
c916351
c4a1766
1d48bfe
c4a1766
 
 
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c4a1766
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c4a1766
 
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
c4a1766
6ffadcd
1d48bfe
 
 
 
 
6ffadcd
 
 
 
1d48bfe
6ffadcd
 
 
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6ffadcd
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
6ffadcd
 
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6ffadcd
a97e72d
bc26f1b
a97e72d
 
6ffadcd
1d48bfe
 
a97e72d
 
 
 
1d48bfe
 
 
a97e72d
1d48bfe
a97e72d
 
 
 
 
 
1d48bfe
 
 
 
 
 
 
 
 
 
 
 
 
a97e72d
 
 
 
1d48bfe
 
 
a97e72d
 
 
c4a1766
 
 
1d48bfe
 
 
a97e72d
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
# Copyright      2022  Xiaomi Corp.        (authors: Fangjun Kuang)
#
# See LICENSE for clarification regarding multiple authors
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

from huggingface_hub import hf_hub_download
from functools import lru_cache
import os

os.system(
    "cp -v /home/user/.local/lib/python3.8/site-packages/k2/lib/*.so /home/user/.local/lib/python3.8/site-packages/sherpa/lib/"
)

import k2
import sherpa


sample_rate = 16000


@lru_cache(maxsize=30)
def get_pretrained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
) -> sherpa.OfflineRecognizer:
    if repo_id in chinese_models:
        return chinese_models[repo_id](
            repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
        )
    elif repo_id in english_models:
        return english_models[repo_id](
            repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
        )
    elif repo_id in chinese_english_mixed_models:
        return chinese_english_mixed_models[repo_id](
            repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
        )
    elif repo_id in tibetan_models:
        return tibetan_models[repo_id](
            repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
        )
    elif repo_id in arabic_models:
        return arabic_models[repo_id](
            repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
        )
    elif repo_id in german_models:
        return german_models[repo_id](
            repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
        )
    else:
        raise ValueError(f"Unsupported repo_id: {repo_id}")


def _get_nn_model_filename(
    repo_id: str,
    filename: str,
    subfolder: str = "exp",
) -> str:
    nn_model_filename = hf_hub_download(
        repo_id=repo_id,
        filename=filename,
        subfolder=subfolder,
    )
    return nn_model_filename


def _get_bpe_model_filename(
    repo_id: str,
    filename: str = "bpe.model",
    subfolder: str = "data/lang_bpe_500",
) -> str:
    bpe_model_filename = hf_hub_download(
        repo_id=repo_id,
        filename=filename,
        subfolder=subfolder,
    )
    return bpe_model_filename


def _get_token_filename(
    repo_id: str,
    filename: str = "tokens.txt",
    subfolder: str = "data/lang_char",
) -> str:
    token_filename = hf_hub_download(
        repo_id=repo_id,
        filename=filename,
        subfolder=subfolder,
    )
    return token_filename


@lru_cache(maxsize=10)
def _get_aishell2_pretrained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
) -> sherpa.OfflineRecognizer:
    assert repo_id in [
        # context-size 1
        "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12",  # noqa
        # context-size 2
        "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-B-2022-07-12",  # noqa
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="cpu_jit.pt",
    )
    tokens = _get_token_filename(repo_id=repo_id)

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_gigaspeech_pre_trained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
) -> sherpa.OfflineRecognizer:
    assert repo_id in [
        "wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2",
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="cpu_jit-iter-3488000-avg-20.pt",
    )
    tokens = "./giga-tokens.txt"

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_librispeech_pre_trained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
) -> sherpa.OfflineRecognizer:
    assert repo_id in [
        "WeijiZhuang/icefall-asr-librispeech-pruned-transducer-stateless8-2022-12-02",  # noqa
        "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13",  # noqa
        "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11",  # noqa
        "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless8-2022-11-14",  # noqa
    ], repo_id

    filename = "cpu_jit.pt"
    if (
        repo_id
        == "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11"
    ):
        filename = "cpu_jit-torch-1.10.0.pt"

    if (
        repo_id
        == "WeijiZhuang/icefall-asr-librispeech-pruned-transducer-stateless8-2022-12-02"
    ):
        filename = "cpu_jit-torch-1.10.pt"

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename=filename,
    )
    tokens = _get_token_filename(repo_id=repo_id, subfolder="data/lang_bpe_500")

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_wenetspeech_pre_trained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
):
    assert repo_id in [
        "luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2",
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="cpu_jit_epoch_10_avg_2_torch_1.7.1.pt",
    )
    tokens = _get_token_filename(repo_id=repo_id)

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_tal_csasr_pre_trained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
):
    assert repo_id in [
        "luomingshuang/icefall_asr_tal-csasr_pruned_transducer_stateless5",
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="cpu_jit.pt",
    )
    tokens = _get_token_filename(repo_id=repo_id)

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_alimeeting_pre_trained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
):
    assert repo_id in [
        "luomingshuang/icefall_asr_alimeeting_pruned_transducer_stateless2",
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="cpu_jit_torch_1.7.1.pt",
    )
    tokens = _get_token_filename(repo_id=repo_id)

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_wenet_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
):
    assert repo_id in [
        "csukuangfj/wenet-chinese-model",
        "csukuangfj/wenet-english-model",
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="final.zip",
        subfolder=".",
    )
    tokens = _get_token_filename(
        repo_id=repo_id,
        filename="units.txt",
        subfolder=".",
    )

    feat_config = sherpa.FeatureConfig(normalize_samples=False)
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_aidatatang_200zh_pretrained_mode(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
):
    assert repo_id in [
        "luomingshuang/icefall_asr_aidatatang-200zh_pruned_transducer_stateless2",
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="cpu_jit_torch.1.7.1.pt",
    )
    tokens = _get_token_filename(repo_id=repo_id)

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_tibetan_pre_trained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
):
    assert repo_id in [
        "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless7-2022-12-02",
        "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29",
    ], repo_id

    filename = "cpu_jit.pt"
    if (
        repo_id
        == "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29"
    ):
        filename = "cpu_jit-epoch-28-avg-23-torch-1.10.0.pt"

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename=filename,
    )

    tokens = _get_token_filename(repo_id=repo_id, subfolder="data/lang_bpe_500")

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_arabic_pre_trained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
):
    assert repo_id in [
        "AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06",
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="cpu_jit.pt",
    )

    tokens = _get_token_filename(repo_id=repo_id, subfolder="data/lang_bpe_5000")

    feat_config = sherpa.FeatureConfig()
    feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
    feat_config.fbank_opts.mel_opts.num_bins = 80
    feat_config.fbank_opts.frame_opts.dither = 0

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        feat_config=feat_config,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


@lru_cache(maxsize=10)
def _get_german_pre_trained_model(
    repo_id: str,
    decoding_method: str,
    num_active_paths: int,
):
    assert repo_id in [
        "csukuangfj/wav2vec2.0-torchaudio",
    ], repo_id

    nn_model = _get_nn_model_filename(
        repo_id=repo_id,
        filename="voxpopuli_asr_base_10k_de.pt",
        subfolder=".",
    )

    tokens = _get_token_filename(
        repo_id=repo_id,
        filename="tokens-de.txt",
        subfolder=".",
    )

    config = sherpa.OfflineRecognizerConfig(
        nn_model=nn_model,
        tokens=tokens,
        use_gpu=False,
        decoding_method=decoding_method,
        num_active_paths=num_active_paths,
    )

    recognizer = sherpa.OfflineRecognizer(config)

    return recognizer


chinese_models = {
    "luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2": _get_wenetspeech_pre_trained_model,  # noqa
    "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12": _get_aishell2_pretrained_model,  # noqa
    "yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-B-2022-07-12": _get_aishell2_pretrained_model,  # noqa
    "luomingshuang/icefall_asr_aidatatang-200zh_pruned_transducer_stateless2": _get_aidatatang_200zh_pretrained_mode,  # noqa
    "luomingshuang/icefall_asr_alimeeting_pruned_transducer_stateless2": _get_alimeeting_pre_trained_model,  # noqa
    "csukuangfj/wenet-chinese-model": _get_wenet_model,
}

english_models = {
    "wgb14/icefall-asr-gigaspeech-pruned-transducer-stateless2": _get_gigaspeech_pre_trained_model,  # noqa
    "WeijiZhuang/icefall-asr-librispeech-pruned-transducer-stateless8-2022-12-02": _get_librispeech_pre_trained_model,  # noqa
    "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless8-2022-11-14": _get_librispeech_pre_trained_model,  # noqa
    "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless7-2022-11-11": _get_librispeech_pre_trained_model,  # noqa
    "csukuangfj/icefall-asr-librispeech-pruned-transducer-stateless3-2022-05-13": _get_librispeech_pre_trained_model,  # noqa
    "csukuangfj/wenet-english-model": _get_wenet_model,
}

chinese_english_mixed_models = {
    "luomingshuang/icefall_asr_tal-csasr_pruned_transducer_stateless5": _get_tal_csasr_pre_trained_model,  # noqa
}

tibetan_models = {
    "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless7-2022-12-02": _get_tibetan_pre_trained_model,  # noqa
    "syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29": _get_tibetan_pre_trained_model,  # noqa
}

arabic_models = {
    "AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06": _get_arabic_pre_trained_model,  # noqa
}

german_models = {
    "csukuangfj/wav2vec2.0-torchaudio": _get_german_pre_trained_model,
}

all_models = {
    **chinese_models,
    **english_models,
    **chinese_english_mixed_models,
    **tibetan_models,
    **arabic_models,
    **german_models,
}

language_to_models = {
    "Chinese": list(chinese_models.keys()),
    "English": list(english_models.keys()),
    "Chinese+English": list(chinese_english_mixed_models.keys()),
    "Tibetan": list(tibetan_models.keys()),
    "Arabic": list(arabic_models.keys()),
    "German": list(german_models.keys()),
}