initial commit for zipformer_ctc

data/lang_bpe_500/HLG.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66d43dc28206a4ef33fc8ed5272abc6dc4b064c78cafc67f9f8a5b88f94be0fc
+size 695095775

data/lang_bpe_500/L.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:29b6b298ec932414dedc38e1971893c6d0c710fc0766dc0c495e387702c2e71f
+size 19025793

data/lang_bpe_500/L_disambig.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33745fbef2e2957f39fbb8a563b3735c7204f8398bbed3b59eb43edb1dfce388
+size 20013121

data/lang_bpe_500/Linv.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d76053824dca246b4ac37882f3e96bc8a8ed76038c0b100bba1f57d00246cee9
+size 19025793

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

exp/cpu_jit.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4931cd1eff010cb3d8757b0dd2be404333429f2b7bbbe11c637f9d87de10553c
+size 421440904

exp/pretrained.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9efff0d24646202e3b5506fa3bb0d32e90076d84c28bfa50af294ee96015d2d
+size 344661407

exp/tensorboard/events.out.tfevents.1678202076.r8n04.343595.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fef35df8e5d5f0f32043396ba6ee06e513e9354545ea95d64ea2cf3a4d2ebfdf
+size 1780128

log/attention_decoder/log-decode-2023-03-10-09-58-58

@@ -0,0 +1,11 @@
+2023-03-10 09:58:58,924 INFO [decode.py:641] Decoding started
+2023-03-10 09:58:58,925 INFO [decode.py:642] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_ctc', 'icefall-git-sha1': '11e21f3-dirty', 'icefall-git-date': 'Thu Mar 9 19:58:30 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 99, 'iter': 0, 'avg': 1, 'use_averaged_model': False, 'method': 'attention-decoder', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'num_decoder_layers': 6, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-10 09:58:59,170 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-10 09:58:59,289 INFO [decode.py:653] device: cuda:0
+2023-03-10 09:59:04,720 INFO [decode.py:720] Loading pre-compiled G_4_gram.pt
+2023-03-10 09:59:08,342 INFO [decode.py:741] About to create model
+2023-03-10 09:59:08,810 INFO [zipformer.py:178] 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.
+2023-03-10 09:59:08,866 INFO [checkpoint.py:112] Loading checkpoint from zipformer_ctc/exp/v0/epoch-99.pt
+2023-03-10 09:59:09,414 INFO [decode.py:824] Number of model parameters: 86083707
+2023-03-10 09:59:09,415 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2023-03-10 09:59:09,504 INFO [asr_datamodule.py:450] About to get test-other cuts

log/attention_decoder/log-decode-2023-03-10-10-10-13

@@ -0,0 +1,11 @@
+2023-03-10 10:10:13,940 INFO [decode.py:641] Decoding started
+2023-03-10 10:10:13,940 INFO [decode.py:642] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_ctc', 'icefall-git-sha1': '11e21f3-dirty', 'icefall-git-date': 'Thu Mar 9 19:58:30 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 99, 'iter': 0, 'avg': 1, 'use_averaged_model': False, 'method': 'attention-decoder', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'num_decoder_layers': 6, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-10 10:10:14,177 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-10 10:10:14,284 INFO [decode.py:653] device: cuda:0
+2023-03-10 10:10:19,750 INFO [decode.py:720] Loading pre-compiled G_4_gram.pt
+2023-03-10 10:10:20,968 INFO [decode.py:741] About to create model
+2023-03-10 10:10:21,454 INFO [zipformer.py:178] 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.
+2023-03-10 10:10:21,526 INFO [checkpoint.py:112] Loading checkpoint from zipformer_ctc/exp/v0/epoch-99.pt
+2023-03-10 10:10:22,138 INFO [decode.py:824] Number of model parameters: 86083707
+2023-03-10 10:10:22,138 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2023-03-10 10:10:22,227 INFO [asr_datamodule.py:450] About to get test-other cuts

log/attention_decoder/log-decode-2023-03-10-10-20-00

@@ -0,0 +1,11 @@
+2023-03-10 10:20:00,301 INFO [decode.py:641] Decoding started
+2023-03-10 10:20:00,302 INFO [decode.py:642] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_ctc', 'icefall-git-sha1': '11e21f3-dirty', 'icefall-git-date': 'Thu Mar 9 19:58:30 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 99, 'iter': 0, 'avg': 1, 'use_averaged_model': False, 'method': 'attention-decoder', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'num_decoder_layers': 6, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-10 10:20:00,577 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-10 10:20:00,713 INFO [decode.py:653] device: cuda:0
+2023-03-10 10:20:07,246 INFO [decode.py:720] Loading pre-compiled G_4_gram.pt
+2023-03-10 10:20:08,507 INFO [decode.py:741] About to create model
+2023-03-10 10:20:09,069 INFO [zipformer.py:178] 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.
+2023-03-10 10:20:09,139 INFO [checkpoint.py:112] Loading checkpoint from zipformer_ctc/exp/v0/epoch-99.pt
+2023-03-10 10:20:09,768 INFO [decode.py:824] Number of model parameters: 86083707
+2023-03-10 10:20:09,768 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2023-03-10 10:20:09,900 INFO [asr_datamodule.py:450] About to get test-other cuts

log/attention_decoder/log-decode-2023-03-10-10-56-28

@@ -0,0 +1,11 @@
+2023-03-10 10:56:28,210 INFO [decode.py:642] Decoding started
+2023-03-10 10:56:28,211 INFO [decode.py:643] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_ctc', 'icefall-git-sha1': '11e21f3-dirty', 'icefall-git-date': 'Thu Mar 9 19:58:30 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r8n03', 'IP address': '10.1.8.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 99, 'iter': 0, 'avg': 1, 'use_averaged_model': False, 'method': 'attention-decoder', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'exp/rnnlm', 'rnn_lm_epoch': 99, 'rnn_lm_avg': 1, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'num_decoder_layers': 6, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-10 10:56:28,460 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-10 10:56:28,572 INFO [decode.py:654] device: cuda:0
+2023-03-10 10:56:33,999 INFO [decode.py:721] Loading pre-compiled G_4_gram.pt
+2023-03-10 10:56:35,323 INFO [decode.py:742] About to create model
+2023-03-10 10:56:35,787 INFO [zipformer.py:178] 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.
+2023-03-10 10:56:35,847 INFO [checkpoint.py:112] Loading checkpoint from zipformer_ctc/exp/v0/epoch-99.pt
+2023-03-10 10:56:36,454 INFO [decode.py:825] Number of model parameters: 86083707
+2023-03-10 10:56:36,454 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2023-03-10 10:56:36,531 INFO [asr_datamodule.py:450] About to get test-other cuts

log/attention_decoder/log-decode-2023-03-10-11-58-59

@@ -0,0 +1,11 @@
+2023-03-10 11:58:59,546 INFO [decode.py:642] Decoding started
+2023-03-10 11:58:59,546 INFO [decode.py:643] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_ctc', 'icefall-git-sha1': '11e21f3-dirty', 'icefall-git-date': 'Thu Mar 9 19:58:30 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n04', 'IP address': '10.1.7.4'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 99, 'iter': 0, 'avg': 1, 'use_averaged_model': False, 'method': 'attention-decoder', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'exp/rnnlm', 'rnn_lm_epoch': 99, 'rnn_lm_avg': 1, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'num_decoder_layers': 6, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-10 11:58:59,841 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-10 11:58:59,957 INFO [decode.py:654] device: cuda:0
+2023-03-10 11:59:07,426 INFO [decode.py:721] Loading pre-compiled G_4_gram.pt
+2023-03-10 11:59:12,292 INFO [decode.py:742] About to create model
+2023-03-10 11:59:12,758 INFO [zipformer.py:178] 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.
+2023-03-10 11:59:12,820 INFO [checkpoint.py:112] Loading checkpoint from zipformer_ctc/exp/v0/epoch-99.pt
+2023-03-10 11:59:14,959 INFO [decode.py:825] Number of model parameters: 86083707
+2023-03-10 11:59:14,960 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2023-03-10 11:59:15,047 INFO [asr_datamodule.py:450] About to get test-other cuts

log/attention_decoder/log-decode-2023-03-10-16-18-14

@@ -0,0 +1,1184 @@
+2023-03-10 16:18:14,062 INFO [decode.py:642] Decoding started
+2023-03-10 16:18:14,062 INFO [decode.py:643] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_ctc', 'icefall-git-sha1': '11e21f3-dirty', 'icefall-git-date': 'Thu Mar 9 19:58:30 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 99, 'iter': 0, 'avg': 1, 'use_averaged_model': False, 'method': 'attention-decoder', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'exp/rnnlm', 'rnn_lm_epoch': 99, 'rnn_lm_avg': 1, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'num_decoder_layers': 6, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-10 16:18:14,321 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-10 16:18:14,432 INFO [decode.py:654] device: cuda:0
+2023-03-10 16:18:20,193 INFO [decode.py:721] Loading pre-compiled G_4_gram.pt
+2023-03-10 16:18:21,447 INFO [decode.py:742] About to create model
+2023-03-10 16:18:21,947 INFO [zipformer.py:178] 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.
+2023-03-10 16:18:22,017 INFO [checkpoint.py:112] Loading checkpoint from zipformer_ctc/exp/v0/epoch-99.pt
+2023-03-10 16:18:22,663 INFO [decode.py:825] Number of model parameters: 86083707
+2023-03-10 16:18:22,663 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2023-03-10 16:18:22,787 INFO [asr_datamodule.py:450] About to get test-other cuts
+2023-03-10 16:18:27,802 INFO [decode.py:580] batch 0/?, cuts processed until now is 14
+2023-03-10 16:23:39,552 INFO [decode.py:580] batch 100/?, cuts processed until now is 2293
+2023-03-10 16:26:52,008 INFO [decode.py:626] 
+For test-clean, WER of different settings are:
+ngram_lm_scale_1.2_attention_scale_4.0	2.35	best for test-clean
+ngram_lm_scale_1.3_attention_scale_5.0	2.35
+ngram_lm_scale_1.5_attention_scale_5.0	2.35
+ngram_lm_scale_0.9_attention_scale_4.0	2.36
+ngram_lm_scale_0.9_attention_scale_5.0	2.36
+ngram_lm_scale_1.0_attention_scale_5.0	2.36
+ngram_lm_scale_1.1_attention_scale_4.0	2.36
+ngram_lm_scale_1.1_attention_scale_5.0	2.36
+ngram_lm_scale_1.2_attention_scale_5.0	2.36
+ngram_lm_scale_1.3_attention_scale_4.0	2.36
+ngram_lm_scale_0.6_attention_scale_1.7	2.37
+ngram_lm_scale_0.6_attention_scale_3.0	2.37
+ngram_lm_scale_0.7_attention_scale_1.9	2.37
+ngram_lm_scale_0.7_attention_scale_2.0	2.37
+ngram_lm_scale_0.7_attention_scale_2.1	2.37
+ngram_lm_scale_0.7_attention_scale_2.2	2.37
+ngram_lm_scale_0.7_attention_scale_2.3	2.37
+ngram_lm_scale_0.7_attention_scale_3.0	2.37
+ngram_lm_scale_0.9_attention_scale_3.0	2.37
+ngram_lm_scale_1.0_attention_scale_3.0	2.37
+ngram_lm_scale_1.0_attention_scale_4.0	2.37
+ngram_lm_scale_1.1_attention_scale_3.0	2.37
+ngram_lm_scale_1.7_attention_scale_4.0	2.37
+ngram_lm_scale_1.7_attention_scale_5.0	2.37
+ngram_lm_scale_0.08_attention_scale_4.0	2.38
+ngram_lm_scale_0.1_attention_scale_4.0	2.38
+ngram_lm_scale_0.3_attention_scale_4.0	2.38
+ngram_lm_scale_0.5_attention_scale_3.0	2.38
+ngram_lm_scale_0.5_attention_scale_5.0	2.38
+ngram_lm_scale_0.6_attention_scale_1.9	2.38
+ngram_lm_scale_0.6_attention_scale_2.0	2.38
+ngram_lm_scale_0.6_attention_scale_2.1	2.38
+ngram_lm_scale_0.6_attention_scale_2.2	2.38
+ngram_lm_scale_0.6_attention_scale_2.3	2.38
+ngram_lm_scale_0.6_attention_scale_2.5	2.38
+ngram_lm_scale_0.6_attention_scale_4.0	2.38
+ngram_lm_scale_0.6_attention_scale_5.0	2.38
+ngram_lm_scale_0.7_attention_scale_2.5	2.38
+ngram_lm_scale_0.7_attention_scale_4.0	2.38
+ngram_lm_scale_0.7_attention_scale_5.0	2.38
+ngram_lm_scale_0.9_attention_scale_2.2	2.38
+ngram_lm_scale_0.9_attention_scale_2.5	2.38
+ngram_lm_scale_1.1_attention_scale_1.9	2.38
+ngram_lm_scale_1.1_attention_scale_2.0	2.38
+ngram_lm_scale_1.1_attention_scale_2.1	2.38
+ngram_lm_scale_1.1_attention_scale_2.2	2.38
+ngram_lm_scale_1.1_attention_scale_2.3	2.38
+ngram_lm_scale_1.1_attention_scale_2.5	2.38
+ngram_lm_scale_1.2_attention_scale_2.5	2.38
+ngram_lm_scale_1.2_attention_scale_3.0	2.38
+ngram_lm_scale_1.3_attention_scale_3.0	2.38
+ngram_lm_scale_1.5_attention_scale_4.0	2.38
+ngram_lm_scale_1.9_attention_scale_5.0	2.38
+ngram_lm_scale_2.0_attention_scale_5.0	2.38
+ngram_lm_scale_0.01_attention_scale_0.7	2.39
+ngram_lm_scale_0.01_attention_scale_1.0	2.39
+ngram_lm_scale_0.01_attention_scale_1.2	2.39
+ngram_lm_scale_0.01_attention_scale_1.3	2.39
+ngram_lm_scale_0.01_attention_scale_1.5	2.39
+ngram_lm_scale_0.01_attention_scale_4.0	2.39
+ngram_lm_scale_0.01_attention_scale_5.0	2.39
+ngram_lm_scale_0.05_attention_scale_0.5	2.39
+ngram_lm_scale_0.05_attention_scale_0.6	2.39
+ngram_lm_scale_0.05_attention_scale_0.7	2.39
+ngram_lm_scale_0.05_attention_scale_0.9	2.39
+ngram_lm_scale_0.05_attention_scale_1.0	2.39
+ngram_lm_scale_0.05_attention_scale_1.1	2.39
+ngram_lm_scale_0.05_attention_scale_1.2	2.39
+ngram_lm_scale_0.05_attention_scale_1.3	2.39
+ngram_lm_scale_0.05_attention_scale_1.5	2.39
+ngram_lm_scale_0.05_attention_scale_1.9	2.39
+ngram_lm_scale_0.05_attention_scale_3.0	2.39
+ngram_lm_scale_0.05_attention_scale_4.0	2.39
+ngram_lm_scale_0.05_attention_scale_5.0	2.39
+ngram_lm_scale_0.08_attention_scale_0.5	2.39
+ngram_lm_scale_0.08_attention_scale_0.6	2.39
+ngram_lm_scale_0.08_attention_scale_0.7	2.39
+ngram_lm_scale_0.08_attention_scale_0.9	2.39
+ngram_lm_scale_0.08_attention_scale_1.0	2.39
+ngram_lm_scale_0.08_attention_scale_1.1	2.39
+ngram_lm_scale_0.08_attention_scale_1.2	2.39
+ngram_lm_scale_0.08_attention_scale_1.3	2.39
+ngram_lm_scale_0.08_attention_scale_1.7	2.39
+ngram_lm_scale_0.08_attention_scale_1.9	2.39
+ngram_lm_scale_0.08_attention_scale_2.1	2.39
+ngram_lm_scale_0.08_attention_scale_2.2	2.39
+ngram_lm_scale_0.08_attention_scale_2.3	2.39
+ngram_lm_scale_0.08_attention_scale_3.0	2.39
+ngram_lm_scale_0.08_attention_scale_5.0	2.39
+ngram_lm_scale_0.1_attention_scale_0.5	2.39
+ngram_lm_scale_0.1_attention_scale_0.6	2.39
+ngram_lm_scale_0.1_attention_scale_1.0	2.39
+ngram_lm_scale_0.1_attention_scale_1.7	2.39
+ngram_lm_scale_0.1_attention_scale_1.9	2.39
+ngram_lm_scale_0.1_attention_scale_3.0	2.39
+ngram_lm_scale_0.1_attention_scale_5.0	2.39
+ngram_lm_scale_0.3_attention_scale_1.2	2.39
+ngram_lm_scale_0.3_attention_scale_1.3	2.39
+ngram_lm_scale_0.3_attention_scale_2.1	2.39
+ngram_lm_scale_0.3_attention_scale_2.2	2.39
+ngram_lm_scale_0.3_attention_scale_3.0	2.39
+ngram_lm_scale_0.3_attention_scale_5.0	2.39
+ngram_lm_scale_0.5_attention_scale_1.3	2.39
+ngram_lm_scale_0.5_attention_scale_1.7	2.39
+ngram_lm_scale_0.5_attention_scale_1.9	2.39
+ngram_lm_scale_0.5_attention_scale_2.0	2.39
+ngram_lm_scale_0.5_attention_scale_2.1	2.39
+ngram_lm_scale_0.5_attention_scale_2.2	2.39
+ngram_lm_scale_0.5_attention_scale_2.3	2.39
+ngram_lm_scale_0.5_attention_scale_2.5	2.39
+ngram_lm_scale_0.5_attention_scale_4.0	2.39
+ngram_lm_scale_0.6_attention_scale_1.3	2.39
+ngram_lm_scale_0.6_attention_scale_1.5	2.39
+ngram_lm_scale_0.7_attention_scale_1.3	2.39
+ngram_lm_scale_0.7_attention_scale_1.5	2.39
+ngram_lm_scale_0.7_attention_scale_1.7	2.39
+ngram_lm_scale_0.9_attention_scale_1.7	2.39
+ngram_lm_scale_0.9_attention_scale_2.0	2.39
+ngram_lm_scale_0.9_attention_scale_2.1	2.39
+ngram_lm_scale_0.9_attention_scale_2.3	2.39
+ngram_lm_scale_1.0_attention_scale_1.9	2.39
+ngram_lm_scale_1.0_attention_scale_2.0	2.39
+ngram_lm_scale_1.0_attention_scale_2.1	2.39
+ngram_lm_scale_1.0_attention_scale_2.2	2.39
+ngram_lm_scale_1.0_attention_scale_2.3	2.39
+ngram_lm_scale_1.0_attention_scale_2.5	2.39
+ngram_lm_scale_1.1_attention_scale_1.7	2.39
+ngram_lm_scale_1.2_attention_scale_1.9	2.39
+ngram_lm_scale_1.2_attention_scale_2.0	2.39
+ngram_lm_scale_1.2_attention_scale_2.1	2.39
+ngram_lm_scale_1.2_attention_scale_2.2	2.39
+ngram_lm_scale_1.2_attention_scale_2.3	2.39
+ngram_lm_scale_1.3_attention_scale_2.3	2.39
+ngram_lm_scale_1.3_attention_scale_2.5	2.39
+ngram_lm_scale_1.5_attention_scale_3.0	2.39
+ngram_lm_scale_1.9_attention_scale_4.0	2.39
+ngram_lm_scale_2.1_attention_scale_5.0	2.39
+ngram_lm_scale_2.2_attention_scale_5.0	2.39
+ngram_lm_scale_2.3_attention_scale_5.0	2.39
+ngram_lm_scale_0.01_attention_scale_0.5	2.4
+ngram_lm_scale_0.01_attention_scale_0.6	2.4
+ngram_lm_scale_0.01_attention_scale_0.9	2.4
+ngram_lm_scale_0.01_attention_scale_1.1	2.4
+ngram_lm_scale_0.01_attention_scale_1.7	2.4
+ngram_lm_scale_0.01_attention_scale_1.9	2.4
+ngram_lm_scale_0.01_attention_scale_2.0	2.4
+ngram_lm_scale_0.01_attention_scale_2.1	2.4
+ngram_lm_scale_0.01_attention_scale_2.2	2.4
+ngram_lm_scale_0.01_attention_scale_2.3	2.4
+ngram_lm_scale_0.01_attention_scale_2.5	2.4
+ngram_lm_scale_0.01_attention_scale_3.0	2.4
+ngram_lm_scale_0.05_attention_scale_1.7	2.4
+ngram_lm_scale_0.05_attention_scale_2.0	2.4
+ngram_lm_scale_0.05_attention_scale_2.1	2.4
+ngram_lm_scale_0.05_attention_scale_2.2	2.4
+ngram_lm_scale_0.05_attention_scale_2.3	2.4
+ngram_lm_scale_0.05_attention_scale_2.5	2.4
+ngram_lm_scale_0.08_attention_scale_1.5	2.4
+ngram_lm_scale_0.08_attention_scale_2.0	2.4
+ngram_lm_scale_0.08_attention_scale_2.5	2.4
+ngram_lm_scale_0.1_attention_scale_0.7	2.4
+ngram_lm_scale_0.1_attention_scale_0.9	2.4
+ngram_lm_scale_0.1_attention_scale_1.1	2.4
+ngram_lm_scale_0.1_attention_scale_1.2	2.4
+ngram_lm_scale_0.1_attention_scale_1.3	2.4
+ngram_lm_scale_0.1_attention_scale_1.5	2.4
+ngram_lm_scale_0.1_attention_scale_2.0	2.4
+ngram_lm_scale_0.1_attention_scale_2.1	2.4
+ngram_lm_scale_0.1_attention_scale_2.2	2.4
+ngram_lm_scale_0.1_attention_scale_2.3	2.4
+ngram_lm_scale_0.1_attention_scale_2.5	2.4
+ngram_lm_scale_0.3_attention_scale_0.3	2.4
+ngram_lm_scale_0.3_attention_scale_0.5	2.4
+ngram_lm_scale_0.3_attention_scale_0.6	2.4
+ngram_lm_scale_0.3_attention_scale_0.7	2.4
+ngram_lm_scale_0.3_attention_scale_1.1	2.4
+ngram_lm_scale_0.3_attention_scale_1.5	2.4
+ngram_lm_scale_0.3_attention_scale_1.7	2.4
+ngram_lm_scale_0.3_attention_scale_1.9	2.4
+ngram_lm_scale_0.3_attention_scale_2.0	2.4
+ngram_lm_scale_0.3_attention_scale_2.3	2.4
+ngram_lm_scale_0.3_attention_scale_2.5	2.4
+ngram_lm_scale_0.5_attention_scale_0.9	2.4
+ngram_lm_scale_0.5_attention_scale_1.0	2.4
+ngram_lm_scale_0.5_attention_scale_1.1	2.4
+ngram_lm_scale_0.5_attention_scale_1.2	2.4
+ngram_lm_scale_0.5_attention_scale_1.5	2.4
+ngram_lm_scale_0.6_attention_scale_1.0	2.4
+ngram_lm_scale_0.6_attention_scale_1.1	2.4
+ngram_lm_scale_0.6_attention_scale_1.2	2.4
+ngram_lm_scale_0.7_attention_scale_1.2	2.4
+ngram_lm_scale_0.9_attention_scale_1.5	2.4
+ngram_lm_scale_0.9_attention_scale_1.9	2.4
+ngram_lm_scale_1.0_attention_scale_1.5	2.4
+ngram_lm_scale_1.0_attention_scale_1.7	2.4
+ngram_lm_scale_1.3_attention_scale_2.2	2.4
+ngram_lm_scale_2.0_attention_scale_4.0	2.4
+ngram_lm_scale_0.05_attention_scale_0.3	2.41
+ngram_lm_scale_0.08_attention_scale_0.3	2.41
+ngram_lm_scale_0.3_attention_scale_1.0	2.41
+ngram_lm_scale_0.5_attention_scale_0.6	2.41
+ngram_lm_scale_0.5_attention_scale_0.7	2.41
+ngram_lm_scale_0.6_attention_scale_0.9	2.41
+ngram_lm_scale_0.7_attention_scale_1.0	2.41
+ngram_lm_scale_0.7_attention_scale_1.1	2.41
+ngram_lm_scale_0.9_attention_scale_1.2	2.41
+ngram_lm_scale_1.3_attention_scale_2.0	2.41
+ngram_lm_scale_1.3_attention_scale_2.1	2.41
+ngram_lm_scale_1.5_attention_scale_2.5	2.41
+ngram_lm_scale_1.7_attention_scale_3.0	2.41
+ngram_lm_scale_2.5_attention_scale_5.0	2.41
+ngram_lm_scale_0.01_attention_scale_0.3	2.42
+ngram_lm_scale_0.1_attention_scale_0.3	2.42
+ngram_lm_scale_0.3_attention_scale_0.9	2.42
+ngram_lm_scale_0.6_attention_scale_0.6	2.42
+ngram_lm_scale_0.6_attention_scale_0.7	2.42
+ngram_lm_scale_0.7_attention_scale_0.9	2.42
+ngram_lm_scale_0.9_attention_scale_1.3	2.42
+ngram_lm_scale_1.0_attention_scale_1.3	2.42
+ngram_lm_scale_2.1_attention_scale_4.0	2.42
+ngram_lm_scale_0.05_attention_scale_0.01	2.43
+ngram_lm_scale_0.05_attention_scale_0.1	2.43
+ngram_lm_scale_0.08_attention_scale_0.05	2.43
+ngram_lm_scale_0.08_attention_scale_0.08	2.43
+ngram_lm_scale_0.08_attention_scale_0.1	2.43
+ngram_lm_scale_0.3_attention_scale_0.1	2.43
+ngram_lm_scale_0.5_attention_scale_0.5	2.43
+ngram_lm_scale_0.7_attention_scale_0.7	2.43
+ngram_lm_scale_0.9_attention_scale_1.1	2.43
+ngram_lm_scale_1.0_attention_scale_1.2	2.43
+ngram_lm_scale_1.1_attention_scale_1.5	2.43
+ngram_lm_scale_1.2_attention_scale_1.7	2.43
+ngram_lm_scale_1.3_attention_scale_1.9	2.43
+ngram_lm_scale_2.2_attention_scale_4.0	2.43
+ngram_lm_scale_0.01_attention_scale_0.01	2.44
+ngram_lm_scale_0.01_attention_scale_0.05	2.44
+ngram_lm_scale_0.01_attention_scale_0.08	2.44
+ngram_lm_scale_0.01_attention_scale_0.1	2.44
+ngram_lm_scale_0.05_attention_scale_0.05	2.44
+ngram_lm_scale_0.05_attention_scale_0.08	2.44
+ngram_lm_scale_0.08_attention_scale_0.01	2.44
+ngram_lm_scale_0.1_attention_scale_0.05	2.44
+ngram_lm_scale_0.1_attention_scale_0.08	2.44
+ngram_lm_scale_0.1_attention_scale_0.1	2.44
+ngram_lm_scale_0.3_attention_scale_0.05	2.44
+ngram_lm_scale_0.3_attention_scale_0.08	2.44
+ngram_lm_scale_0.5_attention_scale_0.3	2.44
+ngram_lm_scale_0.6_attention_scale_0.5	2.44
+ngram_lm_scale_0.1_attention_scale_0.01	2.45
+ngram_lm_scale_0.9_attention_scale_1.0	2.45
+ngram_lm_scale_1.5_attention_scale_2.2	2.45
+ngram_lm_scale_1.5_attention_scale_2.3	2.45
+ngram_lm_scale_0.3_attention_scale_0.01	2.46
+ngram_lm_scale_0.7_attention_scale_0.6	2.46
+ngram_lm_scale_0.9_attention_scale_0.9	2.46
+ngram_lm_scale_1.0_attention_scale_1.1	2.46
+ngram_lm_scale_1.1_attention_scale_1.3	2.47
+ngram_lm_scale_1.2_attention_scale_1.5	2.47
+ngram_lm_scale_1.3_attention_scale_1.7	2.47
+ngram_lm_scale_1.5_attention_scale_2.1	2.47
+ngram_lm_scale_0.7_attention_scale_0.5	2.48
+ngram_lm_scale_1.9_attention_scale_3.0	2.48
+ngram_lm_scale_2.3_attention_scale_4.0	2.48
+ngram_lm_scale_1.1_attention_scale_1.2	2.49
+ngram_lm_scale_1.7_attention_scale_2.5	2.49
+ngram_lm_scale_0.6_attention_scale_0.3	2.5
+ngram_lm_scale_1.0_attention_scale_1.0	2.5
+ngram_lm_scale_1.5_attention_scale_2.0	2.5
+ngram_lm_scale_1.3_attention_scale_1.5	2.51
+ngram_lm_scale_1.5_attention_scale_1.9	2.51
+ngram_lm_scale_2.0_attention_scale_3.0	2.51
+ngram_lm_scale_1.1_attention_scale_1.1	2.52
+ngram_lm_scale_1.2_attention_scale_1.3	2.52
+ngram_lm_scale_1.7_attention_scale_2.3	2.53
+ngram_lm_scale_0.5_attention_scale_0.08	2.54
+ngram_lm_scale_0.5_attention_scale_0.1	2.54
+ngram_lm_scale_1.0_attention_scale_0.9	2.54
+ngram_lm_scale_2.5_attention_scale_4.0	2.54
+ngram_lm_scale_0.9_attention_scale_0.7	2.55
+ngram_lm_scale_1.7_attention_scale_2.2	2.55
+ngram_lm_scale_3.0_attention_scale_5.0	2.55
+ngram_lm_scale_1.2_attention_scale_1.2	2.56
+ngram_lm_scale_0.5_attention_scale_0.05	2.57
+ngram_lm_scale_1.1_attention_scale_1.0	2.58
+ngram_lm_scale_2.1_attention_scale_3.0	2.58
+ngram_lm_scale_0.7_attention_scale_0.3	2.59
+ngram_lm_scale_1.2_attention_scale_1.1	2.59
+ngram_lm_scale_1.3_attention_scale_1.3	2.59
+ngram_lm_scale_1.5_attention_scale_1.7	2.6
+ngram_lm_scale_0.5_attention_scale_0.01	2.61
+ngram_lm_scale_1.7_attention_scale_2.1	2.61
+ngram_lm_scale_1.9_attention_scale_2.5	2.61
+ngram_lm_scale_0.9_attention_scale_0.6	2.62
+ngram_lm_scale_1.1_attention_scale_0.9	2.62
+ngram_lm_scale_1.7_attention_scale_2.0	2.64
+ngram_lm_scale_1.0_attention_scale_0.7	2.65
+ngram_lm_scale_2.2_attention_scale_3.0	2.65
+ngram_lm_scale_0.6_attention_scale_0.1	2.68
+ngram_lm_scale_1.3_attention_scale_1.2	2.68
+ngram_lm_scale_0.9_attention_scale_0.5	2.69
+ngram_lm_scale_1.2_attention_scale_1.0	2.69
+ngram_lm_scale_1.7_attention_scale_1.9	2.69
+ngram_lm_scale_0.6_attention_scale_0.08	2.7
+ngram_lm_scale_1.9_attention_scale_2.3	2.7
+ngram_lm_scale_2.0_attention_scale_2.5	2.7
+ngram_lm_scale_1.5_attention_scale_1.5	2.72
+ngram_lm_scale_0.6_attention_scale_0.05	2.74
+ngram_lm_scale_1.3_attention_scale_1.1	2.75
+ngram_lm_scale_2.3_attention_scale_3.0	2.75
+ngram_lm_scale_1.0_attention_scale_0.6	2.76
+ngram_lm_scale_1.9_attention_scale_2.2	2.76
+ngram_lm_scale_1.2_attention_scale_0.9	2.77
+ngram_lm_scale_0.6_attention_scale_0.01	2.8
+ngram_lm_scale_1.1_attention_scale_0.7	2.81
+ngram_lm_scale_2.1_attention_scale_2.5	2.83
+ngram_lm_scale_1.9_attention_scale_2.1	2.84
+ngram_lm_scale_2.0_attention_scale_2.3	2.84
+ngram_lm_scale_1.3_attention_scale_1.0	2.85
+ngram_lm_scale_1.7_attention_scale_1.7	2.85
+ngram_lm_scale_0.7_attention_scale_0.1	2.86
+ngram_lm_scale_1.0_attention_scale_0.5	2.89
+ngram_lm_scale_1.5_attention_scale_1.3	2.92
+ngram_lm_scale_0.7_attention_scale_0.08	2.93
+ngram_lm_scale_2.0_attention_scale_2.2	2.93
+ngram_lm_scale_1.9_attention_scale_2.0	2.95
+ngram_lm_scale_3.0_attention_scale_4.0	2.96
+ngram_lm_scale_2.2_attention_scale_2.5	2.99
+ngram_lm_scale_1.1_attention_scale_0.6	3.0
+ngram_lm_scale_2.0_attention_scale_2.1	3.01
+ngram_lm_scale_2.1_attention_scale_2.3	3.01
+ngram_lm_scale_0.7_attention_scale_0.05	3.02
+ngram_lm_scale_2.5_attention_scale_3.0	3.02
+ngram_lm_scale_1.9_attention_scale_1.9	3.04
+ngram_lm_scale_1.3_attention_scale_0.9	3.05
+ngram_lm_scale_0.9_attention_scale_0.3	3.06
+ngram_lm_scale_2.1_attention_scale_2.2	3.09
+ngram_lm_scale_1.5_attention_scale_1.2	3.1
+ngram_lm_scale_1.2_attention_scale_0.7	3.12
+ngram_lm_scale_2.0_attention_scale_2.0	3.12
+ngram_lm_scale_1.7_attention_scale_1.5	3.13
+ngram_lm_scale_0.7_attention_scale_0.01	3.15
+ngram_lm_scale_2.3_attention_scale_2.5	3.15
+ngram_lm_scale_2.2_attention_scale_2.3	3.16
+ngram_lm_scale_2.1_attention_scale_2.1	3.23
+ngram_lm_scale_1.1_attention_scale_0.5	3.29
+ngram_lm_scale_2.0_attention_scale_1.9	3.3
+ngram_lm_scale_1.5_attention_scale_1.1	3.31
+ngram_lm_scale_2.2_attention_scale_2.2	3.33
+ngram_lm_scale_1.9_attention_scale_1.7	3.36
+ngram_lm_scale_4.0_attention_scale_5.0	3.37
+ngram_lm_scale_2.1_attention_scale_2.0	3.4
+ngram_lm_scale_2.3_attention_scale_2.3	3.41
+ngram_lm_scale_1.2_attention_scale_0.6	3.46
+ngram_lm_scale_2.2_attention_scale_2.1	3.49
+ngram_lm_scale_2.5_attention_scale_2.5	3.55
+ngram_lm_scale_1.0_attention_scale_0.3	3.56
+ngram_lm_scale_2.1_attention_scale_1.9	3.56
+ngram_lm_scale_2.3_attention_scale_2.2	3.56
+ngram_lm_scale_1.7_attention_scale_1.3	3.58
+ngram_lm_scale_1.5_attention_scale_1.0	3.59
+ngram_lm_scale_1.3_attention_scale_0.7	3.62
+ngram_lm_scale_2.2_attention_scale_2.0	3.65
+ngram_lm_scale_2.0_attention_scale_1.7	3.67
+ngram_lm_scale_2.3_attention_scale_2.1	3.74
+ngram_lm_scale_1.9_attention_scale_1.5	3.82
+ngram_lm_scale_2.2_attention_scale_1.9	3.86
+ngram_lm_scale_1.2_attention_scale_0.5	3.88
+ngram_lm_scale_1.7_attention_scale_1.2	3.9
+ngram_lm_scale_3.0_attention_scale_3.0	3.91
+ngram_lm_scale_0.9_attention_scale_0.1	3.92
+ngram_lm_scale_2.5_attention_scale_2.3	3.94
+ngram_lm_scale_1.5_attention_scale_0.9	3.98
+ngram_lm_scale_2.3_attention_scale_2.0	4.0
+ngram_lm_scale_0.9_attention_scale_0.08	4.05
+ngram_lm_scale_2.1_attention_scale_1.7	4.08
+ngram_lm_scale_1.3_attention_scale_0.6	4.1
+ngram_lm_scale_2.5_attention_scale_2.2	4.19
+ngram_lm_scale_1.1_attention_scale_0.3	4.26
+ngram_lm_scale_0.9_attention_scale_0.05	4.27
+ngram_lm_scale_1.7_attention_scale_1.1	4.27
+ngram_lm_scale_2.3_attention_scale_1.9	4.27
+ngram_lm_scale_2.0_attention_scale_1.5	4.28
+ngram_lm_scale_2.5_attention_scale_2.1	4.47
+ngram_lm_scale_2.2_attention_scale_1.7	4.49
+ngram_lm_scale_1.9_attention_scale_1.3	4.53
+ngram_lm_scale_4.0_attention_scale_4.0	4.55
+ngram_lm_scale_0.9_attention_scale_0.01	4.58
+ngram_lm_scale_1.3_attention_scale_0.5	4.65
+ngram_lm_scale_1.7_attention_scale_1.0	4.76
+ngram_lm_scale_2.1_attention_scale_1.5	4.8
+ngram_lm_scale_2.5_attention_scale_2.0	4.83
+ngram_lm_scale_1.0_attention_scale_0.1	4.89
+ngram_lm_scale_1.9_attention_scale_1.2	5.04
+ngram_lm_scale_2.3_attention_scale_1.7	5.04
+ngram_lm_scale_1.0_attention_scale_0.08	5.06
+ngram_lm_scale_1.5_attention_scale_0.7	5.06
+ngram_lm_scale_5.0_attention_scale_5.0	5.08
+ngram_lm_scale_2.0_attention_scale_1.3	5.2
+ngram_lm_scale_2.5_attention_scale_1.9	5.26
+ngram_lm_scale_1.2_attention_scale_0.3	5.29
+ngram_lm_scale_3.0_attention_scale_2.5	5.31
+ngram_lm_scale_1.0_attention_scale_0.05	5.33
+ngram_lm_scale_1.7_attention_scale_0.9	5.39
+ngram_lm_scale_2.2_attention_scale_1.5	5.44
+ngram_lm_scale_1.9_attention_scale_1.1	5.71
+ngram_lm_scale_1.0_attention_scale_0.01	5.75
+ngram_lm_scale_1.5_attention_scale_0.6	5.8
+ngram_lm_scale_2.0_attention_scale_1.2	5.81
+ngram_lm_scale_2.1_attention_scale_1.3	5.91
+ngram_lm_scale_1.1_attention_scale_0.1	6.0
+ngram_lm_scale_2.3_attention_scale_1.5	6.07
+ngram_lm_scale_3.0_attention_scale_2.3	6.12
+ngram_lm_scale_1.1_attention_scale_0.08	6.18
+ngram_lm_scale_2.5_attention_scale_1.7	6.23
+ngram_lm_scale_1.9_attention_scale_1.0	6.27
+ngram_lm_scale_1.3_attention_scale_0.3	6.3
+ngram_lm_scale_2.0_attention_scale_1.1	6.36
+ngram_lm_scale_1.1_attention_scale_0.05	6.44
+ngram_lm_scale_2.1_attention_scale_1.2	6.45
+ngram_lm_scale_2.2_attention_scale_1.3	6.52
+ngram_lm_scale_1.5_attention_scale_0.5	6.55
+ngram_lm_scale_3.0_attention_scale_2.2	6.58
+ngram_lm_scale_1.7_attention_scale_0.7	6.68
+ngram_lm_scale_1.1_attention_scale_0.01	6.79
+ngram_lm_scale_1.9_attention_scale_0.9	6.84
+ngram_lm_scale_2.0_attention_scale_1.0	6.91
+ngram_lm_scale_1.2_attention_scale_0.1	6.97
+ngram_lm_scale_2.1_attention_scale_1.1	6.99
+ngram_lm_scale_3.0_attention_scale_2.1	7.07
+ngram_lm_scale_2.2_attention_scale_1.2	7.08
+ngram_lm_scale_1.2_attention_scale_0.08	7.14
+ngram_lm_scale_2.3_attention_scale_1.3	7.14
+ngram_lm_scale_2.5_attention_scale_1.5	7.24
+ngram_lm_scale_1.7_attention_scale_0.6	7.36
+ngram_lm_scale_1.2_attention_scale_0.05	7.4
+ngram_lm_scale_3.0_attention_scale_2.0	7.43
+ngram_lm_scale_2.0_attention_scale_0.9	7.47
+ngram_lm_scale_2.1_attention_scale_1.0	7.51
+ngram_lm_scale_2.2_attention_scale_1.1	7.56
+ngram_lm_scale_2.3_attention_scale_1.2	7.58
+ngram_lm_scale_4.0_attention_scale_3.0	7.68
+ngram_lm_scale_1.2_attention_scale_0.01	7.73
+ngram_lm_scale_3.0_attention_scale_1.9	7.78
+ngram_lm_scale_1.3_attention_scale_0.1	7.81
+ngram_lm_scale_5.0_attention_scale_4.0	7.81
+ngram_lm_scale_1.5_attention_scale_0.3	7.87
+ngram_lm_scale_1.7_attention_scale_0.5	7.89
+ngram_lm_scale_1.9_attention_scale_0.7	7.91
+ngram_lm_scale_1.3_attention_scale_0.08	7.92
+ngram_lm_scale_2.1_attention_scale_0.9	7.94
+ngram_lm_scale_2.2_attention_scale_1.0	7.96
+ngram_lm_scale_2.3_attention_scale_1.1	7.97
+ngram_lm_scale_2.5_attention_scale_1.3	7.98
+ngram_lm_scale_1.3_attention_scale_0.05	8.07
+ngram_lm_scale_2.5_attention_scale_1.2	8.26
+ngram_lm_scale_3.0_attention_scale_1.7	8.26
+ngram_lm_scale_1.3_attention_scale_0.01	8.27
+ngram_lm_scale_2.3_attention_scale_1.0	8.29
+ngram_lm_scale_2.2_attention_scale_0.9	8.3
+ngram_lm_scale_2.0_attention_scale_0.7	8.31
+ngram_lm_scale_1.9_attention_scale_0.6	8.32
+ngram_lm_scale_4.0_attention_scale_2.5	8.41
+ngram_lm_scale_2.5_attention_scale_1.1	8.45
+ngram_lm_scale_2.3_attention_scale_0.9	8.48
+ngram_lm_scale_3.0_attention_scale_1.5	8.51
+ngram_lm_scale_2.1_attention_scale_0.7	8.52
+ngram_lm_scale_2.0_attention_scale_0.6	8.55
+ngram_lm_scale_1.9_attention_scale_0.5	8.56
+ngram_lm_scale_4.0_attention_scale_2.3	8.58
+ngram_lm_scale_2.5_attention_scale_1.0	8.61
+ngram_lm_scale_1.7_attention_scale_0.3	8.62
+ngram_lm_scale_4.0_attention_scale_2.2	8.65
+ngram_lm_scale_5.0_attention_scale_3.0	8.67
+ngram_lm_scale_1.5_attention_scale_0.1	8.68
+ngram_lm_scale_2.2_attention_scale_0.7	8.68
+ngram_lm_scale_2.1_attention_scale_0.6	8.72
+ngram_lm_scale_4.0_attention_scale_2.1	8.73
+ngram_lm_scale_3.0_attention_scale_1.3	8.74
+ngram_lm_scale_1.5_attention_scale_0.08	8.75
+ngram_lm_scale_2.0_attention_scale_0.5	8.76
+ngram_lm_scale_2.5_attention_scale_0.9	8.77
+ngram_lm_scale_1.5_attention_scale_0.05	8.83
+ngram_lm_scale_2.3_attention_scale_0.7	8.83
+ngram_lm_scale_4.0_attention_scale_2.0	8.83
+ngram_lm_scale_2.2_attention_scale_0.6	8.86
+ngram_lm_scale_3.0_attention_scale_1.2	8.86
+ngram_lm_scale_2.1_attention_scale_0.5	8.89
+ngram_lm_scale_1.5_attention_scale_0.01	8.91
+ngram_lm_scale_4.0_attention_scale_1.9	8.91
+ngram_lm_scale_2.3_attention_scale_0.6	8.95
+ngram_lm_scale_3.0_attention_scale_1.1	8.95
+ngram_lm_scale_1.9_attention_scale_0.3	8.96
+ngram_lm_scale_2.2_attention_scale_0.5	8.98
+ngram_lm_scale_2.5_attention_scale_0.7	9.0
+ngram_lm_scale_5.0_attention_scale_2.5	9.0
+ngram_lm_scale_1.7_attention_scale_0.1	9.01
+ngram_lm_scale_1.7_attention_scale_0.08	9.04
+ngram_lm_scale_3.0_attention_scale_1.0	9.04
+ngram_lm_scale_2.0_attention_scale_0.3	9.05
+ngram_lm_scale_2.3_attention_scale_0.5	9.05
+ngram_lm_scale_4.0_attention_scale_1.7	9.05
+ngram_lm_scale_2.5_attention_scale_0.6	9.09
+ngram_lm_scale_5.0_attention_scale_2.3	9.1
+ngram_lm_scale_1.7_attention_scale_0.05	9.11
+ngram_lm_scale_3.0_attention_scale_0.9	9.11
+ngram_lm_scale_2.1_attention_scale_0.3	9.12
+ngram_lm_scale_5.0_attention_scale_2.2	9.13
+ngram_lm_scale_4.0_attention_scale_1.5	9.15
+ngram_lm_scale_5.0_attention_scale_2.1	9.17
+ngram_lm_scale_1.7_attention_scale_0.01	9.18
+ngram_lm_scale_2.2_attention_scale_0.3	9.19
+ngram_lm_scale_2.5_attention_scale_0.5	9.19
+ngram_lm_scale_1.9_attention_scale_0.1	9.2
+ngram_lm_scale_5.0_attention_scale_2.0	9.21
+ngram_lm_scale_1.9_attention_scale_0.08	9.22
+ngram_lm_scale_3.0_attention_scale_0.7	9.23
+ngram_lm_scale_2.3_attention_scale_0.3	9.24
+ngram_lm_scale_4.0_attention_scale_1.3	9.24
+ngram_lm_scale_1.9_attention_scale_0.05	9.25
+ngram_lm_scale_2.0_attention_scale_0.1	9.25
+ngram_lm_scale_5.0_attention_scale_1.9	9.25
+ngram_lm_scale_1.9_attention_scale_0.01	9.26
+ngram_lm_scale_2.0_attention_scale_0.08	9.26
+ngram_lm_scale_2.0_attention_scale_0.05	9.28
+ngram_lm_scale_2.1_attention_scale_0.1	9.28
+ngram_lm_scale_4.0_attention_scale_1.2	9.28
+ngram_lm_scale_5.0_attention_scale_1.7	9.29
+ngram_lm_scale_2.1_attention_scale_0.08	9.3
+ngram_lm_scale_2.5_attention_scale_0.3	9.3
+ngram_lm_scale_3.0_attention_scale_0.6	9.3
+ngram_lm_scale_4.0_attention_scale_1.1	9.3
+ngram_lm_scale_2.0_attention_scale_0.01	9.31
+ngram_lm_scale_3.0_attention_scale_0.5	9.32
+ngram_lm_scale_4.0_attention_scale_1.0	9.32
+ngram_lm_scale_5.0_attention_scale_1.5	9.32
+ngram_lm_scale_2.1_attention_scale_0.05	9.33
+ngram_lm_scale_2.2_attention_scale_0.1	9.33
+ngram_lm_scale_2.1_attention_scale_0.01	9.35
+ngram_lm_scale_2.2_attention_scale_0.08	9.35
+ngram_lm_scale_2.3_attention_scale_0.1	9.35
+ngram_lm_scale_4.0_attention_scale_0.9	9.35
+ngram_lm_scale_5.0_attention_scale_1.3	9.35
+ngram_lm_scale_2.2_attention_scale_0.05	9.36
+ngram_lm_scale_2.3_attention_scale_0.08	9.37
+ngram_lm_scale_5.0_attention_scale_1.2	9.37
+ngram_lm_scale_2.2_attention_scale_0.01	9.38
+ngram_lm_scale_2.3_attention_scale_0.05	9.38
+ngram_lm_scale_5.0_attention_scale_1.1	9.38
+ngram_lm_scale_4.0_attention_scale_0.7	9.39
+ngram_lm_scale_2.3_attention_scale_0.01	9.4
+ngram_lm_scale_2.5_attention_scale_0.08	9.4
+ngram_lm_scale_2.5_attention_scale_0.1	9.4
+ngram_lm_scale_3.0_attention_scale_0.3	9.4
+ngram_lm_scale_4.0_attention_scale_0.6	9.4
+ngram_lm_scale_5.0_attention_scale_0.9	9.4
+ngram_lm_scale_5.0_attention_scale_1.0	9.4
+ngram_lm_scale_2.5_attention_scale_0.05	9.42
+ngram_lm_scale_4.0_attention_scale_0.5	9.42
+ngram_lm_scale_5.0_attention_scale_0.7	9.42
+ngram_lm_scale_2.5_attention_scale_0.01	9.43
+ngram_lm_scale_5.0_attention_scale_0.6	9.43
+ngram_lm_scale_4.0_attention_scale_0.3	9.44
+ngram_lm_scale_5.0_attention_scale_0.5	9.45
+ngram_lm_scale_3.0_attention_scale_0.01	9.46
+ngram_lm_scale_3.0_attention_scale_0.05	9.46
+ngram_lm_scale_3.0_attention_scale_0.08	9.46
+ngram_lm_scale_3.0_attention_scale_0.1	9.46
+ngram_lm_scale_5.0_attention_scale_0.3	9.46
+ngram_lm_scale_4.0_attention_scale_0.05	9.48
+ngram_lm_scale_4.0_attention_scale_0.08	9.48
+ngram_lm_scale_4.0_attention_scale_0.1	9.48
+ngram_lm_scale_4.0_attention_scale_0.01	9.49
+ngram_lm_scale_5.0_attention_scale_0.1	9.49
+ngram_lm_scale_5.0_attention_scale_0.08	9.5
+ngram_lm_scale_5.0_attention_scale_0.01	9.51
+ngram_lm_scale_5.0_attention_scale_0.05	9.51
+
+2023-03-10 16:26:56,393 INFO [decode.py:580] batch 0/?, cuts processed until now is 17
+2023-03-10 16:29:21,300 INFO [zipformer.py:1455] attn_weights_entropy = tensor([3.7346, 3.8901, 3.6083, 3.9171, 3.5037, 3.9391, 3.9041, 3.8373],
+       device='cuda:0'), covar=tensor([0.0486, 0.0328, 0.0562, 0.0435, 0.0383, 0.0169, 0.0287, 0.0198],
+       device='cuda:0'), in_proj_covar=tensor([0.0390, 0.0327, 0.0367, 0.0362, 0.0325, 0.0234, 0.0307, 0.0288],
+       device='cuda:0'), out_proj_covar=tensor([0.0006, 0.0006, 0.0005, 0.0006, 0.0005, 0.0004, 0.0005, 0.0005],
+       device='cuda:0')
+2023-03-10 16:32:59,546 INFO [decode.py:580] batch 100/?, cuts processed until now is 2560
+2023-03-10 16:33:13,559 INFO [zipformer.py:1455] attn_weights_entropy = tensor([2.5211, 2.4382, 2.4224, 2.3565, 2.4605, 2.2951, 2.5959, 1.9337],
+       device='cuda:0'), covar=tensor([0.0807, 0.1321, 0.2179, 0.3759, 0.1536, 0.1748, 0.1180, 0.3484],
+       device='cuda:0'), in_proj_covar=tensor([0.0186, 0.0193, 0.0209, 0.0260, 0.0168, 0.0270, 0.0191, 0.0219],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-10 16:36:32,675 INFO [decode.py:626] 
+For test-other, WER of different settings are:
+ngram_lm_scale_0.1_attention_scale_5.0	5.16	best for test-other
+ngram_lm_scale_0.3_attention_scale_5.0	5.16
+ngram_lm_scale_0.08_attention_scale_5.0	5.17
+ngram_lm_scale_0.01_attention_scale_2.2	5.18
+ngram_lm_scale_0.01_attention_scale_2.3	5.18
+ngram_lm_scale_0.01_attention_scale_2.5	5.18
+ngram_lm_scale_0.01_attention_scale_3.0	5.18
+ngram_lm_scale_0.01_attention_scale_4.0	5.18
+ngram_lm_scale_0.01_attention_scale_5.0	5.18
+ngram_lm_scale_0.05_attention_scale_1.9	5.18
+ngram_lm_scale_0.05_attention_scale_2.2	5.18
+ngram_lm_scale_0.05_attention_scale_2.3	5.18
+ngram_lm_scale_0.05_attention_scale_2.5	5.18
+ngram_lm_scale_0.05_attention_scale_3.0	5.18
+ngram_lm_scale_0.05_attention_scale_4.0	5.18
+ngram_lm_scale_0.05_attention_scale_5.0	5.18
+ngram_lm_scale_0.08_attention_scale_2.5	5.18
+ngram_lm_scale_0.08_attention_scale_3.0	5.18
+ngram_lm_scale_0.08_attention_scale_4.0	5.18
+ngram_lm_scale_0.1_attention_scale_2.5	5.18
+ngram_lm_scale_0.1_attention_scale_3.0	5.18
+ngram_lm_scale_0.1_attention_scale_4.0	5.18
+ngram_lm_scale_0.3_attention_scale_4.0	5.18
+ngram_lm_scale_0.01_attention_scale_1.1	5.19
+ngram_lm_scale_0.01_attention_scale_1.2	5.19
+ngram_lm_scale_0.01_attention_scale_1.3	5.19
+ngram_lm_scale_0.01_attention_scale_1.5	5.19
+ngram_lm_scale_0.01_attention_scale_1.7	5.19
+ngram_lm_scale_0.01_attention_scale_1.9	5.19
+ngram_lm_scale_0.01_attention_scale_2.0	5.19
+ngram_lm_scale_0.01_attention_scale_2.1	5.19
+ngram_lm_scale_0.05_attention_scale_1.2	5.19
+ngram_lm_scale_0.05_attention_scale_1.3	5.19
+ngram_lm_scale_0.05_attention_scale_1.5	5.19
+ngram_lm_scale_0.05_attention_scale_1.7	5.19
+ngram_lm_scale_0.05_attention_scale_2.0	5.19
+ngram_lm_scale_0.05_attention_scale_2.1	5.19
+ngram_lm_scale_0.08_attention_scale_1.3	5.19
+ngram_lm_scale_0.08_attention_scale_1.5	5.19
+ngram_lm_scale_0.08_attention_scale_1.7	5.19
+ngram_lm_scale_0.08_attention_scale_2.0	5.19
+ngram_lm_scale_0.08_attention_scale_2.1	5.19
+ngram_lm_scale_0.08_attention_scale_2.2	5.19
+ngram_lm_scale_0.08_attention_scale_2.3	5.19
+ngram_lm_scale_0.1_attention_scale_1.0	5.19
+ngram_lm_scale_0.1_attention_scale_1.1	5.19
+ngram_lm_scale_0.1_attention_scale_1.2	5.19
+ngram_lm_scale_0.1_attention_scale_1.5	5.19
+ngram_lm_scale_0.1_attention_scale_2.1	5.19
+ngram_lm_scale_0.1_attention_scale_2.2	5.19
+ngram_lm_scale_0.1_attention_scale_2.3	5.19
+ngram_lm_scale_0.5_attention_scale_5.0	5.19
+ngram_lm_scale_0.7_attention_scale_5.0	5.19
+ngram_lm_scale_0.01_attention_scale_0.9	5.2
+ngram_lm_scale_0.05_attention_scale_0.9	5.2
+ngram_lm_scale_0.05_attention_scale_1.0	5.2
+ngram_lm_scale_0.05_attention_scale_1.1	5.2
+ngram_lm_scale_0.08_attention_scale_0.9	5.2
+ngram_lm_scale_0.08_attention_scale_1.0	5.2
+ngram_lm_scale_0.08_attention_scale_1.1	5.2
+ngram_lm_scale_0.08_attention_scale_1.2	5.2
+ngram_lm_scale_0.08_attention_scale_1.9	5.2
+ngram_lm_scale_0.1_attention_scale_0.9	5.2
+ngram_lm_scale_0.1_attention_scale_1.3	5.2
+ngram_lm_scale_0.1_attention_scale_1.7	5.2
+ngram_lm_scale_0.1_attention_scale_1.9	5.2
+ngram_lm_scale_0.1_attention_scale_2.0	5.2
+ngram_lm_scale_0.3_attention_scale_3.0	5.2
+ngram_lm_scale_0.5_attention_scale_4.0	5.2
+ngram_lm_scale_0.6_attention_scale_5.0	5.2
+ngram_lm_scale_0.7_attention_scale_4.0	5.2
+ngram_lm_scale_0.9_attention_scale_5.0	5.2
+ngram_lm_scale_0.01_attention_scale_0.7	5.21
+ngram_lm_scale_0.01_attention_scale_1.0	5.21
+ngram_lm_scale_0.5_attention_scale_1.3	5.21
+ngram_lm_scale_0.6_attention_scale_4.0	5.21
+ngram_lm_scale_0.7_attention_scale_3.0	5.21
+ngram_lm_scale_0.9_attention_scale_4.0	5.21
+ngram_lm_scale_1.0_attention_scale_5.0	5.21
+ngram_lm_scale_0.05_attention_scale_0.6	5.22
+ngram_lm_scale_0.05_attention_scale_0.7	5.22
+ngram_lm_scale_0.08_attention_scale_0.7	5.22
+ngram_lm_scale_0.1_attention_scale_0.6	5.22
+ngram_lm_scale_0.1_attention_scale_0.7	5.22
+ngram_lm_scale_0.3_attention_scale_0.9	5.22
+ngram_lm_scale_0.3_attention_scale_1.0	5.22
+ngram_lm_scale_0.3_attention_scale_1.1	5.22
+ngram_lm_scale_0.3_attention_scale_1.2	5.22
+ngram_lm_scale_0.3_attention_scale_1.3	5.22
+ngram_lm_scale_0.3_attention_scale_1.5	5.22
+ngram_lm_scale_0.3_attention_scale_2.1	5.22
+ngram_lm_scale_0.3_attention_scale_2.2	5.22
+ngram_lm_scale_0.3_attention_scale_2.3	5.22
+ngram_lm_scale_0.3_attention_scale_2.5	5.22
+ngram_lm_scale_0.5_attention_scale_1.2	5.22
+ngram_lm_scale_0.5_attention_scale_1.5	5.22
+ngram_lm_scale_0.5_attention_scale_2.1	5.22
+ngram_lm_scale_0.5_attention_scale_2.2	5.22
+ngram_lm_scale_0.5_attention_scale_2.5	5.22
+ngram_lm_scale_0.5_attention_scale_3.0	5.22
+ngram_lm_scale_0.6_attention_scale_1.7	5.22
+ngram_lm_scale_0.6_attention_scale_2.0	5.22
+ngram_lm_scale_0.6_attention_scale_2.1	5.22
+ngram_lm_scale_0.6_attention_scale_2.2	5.22
+ngram_lm_scale_0.6_attention_scale_2.5	5.22
+ngram_lm_scale_0.7_attention_scale_2.1	5.22
+ngram_lm_scale_0.7_attention_scale_2.5	5.22
+ngram_lm_scale_1.0_attention_scale_4.0	5.22
+ngram_lm_scale_1.1_attention_scale_5.0	5.22
+ngram_lm_scale_1.2_attention_scale_5.0	5.22
+ngram_lm_scale_0.01_attention_scale_0.6	5.23
+ngram_lm_scale_0.05_attention_scale_0.5	5.23
+ngram_lm_scale_0.08_attention_scale_0.5	5.23
+ngram_lm_scale_0.08_attention_scale_0.6	5.23
+ngram_lm_scale_0.1_attention_scale_0.5	5.23
+ngram_lm_scale_0.3_attention_scale_1.7	5.23
+ngram_lm_scale_0.3_attention_scale_1.9	5.23
+ngram_lm_scale_0.3_attention_scale_2.0	5.23
+ngram_lm_scale_0.5_attention_scale_1.1	5.23
+ngram_lm_scale_0.5_attention_scale_1.7	5.23
+ngram_lm_scale_0.5_attention_scale_1.9	5.23
+ngram_lm_scale_0.5_attention_scale_2.0	5.23
+ngram_lm_scale_0.5_attention_scale_2.3	5.23
+ngram_lm_scale_0.6_attention_scale_1.5	5.23
+ngram_lm_scale_0.6_attention_scale_1.9	5.23
+ngram_lm_scale_0.6_attention_scale_2.3	5.23
+ngram_lm_scale_0.6_attention_scale_3.0	5.23
+ngram_lm_scale_0.7_attention_scale_1.7	5.23
+ngram_lm_scale_0.7_attention_scale_1.9	5.23
+ngram_lm_scale_0.7_attention_scale_2.0	5.23
+ngram_lm_scale_0.7_attention_scale_2.2	5.23
+ngram_lm_scale_0.7_attention_scale_2.3	5.23
+ngram_lm_scale_1.3_attention_scale_5.0	5.23
+ngram_lm_scale_0.01_attention_scale_0.5	5.24
+ngram_lm_scale_0.05_attention_scale_0.3	5.24
+ngram_lm_scale_0.6_attention_scale_1.3	5.24
+ngram_lm_scale_0.7_attention_scale_1.5	5.24
+ngram_lm_scale_0.9_attention_scale_2.5	5.24
+ngram_lm_scale_0.9_attention_scale_3.0	5.24
+ngram_lm_scale_1.0_attention_scale_3.0	5.24
+ngram_lm_scale_1.1_attention_scale_4.0	5.24
+ngram_lm_scale_1.2_attention_scale_4.0	5.24
+ngram_lm_scale_1.3_attention_scale_4.0	5.24
+ngram_lm_scale_0.08_attention_scale_0.3	5.25
+ngram_lm_scale_0.3_attention_scale_0.7	5.25
+ngram_lm_scale_0.5_attention_scale_1.0	5.25
+ngram_lm_scale_0.7_attention_scale_1.3	5.25
+ngram_lm_scale_0.9_attention_scale_2.2	5.25
+ngram_lm_scale_1.0_attention_scale_2.5	5.25
+ngram_lm_scale_1.1_attention_scale_3.0	5.25
+ngram_lm_scale_1.5_attention_scale_5.0	5.25
+ngram_lm_scale_0.01_attention_scale_0.3	5.26
+ngram_lm_scale_0.5_attention_scale_0.9	5.26
+ngram_lm_scale_0.6_attention_scale_1.1	5.26
+ngram_lm_scale_0.6_attention_scale_1.2	5.26
+ngram_lm_scale_0.9_attention_scale_2.1	5.26
+ngram_lm_scale_0.9_attention_scale_2.3	5.26
+ngram_lm_scale_1.0_attention_scale_2.3	5.26
+ngram_lm_scale_1.7_attention_scale_5.0	5.26
+ngram_lm_scale_0.1_attention_scale_0.3	5.27
+ngram_lm_scale_0.6_attention_scale_0.9	5.27
+ngram_lm_scale_0.6_attention_scale_1.0	5.27
+ngram_lm_scale_0.7_attention_scale_1.2	5.27
+ngram_lm_scale_0.7_attention_scale_1.1	5.28
+ngram_lm_scale_0.9_attention_scale_1.7	5.28
+ngram_lm_scale_0.9_attention_scale_1.9	5.28
+ngram_lm_scale_0.9_attention_scale_2.0	5.28
+ngram_lm_scale_1.0_attention_scale_2.2	5.28
+ngram_lm_scale_1.1_attention_scale_2.5	5.28
+ngram_lm_scale_1.2_attention_scale_3.0	5.28
+ngram_lm_scale_0.3_attention_scale_0.5	5.29
+ngram_lm_scale_0.3_attention_scale_0.6	5.29
+ngram_lm_scale_0.5_attention_scale_0.7	5.29
+ngram_lm_scale_1.0_attention_scale_2.1	5.29
+ngram_lm_scale_1.5_attention_scale_4.0	5.29
+ngram_lm_scale_1.3_attention_scale_3.0	5.3
+ngram_lm_scale_1.9_attention_scale_5.0	5.3
+ngram_lm_scale_0.01_attention_scale_0.1	5.31
+ngram_lm_scale_0.05_attention_scale_0.08	5.31
+ngram_lm_scale_0.05_attention_scale_0.1	5.31
+ngram_lm_scale_0.5_attention_scale_0.6	5.31
+ngram_lm_scale_0.7_attention_scale_1.0	5.31
+ngram_lm_scale_1.0_attention_scale_2.0	5.31
+ngram_lm_scale_2.0_attention_scale_5.0	5.31
+ngram_lm_scale_0.01_attention_scale_0.08	5.32
+ngram_lm_scale_0.08_attention_scale_0.1	5.32
+ngram_lm_scale_0.5_attention_scale_0.5	5.32
+ngram_lm_scale_0.6_attention_scale_0.7	5.32
+ngram_lm_scale_1.1_attention_scale_2.2	5.32
+ngram_lm_scale_1.1_attention_scale_2.3	5.32
+ngram_lm_scale_0.08_attention_scale_0.08	5.33
+ngram_lm_scale_0.1_attention_scale_0.1	5.33
+ngram_lm_scale_0.3_attention_scale_0.3	5.33
+ngram_lm_scale_1.0_attention_scale_1.9	5.33
+ngram_lm_scale_1.2_attention_scale_2.5	5.33
+ngram_lm_scale_1.7_attention_scale_4.0	5.33
+ngram_lm_scale_0.05_attention_scale_0.05	5.34
+ngram_lm_scale_0.08_attention_scale_0.05	5.34
+ngram_lm_scale_0.1_attention_scale_0.08	5.34
+ngram_lm_scale_0.7_attention_scale_0.9	5.34
+ngram_lm_scale_0.9_attention_scale_1.5	5.34
+ngram_lm_scale_1.1_attention_scale_2.1	5.34
+ngram_lm_scale_2.1_attention_scale_5.0	5.34
+ngram_lm_scale_0.01_attention_scale_0.05	5.35
+ngram_lm_scale_0.6_attention_scale_0.6	5.35
+ngram_lm_scale_1.0_attention_scale_1.7	5.35
+ngram_lm_scale_1.1_attention_scale_2.0	5.35
+ngram_lm_scale_1.2_attention_scale_2.3	5.35
+ngram_lm_scale_0.3_attention_scale_0.1	5.36
+ngram_lm_scale_2.2_attention_scale_5.0	5.36
+ngram_lm_scale_0.1_attention_scale_0.05	5.37
+ngram_lm_scale_0.9_attention_scale_1.2	5.37
+ngram_lm_scale_0.9_attention_scale_1.3	5.37
+ngram_lm_scale_1.1_attention_scale_1.9	5.37
+ngram_lm_scale_1.2_attention_scale_2.2	5.37
+ngram_lm_scale_1.3_attention_scale_2.5	5.37
+ngram_lm_scale_0.01_attention_scale_0.01	5.38
+ngram_lm_scale_0.05_attention_scale_0.01	5.38
+ngram_lm_scale_0.08_attention_scale_0.01	5.38
+ngram_lm_scale_0.7_attention_scale_0.7	5.38
+ngram_lm_scale_1.0_attention_scale_1.5	5.38
+ngram_lm_scale_1.2_attention_scale_2.1	5.38
+ngram_lm_scale_0.1_attention_scale_0.01	5.39
+ngram_lm_scale_0.3_attention_scale_0.08	5.39
+ngram_lm_scale_0.5_attention_scale_0.3	5.39
+ngram_lm_scale_0.6_attention_scale_0.5	5.39
+ngram_lm_scale_1.1_attention_scale_1.7	5.39
+ngram_lm_scale_0.3_attention_scale_0.05	5.4
+ngram_lm_scale_0.9_attention_scale_1.1	5.4
+ngram_lm_scale_1.2_attention_scale_2.0	5.4
+ngram_lm_scale_1.3_attention_scale_2.2	5.4
+ngram_lm_scale_1.3_attention_scale_2.3	5.4
+ngram_lm_scale_1.5_attention_scale_3.0	5.4
+ngram_lm_scale_1.9_attention_scale_4.0	5.4
+ngram_lm_scale_0.7_attention_scale_0.6	5.41
+ngram_lm_scale_1.0_attention_scale_1.3	5.41
+ngram_lm_scale_1.2_attention_scale_1.9	5.41
+ngram_lm_scale_0.9_attention_scale_1.0	5.42
+ngram_lm_scale_1.3_attention_scale_2.1	5.42
+ngram_lm_scale_2.3_attention_scale_5.0	5.42
+ngram_lm_scale_0.3_attention_scale_0.01	5.43
+ngram_lm_scale_1.1_attention_scale_1.5	5.43
+ngram_lm_scale_1.2_attention_scale_1.7	5.43
+ngram_lm_scale_1.3_attention_scale_2.0	5.43
+ngram_lm_scale_1.0_attention_scale_1.2	5.44
+ngram_lm_scale_1.3_attention_scale_1.9	5.44
+ngram_lm_scale_1.5_attention_scale_2.5	5.44
+ngram_lm_scale_1.7_attention_scale_3.0	5.44
+ngram_lm_scale_2.0_attention_scale_4.0	5.44
+ngram_lm_scale_2.1_attention_scale_4.0	5.46
+ngram_lm_scale_0.9_attention_scale_0.9	5.47
+ngram_lm_scale_1.0_attention_scale_1.1	5.47
+ngram_lm_scale_1.5_attention_scale_2.3	5.47
+ngram_lm_scale_0.7_attention_scale_0.5	5.48
+ngram_lm_scale_1.1_attention_scale_1.3	5.48
+ngram_lm_scale_2.5_attention_scale_5.0	5.48
+ngram_lm_scale_1.2_attention_scale_1.5	5.49
+ngram_lm_scale_1.5_attention_scale_2.2	5.49
+ngram_lm_scale_2.2_attention_scale_4.0	5.49
+ngram_lm_scale_0.6_attention_scale_0.3	5.51
+ngram_lm_scale_1.3_attention_scale_1.7	5.51
+ngram_lm_scale_1.5_attention_scale_2.1	5.52
+ngram_lm_scale_1.1_attention_scale_1.2	5.54
+ngram_lm_scale_2.3_attention_scale_4.0	5.54
+ngram_lm_scale_1.0_attention_scale_1.0	5.55
+ngram_lm_scale_1.9_attention_scale_3.0	5.55
+ngram_lm_scale_1.7_attention_scale_2.5	5.56
+ngram_lm_scale_1.5_attention_scale_2.0	5.57
+ngram_lm_scale_0.5_attention_scale_0.1	5.6
+ngram_lm_scale_1.3_attention_scale_1.5	5.61
+ngram_lm_scale_1.1_attention_scale_1.1	5.62
+ngram_lm_scale_1.2_attention_scale_1.3	5.62
+ngram_lm_scale_0.5_attention_scale_0.08	5.63
+ngram_lm_scale_1.0_attention_scale_0.9	5.63
+ngram_lm_scale_1.5_attention_scale_1.9	5.64
+ngram_lm_scale_2.0_attention_scale_3.0	5.65
+ngram_lm_scale_0.9_attention_scale_0.7	5.66
+ngram_lm_scale_0.5_attention_scale_0.05	5.67
+ngram_lm_scale_1.7_attention_scale_2.3	5.67
+ngram_lm_scale_0.7_attention_scale_0.3	5.73
+ngram_lm_scale_1.2_attention_scale_1.2	5.73
+ngram_lm_scale_2.5_attention_scale_4.0	5.73
+ngram_lm_scale_1.1_attention_scale_1.0	5.74
+ngram_lm_scale_0.5_attention_scale_0.01	5.75
+ngram_lm_scale_1.7_attention_scale_2.2	5.77
+ngram_lm_scale_0.9_attention_scale_0.6	5.79
+ngram_lm_scale_2.1_attention_scale_3.0	5.82
+ngram_lm_scale_0.6_attention_scale_0.1	5.83
+ngram_lm_scale_1.2_attention_scale_1.1	5.83
+ngram_lm_scale_3.0_attention_scale_5.0	5.83
+ngram_lm_scale_1.3_attention_scale_1.3	5.84
+ngram_lm_scale_1.5_attention_scale_1.7	5.87
+ngram_lm_scale_0.6_attention_scale_0.08	5.88
+ngram_lm_scale_1.1_attention_scale_0.9	5.88
+ngram_lm_scale_1.7_attention_scale_2.1	5.89
+ngram_lm_scale_1.0_attention_scale_0.7	5.9
+ngram_lm_scale_1.9_attention_scale_2.5	5.91
+ngram_lm_scale_0.9_attention_scale_0.5	5.95
+ngram_lm_scale_0.6_attention_scale_0.05	5.97
+ngram_lm_scale_1.7_attention_scale_2.0	6.0
+ngram_lm_scale_1.3_attention_scale_1.2	6.01
+ngram_lm_scale_1.2_attention_scale_1.0	6.02
+ngram_lm_scale_2.2_attention_scale_3.0	6.04
+ngram_lm_scale_0.6_attention_scale_0.01	6.09
+ngram_lm_scale_2.0_attention_scale_2.5	6.11
+ngram_lm_scale_1.0_attention_scale_0.6	6.12
+ngram_lm_scale_1.7_attention_scale_1.9	6.13
+ngram_lm_scale_1.9_attention_scale_2.3	6.13
+ngram_lm_scale_1.5_attention_scale_1.5	6.16
+ngram_lm_scale_1.3_attention_scale_1.1	6.17
+ngram_lm_scale_1.2_attention_scale_0.9	6.21
+ngram_lm_scale_0.7_attention_scale_0.1	6.23
+ngram_lm_scale_1.1_attention_scale_0.7	6.26
+ngram_lm_scale_1.9_attention_scale_2.2	6.29
+ngram_lm_scale_2.3_attention_scale_3.0	6.3
+ngram_lm_scale_0.7_attention_scale_0.08	6.32
+ngram_lm_scale_1.0_attention_scale_0.5	6.35
+ngram_lm_scale_1.3_attention_scale_1.0	6.38
+ngram_lm_scale_2.0_attention_scale_2.3	6.43
+ngram_lm_scale_2.1_attention_scale_2.5	6.43
+ngram_lm_scale_1.9_attention_scale_2.1	6.46
+ngram_lm_scale_0.7_attention_scale_0.05	6.47
+ngram_lm_scale_1.7_attention_scale_1.7	6.49
+ngram_lm_scale_1.5_attention_scale_1.3	6.56
+ngram_lm_scale_0.9_attention_scale_0.3	6.59
+ngram_lm_scale_1.1_attention_scale_0.6	6.61
+ngram_lm_scale_1.9_attention_scale_2.0	6.63
+ngram_lm_scale_2.0_attention_scale_2.2	6.63
+ngram_lm_scale_0.7_attention_scale_0.01	6.69
+ngram_lm_scale_1.3_attention_scale_0.9	6.71
+ngram_lm_scale_2.2_attention_scale_2.5	6.72
+ngram_lm_scale_3.0_attention_scale_4.0	6.73
+ngram_lm_scale_2.1_attention_scale_2.3	6.76
+ngram_lm_scale_2.5_attention_scale_3.0	6.79
+ngram_lm_scale_2.0_attention_scale_2.1	6.81
+ngram_lm_scale_1.9_attention_scale_1.9	6.84
+ngram_lm_scale_1.5_attention_scale_1.2	6.85
+ngram_lm_scale_1.2_attention_scale_0.7	6.87
+ngram_lm_scale_2.1_attention_scale_2.2	6.92
+ngram_lm_scale_1.7_attention_scale_1.5	7.0
+ngram_lm_scale_2.0_attention_scale_2.0	7.0
+ngram_lm_scale_2.3_attention_scale_2.5	7.0
+ngram_lm_scale_2.2_attention_scale_2.3	7.06
+ngram_lm_scale_1.1_attention_scale_0.5	7.12
+ngram_lm_scale_2.1_attention_scale_2.1	7.15
+ngram_lm_scale_2.0_attention_scale_1.9	7.23
+ngram_lm_scale_1.5_attention_scale_1.1	7.24
+ngram_lm_scale_2.2_attention_scale_2.2	7.31
+ngram_lm_scale_1.9_attention_scale_1.7	7.32
+ngram_lm_scale_1.2_attention_scale_0.6	7.35
+ngram_lm_scale_4.0_attention_scale_5.0	7.36
+ngram_lm_scale_2.1_attention_scale_2.0	7.39
+ngram_lm_scale_2.3_attention_scale_2.3	7.42
+ngram_lm_scale_1.0_attention_scale_0.3	7.44
+ngram_lm_scale_2.2_attention_scale_2.1	7.53
+ngram_lm_scale_2.3_attention_scale_2.2	7.63
+ngram_lm_scale_1.3_attention_scale_0.7	7.64
+ngram_lm_scale_1.7_attention_scale_1.3	7.64
+ngram_lm_scale_1.5_attention_scale_1.0	7.65
+ngram_lm_scale_2.5_attention_scale_2.5	7.65
+ngram_lm_scale_2.1_attention_scale_1.9	7.66
+ngram_lm_scale_2.2_attention_scale_2.0	7.76
+ngram_lm_scale_2.0_attention_scale_1.7	7.81
+ngram_lm_scale_2.3_attention_scale_2.1	7.91
+ngram_lm_scale_0.9_attention_scale_0.1	7.92
+ngram_lm_scale_1.2_attention_scale_0.5	7.96
+ngram_lm_scale_1.9_attention_scale_1.5	7.99
+ngram_lm_scale_1.7_attention_scale_1.2	8.02
+ngram_lm_scale_2.2_attention_scale_1.9	8.09
+ngram_lm_scale_0.9_attention_scale_0.08	8.11
+ngram_lm_scale_1.5_attention_scale_0.9	8.12
+ngram_lm_scale_2.5_attention_scale_2.3	8.17
+ngram_lm_scale_3.0_attention_scale_3.0	8.18
+ngram_lm_scale_2.3_attention_scale_2.0	8.23
+ngram_lm_scale_1.3_attention_scale_0.6	8.26
+ngram_lm_scale_2.1_attention_scale_1.7	8.29
+ngram_lm_scale_0.9_attention_scale_0.05	8.41
+ngram_lm_scale_1.1_attention_scale_0.3	8.43
+ngram_lm_scale_2.5_attention_scale_2.2	8.49
+ngram_lm_scale_2.0_attention_scale_1.5	8.58
+ngram_lm_scale_1.7_attention_scale_1.1	8.59
+ngram_lm_scale_2.3_attention_scale_1.9	8.59
+ngram_lm_scale_0.9_attention_scale_0.01	8.83
+ngram_lm_scale_2.5_attention_scale_2.1	8.83
+ngram_lm_scale_2.2_attention_scale_1.7	8.87
+ngram_lm_scale_4.0_attention_scale_4.0	8.91
+ngram_lm_scale_1.9_attention_scale_1.3	8.92
+ngram_lm_scale_1.3_attention_scale_0.5	9.01
+ngram_lm_scale_1.7_attention_scale_1.0	9.14
+ngram_lm_scale_2.1_attention_scale_1.5	9.19
+ngram_lm_scale_2.5_attention_scale_2.0	9.19
+ngram_lm_scale_1.0_attention_scale_0.1	9.23
+ngram_lm_scale_1.0_attention_scale_0.08	9.4
+ngram_lm_scale_2.3_attention_scale_1.7	9.44
+ngram_lm_scale_1.5_attention_scale_0.7	9.48
+ngram_lm_scale_1.9_attention_scale_1.2	9.48
+ngram_lm_scale_5.0_attention_scale_5.0	9.51
+ngram_lm_scale_2.0_attention_scale_1.3	9.63
+ngram_lm_scale_1.2_attention_scale_0.3	9.69
+ngram_lm_scale_2.5_attention_scale_1.9	9.69
+ngram_lm_scale_3.0_attention_scale_2.5	9.71
+ngram_lm_scale_1.0_attention_scale_0.05	9.74
+ngram_lm_scale_1.7_attention_scale_0.9	9.86
+ngram_lm_scale_2.2_attention_scale_1.5	9.9
+ngram_lm_scale_1.9_attention_scale_1.1	10.18
+ngram_lm_scale_1.0_attention_scale_0.01	10.22
+ngram_lm_scale_2.0_attention_scale_1.2	10.35
+ngram_lm_scale_1.5_attention_scale_0.6	10.39
+ngram_lm_scale_2.1_attention_scale_1.3	10.42
+ngram_lm_scale_1.1_attention_scale_0.1	10.54
+ngram_lm_scale_2.3_attention_scale_1.5	10.67
+ngram_lm_scale_1.1_attention_scale_0.08	10.77
+ngram_lm_scale_3.0_attention_scale_2.3	10.77
+ngram_lm_scale_2.5_attention_scale_1.7	10.85
+ngram_lm_scale_1.9_attention_scale_1.0	10.9
+ngram_lm_scale_1.3_attention_scale_0.3	10.92
+ngram_lm_scale_2.0_attention_scale_1.1	11.01
+ngram_lm_scale_1.1_attention_scale_0.05	11.09
+ngram_lm_scale_2.1_attention_scale_1.2	11.09
+ngram_lm_scale_2.2_attention_scale_1.3	11.19
+ngram_lm_scale_1.5_attention_scale_0.5	11.22
+ngram_lm_scale_3.0_attention_scale_2.2	11.23
+ngram_lm_scale_1.7_attention_scale_0.7	11.48
+ngram_lm_scale_1.1_attention_scale_0.01	11.56
+ngram_lm_scale_1.9_attention_scale_0.9	11.59
+ngram_lm_scale_2.0_attention_scale_1.0	11.65
+ngram_lm_scale_3.0_attention_scale_2.1	11.68
+ngram_lm_scale_2.1_attention_scale_1.1	11.71
+ngram_lm_scale_1.2_attention_scale_0.1	11.75
+ngram_lm_scale_2.2_attention_scale_1.2	11.75
+ngram_lm_scale_2.3_attention_scale_1.3	11.8
+ngram_lm_scale_2.5_attention_scale_1.5	11.92
+ngram_lm_scale_1.2_attention_scale_0.08	11.96
+ngram_lm_scale_3.0_attention_scale_2.0	12.05
+ngram_lm_scale_1.7_attention_scale_0.6	12.14
+ngram_lm_scale_1.2_attention_scale_0.05	12.24
+ngram_lm_scale_4.0_attention_scale_3.0	12.26
+ngram_lm_scale_2.0_attention_scale_0.9	12.28
+ngram_lm_scale_2.1_attention_scale_1.0	12.3
+ngram_lm_scale_2.2_attention_scale_1.1	12.32
+ngram_lm_scale_2.3_attention_scale_1.2	12.33
+ngram_lm_scale_3.0_attention_scale_1.9	12.41
+ngram_lm_scale_5.0_attention_scale_4.0	12.42
+ngram_lm_scale_1.2_attention_scale_0.01	12.54
+ngram_lm_scale_1.3_attention_scale_0.1	12.66
+ngram_lm_scale_1.5_attention_scale_0.3	12.66
+ngram_lm_scale_1.7_attention_scale_0.5	12.68
+ngram_lm_scale_1.9_attention_scale_0.7	12.74
+ngram_lm_scale_2.1_attention_scale_0.9	12.75
+ngram_lm_scale_2.2_attention_scale_1.0	12.76
+ngram_lm_scale_2.3_attention_scale_1.1	12.76
+ngram_lm_scale_2.5_attention_scale_1.3	12.76
+ngram_lm_scale_1.3_attention_scale_0.08	12.77
+ngram_lm_scale_1.3_attention_scale_0.05	12.91
+ngram_lm_scale_3.0_attention_scale_1.7	12.91
+ngram_lm_scale_2.5_attention_scale_1.2	13.0
+ngram_lm_scale_2.3_attention_scale_1.0	13.03
+ngram_lm_scale_2.2_attention_scale_0.9	13.04
+ngram_lm_scale_4.0_attention_scale_2.5	13.07
+ngram_lm_scale_2.0_attention_scale_0.7	13.08
+ngram_lm_scale_1.9_attention_scale_0.6	13.1
+ngram_lm_scale_1.3_attention_scale_0.01	13.18
+ngram_lm_scale_2.5_attention_scale_1.1	13.2
+ngram_lm_scale_3.0_attention_scale_1.5	13.24
+ngram_lm_scale_2.3_attention_scale_0.9	13.26
+ngram_lm_scale_2.1_attention_scale_0.7	13.31
+ngram_lm_scale_4.0_attention_scale_2.3	13.31
+ngram_lm_scale_2.0_attention_scale_0.6	13.36
+ngram_lm_scale_1.9_attention_scale_0.5	13.37
+ngram_lm_scale_2.5_attention_scale_1.0	13.38
+ngram_lm_scale_4.0_attention_scale_2.2	13.4
+ngram_lm_scale_5.0_attention_scale_3.0	13.41
+ngram_lm_scale_1.7_attention_scale_0.3	13.45
+ngram_lm_scale_2.2_attention_scale_0.7	13.5
+ngram_lm_scale_4.0_attention_scale_2.1	13.5
+ngram_lm_scale_3.0_attention_scale_1.3	13.51
+ngram_lm_scale_2.1_attention_scale_0.6	13.52
+ngram_lm_scale_1.5_attention_scale_0.1	13.54
+ngram_lm_scale_2.0_attention_scale_0.5	13.56
+ngram_lm_scale_2.5_attention_scale_0.9	13.56
+ngram_lm_scale_4.0_attention_scale_2.0	13.56
+ngram_lm_scale_1.5_attention_scale_0.08	13.58
+ngram_lm_scale_2.3_attention_scale_0.7	13.64
+ngram_lm_scale_3.0_attention_scale_1.2	13.64
+ngram_lm_scale_4.0_attention_scale_1.9	13.64
+ngram_lm_scale_1.5_attention_scale_0.05	13.66
+ngram_lm_scale_2.2_attention_scale_0.6	13.67
+ngram_lm_scale_2.1_attention_scale_0.5	13.69
+ngram_lm_scale_5.0_attention_scale_2.5	13.74
+ngram_lm_scale_1.5_attention_scale_0.01	13.76
+ngram_lm_scale_2.3_attention_scale_0.6	13.77
+ngram_lm_scale_3.0_attention_scale_1.1	13.77
+ngram_lm_scale_1.9_attention_scale_0.3	13.8
+ngram_lm_scale_2.2_attention_scale_0.5	13.8
+ngram_lm_scale_2.5_attention_scale_0.7	13.84
+ngram_lm_scale_4.0_attention_scale_1.7	13.86
+ngram_lm_scale_5.0_attention_scale_2.3	13.88
+ngram_lm_scale_1.7_attention_scale_0.1	13.89
+ngram_lm_scale_2.0_attention_scale_0.3	13.89
+ngram_lm_scale_1.7_attention_scale_0.08	13.9
+ngram_lm_scale_3.0_attention_scale_1.0	13.9
+ngram_lm_scale_2.3_attention_scale_0.5	13.92
+ngram_lm_scale_2.5_attention_scale_0.6	13.94
+ngram_lm_scale_5.0_attention_scale_2.2	13.95
+ngram_lm_scale_1.7_attention_scale_0.05	13.96
+ngram_lm_scale_3.0_attention_scale_0.9	13.97
+ngram_lm_scale_2.1_attention_scale_0.3	13.98
+ngram_lm_scale_4.0_attention_scale_1.5	13.99
+ngram_lm_scale_5.0_attention_scale_2.1	14.0
+ngram_lm_scale_1.7_attention_scale_0.01	14.01
+ngram_lm_scale_2.5_attention_scale_0.5	14.04
+ngram_lm_scale_2.2_attention_scale_0.3	14.05
+ngram_lm_scale_5.0_attention_scale_2.0	14.06
+ngram_lm_scale_1.9_attention_scale_0.1	14.08
+ngram_lm_scale_2.3_attention_scale_0.3	14.09
+ngram_lm_scale_5.0_attention_scale_1.9	14.09
+ngram_lm_scale_1.9_attention_scale_0.08	14.1
+ngram_lm_scale_3.0_attention_scale_0.7	14.1
+ngram_lm_scale_4.0_attention_scale_1.3	14.11
+ngram_lm_scale_1.9_attention_scale_0.05	14.13
+ngram_lm_scale_2.0_attention_scale_0.1	14.13
+ngram_lm_scale_2.0_attention_scale_0.08	14.14
+ngram_lm_scale_1.9_attention_scale_0.01	14.15
+ngram_lm_scale_4.0_attention_scale_1.2	14.15
+ngram_lm_scale_2.0_attention_scale_0.05	14.16
+ngram_lm_scale_5.0_attention_scale_1.7	14.16
+ngram_lm_scale_2.5_attention_scale_0.3	14.17
+ngram_lm_scale_3.0_attention_scale_0.6	14.17
+ngram_lm_scale_4.0_attention_scale_1.1	14.17
+ngram_lm_scale_2.1_attention_scale_0.1	14.18
+ngram_lm_scale_2.0_attention_scale_0.01	14.2
+ngram_lm_scale_2.1_attention_scale_0.08	14.2
+ngram_lm_scale_4.0_attention_scale_1.0	14.22
+ngram_lm_scale_2.2_attention_scale_0.1	14.23
+ngram_lm_scale_3.0_attention_scale_0.5	14.23
+ngram_lm_scale_5.0_attention_scale_1.5	14.23
+ngram_lm_scale_2.1_attention_scale_0.05	14.24
+ngram_lm_scale_2.2_attention_scale_0.08	14.25
+ngram_lm_scale_2.1_attention_scale_0.01	14.26
+ngram_lm_scale_2.2_attention_scale_0.05	14.28
+ngram_lm_scale_2.3_attention_scale_0.1	14.28
+ngram_lm_scale_4.0_attention_scale_0.9	14.28
+ngram_lm_scale_5.0_attention_scale_1.3	14.28
+ngram_lm_scale_2.3_attention_scale_0.08	14.29
+ngram_lm_scale_2.2_attention_scale_0.01	14.31
+ngram_lm_scale_2.3_attention_scale_0.05	14.31
+ngram_lm_scale_5.0_attention_scale_1.2	14.31
+ngram_lm_scale_2.3_attention_scale_0.01	14.33
+ngram_lm_scale_2.5_attention_scale_0.1	14.34
+ngram_lm_scale_3.0_attention_scale_0.3	14.34
+ngram_lm_scale_4.0_attention_scale_0.7	14.34
+ngram_lm_scale_2.5_attention_scale_0.08	14.35
+ngram_lm_scale_5.0_attention_scale_1.1	14.35
+ngram_lm_scale_2.5_attention_scale_0.05	14.36
+ngram_lm_scale_4.0_attention_scale_0.6	14.36
+ngram_lm_scale_5.0_attention_scale_1.0	14.36
+ngram_lm_scale_2.5_attention_scale_0.01	14.38
+ngram_lm_scale_5.0_attention_scale_0.9	14.38
+ngram_lm_scale_3.0_attention_scale_0.08	14.39
+ngram_lm_scale_3.0_attention_scale_0.1	14.39
+ngram_lm_scale_4.0_attention_scale_0.5	14.39
+ngram_lm_scale_3.0_attention_scale_0.05	14.41
+ngram_lm_scale_5.0_attention_scale_0.7	14.41
+ngram_lm_scale_3.0_attention_scale_0.01	14.42
+ngram_lm_scale_4.0_attention_scale_0.3	14.42
+ngram_lm_scale_5.0_attention_scale_0.6	14.42
+ngram_lm_scale_5.0_attention_scale_0.5	14.43
+ngram_lm_scale_4.0_attention_scale_0.08	14.45
+ngram_lm_scale_4.0_attention_scale_0.1	14.45
+ngram_lm_scale_5.0_attention_scale_0.3	14.46
+ngram_lm_scale_4.0_attention_scale_0.01	14.47
+ngram_lm_scale_4.0_attention_scale_0.05	14.47
+ngram_lm_scale_5.0_attention_scale_0.05	14.48
+ngram_lm_scale_5.0_attention_scale_0.08	14.48
+ngram_lm_scale_5.0_attention_scale_0.1	14.48
+ngram_lm_scale_5.0_attention_scale_0.01	14.5
+
+2023-03-10 16:36:32,676 INFO [decode.py:882] Done!

log/ctc_decoding/log-decode-2023-03-09-01-20-40

@@ -0,0 +1,74 @@
+2023-03-09 01:20:40,285 INFO [decode.py:657] Decoding started
+2023-03-09 01:20:40,285 INFO [decode.py:658] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'surt', 'icefall-git-sha1': 'e9931b7-dirty', 'icefall-git-date': 'Fri Mar 3 16:27:17 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r8n03', 'IP address': '10.1.8.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 21, 'iter': 0, 'avg': 5, 'use_averaged_model': True, 'method': 'ctc-decoding', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc_att/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'num_decoder_layers': 6, 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-09 01:20:40,464 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-09 01:20:40,857 INFO [decode.py:669] device: cuda:0
+2023-03-09 01:20:45,992 INFO [decode.py:757] About to create model
+2023-03-09 01:20:46,623 INFO [zipformer.py:178] 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.
+2023-03-09 01:20:46,709 INFO [decode.py:824] Calculating the averaged model over epoch range from 16 (excluded) to 21
+2023-03-09 01:21:10,542 INFO [decode.py:840] Number of model parameters: 86083707
+2023-03-09 01:21:10,542 INFO [asr_datamodule.py:440] About to get dev-clean cuts
+2023-03-09 01:21:10,609 INFO [asr_datamodule.py:454] About to get dev-other cuts
+2023-03-09 01:21:10,610 INFO [asr_datamodule.py:468] About to get test-clean cuts
+2023-03-09 01:21:10,610 INFO [asr_datamodule.py:482] About to get test-other cuts
+2023-03-09 01:21:12,241 INFO [decode.py:595] batch 0/?, cuts processed until now is 16
+2023-03-09 01:22:03,615 INFO [decode.py:595] batch 100/?, cuts processed until now is 2335
+2023-03-09 01:22:10,131 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-dev-clean-ctc-decoding.txt
+2023-03-09 01:22:10,200 INFO [utils.py:538] [dev-clean-ctc-decoding] %WER 2.48% [1350 / 54402, 121 ins, 114 del, 1115 sub ]
+2023-03-09 01:22:10,350 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-dev-clean-ctc-decoding.txt
+2023-03-09 01:22:10,351 INFO [decode.py:641] 
+For dev-clean, WER of different settings are:
+ctc-decoding	2.48	best for dev-clean
+
+2023-03-09 01:22:11,351 INFO [decode.py:595] batch 0/?, cuts processed until now is 18
+2023-03-09 01:22:13,439 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.2532, 2.9490, 3.3925, 4.3052, 3.8906, 3.8377, 2.8715, 2.3639],
+       device='cuda:0'), covar=tensor([0.0741, 0.2031, 0.0892, 0.0517, 0.0864, 0.0483, 0.1706, 0.2449],
+       device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0208, 0.0181, 0.0204, 0.0209, 0.0165, 0.0193, 0.0180],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-09 01:22:41,747 INFO [zipformer.py:1447] attn_weights_entropy = tensor([5.9773, 6.1870, 5.7736, 5.9844, 5.8950, 5.6403, 5.6171, 5.5847],
+       device='cuda:0'), covar=tensor([0.0838, 0.0432, 0.0575, 0.0440, 0.0393, 0.0988, 0.1716, 0.1633],
+       device='cuda:0'), in_proj_covar=tensor([0.0494, 0.0577, 0.0433, 0.0428, 0.0406, 0.0445, 0.0587, 0.0504],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0003],
+       device='cuda:0')
+2023-03-09 01:22:53,505 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.2329, 4.1956, 4.1419, 4.0666, 4.6500, 4.2362, 4.0969, 2.4006],
+       device='cuda:0'), covar=tensor([0.0272, 0.0398, 0.0414, 0.0302, 0.0936, 0.0228, 0.0368, 0.1957],
+       device='cuda:0'), in_proj_covar=tensor([0.0140, 0.0161, 0.0165, 0.0181, 0.0356, 0.0137, 0.0152, 0.0208],
+       device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-09 01:22:55,821 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.8753, 4.1602, 3.9796, 3.9890, 4.1755, 3.9942, 3.1298, 4.0274],
+       device='cuda:0'), covar=tensor([0.0157, 0.0136, 0.0176, 0.0131, 0.0120, 0.0147, 0.0660, 0.0244],
+       device='cuda:0'), in_proj_covar=tensor([0.0086, 0.0082, 0.0104, 0.0064, 0.0069, 0.0081, 0.0099, 0.0103],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002, 0.0003, 0.0003],
+       device='cuda:0')
+2023-03-09 01:23:01,448 INFO [decode.py:595] batch 100/?, cuts processed until now is 2625
+2023-03-09 01:23:05,727 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-dev-other-ctc-decoding.txt
+2023-03-09 01:23:05,787 INFO [utils.py:538] [dev-other-ctc-decoding] %WER 6.36% [3238 / 50948, 275 ins, 255 del, 2708 sub ]
+2023-03-09 01:23:05,928 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-dev-other-ctc-decoding.txt
+2023-03-09 01:23:05,929 INFO [decode.py:641] 
+For dev-other, WER of different settings are:
+ctc-decoding	6.36	best for dev-other
+
+2023-03-09 01:23:07,031 INFO [decode.py:595] batch 0/?, cuts processed until now is 14
+2023-03-09 01:23:59,819 INFO [decode.py:595] batch 100/?, cuts processed until now is 2293
+2023-03-09 01:24:03,002 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.2997, 4.3643, 4.0564, 2.7896, 4.1000, 4.1336, 3.6739, 2.6991],
+       device='cuda:0'), covar=tensor([0.0108, 0.0111, 0.0289, 0.0952, 0.0129, 0.0232, 0.0338, 0.1326],
+       device='cuda:0'), in_proj_covar=tensor([0.0069, 0.0096, 0.0096, 0.0107, 0.0079, 0.0105, 0.0094, 0.0101],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0003, 0.0003, 0.0004, 0.0003, 0.0004, 0.0004, 0.0004],
+       device='cuda:0')
+2023-03-09 01:24:06,967 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-clean-ctc-decoding.txt
+2023-03-09 01:24:07,032 INFO [utils.py:538] [test-clean-ctc-decoding] %WER 2.68% [1410 / 52576, 136 ins, 114 del, 1160 sub ]
+2023-03-09 01:24:07,178 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-clean-ctc-decoding.txt
+2023-03-09 01:24:07,179 INFO [decode.py:641] 
+For test-clean, WER of different settings are:
+ctc-decoding	2.68	best for test-clean
+
+2023-03-09 01:24:08,272 INFO [decode.py:595] batch 0/?, cuts processed until now is 17
+2023-03-09 01:24:59,720 INFO [decode.py:595] batch 100/?, cuts processed until now is 2560
+2023-03-09 01:25:06,602 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-other-ctc-decoding.txt
+2023-03-09 01:25:06,673 INFO [utils.py:538] [test-other-ctc-decoding] %WER 6.42% [3362 / 52343, 305 ins, 277 del, 2780 sub ]
+2023-03-09 01:25:06,823 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-other-ctc-decoding.txt
+2023-03-09 01:25:06,824 INFO [decode.py:641] 
+For test-other, WER of different settings are:
+ctc-decoding	6.42	best for test-other
+
+2023-03-09 01:25:06,824 INFO [decode.py:901] Done!

log/ctc_decoding/log-decode-2023-03-09-04-23-24

@@ -0,0 +1,89 @@
+2023-03-09 04:23:24,501 INFO [decode.py:657] Decoding started
+2023-03-09 04:23:24,501 INFO [decode.py:658] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'surt', 'icefall-git-sha1': 'e9931b7-dirty', 'icefall-git-date': 'Fri Mar 3 16:27:17 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r8n03', 'IP address': '10.1.8.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 23, 'iter': 0, 'avg': 5, 'use_averaged_model': True, 'method': 'ctc-decoding', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc_att/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'num_decoder_layers': 6, 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-09 04:23:24,610 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-09 04:23:24,725 INFO [decode.py:669] device: cuda:0
+2023-03-09 04:23:29,169 INFO [decode.py:757] About to create model
+2023-03-09 04:23:29,625 INFO [zipformer.py:178] 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.
+2023-03-09 04:23:29,683 INFO [decode.py:824] Calculating the averaged model over epoch range from 18 (excluded) to 23
+2023-03-09 04:23:59,034 INFO [decode.py:840] Number of model parameters: 86083707
+2023-03-09 04:23:59,035 INFO [asr_datamodule.py:440] About to get dev-clean cuts
+2023-03-09 04:23:59,102 INFO [asr_datamodule.py:454] About to get dev-other cuts
+2023-03-09 04:23:59,103 INFO [asr_datamodule.py:468] About to get test-clean cuts
+2023-03-09 04:23:59,104 INFO [asr_datamodule.py:482] About to get test-other cuts
+2023-03-09 04:24:00,793 INFO [decode.py:595] batch 0/?, cuts processed until now is 16
+2023-03-09 04:24:20,887 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.5097, 2.4544, 3.7387, 3.2820, 2.6631, 3.5698, 3.6508, 3.5118],
+       device='cuda:0'), covar=tensor([0.0233, 0.1487, 0.0209, 0.0739, 0.1421, 0.0285, 0.0247, 0.0240],
+       device='cuda:0'), in_proj_covar=tensor([0.0183, 0.0236, 0.0176, 0.0306, 0.0256, 0.0204, 0.0164, 0.0194],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-09 04:24:23,072 INFO [zipformer.py:1447] attn_weights_entropy = tensor([2.5664, 2.2849, 2.5186, 3.0238, 2.7956, 2.9528, 2.4076, 2.0466],
+       device='cuda:0'), covar=tensor([0.0793, 0.2076, 0.0952, 0.0847, 0.0986, 0.0889, 0.1785, 0.2259],
+       device='cuda:0'), in_proj_covar=tensor([0.0173, 0.0209, 0.0183, 0.0211, 0.0217, 0.0170, 0.0196, 0.0182],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0002, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-09 04:24:49,384 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.8208, 5.4436, 5.4218, 4.8623, 5.9272, 4.8507, 5.3783, 3.3480],
+       device='cuda:0'), covar=tensor([0.0181, 0.0106, 0.0117, 0.0305, 0.0505, 0.0170, 0.0146, 0.1411],
+       device='cuda:0'), in_proj_covar=tensor([0.0151, 0.0176, 0.0177, 0.0193, 0.0361, 0.0147, 0.0166, 0.0209],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0003, 0.0001, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-09 04:24:52,404 INFO [decode.py:595] batch 100/?, cuts processed until now is 2335
+2023-03-09 04:24:59,003 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-dev-clean-ctc-decoding.txt
+2023-03-09 04:24:59,072 INFO [utils.py:538] [dev-clean-ctc-decoding] %WER 2.44% [1325 / 54402, 115 ins, 113 del, 1097 sub ]
+2023-03-09 04:24:59,222 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-dev-clean-ctc-decoding.txt
+2023-03-09 04:24:59,223 INFO [decode.py:641] 
+For dev-clean, WER of different settings are:
+ctc-decoding	2.44	best for dev-clean
+
+2023-03-09 04:25:00,276 INFO [decode.py:595] batch 0/?, cuts processed until now is 18
+2023-03-09 04:25:20,446 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.1746, 3.6242, 3.1356, 3.6365, 2.3967, 3.4838, 2.4663, 1.7018],
+       device='cuda:0'), covar=tensor([0.0603, 0.0347, 0.0963, 0.0240, 0.1707, 0.0248, 0.1535, 0.1882],
+       device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0154, 0.0252, 0.0147, 0.0215, 0.0134, 0.0225, 0.0198],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-09 04:25:50,986 INFO [decode.py:595] batch 100/?, cuts processed until now is 2625
+2023-03-09 04:25:55,314 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-dev-other-ctc-decoding.txt
+2023-03-09 04:25:55,383 INFO [utils.py:538] [dev-other-ctc-decoding] %WER 6.20% [3157 / 50948, 254 ins, 245 del, 2658 sub ]
+2023-03-09 04:25:55,534 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-dev-other-ctc-decoding.txt
+2023-03-09 04:25:55,536 INFO [decode.py:641] 
+For dev-other, WER of different settings are:
+ctc-decoding	6.2	best for dev-other
+
+2023-03-09 04:25:56,598 INFO [decode.py:595] batch 0/?, cuts processed until now is 14
+2023-03-09 04:26:01,569 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.4120, 5.0760, 5.2997, 5.1498, 3.5879, 4.8010, 3.7668, 2.6052],
+       device='cuda:0'), covar=tensor([0.0291, 0.0159, 0.0397, 0.0115, 0.1199, 0.0151, 0.1023, 0.1435],
+       device='cuda:0'), in_proj_covar=tensor([0.0182, 0.0154, 0.0252, 0.0147, 0.0215, 0.0134, 0.0225, 0.0198],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-09 04:26:16,340 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.1817, 3.9030, 3.9128, 3.6124, 3.7950, 3.8524, 3.9487, 3.2508],
+       device='cuda:0'), covar=tensor([0.0674, 0.0929, 0.1161, 0.1590, 0.1837, 0.0824, 0.0429, 0.2261],
+       device='cuda:0'), in_proj_covar=tensor([0.0162, 0.0177, 0.0189, 0.0245, 0.0149, 0.0250, 0.0169, 0.0209],
+       device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0001, 0.0002, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-09 04:26:49,057 INFO [decode.py:595] batch 100/?, cuts processed until now is 2293
+2023-03-09 04:26:56,021 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-clean-ctc-decoding.txt
+2023-03-09 04:26:56,086 INFO [utils.py:538] [test-clean-ctc-decoding] %WER 2.58% [1354 / 52576, 134 ins, 101 del, 1119 sub ]
+2023-03-09 04:26:56,224 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-clean-ctc-decoding.txt
+2023-03-09 04:26:56,225 INFO [decode.py:641] 
+For test-clean, WER of different settings are:
+ctc-decoding	2.58	best for test-clean
+
+2023-03-09 04:26:57,298 INFO [decode.py:595] batch 0/?, cuts processed until now is 17
+2023-03-09 04:27:36,606 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.5775, 2.5343, 3.2675, 2.5832, 3.1080, 3.7384, 3.6521, 2.7340],
+       device='cuda:0'), covar=tensor([0.0503, 0.1965, 0.1334, 0.1462, 0.1432, 0.1233, 0.0734, 0.1428],
+       device='cuda:0'), in_proj_covar=tensor([0.0237, 0.0237, 0.0272, 0.0212, 0.0257, 0.0360, 0.0253, 0.0225],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003],
+       device='cuda:0')
+2023-03-09 04:27:44,448 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.1154, 4.9219, 4.8740, 2.4294, 2.1290, 3.0700, 2.4395, 3.8819],
+       device='cuda:0'), covar=tensor([0.0583, 0.0207, 0.0217, 0.5031, 0.5416, 0.2327, 0.3603, 0.1347],
+       device='cuda:0'), in_proj_covar=tensor([0.0346, 0.0263, 0.0258, 0.0238, 0.0335, 0.0328, 0.0247, 0.0356],
+       device='cuda:0'), out_proj_covar=tensor([1.4699e-04, 9.7060e-05, 1.0949e-04, 1.0262e-04, 1.4174e-04, 1.2850e-04,
+        9.9139e-05, 1.4624e-04], device='cuda:0')
+2023-03-09 04:27:48,477 INFO [decode.py:595] batch 100/?, cuts processed until now is 2560
+2023-03-09 04:27:55,425 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-other-ctc-decoding.txt
+2023-03-09 04:27:55,496 INFO [utils.py:538] [test-other-ctc-decoding] %WER 6.25% [3271 / 52343, 306 ins, 273 del, 2692 sub ]
+2023-03-09 04:27:55,651 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-other-ctc-decoding.txt
+2023-03-09 04:27:55,652 INFO [decode.py:641] 
+For test-other, WER of different settings are:
+ctc-decoding	6.25	best for test-other
+
+2023-03-09 04:27:55,652 INFO [decode.py:901] Done!

log/ctc_decoding/log-decode-2023-03-09-16-35-15

@@ -0,0 +1,20 @@
+2023-03-09 16:35:15,181 INFO [decode.py:657] Decoding started
+2023-03-09 16:35:15,181 INFO [decode.py:658] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'surt', 'icefall-git-sha1': 'e9931b7-dirty', 'icefall-git-date': 'Fri Mar 3 16:27:17 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'method': 'ctc-decoding', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc_att/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500_new'), 'num_decoder_layers': 6, 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-09 16:35:15,311 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500_new/Linv.pt
+2023-03-09 16:35:15,402 INFO [decode.py:669] device: cuda:0
+2023-03-09 16:35:19,651 INFO [decode.py:757] About to create model
+2023-03-09 16:35:20,131 INFO [zipformer.py:178] 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.
+2023-03-09 16:35:20,189 INFO [decode.py:824] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2023-03-09 16:35:40,878 INFO [decode.py:840] Number of model parameters: 86083707
+2023-03-09 16:35:40,879 INFO [asr_datamodule.py:468] About to get test-clean cuts
+2023-03-09 16:35:41,777 INFO [asr_datamodule.py:482] About to get test-other cuts
+2023-03-09 16:35:43,833 INFO [decode.py:595] batch 0/?, cuts processed until now is 14
+2023-03-09 16:36:35,032 INFO [decode.py:595] batch 100/?, cuts processed until now is 2293
+2023-03-09 16:36:41,874 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-clean-ctc-decoding.txt
+2023-03-09 16:36:41,975 INFO [utils.py:538] [test-clean-ctc-decoding] %WER 98.49% [51780 / 52576, 6107 ins, 4221 del, 41452 sub ]
+2023-03-09 16:36:42,252 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-clean-ctc-decoding.txt
+2023-03-09 16:36:42,253 INFO [decode.py:641] 
+For test-clean, WER of different settings are:
+ctc-decoding	98.49	best for test-clean
+
+2023-03-09 16:36:43,652 INFO [decode.py:595] batch 0/?, cuts processed until now is 17

log/ctc_decoding/log-decode-2023-03-09-16-37-47

@@ -0,0 +1,54 @@
+2023-03-09 16:37:47,751 INFO [decode.py:657] Decoding started
+2023-03-09 16:37:47,752 INFO [decode.py:658] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'surt', 'icefall-git-sha1': 'e9931b7-dirty', 'icefall-git-date': 'Fri Mar 3 16:27:17 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'method': 'ctc-decoding', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc_att/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'num_decoder_layers': 6, 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-09 16:37:48,092 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-09 16:37:48,491 INFO [decode.py:669] device: cuda:0
+2023-03-09 16:37:52,667 INFO [decode.py:757] About to create model
+2023-03-09 16:37:53,158 INFO [zipformer.py:178] 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.
+2023-03-09 16:37:53,216 INFO [decode.py:824] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2023-03-09 16:37:56,231 INFO [decode.py:840] Number of model parameters: 86083707
+2023-03-09 16:37:56,232 INFO [asr_datamodule.py:468] About to get test-clean cuts
+2023-03-09 16:37:56,325 INFO [asr_datamodule.py:482] About to get test-other cuts
+2023-03-09 16:37:57,801 INFO [decode.py:595] batch 0/?, cuts processed until now is 14
+2023-03-09 16:38:21,953 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.8734, 4.8359, 4.7836, 2.2005, 1.9394, 3.1305, 2.3672, 3.7309],
+       device='cuda:0'), covar=tensor([0.0726, 0.0255, 0.0251, 0.5680, 0.6138, 0.2288, 0.4314, 0.1432],
+       device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0282, 0.0266, 0.0244, 0.0333, 0.0327, 0.0254, 0.0360],
+       device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001],
+       device='cuda:0')
+2023-03-09 16:38:36,815 INFO [zipformer.py:1447] attn_weights_entropy = tensor([2.9556, 3.3917, 2.8934, 3.1775, 3.5526, 3.3754, 2.8833, 3.5053],
+       device='cuda:0'), covar=tensor([0.0973, 0.0680, 0.1121, 0.0759, 0.0776, 0.0769, 0.0865, 0.0496],
+       device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0218, 0.0225, 0.0201, 0.0282, 0.0242, 0.0198, 0.0288],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004],
+       device='cuda:0')
+2023-03-09 16:38:42,569 INFO [zipformer.py:1447] attn_weights_entropy = tensor([3.9828, 4.9623, 4.8838, 2.2937, 1.9373, 3.3023, 2.4284, 3.7700],
+       device='cuda:0'), covar=tensor([0.0650, 0.0265, 0.0272, 0.5707, 0.6273, 0.2079, 0.4437, 0.1474],
+       device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0282, 0.0266, 0.0244, 0.0333, 0.0327, 0.0254, 0.0360],
+       device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001],
+       device='cuda:0')
+2023-03-09 16:38:47,820 INFO [decode.py:595] batch 100/?, cuts processed until now is 2293
+2023-03-09 16:38:54,624 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-clean-ctc-decoding.txt
+2023-03-09 16:38:54,683 INFO [utils.py:538] [test-clean-ctc-decoding] %WER 2.50% [1316 / 52576, 140 ins, 94 del, 1082 sub ]
+2023-03-09 16:38:54,821 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-clean-ctc-decoding.txt
+2023-03-09 16:38:54,822 INFO [decode.py:641] 
+For test-clean, WER of different settings are:
+ctc-decoding	2.5	best for test-clean
+
+2023-03-09 16:38:55,805 INFO [decode.py:595] batch 0/?, cuts processed until now is 17
+2023-03-09 16:39:20,752 INFO [zipformer.py:1447] attn_weights_entropy = tensor([4.9714, 5.3511, 4.9382, 5.1933, 5.0431, 4.6623, 4.8124, 4.6678],
+       device='cuda:0'), covar=tensor([0.1602, 0.0753, 0.0770, 0.0644, 0.0709, 0.1634, 0.2040, 0.2010],
+       device='cuda:0'), in_proj_covar=tensor([0.0535, 0.0618, 0.0469, 0.0458, 0.0433, 0.0468, 0.0622, 0.0530],
+       device='cuda:0'), out_proj_covar=tensor([0.0004, 0.0004, 0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004],
+       device='cuda:0')
+2023-03-09 16:39:36,652 INFO [zipformer.py:1447] attn_weights_entropy = tensor([2.9433, 3.5919, 3.1227, 3.3415, 3.7766, 3.5015, 2.9274, 3.8561],
+       device='cuda:0'), covar=tensor([0.0918, 0.0514, 0.1023, 0.0648, 0.0696, 0.0717, 0.0854, 0.0489],
+       device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0218, 0.0225, 0.0201, 0.0282, 0.0242, 0.0198, 0.0288],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004],
+       device='cuda:0')
+2023-03-09 16:39:44,942 INFO [decode.py:595] batch 100/?, cuts processed until now is 2560
+2023-03-09 16:39:51,805 INFO [decode.py:615] The transcripts are stored in zipformer_ctc_att/exp/v0/recogs-test-other-ctc-decoding.txt
+2023-03-09 16:39:51,868 INFO [utils.py:538] [test-other-ctc-decoding] %WER 5.86% [3068 / 52343, 278 ins, 243 del, 2547 sub ]
+2023-03-09 16:39:52,010 INFO [decode.py:627] Wrote detailed error stats to zipformer_ctc_att/exp/v0/errs-test-other-ctc-decoding.txt
+2023-03-09 16:39:52,011 INFO [decode.py:641] 
+For test-other, WER of different settings are:
+ctc-decoding	5.86	best for test-other
+
+2023-03-09 16:39:52,011 INFO [decode.py:897] Done!

log/whole_lattice_rescoring/log-decode-2023-03-09-16-44-16

@@ -0,0 +1,11 @@
+2023-03-09 16:44:16,315 INFO [decode.py:657] Decoding started
+2023-03-09 16:44:16,315 INFO [decode.py:658] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'surt', 'icefall-git-sha1': 'e9931b7-dirty', 'icefall-git-date': 'Fri Mar 3 16:27:17 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n04', 'IP address': '10.1.7.4'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'method': 'whole-lattice-rescoring', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc_att/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'num_decoder_layers': 6, 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-09 16:44:16,539 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-09 16:44:16,650 INFO [decode.py:669] device: cuda:0
+2023-03-09 16:44:25,215 INFO [decode.py:736] Loading pre-compiled G_4_gram.pt
+2023-03-09 16:44:25,879 INFO [decode.py:757] About to create model
+2023-03-09 16:44:26,383 INFO [zipformer.py:178] 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.
+2023-03-09 16:44:26,451 INFO [decode.py:824] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2023-03-09 16:44:40,859 INFO [decode.py:840] Number of model parameters: 86083707
+2023-03-09 16:44:40,859 INFO [asr_datamodule.py:468] About to get test-clean cuts
+2023-03-09 16:44:40,922 INFO [asr_datamodule.py:482] About to get test-other cuts

log/whole_lattice_rescoring/log-decode-2023-03-10-09-28-13

@@ -0,0 +1,11 @@
+2023-03-10 09:28:13,659 INFO [decode.py:643] Decoding started
+2023-03-10 09:28:13,660 INFO [decode.py:644] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_ctc', 'icefall-git-sha1': '11e21f3-dirty', 'icefall-git-date': 'Thu Mar 9 19:58:30 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 99, 'iter': 0, 'avg': 1, 'use_averaged_model': False, 'method': 'whole-lattice-rescoring', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'num_decoder_layers': 6, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-10 09:28:13,904 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-10 09:28:14,020 INFO [decode.py:655] device: cuda:0
+2023-03-10 09:28:19,411 INFO [decode.py:722] Loading pre-compiled G_4_gram.pt
+2023-03-10 09:28:21,556 INFO [decode.py:743] About to create model
+2023-03-10 09:28:22,044 INFO [zipformer.py:178] 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.
+2023-03-10 09:28:22,105 INFO [checkpoint.py:112] Loading checkpoint from zipformer_ctc/exp/v0/epoch-99.pt
+2023-03-10 09:28:24,568 INFO [decode.py:826] Number of model parameters: 86083707
+2023-03-10 09:28:24,568 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2023-03-10 09:28:24,847 INFO [asr_datamodule.py:450] About to get test-other cuts

log/whole_lattice_rescoring/log-decode-2023-03-10-09-44-58

@@ -0,0 +1,213 @@
+2023-03-10 09:44:58,872 INFO [decode.py:643] Decoding started
+2023-03-10 09:44:58,873 INFO [decode.py:644] {'frame_shift_ms': 10.0, 'allowed_excess_duration_ratio': 0.1, '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': '1.23.3', 'k2-build-type': 'Debug', 'k2-with-cuda': True, 'k2-git-sha1': '3b81ac9686aee539d447bb2085b2cdfc131c7c91', 'k2-git-date': 'Thu Jan 26 20:40:25 2023', 'lhotse-version': '1.9.0.dev+git.97bf4b0.dirty', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'zipformer_ctc', 'icefall-git-sha1': '11e21f3-dirty', 'icefall-git-date': 'Thu Mar 9 19:58:30 2023', 'icefall-path': '/exp/draj/mini_scale_2022/icefall', 'k2-path': '/exp/draj/mini_scale_2022/k2/k2/python/k2/__init__.py', 'lhotse-path': '/exp/draj/mini_scale_2022/lhotse/lhotse/__init__.py', 'hostname': 'r7n03', 'IP address': '10.1.7.3'}, 'beam_size': 10, 'reduction': 'sum', 'use_double_scores': True, 'search_beam': 20, 'output_beam': 8, 'min_active_states': 30, 'max_active_states': 10000, 'epoch': 99, 'iter': 0, 'avg': 1, 'use_averaged_model': False, 'method': 'whole-lattice-rescoring', 'num_paths': 100, 'nbest_scale': 0.5, 'exp_dir': PosixPath('zipformer_ctc/exp/v0'), 'lang_dir': PosixPath('data/lang_bpe_500'), 'lm_dir': PosixPath('data/lm'), 'rnn_lm_exp_dir': 'rnn_lm/exp', 'rnn_lm_epoch': 7, 'rnn_lm_avg': 2, 'rnn_lm_embedding_dim': 2048, 'rnn_lm_hidden_dim': 2048, 'rnn_lm_num_layers': 4, 'rnn_lm_tie_weights': False, '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', 'num_decoder_layers': 6, 'full_libri': True, 'manifest_dir': PosixPath('data/manifests'), 'max_duration': 200.0, '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'}
+2023-03-10 09:44:59,114 INFO [lexicon.py:168] Loading pre-compiled data/lang_bpe_500/Linv.pt
+2023-03-10 09:44:59,229 INFO [decode.py:655] device: cuda:0
+2023-03-10 09:45:04,903 INFO [decode.py:697] Loading G_4_gram.fst.txt
+2023-03-10 09:45:04,904 WARNING [decode.py:698] It may take 8 minutes.
+2023-03-10 09:48:46,802 INFO [decode.py:743] About to create model
+2023-03-10 09:48:47,199 INFO [zipformer.py:178] 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.
+2023-03-10 09:48:47,247 INFO [checkpoint.py:112] Loading checkpoint from zipformer_ctc/exp/v0/epoch-99.pt
+2023-03-10 09:48:47,792 INFO [decode.py:826] Number of model parameters: 86083707
+2023-03-10 09:48:47,792 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2023-03-10 09:48:47,895 INFO [asr_datamodule.py:450] About to get test-other cuts
+2023-03-10 09:48:50,410 INFO [decode.py:581] batch 0/?, cuts processed until now is 14
+2023-03-10 09:48:52,452 INFO [zipformer.py:1455] attn_weights_entropy = tensor([2.4185, 2.8395, 2.4086, 2.6454, 2.8039, 2.8219, 2.4147, 2.4832],
+       device='cuda:0'), covar=tensor([0.1030, 0.0560, 0.1220, 0.0820, 0.0797, 0.0761, 0.0948, 0.0353],
+       device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0218, 0.0225, 0.0201, 0.0282, 0.0242, 0.0198, 0.0288],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004],
+       device='cuda:0')
+2023-03-10 09:50:31,259 INFO [zipformer.py:1455] attn_weights_entropy = tensor([4.1906, 4.5374, 4.0955, 4.3857, 4.7103, 4.4173, 4.2439, 4.9290],
+       device='cuda:0'), covar=tensor([0.0660, 0.0368, 0.0880, 0.0456, 0.0623, 0.0703, 0.0533, 0.0846],
+       device='cuda:0'), in_proj_covar=tensor([0.0201, 0.0218, 0.0225, 0.0201, 0.0282, 0.0242, 0.0198, 0.0288],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004],
+       device='cuda:0')
+2023-03-10 09:50:35,365 INFO [zipformer.py:1455] attn_weights_entropy = tensor([4.4905, 2.8404, 4.9623, 4.1641, 3.0908, 4.2355, 4.7916, 4.6903],
+       device='cuda:0'), covar=tensor([0.0290, 0.1696, 0.0221, 0.0668, 0.1630, 0.0267, 0.0223, 0.0264],
+       device='cuda:0'), in_proj_covar=tensor([0.0209, 0.0240, 0.0202, 0.0315, 0.0260, 0.0221, 0.0192, 0.0219],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-10 09:51:08,487 INFO [decode.py:581] batch 100/?, cuts processed until now is 2293
+2023-03-10 09:51:27,701 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.1.txt
+2023-03-10 09:51:27,766 INFO [utils.py:558] [test-clean-lm_scale_0.1] %WER 2.46% [1291 / 52576, 191 ins, 107 del, 993 sub ]
+2023-03-10 09:51:27,917 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.1.txt
+2023-03-10 09:51:27,941 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.2.txt
+2023-03-10 09:51:28,004 INFO [utils.py:558] [test-clean-lm_scale_0.2] %WER 2.44% [1285 / 52576, 175 ins, 120 del, 990 sub ]
+2023-03-10 09:51:28,157 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.2.txt
+2023-03-10 09:51:28,179 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.3.txt
+2023-03-10 09:51:28,238 INFO [utils.py:558] [test-clean-lm_scale_0.3] %WER 2.46% [1293 / 52576, 158 ins, 146 del, 989 sub ]
+2023-03-10 09:51:28,375 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.3.txt
+2023-03-10 09:51:28,397 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.4.txt
+2023-03-10 09:51:28,453 INFO [utils.py:558] [test-clean-lm_scale_0.4] %WER 2.51% [1321 / 52576, 147 ins, 185 del, 989 sub ]
+2023-03-10 09:51:28,591 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.4.txt
+2023-03-10 09:51:28,611 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.5.txt
+2023-03-10 09:51:28,666 INFO [utils.py:558] [test-clean-lm_scale_0.5] %WER 2.61% [1373 / 52576, 138 ins, 239 del, 996 sub ]
+2023-03-10 09:51:28,803 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.5.txt
+2023-03-10 09:51:28,825 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.6.txt
+2023-03-10 09:51:28,885 INFO [utils.py:558] [test-clean-lm_scale_0.6] %WER 2.80% [1473 / 52576, 125 ins, 351 del, 997 sub ]
+2023-03-10 09:51:29,223 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.6.txt
+2023-03-10 09:51:29,245 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.7.txt
+2023-03-10 09:51:29,304 INFO [utils.py:558] [test-clean-lm_scale_0.7] %WER 3.19% [1678 / 52576, 110 ins, 566 del, 1002 sub ]
+2023-03-10 09:51:29,444 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.7.txt
+2023-03-10 09:51:29,466 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.8.txt
+2023-03-10 09:51:29,528 INFO [utils.py:558] [test-clean-lm_scale_0.8] %WER 3.81% [2001 / 52576, 95 ins, 908 del, 998 sub ]
+2023-03-10 09:51:29,673 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.8.txt
+2023-03-10 09:51:29,695 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_0.9.txt
+2023-03-10 09:51:29,762 INFO [utils.py:558] [test-clean-lm_scale_0.9] %WER 4.81% [2530 / 52576, 87 ins, 1449 del, 994 sub ]
+2023-03-10 09:51:29,915 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_0.9.txt
+2023-03-10 09:51:29,939 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.0.txt
+2023-03-10 09:51:30,006 INFO [utils.py:558] [test-clean-lm_scale_1.0] %WER 6.33% [3328 / 52576, 73 ins, 2273 del, 982 sub ]
+2023-03-10 09:51:30,168 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.0.txt
+2023-03-10 09:51:30,191 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.1.txt
+2023-03-10 09:51:30,248 INFO [utils.py:558] [test-clean-lm_scale_1.1] %WER 8.13% [4272 / 52576, 64 ins, 3237 del, 971 sub ]
+2023-03-10 09:51:30,385 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.1.txt
+2023-03-10 09:51:30,407 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.2.txt
+2023-03-10 09:51:30,467 INFO [utils.py:558] [test-clean-lm_scale_1.2] %WER 10.16% [5340 / 52576, 56 ins, 4321 del, 963 sub ]
+2023-03-10 09:51:30,603 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.2.txt
+2023-03-10 09:51:30,627 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.3.txt
+2023-03-10 09:51:30,690 INFO [utils.py:558] [test-clean-lm_scale_1.3] %WER 12.17% [6396 / 52576, 45 ins, 5400 del, 951 sub ]
+2023-03-10 09:51:31,002 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.3.txt
+2023-03-10 09:51:31,021 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.4.txt
+2023-03-10 09:51:31,078 INFO [utils.py:558] [test-clean-lm_scale_1.4] %WER 13.82% [7264 / 52576, 41 ins, 6282 del, 941 sub ]
+2023-03-10 09:51:31,226 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.4.txt
+2023-03-10 09:51:31,247 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.5.txt
+2023-03-10 09:51:31,318 INFO [utils.py:558] [test-clean-lm_scale_1.5] %WER 15.01% [7893 / 52576, 33 ins, 6939 del, 921 sub ]
+2023-03-10 09:51:31,460 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.5.txt
+2023-03-10 09:51:31,481 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.6.txt
+2023-03-10 09:51:31,541 INFO [utils.py:558] [test-clean-lm_scale_1.6] %WER 15.78% [8295 / 52576, 27 ins, 7352 del, 916 sub ]
+2023-03-10 09:51:31,680 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.6.txt
+2023-03-10 09:51:31,700 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.7.txt
+2023-03-10 09:51:31,758 INFO [utils.py:558] [test-clean-lm_scale_1.7] %WER 16.35% [8596 / 52576, 26 ins, 7653 del, 917 sub ]
+2023-03-10 09:51:31,899 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.7.txt
+2023-03-10 09:51:31,918 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.8.txt
+2023-03-10 09:51:31,972 INFO [utils.py:558] [test-clean-lm_scale_1.8] %WER 16.80% [8835 / 52576, 25 ins, 7886 del, 924 sub ]
+2023-03-10 09:51:32,121 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.8.txt
+2023-03-10 09:51:32,140 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_1.9.txt
+2023-03-10 09:51:32,196 INFO [utils.py:558] [test-clean-lm_scale_1.9] %WER 17.12% [9003 / 52576, 25 ins, 8052 del, 926 sub ]
+2023-03-10 09:51:32,335 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_1.9.txt
+2023-03-10 09:51:32,355 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-clean-lm_scale_2.0.txt
+2023-03-10 09:51:32,412 INFO [utils.py:558] [test-clean-lm_scale_2.0] %WER 17.42% [9161 / 52576, 24 ins, 8205 del, 932 sub ]
+2023-03-10 09:51:32,722 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-clean-lm_scale_2.0.txt
+2023-03-10 09:51:32,723 INFO [decode.py:627] 
+For test-clean, WER of different settings are:
+lm_scale_0.2	2.44	best for test-clean
+lm_scale_0.1	2.46
+lm_scale_0.3	2.46
+lm_scale_0.4	2.51
+lm_scale_0.5	2.61
+lm_scale_0.6	2.8
+lm_scale_0.7	3.19
+lm_scale_0.8	3.81
+lm_scale_0.9	4.81
+lm_scale_1.0	6.33
+lm_scale_1.1	8.13
+lm_scale_1.2	10.16
+lm_scale_1.3	12.17
+lm_scale_1.4	13.82
+lm_scale_1.5	15.01
+lm_scale_1.6	15.78
+lm_scale_1.7	16.35
+lm_scale_1.8	16.8
+lm_scale_1.9	17.12
+lm_scale_2.0	17.42
+
+2023-03-10 09:51:34,860 INFO [decode.py:581] batch 0/?, cuts processed until now is 17
+2023-03-10 09:52:09,534 INFO [zipformer.py:1455] attn_weights_entropy = tensor([3.6229, 3.2254, 3.7114, 4.5989, 4.0629, 3.9987, 3.1179, 2.7495],
+       device='cuda:0'), covar=tensor([0.0634, 0.1673, 0.0805, 0.0403, 0.0807, 0.0389, 0.1392, 0.1981],
+       device='cuda:0'), in_proj_covar=tensor([0.0178, 0.0213, 0.0182, 0.0217, 0.0226, 0.0179, 0.0198, 0.0185],
+       device='cuda:0'), out_proj_covar=tensor([0.0002, 0.0002, 0.0002, 0.0003, 0.0003, 0.0002, 0.0002, 0.0002],
+       device='cuda:0')
+2023-03-10 09:53:06,108 INFO [zipformer.py:1455] attn_weights_entropy = tensor([5.0127, 4.0324, 4.2679, 3.9797, 4.2217, 5.0873, 4.9086, 4.1711],
+       device='cuda:0'), covar=tensor([0.0340, 0.1178, 0.0924, 0.0920, 0.0887, 0.0680, 0.0421, 0.0949],
+       device='cuda:0'), in_proj_covar=tensor([0.0239, 0.0241, 0.0280, 0.0214, 0.0261, 0.0372, 0.0262, 0.0227],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0004, 0.0003, 0.0003],
+       device='cuda:0')
+2023-03-10 09:53:46,341 INFO [zipformer.py:1455] attn_weights_entropy = tensor([4.7764, 5.4848, 5.2078, 3.5640, 3.2814, 4.2858, 3.7191, 4.6062],
+       device='cuda:0'), covar=tensor([0.0498, 0.0254, 0.0225, 0.3783, 0.3929, 0.1468, 0.2853, 0.1094],
+       device='cuda:0'), in_proj_covar=tensor([0.0351, 0.0282, 0.0266, 0.0244, 0.0333, 0.0327, 0.0254, 0.0360],
+       device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001, 0.0001],
+       device='cuda:0')
+2023-03-10 09:53:49,513 INFO [decode.py:581] batch 100/?, cuts processed until now is 2560
+2023-03-10 09:54:08,539 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.1.txt
+2023-03-10 09:54:08,609 INFO [utils.py:558] [test-other-lm_scale_0.1] %WER 5.38% [2814 / 52343, 336 ins, 224 del, 2254 sub ]
+2023-03-10 09:54:08,764 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.1.txt
+2023-03-10 09:54:08,790 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.2.txt
+2023-03-10 09:54:08,856 INFO [utils.py:558] [test-other-lm_scale_0.2] %WER 5.38% [2814 / 52343, 304 ins, 271 del, 2239 sub ]
+2023-03-10 09:54:09,009 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.2.txt
+2023-03-10 09:54:09,032 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.3.txt
+2023-03-10 09:54:09,101 INFO [utils.py:558] [test-other-lm_scale_0.3] %WER 5.42% [2835 / 52343, 274 ins, 342 del, 2219 sub ]
+2023-03-10 09:54:09,249 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.3.txt
+2023-03-10 09:54:09,270 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.4.txt
+2023-03-10 09:54:09,328 INFO [utils.py:558] [test-other-lm_scale_0.4] %WER 5.53% [2895 / 52343, 244 ins, 442 del, 2209 sub ]
+2023-03-10 09:54:09,473 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.4.txt
+2023-03-10 09:54:09,496 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.5.txt
+2023-03-10 09:54:09,555 INFO [utils.py:558] [test-other-lm_scale_0.5] %WER 5.76% [3014 / 52343, 211 ins, 596 del, 2207 sub ]
+2023-03-10 09:54:09,693 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.5.txt
+2023-03-10 09:54:09,714 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.6.txt
+2023-03-10 09:54:09,775 INFO [utils.py:558] [test-other-lm_scale_0.6] %WER 6.10% [3194 / 52343, 184 ins, 832 del, 2178 sub ]
+2023-03-10 09:54:09,926 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.6.txt
+2023-03-10 09:54:09,949 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.7.txt
+2023-03-10 09:54:10,010 INFO [utils.py:558] [test-other-lm_scale_0.7] %WER 6.79% [3552 / 52343, 166 ins, 1234 del, 2152 sub ]
+2023-03-10 09:54:10,166 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.7.txt
+2023-03-10 09:54:10,190 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.8.txt
+2023-03-10 09:54:10,420 INFO [utils.py:558] [test-other-lm_scale_0.8] %WER 7.86% [4113 / 52343, 140 ins, 1849 del, 2124 sub ]
+2023-03-10 09:54:10,578 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.8.txt
+2023-03-10 09:54:10,624 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_0.9.txt
+2023-03-10 09:54:10,681 INFO [utils.py:558] [test-other-lm_scale_0.9] %WER 9.31% [4872 / 52343, 119 ins, 2698 del, 2055 sub ]
+2023-03-10 09:54:10,826 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_0.9.txt
+2023-03-10 09:54:10,847 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.0.txt
+2023-03-10 09:54:10,906 INFO [utils.py:558] [test-other-lm_scale_1.0] %WER 11.30% [5916 / 52343, 94 ins, 3842 del, 1980 sub ]
+2023-03-10 09:54:11,073 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.0.txt
+2023-03-10 09:54:11,097 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.1.txt
+2023-03-10 09:54:11,167 INFO [utils.py:558] [test-other-lm_scale_1.1] %WER 13.80% [7221 / 52343, 78 ins, 5220 del, 1923 sub ]
+2023-03-10 09:54:11,330 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.1.txt
+2023-03-10 09:54:11,352 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.2.txt
+2023-03-10 09:54:11,426 INFO [utils.py:558] [test-other-lm_scale_1.2] %WER 16.36% [8564 / 52343, 62 ins, 6654 del, 1848 sub ]
+2023-03-10 09:54:11,580 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.2.txt
+2023-03-10 09:54:11,606 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.3.txt
+2023-03-10 09:54:11,665 INFO [utils.py:558] [test-other-lm_scale_1.3] %WER 18.62% [9745 / 52343, 46 ins, 7906 del, 1793 sub ]
+2023-03-10 09:54:11,849 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.3.txt
+2023-03-10 09:54:11,872 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.4.txt
+2023-03-10 09:54:11,933 INFO [utils.py:558] [test-other-lm_scale_1.4] %WER 20.37% [10662 / 52343, 37 ins, 8865 del, 1760 sub ]
+2023-03-10 09:54:12,092 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.4.txt
+2023-03-10 09:54:12,117 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.5.txt
+2023-03-10 09:54:12,183 INFO [utils.py:558] [test-other-lm_scale_1.5] %WER 21.53% [11270 / 52343, 31 ins, 9513 del, 1726 sub ]
+2023-03-10 09:54:12,533 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.5.txt
+2023-03-10 09:54:12,571 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.6.txt
+2023-03-10 09:54:12,634 INFO [utils.py:558] [test-other-lm_scale_1.6] %WER 22.38% [11714 / 52343, 26 ins, 9977 del, 1711 sub ]
+2023-03-10 09:54:12,800 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.6.txt
+2023-03-10 09:54:12,819 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.7.txt
+2023-03-10 09:54:12,881 INFO [utils.py:558] [test-other-lm_scale_1.7] %WER 22.98% [12027 / 52343, 22 ins, 10317 del, 1688 sub ]
+2023-03-10 09:54:13,031 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.7.txt
+2023-03-10 09:54:13,051 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.8.txt
+2023-03-10 09:54:13,116 INFO [utils.py:558] [test-other-lm_scale_1.8] %WER 23.46% [12278 / 52343, 21 ins, 10578 del, 1679 sub ]
+2023-03-10 09:54:13,280 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.8.txt
+2023-03-10 09:54:13,301 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_1.9.txt
+2023-03-10 09:54:13,362 INFO [utils.py:558] [test-other-lm_scale_1.9] %WER 23.87% [12494 / 52343, 21 ins, 10793 del, 1680 sub ]
+2023-03-10 09:54:13,530 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_1.9.txt
+2023-03-10 09:54:13,552 INFO [decode.py:601] The transcripts are stored in zipformer_ctc/exp/v0/recogs-test-other-lm_scale_2.0.txt
+2023-03-10 09:54:13,613 INFO [utils.py:558] [test-other-lm_scale_2.0] %WER 24.19% [12662 / 52343, 20 ins, 10954 del, 1688 sub ]
+2023-03-10 09:54:13,774 INFO [decode.py:613] Wrote detailed error stats to zipformer_ctc/exp/v0/errs-test-other-lm_scale_2.0.txt
+2023-03-10 09:54:13,775 INFO [decode.py:627] 
+For test-other, WER of different settings are:
+lm_scale_0.1	5.38	best for test-other
+lm_scale_0.2	5.38
+lm_scale_0.3	5.42
+lm_scale_0.4	5.53
+lm_scale_0.5	5.76
+lm_scale_0.6	6.1
+lm_scale_0.7	6.79
+lm_scale_0.8	7.86
+lm_scale_0.9	9.31
+lm_scale_1.0	11.3
+lm_scale_1.1	13.8
+lm_scale_1.2	16.36
+lm_scale_1.3	18.62
+lm_scale_1.4	20.37
+lm_scale_1.5	21.53
+lm_scale_1.6	22.38
+lm_scale_1.7	22.98
+lm_scale_1.8	23.46
+lm_scale_1.9	23.87
+lm_scale_2.0	24.19
+
+2023-03-10 09:54:13,775 INFO [decode.py:883] Done!

test_wavs/trans.txt

@@ -0,0 +1,3 @@
+1089-134686-0001 AFTER EARLY NIGHTFALL THE YELLOW LAMPS WOULD LIGHT UP HERE AND THERE THE SQUALID QUARTER OF THE BROTHELS
+1221-135766-0001 GOD AS A DIRECT CONSEQUENCE OF THE SIN WHICH MAN THUS PUNISHED HAD GIVEN HER A LOVELY CHILD WHOSE PLACE WAS ON THAT SAME DISHONOURED BOSOM TO CONNECT HER PARENT FOR EVER WITH THE RACE AND DESCENT OF MORTALS AND TO BE FINALLY A BLESSED SOUL IN HEAVEN
+1221-135766-0002 YET THESE THOUGHTS AFFECTED HESTER PRYNNE LESS WITH HOPE THAN APPREHENSION