upload files

README.md

@@ -0,0 +1 @@
+See https://github.com/k2-fsa/icefall/pull/1111

data/lang_bpe_500/HLG.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b5dbbe8b485c0cb37d11e07e8e734990f1e40a2d00fe9689d8da2e7b6fe72883
+size 845007583

data/lang_bpe_500/L.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b1b88996f918737fba67fbd29152018b51a537c16ce0718a2b43d5140583224e
+size 19025703

data/lang_bpe_500/LG.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3bb9f021c7aad79d45dc275ba8154a430c4f660a319dcb872cd52500f25553d6
+size 249852195

data/lang_bpe_500/Linv.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cbc8b3687a1b8f0811a84106b3b310642566c7b1bc282a929878f9269507a2c6
+size 19025703

data/lang_bpe_500/bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c53433de083c4a6ad12d034550ef22de68cec62c4f58932a7b6b8b2f1e743fa5
+size 244865

data/lang_bpe_500/tokens.txt

@@ -0,0 +1,502 @@
+<blk> 0
+<sos/eos> 1
+<unk> 2
+S 3
+▁THE 4
+▁A 5
+T 6
+▁AND 7
+ED 8
+▁OF 9
+▁TO 10
+E 11
+D 12
+N 13
+ING 14
+▁IN 15
+Y 16
+M 17
+C 18
+▁I 19
+A 20
+P 21
+▁HE 22
+R 23
+O 24
+L 25
+RE 26
+I 27
+U 28
+ER 29
+▁IT 30
+LY 31
+▁THAT 32
+▁WAS 33
+▁ 34
+▁S 35
+AR 36
+▁BE 37
+F 38
+▁C 39
+IN 40
+B 41
+▁FOR 42
+OR 43
+LE 44
+' 45
+▁HIS 46
+▁YOU 47
+AL 48
+▁RE 49
+V 50
+▁B 51
+G 52
+RI 53
+▁E 54
+▁WITH 55
+▁T 56
+▁AS 57
+LL 58
+▁P 59
+▁HER 60
+ST 61
+▁HAD 62
+▁SO 63
+▁F 64
+W 65
+CE 66
+▁IS 67
+ND 68
+▁NOT 69
+TH 70
+▁BUT 71
+EN 72
+▁SHE 73
+▁ON 74
+VE 75
+ON 76
+SE 77
+▁DE 78
+UR 79
+▁G 80
+CH 81
+K 82
+TER 83
+▁AT 84
+IT 85
+▁ME 86
+RO 87
+NE 88
+RA 89
+ES 90
+IL 91
+NG 92
+IC 93
+▁NO 94
+▁HIM 95
+ENT 96
+IR 97
+▁WE 98
+H 99
+▁DO 100
+▁ALL 101
+▁HAVE 102
+LO 103
+▁BY 104
+▁MY 105
+▁MO 106
+▁THIS 107
+LA 108
+▁ST 109
+▁WHICH 110
+▁CON 111
+▁THEY 112
+CK 113
+TE 114
+▁SAID 115
+▁FROM 116
+▁GO 117
+▁WHO 118
+▁TH 119
+▁OR 120
+▁D 121
+▁W 122
+VER 123
+LI 124
+▁SE 125
+▁ONE 126
+▁CA 127
+▁AN 128
+▁LA 129
+▁WERE 130
+EL 131
+▁HA 132
+▁MAN 133
+▁FA 134
+▁EX 135
+AD 136
+▁SU 137
+RY 138
+▁MI 139
+AT 140
+▁BO 141
+▁WHEN 142
+AN 143
+THER 144
+PP 145
+ATION 146
+▁FI 147
+▁WOULD 148
+▁PRO 149
+OW 150
+ET 151
+▁O 152
+▁THERE 153
+▁HO 154
+ION 155
+▁WHAT 156
+▁FE 157
+▁PA 158
+US 159
+MENT 160
+▁MA 161
+UT 162
+▁OUT 163
+▁THEIR 164
+▁IF 165
+▁LI 166
+▁K 167
+▁WILL 168
+▁ARE 169
+ID 170
+▁RO 171
+DE 172
+TION 173
+▁WA 174
+PE 175
+▁UP 176
+▁SP 177
+▁PO 178
+IGHT 179
+▁UN 180
+RU 181
+▁LO 182
+AS 183
+OL 184
+▁LE 185
+▁BEEN 186
+▁SH 187
+▁RA 188
+▁SEE 189
+KE 190
+UL 191
+TED 192
+▁SA 193
+UN 194
+UND 195
+ANT 196
+▁NE 197
+IS 198
+▁THEM 199
+CI 200
+GE 201
+▁COULD 202
+▁DIS 203
+OM 204
+ISH 205
+HE 206
+EST 207
+▁SOME 208
+ENCE 209
+ITY 210
+IVE 211
+▁US 212
+▁MORE 213
+▁EN 214
+ARD 215
+ATE 216
+▁YOUR 217
+▁INTO 218
+▁KNOW 219
+▁CO 220
+ANCE 221
+▁TIME 222
+▁WI 223
+▁YE 224
+AGE 225
+▁NOW 226
+TI 227
+FF 228
+ABLE 229
+▁VERY 230
+▁LIKE 231
+AM 232
+HI 233
+Z 234
+▁OTHER 235
+▁THAN 236
+▁LITTLE 237
+▁DID 238
+▁LOOK 239
+TY 240
+ERS 241
+▁CAN 242
+▁CHA 243
+▁AR 244
+X 245
+FUL 246
+UGH 247
+▁BA 248
+▁DAY 249
+▁ABOUT 250
+TEN 251
+IM 252
+▁ANY 253
+▁PRE 254
+▁OVER 255
+IES 256
+NESS 257
+ME 258
+BLE 259
+▁M 260
+ROW 261
+▁HAS 262
+▁GREAT 263
+▁VI 264
+TA 265
+▁AFTER 266
+PER 267
+▁AGAIN 268
+HO 269
+SH 270
+▁UPON 271
+▁DI 272
+▁HAND 273
+▁COM 274
+IST 275
+TURE 276
+▁STA 277
+▁THEN 278
+▁SHOULD 279
+▁GA 280
+OUS 281
+OUR 282
+▁WELL 283
+▁ONLY 284
+MAN 285
+▁GOOD 286
+▁TWO 287
+▁MAR 288
+▁SAY 289
+▁HU 290
+TING 291
+▁OUR 292
+RESS 293
+▁DOWN 294
+IOUS 295
+▁BEFORE 296
+▁DA 297
+▁NA 298
+QUI 299
+▁MADE 300
+▁EVERY 301
+▁OLD 302
+▁EVEN 303
+IG 304
+▁COME 305
+▁GRA 306
+▁RI 307
+▁LONG 308
+OT 309
+SIDE 310
+WARD 311
+▁FO 312
+▁WHERE 313
+MO 314
+LESS 315
+▁SC 316
+▁MUST 317
+▁NEVER 318
+▁HOW 319
+▁CAME 320
+▁SUCH 321
+▁RU 322
+▁TAKE 323
+▁WO 324
+▁CAR 325
+UM 326
+AK 327
+▁THINK 328
+▁MUCH 329
+▁MISTER 330
+▁MAY 331
+▁JO 332
+▁WAY 333
+▁COMP 334
+▁THOUGHT 335
+▁STO 336
+▁MEN 337
+▁BACK 338
+▁DON 339
+J 340
+▁LET 341
+▁TRA 342
+▁FIRST 343
+▁JUST 344
+▁VA 345
+▁OWN 346
+▁PLA 347
+▁MAKE 348
+ATED 349
+▁HIMSELF 350
+▁WENT 351
+▁PI 352
+GG 353
+RING 354
+▁DU 355
+▁MIGHT 356
+▁PART 357
+▁GIVE 358
+▁IMP 359
+▁BU 360
+▁PER 361
+▁PLACE 362
+▁HOUSE 363
+▁THROUGH 364
+IAN 365
+▁SW 366
+▁UNDER 367
+QUE 368
+▁AWAY 369
+▁LOVE 370
+QUA 371
+▁LIFE 372
+▁GET 373
+▁WITHOUT 374
+▁PASS 375
+▁TURN 376
+IGN 377
+▁HEAD 378
+▁MOST 379
+▁THOSE 380
+▁SHALL 381
+▁EYES 382
+▁COL 383
+▁STILL 384
+▁NIGHT 385
+▁NOTHING 386
+ITION 387
+HA 388
+▁TELL 389
+▁WORK 390
+▁LAST 391
+▁NEW 392
+▁FACE 393
+▁HI 394
+▁WORD 395
+▁FOUND 396
+▁COUNT 397
+▁OB 398
+▁WHILE 399
+▁SHA 400
+▁MEAN 401
+▁SAW 402
+▁PEOPLE 403
+▁FRIEND 404
+▁THREE 405
+▁ROOM 406
+▁SAME 407
+▁THOUGH 408
+▁RIGHT 409
+▁CHILD 410
+▁FATHER 411
+▁ANOTHER 412
+▁HEART 413
+▁WANT 414
+▁TOOK 415
+OOK 416
+▁LIGHT 417
+▁MISSUS 418
+▁OPEN 419
+▁JU 420
+▁ASKED 421
+PORT 422
+▁LEFT 423
+▁JA 424
+▁WORLD 425
+▁HOME 426
+▁WHY 427
+▁ALWAYS 428
+▁ANSWER 429
+▁SEEMED 430
+▁SOMETHING 431
+▁GIRL 432
+▁BECAUSE 433
+▁NAME 434
+▁TOLD 435
+▁NI 436
+▁HIGH 437
+IZE 438
+▁WOMAN 439
+▁FOLLOW 440
+▁RETURN 441
+▁KNEW 442
+▁EACH 443
+▁KIND 444
+▁JE 445
+▁ACT 446
+▁LU 447
+▁CERTAIN 448
+▁YEARS 449
+▁QUITE 450
+▁APPEAR 451
+▁BETTER 452
+▁HALF 453
+▁PRESENT 454
+▁PRINCE 455
+SHIP 456
+▁ALSO 457
+▁BEGAN 458
+▁HAVING 459
+▁ENOUGH 460
+▁PERSON 461
+▁LADY 462
+▁WHITE 463
+▁COURSE 464
+▁VOICE 465
+▁SPEAK 466
+▁POWER 467
+▁MORNING 468
+▁BETWEEN 469
+▁AMONG 470
+▁KEEP 471
+▁WALK 472
+▁MATTER 473
+▁TEA 474
+▁BELIEVE 475
+▁SMALL 476
+▁TALK 477
+▁FELT 478
+▁HORSE 479
+▁MYSELF 480
+▁SIX 481
+▁HOWEVER 482
+▁FULL 483
+▁HERSELF 484
+▁POINT 485
+▁STOOD 486
+▁HUNDRED 487
+▁ALMOST 488
+▁SINCE 489
+▁LARGE 490
+▁LEAVE 491
+▁PERHAPS 492
+▁DARK 493
+▁SUDDEN 494
+▁REPLIED 495
+▁ANYTHING 496
+▁WONDER 497
+▁UNTIL 498
+Q 499
+#0 500
+#1 501

data/lm/G_4_gram.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c91581bc632f9c72c557ccdf726298255b9627a6ac38270c51891459b82630e9
+size 3700956587

decoding_results/1best/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-19-45-46

@@ -0,0 +1,26 @@
+2023-06-13 19:45:46,339 INFO [ctc_decode.py:641] Decoding started
+2023-06-13 19:45:46,340 INFO [ctc_decode.py:647] Device: cuda:0
+2023-06-13 19:45:46,340 INFO [ctc_decode.py:648] {'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': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-dirty', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': '1best', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/my_lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, '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('zipformer/exp-ctc-rnnt/1best'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
+2023-06-13 19:45:46,609 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-06-13 19:45:51,497 INFO [ctc_decode.py:726] About to create model
+2023-06-13 19:45:52,095 INFO [ctc_decode.py:793] Calculating the averaged model over epoch range from 24 (excluded) to 40
+2023-06-13 19:45:54,495 INFO [ctc_decode.py:810] Number of model parameters: 65805511
+2023-06-13 19:45:54,496 INFO [asr_datamodule.py:465] About to get test-clean cuts
+2023-06-13 19:45:54,500 INFO [asr_datamodule.py:472] About to get test-other cuts
+2023-06-13 19:45:55,448 INFO [ctc_decode.py:558] batch 0/?, cuts processed until now is 21
+2023-06-13 19:46:22,063 INFO [ctc_decode.py:572] The transcripts are stored in zipformer/exp-ctc-rnnt/1best/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 19:46:22,133 INFO [utils.py:561] [test-clean-no_rescore] %WER 2.46% [1294 / 52576, 181 ins, 93 del, 1020 sub ]
+2023-06-13 19:46:22,308 INFO [ctc_decode.py:581] Wrote detailed error stats to zipformer/exp-ctc-rnnt/1best/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 19:46:22,309 INFO [ctc_decode.py:595] 
+For test-clean, WER of different settings are:
+no_rescore	2.46	best for test-clean
+
+2023-06-13 19:46:23,023 INFO [ctc_decode.py:558] batch 0/?, cuts processed until now is 26
+2023-06-13 19:46:50,469 INFO [ctc_decode.py:572] The transcripts are stored in zipformer/exp-ctc-rnnt/1best/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 19:46:50,541 INFO [utils.py:561] [test-other-no_rescore] %WER 5.11% [2674 / 52343, 288 ins, 256 del, 2130 sub ]
+2023-06-13 19:46:50,716 INFO [ctc_decode.py:581] Wrote detailed error stats to zipformer/exp-ctc-rnnt/1best/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 19:46:50,717 INFO [ctc_decode.py:595] 
+For test-other, WER of different settings are:
+no_rescore	5.11	best for test-other
+
+2023-06-13 19:46:50,717 INFO [ctc_decode.py:843] Done!

decoding_results/1best/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+no_rescore	2.46

decoding_results/1best/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+no_rescore	5.11

decoding_results/ctc-decoding/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-19-42-20

@@ -0,0 +1,31 @@
+2023-06-13 19:42:20,348 INFO [ctc_decode.py:641] Decoding started
+2023-06-13 19:42:20,349 INFO [ctc_decode.py:647] Device: cuda:0
+2023-06-13 19:42:20,349 INFO [ctc_decode.py:648] {'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': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-dirty', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'ctc-decoding', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/my_lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, '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('zipformer/exp-ctc-rnnt/ctc-decoding'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
+2023-06-13 19:42:20,615 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-06-13 19:42:24,473 INFO [ctc_decode.py:726] About to create model
+2023-06-13 19:42:25,063 INFO [ctc_decode.py:793] Calculating the averaged model over epoch range from 24 (excluded) to 40
+2023-06-13 19:42:27,461 INFO [ctc_decode.py:810] Number of model parameters: 65805511
+2023-06-13 19:42:27,462 INFO [asr_datamodule.py:465] About to get test-clean cuts
+2023-06-13 19:42:27,466 INFO [asr_datamodule.py:472] About to get test-other cuts
+2023-06-13 19:42:28,473 INFO [ctc_decode.py:558] batch 0/?, cuts processed until now is 21
+2023-06-13 19:42:54,474 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.6098, 4.0538, 4.2424, 4.1121], device='cuda:0')
+2023-06-13 19:42:57,548 INFO [ctc_decode.py:572] The transcripts are stored in zipformer/exp-ctc-rnnt/ctc-decoding/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 19:42:57,621 INFO [utils.py:561] [test-clean-ctc-decoding] %WER 2.40% [1262 / 52576, 133 ins, 89 del, 1040 sub ]
+2023-06-13 19:42:57,796 INFO [ctc_decode.py:581] Wrote detailed error stats to zipformer/exp-ctc-rnnt/ctc-decoding/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 19:42:57,796 INFO [ctc_decode.py:595] 
+For test-clean, WER of different settings are:
+ctc-decoding	2.4	best for test-clean
+
+2023-06-13 19:42:58,556 INFO [ctc_decode.py:558] batch 0/?, cuts processed until now is 26
+2023-06-13 19:43:06,459 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.0038, 3.6233, 3.5348, 3.5192], device='cuda:0')
+2023-06-13 19:43:10,237 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.2716, 2.5148, 3.4661, 1.9089], device='cuda:0')
+2023-06-13 19:43:19,177 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([2.3797, 3.3571, 3.0888, 2.5630], device='cuda:0')
+2023-06-13 19:43:25,600 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.5188, 4.0134, 3.8820, 4.2079], device='cuda:0')
+2023-06-13 19:43:27,974 INFO [ctc_decode.py:572] The transcripts are stored in zipformer/exp-ctc-rnnt/ctc-decoding/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 19:43:28,074 INFO [utils.py:561] [test-other-ctc-decoding] %WER 5.66% [2961 / 52343, 278 ins, 224 del, 2459 sub ]
+2023-06-13 19:43:28,356 INFO [ctc_decode.py:581] Wrote detailed error stats to zipformer/exp-ctc-rnnt/ctc-decoding/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 19:43:28,357 INFO [ctc_decode.py:595] 
+For test-other, WER of different settings are:
+ctc-decoding	5.66	best for test-other
+
+2023-06-13 19:43:28,357 INFO [ctc_decode.py:843] Done!

decoding_results/ctc-decoding/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+ctc-decoding	2.4

decoding_results/ctc-decoding/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+ctc-decoding	5.66

decoding_results/nbest-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-24-42

@@ -0,0 +1,6 @@
+2023-06-13 17:24:42,946 INFO [ctc_decode.py:633] Decoding started
+2023-06-13 17:24:42,946 INFO [ctc_decode.py:639] Device: cuda:0
+2023-06-13 17:24:42,946 INFO [ctc_decode.py:640] {'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': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-clean', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'nbest-rescoring', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, '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('zipformer/exp-ctc-rnnt/nbest-rescoring'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
+2023-06-13 17:24:43,219 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-06-13 17:24:48,038 INFO [ctc_decode.py:676] Loading G_4_gram.fst.txt
+2023-06-13 17:24:48,038 WARNING [ctc_decode.py:677] It may take 8 minutes.

decoding_results/nbest-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-28-30

@@ -0,0 +1,180 @@
+2023-06-13 17:28:30,904 INFO [ctc_decode.py:633] Decoding started
+2023-06-13 17:28:30,904 INFO [ctc_decode.py:639] Device: cuda:0
+2023-06-13 17:28:30,904 INFO [ctc_decode.py:640] {'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': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-clean', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'nbest-rescoring', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/my_lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, '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('zipformer/exp-ctc-rnnt/nbest-rescoring'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
+2023-06-13 17:28:31,179 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-06-13 17:28:35,887 INFO [ctc_decode.py:701] Loading pre-compiled G_4_gram.pt
+2023-06-13 17:28:58,167 INFO [ctc_decode.py:718] About to create model
+2023-06-13 17:28:58,818 INFO [ctc_decode.py:785] Calculating the averaged model over epoch range from 24 (excluded) to 40
+2023-06-13 17:29:02,258 INFO [ctc_decode.py:802] Number of model parameters: 65805511
+2023-06-13 17:29:02,259 INFO [asr_datamodule.py:465] About to get test-clean cuts
+2023-06-13 17:29:02,262 INFO [asr_datamodule.py:472] About to get test-other cuts
+2023-06-13 17:29:03,502 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 21
+2023-06-13 17:29:39,698 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.6109, 4.3406, 4.1379, 4.6409], device='cuda:0')
+2023-06-13 17:29:53,312 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:53,387 INFO [utils.py:561] [test-clean-lm_scale_0.1] %WER 2.49% [1309 / 52576, 219 ins, 68 del, 1022 sub ]
+2023-06-13 17:29:53,562 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:53,585 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:53,654 INFO [utils.py:561] [test-clean-lm_scale_0.2] %WER 2.45% [1289 / 52576, 211 ins, 68 del, 1010 sub ]
+2023-06-13 17:29:53,833 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:53,858 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:54,096 INFO [utils.py:561] [test-clean-lm_scale_0.3] %WER 2.42% [1272 / 52576, 202 ins, 72 del, 998 sub ]
+2023-06-13 17:29:54,266 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:54,289 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:54,356 INFO [utils.py:561] [test-clean-lm_scale_0.4] %WER 2.39% [1254 / 52576, 192 ins, 74 del, 988 sub ]
+2023-06-13 17:29:54,529 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:54,552 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:54,622 INFO [utils.py:561] [test-clean-lm_scale_0.5] %WER 2.37% [1248 / 52576, 184 ins, 81 del, 983 sub ]
+2023-06-13 17:29:54,795 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:54,817 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:54,886 INFO [utils.py:561] [test-clean-lm_scale_0.6] %WER 2.38% [1250 / 52576, 178 ins, 87 del, 985 sub ]
+2023-06-13 17:29:55,057 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:55,080 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:55,147 INFO [utils.py:561] [test-clean-lm_scale_0.7] %WER 2.40% [1261 / 52576, 172 ins, 97 del, 992 sub ]
+2023-06-13 17:29:55,320 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:55,342 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:55,409 INFO [utils.py:561] [test-clean-lm_scale_0.8] %WER 2.42% [1272 / 52576, 167 ins, 108 del, 997 sub ]
+2023-06-13 17:29:55,577 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:55,599 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:55,822 INFO [utils.py:561] [test-clean-lm_scale_0.9] %WER 2.44% [1284 / 52576, 158 ins, 123 del, 1003 sub ]
+2023-06-13 17:29:55,994 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:56,016 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:56,082 INFO [utils.py:561] [test-clean-lm_scale_1.0] %WER 2.50% [1312 / 52576, 154 ins, 140 del, 1018 sub ]
+2023-06-13 17:29:56,252 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:56,275 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:56,343 INFO [utils.py:561] [test-clean-lm_scale_1.1] %WER 2.56% [1348 / 52576, 156 ins, 155 del, 1037 sub ]
+2023-06-13 17:29:56,511 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:56,534 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:56,601 INFO [utils.py:561] [test-clean-lm_scale_1.2] %WER 2.60% [1367 / 52576, 153 ins, 162 del, 1052 sub ]
+2023-06-13 17:29:56,773 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:56,795 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:56,864 INFO [utils.py:561] [test-clean-lm_scale_1.3] %WER 2.66% [1396 / 52576, 153 ins, 179 del, 1064 sub ]
+2023-06-13 17:29:57,033 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:57,056 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:57,123 INFO [utils.py:561] [test-clean-lm_scale_1.4] %WER 2.69% [1414 / 52576, 150 ins, 191 del, 1073 sub ]
+2023-06-13 17:29:57,294 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:57,316 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:57,383 INFO [utils.py:561] [test-clean-lm_scale_1.5] %WER 2.73% [1436 / 52576, 149 ins, 203 del, 1084 sub ]
+2023-06-13 17:29:57,707 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:57,730 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:57,798 INFO [utils.py:561] [test-clean-lm_scale_1.6] %WER 2.80% [1471 / 52576, 149 ins, 220 del, 1102 sub ]
+2023-06-13 17:29:57,974 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:57,996 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:58,062 INFO [utils.py:561] [test-clean-lm_scale_1.7] %WER 2.85% [1501 / 52576, 149 ins, 227 del, 1125 sub ]
+2023-06-13 17:29:58,237 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:58,259 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:58,326 INFO [utils.py:561] [test-clean-lm_scale_1.8] %WER 2.89% [1520 / 52576, 149 ins, 233 del, 1138 sub ]
+2023-06-13 17:29:58,494 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:58,516 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:58,583 INFO [utils.py:561] [test-clean-lm_scale_1.9] %WER 2.94% [1545 / 52576, 147 ins, 242 del, 1156 sub ]
+2023-06-13 17:29:58,750 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:58,773 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:58,841 INFO [utils.py:561] [test-clean-lm_scale_2.0] %WER 2.98% [1566 / 52576, 147 ins, 252 del, 1167 sub ]
+2023-06-13 17:29:59,014 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:29:59,015 INFO [ctc_decode.py:587] 
+For test-clean, WER of different settings are:
+lm_scale_0.5	2.37	best for test-clean
+lm_scale_0.6	2.38
+lm_scale_0.4	2.39
+lm_scale_0.7	2.4
+lm_scale_0.3	2.42
+lm_scale_0.8	2.42
+lm_scale_0.9	2.44
+lm_scale_0.2	2.45
+lm_scale_0.1	2.49
+lm_scale_1.0	2.5
+lm_scale_1.1	2.56
+lm_scale_1.2	2.6
+lm_scale_1.3	2.66
+lm_scale_1.4	2.69
+lm_scale_1.5	2.73
+lm_scale_1.6	2.8
+lm_scale_1.7	2.85
+lm_scale_1.8	2.89
+lm_scale_1.9	2.94
+lm_scale_2.0	2.98
+
+2023-06-13 17:30:00,580 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 26
+2023-06-13 17:31:00,699 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:00,770 INFO [utils.py:561] [test-other-lm_scale_0.1] %WER 5.20% [2721 / 52343, 362 ins, 183 del, 2176 sub ]
+2023-06-13 17:31:00,946 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:00,970 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:01,039 INFO [utils.py:561] [test-other-lm_scale_0.2] %WER 5.10% [2670 / 52343, 355 ins, 183 del, 2132 sub ]
+2023-06-13 17:31:01,212 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:01,235 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:01,306 INFO [utils.py:561] [test-other-lm_scale_0.3] %WER 5.04% [2638 / 52343, 342 ins, 187 del, 2109 sub ]
+2023-06-13 17:31:01,481 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:01,503 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:01,573 INFO [utils.py:561] [test-other-lm_scale_0.4] %WER 4.99% [2610 / 52343, 323 ins, 198 del, 2089 sub ]
+2023-06-13 17:31:01,745 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:01,769 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:01,842 INFO [utils.py:561] [test-other-lm_scale_0.5] %WER 4.95% [2592 / 52343, 316 ins, 214 del, 2062 sub ]
+2023-06-13 17:31:02,020 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:02,042 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:02,110 INFO [utils.py:561] [test-other-lm_scale_0.6] %WER 4.93% [2583 / 52343, 290 ins, 234 del, 2059 sub ]
+2023-06-13 17:31:02,283 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:02,306 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:02,378 INFO [utils.py:561] [test-other-lm_scale_0.7] %WER 4.94% [2584 / 52343, 279 ins, 250 del, 2055 sub ]
+2023-06-13 17:31:02,550 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:02,573 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:02,641 INFO [utils.py:561] [test-other-lm_scale_0.8] %WER 4.99% [2614 / 52343, 265 ins, 288 del, 2061 sub ]
+2023-06-13 17:31:02,975 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:02,998 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:03,067 INFO [utils.py:561] [test-other-lm_scale_0.9] %WER 5.03% [2635 / 52343, 252 ins, 305 del, 2078 sub ]
+2023-06-13 17:31:03,244 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:03,268 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:03,338 INFO [utils.py:561] [test-other-lm_scale_1.0] %WER 5.08% [2660 / 52343, 250 ins, 331 del, 2079 sub ]
+2023-06-13 17:31:03,513 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:03,537 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:03,608 INFO [utils.py:561] [test-other-lm_scale_1.1] %WER 5.16% [2701 / 52343, 244 ins, 350 del, 2107 sub ]
+2023-06-13 17:31:03,782 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:03,805 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:03,874 INFO [utils.py:561] [test-other-lm_scale_1.2] %WER 5.25% [2748 / 52343, 241 ins, 383 del, 2124 sub ]
+2023-06-13 17:31:04,060 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:04,085 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:04,153 INFO [utils.py:561] [test-other-lm_scale_1.3] %WER 5.30% [2772 / 52343, 237 ins, 404 del, 2131 sub ]
+2023-06-13 17:31:04,327 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:04,350 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:04,418 INFO [utils.py:561] [test-other-lm_scale_1.4] %WER 5.38% [2815 / 52343, 231 ins, 422 del, 2162 sub ]
+2023-06-13 17:31:04,591 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:04,615 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:04,683 INFO [utils.py:561] [test-other-lm_scale_1.5] %WER 5.45% [2855 / 52343, 230 ins, 452 del, 2173 sub ]
+2023-06-13 17:31:05,037 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:05,060 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:05,129 INFO [utils.py:561] [test-other-lm_scale_1.6] %WER 5.52% [2887 / 52343, 226 ins, 469 del, 2192 sub ]
+2023-06-13 17:31:05,304 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:05,327 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:05,395 INFO [utils.py:561] [test-other-lm_scale_1.7] %WER 5.58% [2919 / 52343, 227 ins, 492 del, 2200 sub ]
+2023-06-13 17:31:05,572 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:05,595 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:05,664 INFO [utils.py:561] [test-other-lm_scale_1.8] %WER 5.63% [2947 / 52343, 227 ins, 506 del, 2214 sub ]
+2023-06-13 17:31:05,840 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:05,863 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:05,938 INFO [utils.py:561] [test-other-lm_scale_1.9] %WER 5.70% [2984 / 52343, 227 ins, 520 del, 2237 sub ]
+2023-06-13 17:31:06,113 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:06,136 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:06,204 INFO [utils.py:561] [test-other-lm_scale_2.0] %WER 5.74% [3007 / 52343, 224 ins, 532 del, 2251 sub ]
+2023-06-13 17:31:06,379 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:31:06,380 INFO [ctc_decode.py:587] 
+For test-other, WER of different settings are:
+lm_scale_0.6	4.93	best for test-other
+lm_scale_0.7	4.94
+lm_scale_0.5	4.95
+lm_scale_0.4	4.99
+lm_scale_0.8	4.99
+lm_scale_0.9	5.03
+lm_scale_0.3	5.04
+lm_scale_1.0	5.08
+lm_scale_0.2	5.1
+lm_scale_1.1	5.16
+lm_scale_0.1	5.2
+lm_scale_1.2	5.25
+lm_scale_1.3	5.3
+lm_scale_1.4	5.38
+lm_scale_1.5	5.45
+lm_scale_1.6	5.52
+lm_scale_1.7	5.58
+lm_scale_1.8	5.63
+lm_scale_1.9	5.7
+lm_scale_2.0	5.74
+
+2023-06-13 17:31:06,380 INFO [ctc_decode.py:835] Done!

decoding_results/nbest-rescoring/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,21 @@
+settings	WER
+lm_scale_0.5	2.37
+lm_scale_0.6	2.38
+lm_scale_0.4	2.39
+lm_scale_0.7	2.4
+lm_scale_0.3	2.42
+lm_scale_0.8	2.42
+lm_scale_0.9	2.44
+lm_scale_0.2	2.45
+lm_scale_0.1	2.49
+lm_scale_1.0	2.5
+lm_scale_1.1	2.56
+lm_scale_1.2	2.6
+lm_scale_1.3	2.66
+lm_scale_1.4	2.69
+lm_scale_1.5	2.73
+lm_scale_1.6	2.8
+lm_scale_1.7	2.85
+lm_scale_1.8	2.89
+lm_scale_1.9	2.94
+lm_scale_2.0	2.98

decoding_results/nbest-rescoring/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,21 @@
+settings	WER
+lm_scale_0.6	4.93
+lm_scale_0.7	4.94
+lm_scale_0.5	4.95
+lm_scale_0.4	4.99
+lm_scale_0.8	4.99
+lm_scale_0.9	5.03
+lm_scale_0.3	5.04
+lm_scale_1.0	5.08
+lm_scale_0.2	5.1
+lm_scale_1.1	5.16
+lm_scale_0.1	5.2
+lm_scale_1.2	5.25
+lm_scale_1.3	5.3
+lm_scale_1.4	5.38
+lm_scale_1.5	5.45
+lm_scale_1.6	5.52
+lm_scale_1.7	5.58
+lm_scale_1.8	5.63
+lm_scale_1.9	5.7
+lm_scale_2.0	5.74

decoding_results/nbest/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-21-29

@@ -0,0 +1,35 @@
+2023-06-13 17:21:29,434 INFO [ctc_decode.py:633] Decoding started
+2023-06-13 17:21:29,434 INFO [ctc_decode.py:639] Device: cuda:0
+2023-06-13 17:21:29,434 INFO [ctc_decode.py:640] {'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': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-clean', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'nbest', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, '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('zipformer/exp-ctc-rnnt/nbest'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
+2023-06-13 17:21:29,725 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-06-13 17:21:34,644 INFO [ctc_decode.py:718] About to create model
+2023-06-13 17:21:35,244 INFO [ctc_decode.py:785] Calculating the averaged model over epoch range from 24 (excluded) to 40
+2023-06-13 17:21:39,252 INFO [ctc_decode.py:802] Number of model parameters: 65805511
+2023-06-13 17:21:39,252 INFO [asr_datamodule.py:465] About to get test-clean cuts
+2023-06-13 17:21:39,255 INFO [asr_datamodule.py:472] About to get test-other cuts
+2023-06-13 17:21:40,545 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 21
+2023-06-13 17:21:42,976 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([5.0315, 4.4660, 4.2767, 4.5975], device='cuda:0')
+2023-06-13 17:21:55,627 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.4431, 2.6692, 3.6222, 2.0333], device='cuda:0')
+2023-06-13 17:22:11,994 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([2.2852, 3.4389, 3.4427, 3.4871, 3.2347, 2.9755, 2.4241, 2.8685],
+       device='cuda:0')
+2023-06-13 17:22:16,053 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.5884, 2.7881, 3.7633, 2.0764], device='cuda:0')
+2023-06-13 17:22:20,226 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([3.9332, 4.9863, 4.5516, 4.1826], device='cuda:0')
+2023-06-13 17:22:26,654 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([2.8336, 4.2831, 4.1287, 3.3007, 3.3161, 3.7073, 4.0190, 3.6861],
+       device='cuda:0')
+2023-06-13 17:22:28,090 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:22:28,169 INFO [utils.py:561] [test-clean-no_rescore-nbest-scale-1.0-100] %WER 2.46% [1294 / 52576, 181 ins, 93 del, 1020 sub ]
+2023-06-13 17:22:28,344 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:22:28,345 INFO [ctc_decode.py:587] 
+For test-clean, WER of different settings are:
+no_rescore-nbest-scale-1.0-100	2.46	best for test-clean
+
+2023-06-13 17:22:29,623 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 26
+2023-06-13 17:22:59,061 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([3.7139, 3.3612, 2.7698, 3.3808], device='cuda:0')
+2023-06-13 17:23:25,555 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/nbest/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:23:25,630 INFO [utils.py:561] [test-other-no_rescore-nbest-scale-1.0-100] %WER 5.11% [2674 / 52343, 288 ins, 256 del, 2130 sub ]
+2023-06-13 17:23:25,808 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/nbest/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:23:25,808 INFO [ctc_decode.py:587] 
+For test-other, WER of different settings are:
+no_rescore-nbest-scale-1.0-100	5.11	best for test-other
+
+2023-06-13 17:23:25,808 INFO [ctc_decode.py:835] Done!

decoding_results/nbest/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+no_rescore-nbest-scale-1.0-100	2.46

decoding_results/nbest/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+no_rescore-nbest-scale-1.0-100	5.11

decoding_results/whole-lattice-rescoring/log-decode-epoch-40-avg-16-use-averaged-model-2023-06-13-17-33-26

@@ -0,0 +1,251 @@
+2023-06-13 17:33:26,582 INFO [ctc_decode.py:633] Decoding started
+2023-06-13 17:33:26,582 INFO [ctc_decode.py:639] Device: cuda:0
+2023-06-13 17:33:26,582 INFO [ctc_decode.py:640] {'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': '1.24.1', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'c51a0b9684442a88ee37f3ce0af686a04b66855b', 'k2-git-date': 'Mon May 1 21:38:03 2023', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'new-zipformer-add-ctc', 'icefall-git-sha1': '046b6cb6-clean', 'icefall-git-date': 'Fri Jun 2 15:51:49 2023', 'icefall-path': '/ceph-zw/workspace/zipformer/icefall_zipformer', 'k2-path': '/ceph-zw/workspace/k2/k2/k2/python/k2/__init__.py', 'lhotse-path': '/ceph-zw/workspace/share/lhotse/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-1-1220091118-57c4d55446-mlpzc', 'IP address': '10.177.22.19'}, 'frame_shift_ms': 10, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'use_double_scores': True, 'epoch': 40, 'iter': 0, 'avg': 16, 'use_averaged_model': True, 'exp_dir': PosixPath('zipformer/exp-ctc-rnnt'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'context_size': 2, 'decoding_method': 'whole-lattice-rescoring', 'num_paths': 100, 'nbest_scale': 1.0, 'hlg_scale': 0.6, 'lm_dir': PosixPath('data/my_lm'), 'num_encoder_layers': '2,2,3,4,3,2', 'downsampling_factor': '1,2,4,8,4,2', 'feedforward_dim': '512,768,1024,1536,1024,768', 'num_heads': '4,4,4,8,4,4', 'encoder_dim': '192,256,384,512,384,256', 'query_head_dim': '32', 'value_head_dim': '12', 'pos_head_dim': '4', 'pos_dim': 48, 'encoder_unmasked_dim': '192,192,256,256,256,192', 'cnn_module_kernel': '31,31,15,15,15,31', 'decoder_dim': 512, 'joiner_dim': 512, 'causal': False, 'chunk_size': '16,32,64,-1', 'left_context_frames': '64,128,256,-1', 'use_transducer': True, 'use_ctc': True, 'full_libri': True, 'mini_libri': False, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 300, '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('zipformer/exp-ctc-rnnt/whole-lattice-rescoring'), 'suffix': 'epoch-40-avg-16-use-averaged-model'}
+2023-06-13 17:33:26,852 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-06-13 17:33:31,836 INFO [ctc_decode.py:701] Loading pre-compiled G_4_gram.pt
+2023-06-13 17:33:45,493 INFO [ctc_decode.py:718] About to create model
+2023-06-13 17:33:46,067 INFO [ctc_decode.py:785] Calculating the averaged model over epoch range from 24 (excluded) to 40
+2023-06-13 17:33:48,190 INFO [ctc_decode.py:802] Number of model parameters: 65805511
+2023-06-13 17:33:48,190 INFO [asr_datamodule.py:465] About to get test-clean cuts
+2023-06-13 17:33:48,193 INFO [asr_datamodule.py:472] About to get test-other cuts
+2023-06-13 17:33:49,550 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 21
+2023-06-13 17:34:04,381 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.1588, 2.7219, 2.7286, 3.4695], device='cuda:0')
+2023-06-13 17:34:43,425 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:43,497 INFO [utils.py:561] [test-clean-lm_scale_0.1] %WER 2.51% [1318 / 52576, 228 ins, 66 del, 1024 sub ]
+2023-06-13 17:34:43,669 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:43,693 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:43,761 INFO [utils.py:561] [test-clean-lm_scale_0.2] %WER 2.47% [1299 / 52576, 216 ins, 67 del, 1016 sub ]
+2023-06-13 17:34:44,120 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:44,143 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:44,214 INFO [utils.py:561] [test-clean-lm_scale_0.3] %WER 2.42% [1272 / 52576, 203 ins, 71 del, 998 sub ]
+2023-06-13 17:34:44,391 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:44,413 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:44,482 INFO [utils.py:561] [test-clean-lm_scale_0.4] %WER 2.39% [1254 / 52576, 192 ins, 74 del, 988 sub ]
+2023-06-13 17:34:44,653 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:44,675 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:44,744 INFO [utils.py:561] [test-clean-lm_scale_0.5] %WER 2.37% [1245 / 52576, 184 ins, 81 del, 980 sub ]
+2023-06-13 17:34:44,920 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:44,943 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:45,011 INFO [utils.py:561] [test-clean-lm_scale_0.6] %WER 2.38% [1250 / 52576, 178 ins, 87 del, 985 sub ]
+2023-06-13 17:34:45,180 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:45,202 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:45,272 INFO [utils.py:561] [test-clean-lm_scale_0.7] %WER 2.39% [1259 / 52576, 171 ins, 98 del, 990 sub ]
+2023-06-13 17:34:45,440 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:45,462 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:45,530 INFO [utils.py:561] [test-clean-lm_scale_0.8] %WER 2.43% [1277 / 52576, 164 ins, 113 del, 1000 sub ]
+2023-06-13 17:34:45,698 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:45,720 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:45,790 INFO [utils.py:561] [test-clean-lm_scale_0.9] %WER 2.49% [1307 / 52576, 155 ins, 137 del, 1015 sub ]
+2023-06-13 17:34:45,961 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:45,983 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:46,202 INFO [utils.py:561] [test-clean-lm_scale_1.0] %WER 2.57% [1350 / 52576, 151 ins, 162 del, 1037 sub ]
+2023-06-13 17:34:46,373 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:46,396 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:46,464 INFO [utils.py:561] [test-clean-lm_scale_1.1] %WER 2.72% [1432 / 52576, 151 ins, 202 del, 1079 sub ]
+2023-06-13 17:34:46,635 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:46,657 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:46,726 INFO [utils.py:561] [test-clean-lm_scale_1.2] %WER 2.84% [1493 / 52576, 148 ins, 236 del, 1109 sub ]
+2023-06-13 17:34:46,898 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:46,920 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:46,988 INFO [utils.py:561] [test-clean-lm_scale_1.3] %WER 3.03% [1594 / 52576, 147 ins, 293 del, 1154 sub ]
+2023-06-13 17:34:47,159 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:47,181 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:47,250 INFO [utils.py:561] [test-clean-lm_scale_1.4] %WER 3.22% [1695 / 52576, 144 ins, 360 del, 1191 sub ]
+2023-06-13 17:34:47,421 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:47,443 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:47,513 INFO [utils.py:561] [test-clean-lm_scale_1.5] %WER 3.44% [1810 / 52576, 136 ins, 432 del, 1242 sub ]
+2023-06-13 17:34:47,683 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:47,705 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:47,923 INFO [utils.py:561] [test-clean-lm_scale_1.6] %WER 3.69% [1939 / 52576, 136 ins, 510 del, 1293 sub ]
+2023-06-13 17:34:48,094 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:48,117 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:48,185 INFO [utils.py:561] [test-clean-lm_scale_1.7] %WER 3.93% [2065 / 52576, 133 ins, 576 del, 1356 sub ]
+2023-06-13 17:34:48,360 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:48,382 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:48,451 INFO [utils.py:561] [test-clean-lm_scale_1.8] %WER 4.11% [2163 / 52576, 131 ins, 629 del, 1403 sub ]
+2023-06-13 17:34:48,621 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:48,644 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:48,711 INFO [utils.py:561] [test-clean-lm_scale_1.9] %WER 4.35% [2285 / 52576, 129 ins, 701 del, 1455 sub ]
+2023-06-13 17:34:48,889 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:48,911 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:48,980 INFO [utils.py:561] [test-clean-lm_scale_2.0] %WER 4.58% [2409 / 52576, 130 ins, 767 del, 1512 sub ]
+2023-06-13 17:34:49,152 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-clean-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:34:49,153 INFO [ctc_decode.py:587] 
+For test-clean, WER of different settings are:
+lm_scale_0.5	2.37	best for test-clean
+lm_scale_0.6	2.38
+lm_scale_0.4	2.39
+lm_scale_0.7	2.39
+lm_scale_0.3	2.42
+lm_scale_0.8	2.43
+lm_scale_0.2	2.47
+lm_scale_0.9	2.49
+lm_scale_0.1	2.51
+lm_scale_1.0	2.57
+lm_scale_1.1	2.72
+lm_scale_1.2	2.84
+lm_scale_1.3	3.03
+lm_scale_1.4	3.22
+lm_scale_1.5	3.44
+lm_scale_1.6	3.69
+lm_scale_1.7	3.93
+lm_scale_1.8	4.11
+lm_scale_1.9	4.35
+lm_scale_2.0	4.58
+
+2023-06-13 17:34:50,631 INFO [ctc_decode.py:550] batch 0/?, cuts processed until now is 26
+2023-06-13 17:35:10,848 INFO [decode.py:893] Caught exception:
+CUDA out of memory. Tried to allocate 8.00 GiB (GPU 0; 31.75 GiB total capacity; 23.78 GiB already allocated; 181.69 MiB free; 30.20 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
+Exception raised from malloc at ../c10/cuda/CUDACachingAllocator.cpp:536 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x42 (0x7f34d9c0fd62 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x25358 (0x7f34d9e7b358 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #2: <unknown function> + 0x25d72 (0x7f34d9e7bd72 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #3: <unknown function> + 0x261a2 (0x7f34d9e7c1a2 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #4: k2::PytorchCudaContext::Allocate(unsigned long, void**) + 0x32 (0x7f34328d10d2 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #5: k2::NewRegion(std::shared_ptr<k2::Context>, unsigned long) + 0x112 (0x7f34325f6392 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #6: k2::Hash::Hash(std::shared_ptr<k2::Context>, int, int, int) + 0x2f7 (0x7f34326cb3a7 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #7: k2::Hash::Resize(int, int, int, bool) + 0x1b4 (0x7f34326c0eb4 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #8: k2::DeviceIntersector::ForwardSortedA() + 0x53e (0x7f34326f5fae in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #9: k2::IntersectDevice(k2::Ragged<k2::Arc>&, int, k2::Ragged<k2::Arc>&, int, k2::Array1<int> const&, k2::Array1<int>*, k2::Array1<int>*, bool) + 0x4cd (0x7f34326d81fd in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #10: <unknown function> + 0x75936 (0x7f3433715936 in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
+frame #11: <unknown function> + 0x377ec (0x7f34336d77ec in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
+<omitting python frames>
+frame #34: python3() [0x662c2e]
+frame #39: __libc_start_main + 0xe7 (0x7f353b2a7bf7 in /lib/x86_64-linux-gnu/libc.so.6)
+
+
+2023-06-13 17:35:10,848 INFO [decode.py:897] num_arcs before pruning: 262949
+2023-06-13 17:35:10,848 INFO [decode.py:898] This OOM is not an error. You can ignore it. If your model does not converge well, or --max-duration is too large, or the input sound file is difficult to decode, you will meet this exception.
+2023-06-13 17:35:10,857 INFO [decode.py:909] num_arcs after pruning: 45772
+2023-06-13 17:35:36,889 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([3.4643, 2.3418, 2.5677, 2.0390], device='cuda:0')
+2023-06-13 17:35:37,983 INFO [decode.py:893] Caught exception:
+CUDA out of memory. Tried to allocate 2.00 GiB (GPU 0; 31.75 GiB total capacity; 29.74 GiB already allocated; 27.69 MiB free; 30.35 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
+Exception raised from malloc at ../c10/cuda/CUDACachingAllocator.cpp:536 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x42 (0x7f34d9c0fd62 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x25358 (0x7f34d9e7b358 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #2: <unknown function> + 0x25d72 (0x7f34d9e7bd72 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #3: <unknown function> + 0x261a2 (0x7f34d9e7c1a2 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #4: k2::PytorchCudaContext::Allocate(unsigned long, void**) + 0x32 (0x7f34328d10d2 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #5: k2::Region::Extend(unsigned long) + 0x7a (0x7f34326dcaea in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #6: k2::Array1<k2::intersect_internal::StateInfo>::Resize(int, bool) + 0xb5 (0x7f34326dd435 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #7: void k2::DeviceIntersector::ForwardSortedAOneIter<k2::Hash::PackedAccessor>(int, k2::Array1<int> const&, k2::Array1<int> const&, k2::Array1<int> const&, k2::Array1<int> const&, int) + 0x9da (0x7f34326f4d8a in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #8: k2::DeviceIntersector::ForwardSortedA() + 0xadb (0x7f34326f654b in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #9: k2::IntersectDevice(k2::Ragged<k2::Arc>&, int, k2::Ragged<k2::Arc>&, int, k2::Array1<int> const&, k2::Array1<int>*, k2::Array1<int>*, bool) + 0x4cd (0x7f34326d81fd in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #10: <unknown function> + 0x75936 (0x7f3433715936 in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
+frame #11: <unknown function> + 0x377ec (0x7f34336d77ec in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
+<omitting python frames>
+frame #34: python3() [0x662c2e]
+frame #39: __libc_start_main + 0xe7 (0x7f353b2a7bf7 in /lib/x86_64-linux-gnu/libc.so.6)
+
+
+2023-06-13 17:35:37,983 INFO [decode.py:897] num_arcs before pruning: 305119
+2023-06-13 17:35:37,983 INFO [decode.py:898] This OOM is not an error. You can ignore it. If your model does not converge well, or --max-duration is too large, or the input sound file is difficult to decode, you will meet this exception.
+2023-06-13 17:35:38,002 INFO [decode.py:909] num_arcs after pruning: 45058
+2023-06-13 17:35:49,399 INFO [zipformer.py:1711] name=None, attn_weights_entropy = tensor([4.0697, 2.1705, 3.1589, 1.7420], device='cuda:0')
+2023-06-13 17:35:51,043 INFO [decode.py:893] Caught exception:
+CUDA out of memory. Tried to allocate 2.00 GiB (GPU 0; 31.75 GiB total capacity; 28.59 GiB already allocated; 813.69 MiB free; 29.58 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF
+Exception raised from malloc at ../c10/cuda/CUDACachingAllocator.cpp:536 (most recent call first):
+frame #0: c10::Error::Error(c10::SourceLocation, std::string) + 0x42 (0x7f34d9c0fd62 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10.so)
+frame #1: <unknown function> + 0x25358 (0x7f34d9e7b358 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #2: <unknown function> + 0x25d72 (0x7f34d9e7bd72 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #3: <unknown function> + 0x261a2 (0x7f34d9e7c1a2 in /ceph-jb/yaozengwei/env/k2_icefall/lib/python3.8/site-packages/torch/lib/libc10_cuda.so)
+frame #4: k2::PytorchCudaContext::Allocate(unsigned long, void**) + 0x32 (0x7f34328d10d2 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #5: k2::Region::Extend(unsigned long) + 0x7a (0x7f34326dcaea in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #6: k2::Array1<k2::intersect_internal::StateInfo>::Resize(int, bool) + 0xb5 (0x7f34326dd435 in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #7: void k2::DeviceIntersector::ForwardSortedAOneIter<k2::Hash::PackedAccessor>(int, k2::Array1<int> const&, k2::Array1<int> const&, k2::Array1<int> const&, k2::Array1<int> const&, int) + 0x9da (0x7f34326f4d8a in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #8: k2::DeviceIntersector::ForwardSortedA() + 0xadb (0x7f34326f654b in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #9: k2::IntersectDevice(k2::Ragged<k2::Arc>&, int, k2::Ragged<k2::Arc>&, int, k2::Array1<int> const&, k2::Array1<int>*, k2::Array1<int>*, bool) + 0x4cd (0x7f34326d81fd in /ceph-zw/workspace/k2/k2/build_release/lib/libk2context.so)
+frame #10: <unknown function> + 0x75936 (0x7f3433715936 in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
+frame #11: <unknown function> + 0x377ec (0x7f34336d77ec in /ceph-zw/workspace/k2/k2/build_release/lib/_k2.cpython-38-x86_64-linux-gnu.so)
+<omitting python frames>
+frame #34: python3() [0x662c2e]
+frame #39: __libc_start_main + 0xe7 (0x7f353b2a7bf7 in /lib/x86_64-linux-gnu/libc.so.6)
+
+
+2023-06-13 17:35:51,043 INFO [decode.py:897] num_arcs before pruning: 347421
+2023-06-13 17:35:51,043 INFO [decode.py:898] This OOM is not an error. You can ignore it. If your model does not converge well, or --max-duration is too large, or the input sound file is difficult to decode, you will meet this exception.
+2023-06-13 17:35:51,064 INFO [decode.py:909] num_arcs after pruning: 22726
+2023-06-13 17:35:52,877 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:52,951 INFO [utils.py:561] [test-other-lm_scale_0.1] %WER 5.29% [2768 / 52343, 385 ins, 176 del, 2207 sub ]
+2023-06-13 17:35:53,126 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:53,150 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:53,226 INFO [utils.py:561] [test-other-lm_scale_0.2] %WER 5.13% [2686 / 52343, 365 ins, 176 del, 2145 sub ]
+2023-06-13 17:35:53,401 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:53,424 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:53,493 INFO [utils.py:561] [test-other-lm_scale_0.3] %WER 5.04% [2639 / 52343, 346 ins, 185 del, 2108 sub ]
+2023-06-13 17:35:53,666 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:53,688 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:53,758 INFO [utils.py:561] [test-other-lm_scale_0.4] %WER 4.98% [2608 / 52343, 325 ins, 196 del, 2087 sub ]
+2023-06-13 17:35:53,933 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:53,955 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:54,025 INFO [utils.py:561] [test-other-lm_scale_0.5] %WER 4.93% [2581 / 52343, 315 ins, 212 del, 2054 sub ]
+2023-06-13 17:35:54,198 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:54,221 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:54,291 INFO [utils.py:561] [test-other-lm_scale_0.6] %WER 4.91% [2568 / 52343, 290 ins, 231 del, 2047 sub ]
+2023-06-13 17:35:54,464 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:54,487 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:54,557 INFO [utils.py:561] [test-other-lm_scale_0.7] %WER 4.88% [2556 / 52343, 272 ins, 256 del, 2028 sub ]
+2023-06-13 17:35:54,728 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:54,751 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:54,821 INFO [utils.py:561] [test-other-lm_scale_0.8] %WER 4.97% [2600 / 52343, 259 ins, 303 del, 2038 sub ]
+2023-06-13 17:35:54,995 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:55,017 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:55,086 INFO [utils.py:561] [test-other-lm_scale_0.9] %WER 5.07% [2652 / 52343, 251 ins, 347 del, 2054 sub ]
+2023-06-13 17:35:55,260 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:55,283 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:55,353 INFO [utils.py:561] [test-other-lm_scale_1.0] %WER 5.16% [2703 / 52343, 239 ins, 400 del, 2064 sub ]
+2023-06-13 17:35:55,679 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:55,702 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:55,772 INFO [utils.py:561] [test-other-lm_scale_1.1] %WER 5.36% [2805 / 52343, 230 ins, 461 del, 2114 sub ]
+2023-06-13 17:35:55,945 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:55,968 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:56,039 INFO [utils.py:561] [test-other-lm_scale_1.2] %WER 5.60% [2933 / 52343, 220 ins, 570 del, 2143 sub ]
+2023-06-13 17:35:56,214 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:56,237 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:56,313 INFO [utils.py:561] [test-other-lm_scale_1.3] %WER 5.88% [3079 / 52343, 214 ins, 691 del, 2174 sub ]
+2023-06-13 17:35:56,489 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:56,512 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:56,584 INFO [utils.py:561] [test-other-lm_scale_1.4] %WER 6.16% [3226 / 52343, 214 ins, 780 del, 2232 sub ]
+2023-06-13 17:35:56,759 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:56,782 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:56,854 INFO [utils.py:561] [test-other-lm_scale_1.5] %WER 6.52% [3411 / 52343, 215 ins, 898 del, 2298 sub ]
+2023-06-13 17:35:57,029 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:57,052 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:57,120 INFO [utils.py:561] [test-other-lm_scale_1.6] %WER 6.80% [3559 / 52343, 206 ins, 1009 del, 2344 sub ]
+2023-06-13 17:35:57,294 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:57,317 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:57,386 INFO [utils.py:561] [test-other-lm_scale_1.7] %WER 7.13% [3731 / 52343, 204 ins, 1139 del, 2388 sub ]
+2023-06-13 17:35:57,712 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:57,734 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:57,805 INFO [utils.py:561] [test-other-lm_scale_1.8] %WER 7.42% [3886 / 52343, 198 ins, 1248 del, 2440 sub ]
+2023-06-13 17:35:57,980 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:58,003 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:58,072 INFO [utils.py:561] [test-other-lm_scale_1.9] %WER 7.75% [4059 / 52343, 194 ins, 1359 del, 2506 sub ]
+2023-06-13 17:35:58,245 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:58,268 INFO [ctc_decode.py:564] The transcripts are stored in zipformer/exp-ctc-rnnt/whole-lattice-rescoring/recogs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:58,338 INFO [utils.py:561] [test-other-lm_scale_2.0] %WER 8.04% [4206 / 52343, 190 ins, 1459 del, 2557 sub ]
+2023-06-13 17:35:58,511 INFO [ctc_decode.py:573] Wrote detailed error stats to zipformer/exp-ctc-rnnt/whole-lattice-rescoring/errs-test-other-epoch-40-avg-16-use-averaged-model.txt
+2023-06-13 17:35:58,512 INFO [ctc_decode.py:587] 
+For test-other, WER of different settings are:
+lm_scale_0.7	4.88	best for test-other
+lm_scale_0.6	4.91
+lm_scale_0.5	4.93
+lm_scale_0.8	4.97
+lm_scale_0.4	4.98
+lm_scale_0.3	5.04
+lm_scale_0.9	5.07
+lm_scale_0.2	5.13
+lm_scale_1.0	5.16
+lm_scale_0.1	5.29
+lm_scale_1.1	5.36
+lm_scale_1.2	5.6
+lm_scale_1.3	5.88
+lm_scale_1.4	6.16
+lm_scale_1.5	6.52
+lm_scale_1.6	6.8
+lm_scale_1.7	7.13
+lm_scale_1.8	7.42
+lm_scale_1.9	7.75
+lm_scale_2.0	8.04
+
+2023-06-13 17:35:58,512 INFO [ctc_decode.py:835] Done!

decoding_results/whole-lattice-rescoring/wer-summary-test-clean-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,21 @@
+settings	WER
+lm_scale_0.5	2.37
+lm_scale_0.6	2.38
+lm_scale_0.4	2.39
+lm_scale_0.7	2.39
+lm_scale_0.3	2.42
+lm_scale_0.8	2.43
+lm_scale_0.2	2.47
+lm_scale_0.9	2.49
+lm_scale_0.1	2.51
+lm_scale_1.0	2.57
+lm_scale_1.1	2.72
+lm_scale_1.2	2.84
+lm_scale_1.3	3.03
+lm_scale_1.4	3.22
+lm_scale_1.5	3.44
+lm_scale_1.6	3.69
+lm_scale_1.7	3.93
+lm_scale_1.8	4.11
+lm_scale_1.9	4.35
+lm_scale_2.0	4.58

decoding_results/whole-lattice-rescoring/wer-summary-test-other-epoch-40-avg-16-use-averaged-model.txt

@@ -0,0 +1,21 @@
+settings	WER
+lm_scale_0.7	4.88
+lm_scale_0.6	4.91
+lm_scale_0.5	4.93
+lm_scale_0.8	4.97
+lm_scale_0.4	4.98
+lm_scale_0.3	5.04
+lm_scale_0.9	5.07
+lm_scale_0.2	5.13
+lm_scale_1.0	5.16
+lm_scale_0.1	5.29
+lm_scale_1.1	5.36
+lm_scale_1.2	5.6
+lm_scale_1.3	5.88
+lm_scale_1.4	6.16
+lm_scale_1.5	6.52
+lm_scale_1.6	6.8
+lm_scale_1.7	7.13
+lm_scale_1.8	7.42
+lm_scale_1.9	7.75
+lm_scale_2.0	8.04

exp/decode.sh

@@ -0,0 +1,16 @@
+export CUDA_VISIBLE_DEVICES="0"
+for m in ctc-decoding 1best nbest nbest-rescoring whole-lattice-rescoring; do
+  ./zipformer/ctc_decode.py \
+      --epoch 40 \
+      --avg 16 \
+      --exp-dir zipformer/exp-ctc-rnnt \
+      --use-transducer 1 \
+      --use-ctc 1 \
+      --max-duration 300 \
+      --causal 0 \
+      --num-paths 100 \
+      --nbest-scale 1.0 \
+      --hlg-scale 0.6 \
+      --decoding-method $m
+done
+

exp/epoch-40.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fee348148cdb5910ad116ad0fcfa6469f5976e0b1552d9b1899a3b3093c18d39
+size 1053871436

exp/export.sh

@@ -0,0 +1,10 @@
+./zipformer/export.py \
+  --exp-dir ./zipformer/exp-ctc-rnnt \
+  --use-transducer 1 \
+  --use-ctc 1 \
+  --bpe-model data/lang_bpe_500/bpe.model \
+  --epoch 40 \
+  --avg 16 \
+  --jit 1
+
+

exp/jit_script.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a877a5d18386311fe069dec68bb7e0ca956129b90f8a4ce4e956dad1ac98e154
+size 265955146