Spaces:
Runtime error
Runtime error
Commit
·
72452c7
1
Parent(s):
cc12677
add arabic models
Browse files- examples.py +19 -0
- model.py +46 -0
- test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0051454:0052244.wav +0 -0
- test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0052244:0053004.wav +0 -0
- test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0053813:0054281.wav +0 -0
- test_wavs/arabic/trans.txt +3 -0
examples.py
CHANGED
|
@@ -197,4 +197,23 @@ examples = [
|
|
| 197 |
4,
|
| 198 |
"./test_wavs/tibetan/a_0_cacm-A70_31118.wav",
|
| 199 |
],
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 200 |
]
|
|
|
|
| 197 |
4,
|
| 198 |
"./test_wavs/tibetan/a_0_cacm-A70_31118.wav",
|
| 199 |
],
|
| 200 |
+
# arabic
|
| 201 |
+
[
|
| 202 |
+
"Arabic",
|
| 203 |
+
"AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06" "greedy_search",
|
| 204 |
+
4,
|
| 205 |
+
"./test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0051454:0052244.wav",
|
| 206 |
+
],
|
| 207 |
+
[
|
| 208 |
+
"Arabic",
|
| 209 |
+
"AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06" "greedy_search",
|
| 210 |
+
4,
|
| 211 |
+
"./test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0052244:0053004.wav",
|
| 212 |
+
],
|
| 213 |
+
[
|
| 214 |
+
"Arabic",
|
| 215 |
+
"AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06" "greedy_search",
|
| 216 |
+
4,
|
| 217 |
+
"./test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0053813:0054281.wav",
|
| 218 |
+
],
|
| 219 |
]
|
model.py
CHANGED
|
@@ -51,6 +51,10 @@ def get_pretrained_model(
|
|
| 51 |
return tibetan_models[repo_id](
|
| 52 |
repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
|
| 53 |
)
|
|
|
|
|
|
|
|
|
|
|
|
|
| 54 |
else:
|
| 55 |
raise ValueError(f"Unsupported repo_id: {repo_id}")
|
| 56 |
|
|
@@ -402,6 +406,42 @@ def _get_tibetan_pre_trained_model(
|
|
| 402 |
return recognizer
|
| 403 |
|
| 404 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 405 |
chinese_models = {
|
| 406 |
"luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2": _get_wenetspeech_pre_trained_model, # noqa
|
| 407 |
"yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12": _get_aishell2_pretrained_model, # noqa
|
|
@@ -427,11 +467,16 @@ tibetan_models = {
|
|
| 427 |
"syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29": _get_tibetan_pre_trained_model, # noqa
|
| 428 |
}
|
| 429 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 430 |
all_models = {
|
| 431 |
**chinese_models,
|
| 432 |
**english_models,
|
| 433 |
**chinese_english_mixed_models,
|
| 434 |
**tibetan_models,
|
|
|
|
| 435 |
}
|
| 436 |
|
| 437 |
language_to_models = {
|
|
@@ -439,4 +484,5 @@ language_to_models = {
|
|
| 439 |
"English": list(english_models.keys()),
|
| 440 |
"Chinese+English": list(chinese_english_mixed_models.keys()),
|
| 441 |
"Tibetan": list(tibetan_models.keys()),
|
|
|
|
| 442 |
}
|
|
|
|
| 51 |
return tibetan_models[repo_id](
|
| 52 |
repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
|
| 53 |
)
|
| 54 |
+
elif repo_id in arabic_models:
|
| 55 |
+
return arabic_models[repo_id](
|
| 56 |
+
repo_id, decoding_method=decoding_method, num_active_paths=num_active_paths
|
| 57 |
+
)
|
| 58 |
else:
|
| 59 |
raise ValueError(f"Unsupported repo_id: {repo_id}")
|
| 60 |
|
|
|
|
| 406 |
return recognizer
|
| 407 |
|
| 408 |
|
| 409 |
+
@lru_cache(maxsize=10)
|
| 410 |
+
def _get_arabic_pre_trained_model(
|
| 411 |
+
repo_id: str,
|
| 412 |
+
decoding_method: str,
|
| 413 |
+
num_active_paths: int,
|
| 414 |
+
):
|
| 415 |
+
assert repo_id in [
|
| 416 |
+
"AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06",
|
| 417 |
+
], repo_id
|
| 418 |
+
|
| 419 |
+
nn_model = _get_nn_model_filename(
|
| 420 |
+
repo_id=repo_id,
|
| 421 |
+
filename=filename,
|
| 422 |
+
)
|
| 423 |
+
|
| 424 |
+
tokens = _get_token_filename(repo_id=repo_id, subfolder="data/lang_bpe_5000")
|
| 425 |
+
|
| 426 |
+
feat_config = sherpa.FeatureConfig()
|
| 427 |
+
feat_config.fbank_opts.frame_opts.samp_freq = sample_rate
|
| 428 |
+
feat_config.fbank_opts.mel_opts.num_bins = 80
|
| 429 |
+
feat_config.fbank_opts.frame_opts.dither = 0
|
| 430 |
+
|
| 431 |
+
config = sherpa.OfflineRecognizerConfig(
|
| 432 |
+
nn_model=nn_model,
|
| 433 |
+
tokens=tokens,
|
| 434 |
+
use_gpu=False,
|
| 435 |
+
feat_config=feat_config,
|
| 436 |
+
decoding_method=decoding_method,
|
| 437 |
+
num_active_paths=num_active_paths,
|
| 438 |
+
)
|
| 439 |
+
|
| 440 |
+
recognizer = sherpa.OfflineRecognizer(config)
|
| 441 |
+
|
| 442 |
+
return recognizer
|
| 443 |
+
|
| 444 |
+
|
| 445 |
chinese_models = {
|
| 446 |
"luomingshuang/icefall_asr_wenetspeech_pruned_transducer_stateless2": _get_wenetspeech_pre_trained_model, # noqa
|
| 447 |
"yuekai/icefall-asr-aishell2-pruned-transducer-stateless5-A-2022-07-12": _get_aishell2_pretrained_model, # noqa
|
|
|
|
| 467 |
"syzym/icefall-asr-xbmu-amdo31-pruned-transducer-stateless5-2022-11-29": _get_tibetan_pre_trained_model, # noqa
|
| 468 |
}
|
| 469 |
|
| 470 |
+
arabic_models = {
|
| 471 |
+
"AmirHussein/icefall-asr-mgb2-conformer_ctc-2022-27-06": _get_arabic_pre_trained_model, # noqa
|
| 472 |
+
}
|
| 473 |
+
|
| 474 |
all_models = {
|
| 475 |
**chinese_models,
|
| 476 |
**english_models,
|
| 477 |
**chinese_english_mixed_models,
|
| 478 |
**tibetan_models,
|
| 479 |
+
**arabic_models,
|
| 480 |
}
|
| 481 |
|
| 482 |
language_to_models = {
|
|
|
|
| 484 |
"English": list(english_models.keys()),
|
| 485 |
"Chinese+English": list(chinese_english_mixed_models.keys()),
|
| 486 |
"Tibetan": list(tibetan_models.keys()),
|
| 487 |
+
"Arabic": list(arabic_models.keys()),
|
| 488 |
}
|
test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0051454:0052244.wav
ADDED
|
Binary file (253 kB). View file
|
|
|
test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0052244:0053004.wav
ADDED
|
Binary file (243 kB). View file
|
|
|
test_wavs/arabic/94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0053813:0054281.wav
ADDED
|
Binary file (150 kB). View file
|
|
|
test_wavs/arabic/trans.txt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0053813:0054281 بعد أن عجز وبدأ يصدر مشكلات شعبه ومشكلات مصر
|
| 2 |
+
94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0051454:0052244 وهؤلاء أولياء الشيطان ها هو ذا أحدهم الآن ضيفا عليكم على قناة الجزيرة ولا يستحي في ذلك
|
| 3 |
+
94D37D38-B203-4FC0-9F3A-538F5C174920_spk-0001_seg-0052244:0053004 عندما استغاث الليبيون بالعالم استغاثوا لرفع الظلم وليس لقهر إرادة الأمة ومصادرة الحياة الدستورية
|