first commit

.gitattributes

@@ -32,3 +32,7 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+data filter=lfs diff=lfs merge=lfs -text
+exp filter=lfs diff=lfs merge=lfs -text
+decoding-result filter=lfs diff=lfs merge=lfs -text
+test_wavs filter=lfs diff=lfs merge=lfs -text

data/lang_bpe_500/L.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90ba23c27471509a03c982f2ceb55c2622a84d2d7d31162cffada8efac16c945
+size 330023

data/lang_bpe_500/LG.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8857024d33fc2f06133c02bb3694ec6e716a06f3853eab1d9bfac1fdfac5f62c
+size 64875683

data/lang_bpe_500/Linv.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ccca9f0cced910c93d7fe8e3b5fc34c85d2f440fc2d6d2cfe1f50e7dac5e9acf
+size 330023

data/lang_bpe_500/bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c029f22a5cdd87967db44988bf93f4f33452bdaa4c6fdd8ec5f4deca1b9bf96
+size 247505

data/lang_bpe_500/tokens.txt

@@ -0,0 +1,503 @@
+<blk> 0
+<sos/eos> 1
+<unk> 2
+▁ 3
+▁དང 4
+ས 5
+▁པ 6
+ར 7
+ང 8
+ན 9
+ད 10
+▁བ 11
+ལ 12
+▁ལ 13
+ག 14
+▁པའི 15
+འི 16
+▁ཀྱི 17
+▁མ 18
+▁མི 19
+▁གི 20
+▁བྱེད 21
+▁དུ 22
+▁རྒྱུ 23
+▁དེ 24
+▁བྱ 25
+▁ཡིན 26
+▁དགོས 27
+▁ན 28
+▁ཡོད 29
+▁རྒྱ 30
+▁ལས 31
+▁བའི 32
+▁པོ 33
+ེ 34
+▁གྱི 35
+▁ས 36
+མ 37
+▁ང 38
+བ 39
+▁ནི 40
+▁ག 41
+▁ནས 42
+▁རིགས 43
+▁ལུགས 44
+▁པར 45
+▁རང 46
+ོས 47
+▁ཚོགས 48
+▁ད 49
+▁ཆེ 50
+▁ཚོ 51
+ི 52
+▁བར 53
+▁ཞིག 54
+ུ 55
+▁གཏོང 56
+▁སྤྱི 57
+▁དོན 58
+▁ཏ 59
+▁བྱས 60
+▁དག 61
+▁ཆ 62
+▁རེད 63
+▁ཁ 64
+▁སུ 65
+▁མེད 66
+▁གཅི 67
+▁གནས 68
+▁ར 69
+▁མང 70
+▁བྱུང 71
+▁ཆེན 72
+▁འདི 73
+▁ཡང 74
+▁རིང 75
+▁འགྱུར 76
+▁ནང 77
+▁འ 78
+▁ཏུ 79
+▁ཐུབ 80
+▁སྐྱེ 81
+གས 82
+▁རྩ 83
+▁སྲིད 84
+▁གས 85
+ོ 86
+▁ཕྱོགས 87
+▁ཤུགས 88
+▁མོ 89
+▁དམངས 90
+▁ཆོས 91
+▁གང 92
+▁དཔ 93
+▁སྤེལ 94
+▁གིས 95
+▁དུས 96
+▁གཞི 97
+བས 98
+▁འཕེལ 99
+▁དབ 100
+▁ལྟ 101
+▁རིམ 102
+▁བས 103
+▁གོ 104
+▁ཐོག 105
+▁དམ 106
+▁བོ 107
+▁སྒོ 108
+▁མཐུན 109
+▁ཚང 110
+▁འདྲ 111
+▁ལོ 112
+▁བཞིན 113
+▁ཡ 114
+▁ཀྲུ 115
+▁ཡོང 116
+▁སློ 117
+▁ལམ 118
+▁ལག 119
+▁ཁབ 120
+ོད 121
+▁ཀྱིས 122
+འ 123
+▁ངེས 124
+▁ཐོན 125
+▁འཛུགས 126
+ངས 127
+▁ཚད 128
+▁འཛིན 129
+▁རིག 130
+▁གོང 131
+ོག 132
+▁ཀྱང 133
+▁རེ 134
+▁གཞན 135
+▁ཞིང 136
+▁ཚུལ 137
+▁ལེགས 138
+▁རུ 139
+▁སྐྱོང 140
+▁འབྲེལ 141
+▁རྐྱེན 142
+▁ལྡན 143
+▁ཁག 144
+▁ཀ 145
+▁ཅན 146
+▁ནུས 147
+▁རྒྱས 148
+▁ཧ 149
+▁གཉིས 150
+▁ཁོང 151
+ུང 152
+▁ཚན 153
+▁གྱིས 154
+▁ཞི 155
+ོན 156
+▁ལེན 157
+▁སྤྱོད 158
+▁ཐག 159
+▁གསོ 160
+▁སྔ 161
+▁ཕན 162
+▁འཚོ 163
+▁ཟ 164
+▁དོ 165
+▁གནད 166
+▁བསམ 167
+▁བཏང 168
+▁སོ 169
+▁ཡོངས 170
+▁ཁྱོ 171
+▁འགོ 172
+▁ངོ 173
+▁ཏེ 174
+▁བཟོ 175
+▁ཤེས 176
+▁ཐ 177
+▁བློ 178
+▁ཉི 179
+▁བདེ 180
+▁ཁེ 181
+▁ཤ 182
+▁འགྲོ 183
+▁འབྱོ 184
+▁བུ 185
+▁ཅ 186
+▁སྣ 187
+▁ཙ 188
+▁ཁོ 189
+▁སྒྲིག 190
+▁གཏ 191
+▁གཞུང 192
+▁སྐུལ 193
+ུག 194
+▁ཁྲིམས 195
+▁ཆོག 196
+▁གྲོ 197
+▁ཁྱ 198
+▁སྲུང 199
+▁ལྟར 200
+▁རབས 201
+▁འགན 202
+▁ཁུ 203
+▁ཕྱི 204
+▁སྐབས 205
+▁བཅོས 206
+▁གཙོ 207
+▁ཐབས 208
+▁སྐ 209
+▁སྔོན 210
+▁གླིང 211
+▁འཛ 212
+▁བཅ 213
+▁ཨེ 214
+▁གྱུར 215
+▁ཕ 216
+▁དབྱི 217
+▁འབ 218
+▁བྲ 219
+▁ཆགས 220
+▁མཐའ 221
+དག 222
+▁ཟིག 223
+▁འོ 224
+▁ཁང 225
+▁མཐོ 226
+▁གཅོད 227
+▁སྐྲུན 228
+▁འཐབ 229
+▁ཤས 230
+མས 231
+ཉམ 232
+▁གྲུབ 233
+▁སྡེ 234
+▁འཐུས 235
+▁ཁྲིད 236
+▁ཚ 237
+▁འཇུག 238
+▁རྩལ 239
+▁མཁན 240
+▁ཡི 241
+▁དེའི 242
+▁རྣམས 243
+▁གསལ 244
+▁བཤད 245
+▁སྟོབས 246
+▁ཆུང 247
+▁བཀ 248
+▁ཇུས 249
+▁ཅི 250
+▁སོང 251
+▁ཤིང 252
+▁ཉེ 253
+▁གསུམ 254
+▁ཆུ 255
+ཇེ 256
+▁དུང 257
+▁ཟད 258
+▁ཆད 259
+▁རྣམ 260
+▁གྲ 261
+▁འབྱུང 262
+▁མཐོང 263
+▁ཐད 264
+▁མེ 265
+▁བརྟ 266
+ྭ 267
+ུལ 268
+▁འགྲ 269
+▁ཡུ 270
+▁མིན 271
+འུ 272
+▁ཐུ 273
+▁རི 274
+▁མྱོང 275
+▁འཁྱོངས 276
+▁འཇོག 277
+▁སྨ 278
+▁ཨ 279
+▁སྟེང 280
+▁ཡུལ 281
+▁མཚན 282
+▁ཡོན 283
+▁ལྷག 284
+▁ཤི 285
+▁ནའང 286
+▁རྗེས 287
+ྒ 288
+▁གྲངས 289
+▁འགོག 290
+▁དགོ 291
+▁འད 292
+▁ཕུ 293
+▁ཁྱབ 294
+▁ཅིག 295
+▁ཐང 296
+▁བསྒྱུར 297
+▁ཆི 298
+▁ཕྲ 299
+▁གན 300
+▁ཐོབ 301
+▁ཁྲ 302
+▁ཉུང 303
+▁རྩི 304
+▁མངོན 305
+▁འོག 306
+▁འདོན 307
+ེམས 308
+ེད 309
+▁ཁྲོད 310
+▁རུང 311
+▁འབྲས 312
+▁ཁུངས 313
+▁པས 314
+▁བཙ 315
+▁སྐོར 316
+འོ 317
+▁ལྡོག 318
+▁དཀའ 319
+▁དམིགས 320
+▁ཙི 321
+▁སྲོ 322
+▁མཚུངས 323
+▁ཕྱིར 324
+▁བཅུ 325
+▁སྒྲུབ 326
+▁ཚེ 327
+▁སྟེ 328
+▁སྐྱེད 329
+▁བསྟར 330
+▁རོ 331
+▁བརྩ 332
+▁རྨ 333
+▁ཆབ 334
+▁སྲོལ 335
+▁དངུལ 336
+▁བརྒྱུད 337
+▁མཐུ 338
+▁སྤྲོ 339
+▁བཀོད 340
+▁ཞུ 341
+▁སྲ 342
+▁རྒྱག 343
+▁སྣང 344
+▁འདུ 345
+▁ཤིག 346
+▁ཞུགས 347
+▁མཚ 348
+▁འབུ 349
+▁འདུག 350
+▁སྣོན 351
+▁བརྟེན 352
+▁མཚོན 353
+▁བཟུང 354
+▁དཔེ 355
+▁འཆར 356
+▁ཁྱི 357
+▁ཚོང 358
+▁བཟང 359
+▁སྦྲ 360
+▁འཐ 361
+▁མགྱོགས 362
+▁རྟ 363
+▁བཞག 364
+▁བབ 365
+▁ནོ 366
+▁རྩོད 367
+▁འཕྲ 368
+▁གཏོད 369
+▁བསལ 370
+▁འདེམས 371
+▁ཚིག 372
+བྱེད 373
+▁ཟབ 374
+▁སོགས 375
+▁གཤིས 376
+▁གཟུགས 377
+▁སྡོ 378
+▁བརྩི 379
+▁ཧྲ 380
+▁བརྒྱ 381
+▁རྩོ 382
+▁མུ 383
+▁འདེད 384
+▁སྦྱོང 385
+▁འགུ 386
+▁དྲག 387
+▁བསྐྱ 388
+▁རྒོལ 389
+▁བལྟ 390
+▁ཕྱིན 391
+▁རླ 392
+▁འགྱོ 393
+▁ཉམས 394
+▁གྲྭ 395
+▁ཡིག 396
+▁འགལ 397
+▁དབྱང 398
+▁སྟངས 399
+▁རྟག 400
+▁གུ 401
+▁སྟོན 402
+ུབ 403
+▁ལུས 404
+▁གནོད 405
+▁སྙི 406
+▁འཇིག 407
+▁ཨུ 408
+▁ཕོ 409
+▁ཐེབས 410
+▁བླང 411
+▁སི 412
+▁མཚོ 413
+▁དད 414
+▁ཁྱེ 415
+▁ཅུ 416
+▁རྒྱུས 417
+▁ཞེ 418
+▁འདྲེ 419
+▁དགའ 420
+▁ཆེད 421
+▁ནོར 422
+▁མཐར 423
+▁སྤྱད 424
+▁ཚུན 425
+▁རོག 426
+▁སྒྲིལ 427
+▁བཞི 428
+▁འགའ 429
+▁རྫས 430
+▁ཟེར 431
+▁ཇི 432
+▁ནམ 433
+▁འགོས 434
+▁སྡུ 435
+▁སྐྱོ 436
+▁ཧུ 437
+▁སྐྱོན 438
+▁བརྗེ 439
+ཏིང 440
+▁ཐེང 441
+▁ཡག 442
+ྤ 443
+▁གོམ 444
+▁འཁོར 445
+▁སྤོ 446
+▁ཐོ 447
+▁ཟིན 448
+▁བཏོན 449
+▁སྦྱོ 450
+▁འདོད 451
+▁གཞག 452
+▁བསྒྲུབ 453
+▁མདུ 454
+▁ཐུག 455
+▁ཅིང 456
+▁གཞོན 457
+ཱ 458
+ཕ 459
+ྟ 460
+༄ 461
+༅ 462
+ཎ 463
+ྥ 464
+ཝ 465
+ྕ 466
+ྫ 467
+ྨ 468
+ྙ 469
+ྷ 470
+ྦ 471
+ྗ 472
+ཨ 473
+ཧ 474
+ྔ 475
+ྣ 476
+ཙ 477
+ཇ 478
+ཛ 479
+ྡ 480
+ྩ 481
+ཉ 482
+ླ 483
+ཟ 484
+ཤ 485
+ཅ 486
+ྐ 487
+ཏ 488
+ཞ 489
+ཆ 490
+ཀ 491
+ཐ 492
+ཚ 493
+ཁ 494
+ཡ 495
+ྲ 496
+པ 497
+ྰ 498
+ྱ 499
+#0 500
+#1 501
+#2 502

decoding-result/beam_search/log-decode-epoch-23-avg-11-beam_search-beam-size-4-use-averaged-model-2022-12-02-13-41-06

@@ -0,0 +1,18 @@
+2022-12-02 13:41:06,396 INFO [decode.py:682] Decoding started
+2022-12-02 13:41:06,397 INFO [decode.py:688] Device: cuda:0
+2022-12-02 13:41:06,399 INFO [decode.py:703] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.22', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '', 'k2-git-date': '', 'lhotse-version': '1.10.0', 'torch-version': '1.12.1', 'torch-cuda-available': True, 'torch-cuda-version': '11.6', 'python-version': '3.9', 'icefall-git-branch': 'master', 'icefall-git-sha1': 'e5d9426-dirty', 'icefall-git-date': 'Tue Nov 22 11:45:03 2022', 'icefall-path': '/root/workspace/icefall', 'k2-path': '/root/workspace/k2/k2/python/k2/__init__.py', 'lhotse-path': '/root/miniconda3/envs/k2/lib/python3.9/site-packages/lhotse/__init__.py', 'hostname': 'VM-0-13-centos', 'IP address': '127.0.0.1'}, 'epoch': 23, 'iter': 0, 'avg': 11, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'beam_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'simulate_streaming': False, 'decode_chunk_size': 16, 'left_context': 64, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7/exp/beam_search'), 'suffix': 'epoch-23-avg-11-beam_search-beam-size-4-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-02 13:41:06,399 INFO [decode.py:705] About to create model
+2022-12-02 13:41:06,854 INFO [zipformer.py:179] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-02 13:41:06,864 INFO [decode.py:772] Calculating the averaged model over epoch range from 12 (excluded) to 23
+2022-12-02 13:41:10,607 INFO [decode.py:806] Number of model parameters: 70369391
+2022-12-02 13:41:10,607 INFO [asr_datamodule.py:408] About to get test cuts from data/fbank/xbmu_amdo31_cuts_test.jsonl.gz
+2022-12-02 13:42:14,866 INFO [decode.py:585] batch 0/?, cuts processed until now is 99
+2022-12-02 13:58:36,011 INFO [decode.py:585] batch 20/?, cuts processed until now is 2000
+2022-12-02 13:59:08,068 INFO [decode.py:601] The transcripts are stored in pruned_transducer_stateless7/exp/beam_search/recogs-test-beam_size_4-epoch-23-avg-11-beam_search-beam-size-4-use-averaged-model.txt
+2022-12-02 13:59:08,122 INFO [utils.py:530] [test-beam_size_4] %WER 9.77% [3285 / 33628, 298 ins, 292 del, 2695 sub ]
+2022-12-02 13:59:08,227 INFO [decode.py:614] Wrote detailed error stats to pruned_transducer_stateless7/exp/beam_search/errs-test-beam_size_4-epoch-23-avg-11-beam_search-beam-size-4-use-averaged-model.txt
+2022-12-02 13:59:08,228 INFO [decode.py:630] 
+For test, WER of different settings are:
+beam_size_4	9.77	best for test
+
+2022-12-02 13:59:08,228 INFO [decode.py:835] Done!

decoding-result/greedy_search/log-decode-epoch-23-avg-11-context-2-max-sym-per-frame-1-use-averaged-model-2022-12-02-11-43-34

@@ -0,0 +1,17 @@
+2022-12-02 11:43:34,585 INFO [decode.py:682] Decoding started
+2022-12-02 11:43:34,585 INFO [decode.py:688] Device: cuda:0
+2022-12-02 11:43:34,587 INFO [decode.py:703] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.22', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '', 'k2-git-date': '', 'lhotse-version': '1.10.0', 'torch-version': '1.12.1', 'torch-cuda-available': True, 'torch-cuda-version': '11.6', 'python-version': '3.9', 'icefall-git-branch': 'master', 'icefall-git-sha1': 'e5d9426-dirty', 'icefall-git-date': 'Tue Nov 22 11:45:03 2022', 'icefall-path': '/root/workspace/icefall', 'k2-path': '/root/workspace/k2/k2/python/k2/__init__.py', 'lhotse-path': '/root/miniconda3/envs/k2/lib/python3.9/site-packages/lhotse/__init__.py', 'hostname': 'VM-0-13-centos', 'IP address': '127.0.0.1'}, 'epoch': 23, 'iter': 0, 'avg': 11, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'greedy_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'simulate_streaming': False, 'decode_chunk_size': 16, 'left_context': 64, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7/exp/greedy_search'), 'suffix': 'epoch-23-avg-11-context-2-max-sym-per-frame-1-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-02 11:43:34,588 INFO [decode.py:705] About to create model
+2022-12-02 11:43:35,054 INFO [zipformer.py:179] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-02 11:43:35,064 INFO [decode.py:772] Calculating the averaged model over epoch range from 12 (excluded) to 23
+2022-12-02 11:43:38,758 INFO [decode.py:806] Number of model parameters: 70369391
+2022-12-02 11:43:38,758 INFO [asr_datamodule.py:408] About to get test cuts from data/fbank/xbmu_amdo31_cuts_test.jsonl.gz
+2022-12-02 11:43:42,144 INFO [decode.py:585] batch 0/?, cuts processed until now is 99
+2022-12-02 11:43:59,551 INFO [decode.py:601] The transcripts are stored in pruned_transducer_stateless7/exp/greedy_search/recogs-test-greedy_search-epoch-23-avg-11-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-02 11:43:59,603 INFO [utils.py:530] [test-greedy_search] %WER 10.13% [3405 / 33628, 260 ins, 396 del, 2749 sub ]
+2022-12-02 11:43:59,709 INFO [decode.py:614] Wrote detailed error stats to pruned_transducer_stateless7/exp/greedy_search/errs-test-greedy_search-epoch-23-avg-11-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-02 11:43:59,709 INFO [decode.py:630] 
+For test, WER of different settings are:
+greedy_search	10.13	best for test
+
+2022-12-02 11:43:59,709 INFO [decode.py:835] Done!

decoding-result/modified_beam_search/log-decode-epoch-23-avg-11-modified_beam_search-beam-size-4-use-averaged-model-2022-12-02-12-39-55

@@ -0,0 +1,18 @@
+2022-12-02 12:39:55,315 INFO [decode.py:682] Decoding started
+2022-12-02 12:39:55,315 INFO [decode.py:688] Device: cuda:0
+2022-12-02 12:39:55,317 INFO [decode.py:703] {'best_train_loss': inf, 'best_valid_loss': inf, 'best_train_epoch': -1, 'best_valid_epoch': -1, 'batch_idx_train': 0, 'log_interval': 50, 'reset_interval': 200, 'valid_interval': 3000, 'feature_dim': 80, 'subsampling_factor': 4, 'warm_step': 2000, 'env_info': {'k2-version': '1.22', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': '', 'k2-git-date': '', 'lhotse-version': '1.10.0', 'torch-version': '1.12.1', 'torch-cuda-available': True, 'torch-cuda-version': '11.6', 'python-version': '3.9', 'icefall-git-branch': 'master', 'icefall-git-sha1': 'e5d9426-dirty', 'icefall-git-date': 'Tue Nov 22 11:45:03 2022', 'icefall-path': '/root/workspace/icefall', 'k2-path': '/root/workspace/k2/k2/python/k2/__init__.py', 'lhotse-path': '/root/miniconda3/envs/k2/lib/python3.9/site-packages/lhotse/__init__.py', 'hostname': 'VM-0-13-centos', 'IP address': '127.0.0.1'}, 'epoch': 23, 'iter': 0, 'avg': 11, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7/exp'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'modified_beam_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'simulate_streaming': False, 'decode_chunk_size': 16, 'left_context': 64, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7/exp/modified_beam_search'), 'suffix': 'epoch-23-avg-11-modified_beam_search-beam-size-4-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-02 12:39:55,318 INFO [decode.py:705] About to create model
+2022-12-02 12:39:55,768 INFO [zipformer.py:179] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-02 12:39:55,778 INFO [decode.py:772] Calculating the averaged model over epoch range from 12 (excluded) to 23
+2022-12-02 12:39:59,502 INFO [decode.py:806] Number of model parameters: 70369391
+2022-12-02 12:39:59,502 INFO [asr_datamodule.py:408] About to get test cuts from data/fbank/xbmu_amdo31_cuts_test.jsonl.gz
+2022-12-02 12:40:05,457 INFO [decode.py:585] batch 0/?, cuts processed until now is 99
+2022-12-02 12:41:03,134 INFO [decode.py:585] batch 20/?, cuts processed until now is 2000
+2022-12-02 12:41:05,419 INFO [decode.py:601] The transcripts are stored in pruned_transducer_stateless7/exp/modified_beam_search/recogs-test-beam_size_4-epoch-23-avg-11-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-02 12:41:05,471 INFO [utils.py:530] [test-beam_size_4] %WER 9.70% [3262 / 33628, 283 ins, 292 del, 2687 sub ]
+2022-12-02 12:41:05,578 INFO [decode.py:614] Wrote detailed error stats to pruned_transducer_stateless7/exp/modified_beam_search/errs-test-beam_size_4-epoch-23-avg-11-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-02 12:41:05,578 INFO [decode.py:630] 
+For test, WER of different settings are:
+beam_size_4	9.7	best for test
+
+2022-12-02 12:41:05,578 INFO [decode.py:835] Done!

exp/cpu_jit.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:95bb54eaab329da342450f3248a038d5c1259a1c0c94857b42047bbb78cb4b6f
+size 281740798

exp/epoch-21.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:319cd1be0acf7a8a19ea06156db5dbbda5c0d61fa5bdddaf12ff862a20fc8e27
+size 1126566559

exp/epoch-22.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c0fd0dad1e34afd1bfa57bda22b58c6a7cc8a4fac48143d07c09c085981e3198
+size 1126566623

exp/epoch-23.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f112ea17965749d3e6d616ab2dce27129709f10adc1591975bef31bd386fb247
+size 1126566623

exp/pretrained.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9906a4ee24e4497d57b16a2405888b81cb163d941cf6fb77b759af3a0b7acee1
+size 281766253

exp/tensorboard/events.out.tfevents.1669893512.VM-0-13-centos.31587.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eae516765c8ebc2428e7cc76179701315be5f24b0c72499c9dd866f6ab6454dd
+size 372641

test_wavs/trans.txt

@@ -0,0 +1,3 @@
+a_0_cacm-A70_31116.wav ལོ བཅུ ཙམ མ འདང བའི དུས སྐབས ནང
+a_0_cacm-A70_31117.wav དྲག པོའི ངོ ལོག ཟིང འཁྲུག སྒྲིག འཛུགས དང ངན བཀོད བྱས ཡོད
+a_0_cacm-A70_31118.wav གནས བབ འདིའི རིགས གང མགྱོགས འགྱུར བ གཏོང དགོས