upload files

README.md

@@ -0,0 +1,3 @@
+This repo contains pre-trained models, checkpoints, training logs and decoding results of the following pull-request:
+
+https://github.com/k2-fsa/icefall/pull/787

data/lang_bpe_500/HLG.pt

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

data/lang_bpe_500/L.pt

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

data/lang_bpe_500/LG.pt

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

data/lang_bpe_500/Linv.pt

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

data/lang_bpe_500/bpe.model

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

data/lang_bpe_500/tokens.txt

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

data/lm/G_4_gram.pt

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

decoding_results/fast_beam_search/log-decode-epoch-30-avg-9-streaming-chunk-size-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-12-26-15-53-41

@@ -0,0 +1,45 @@
+2022-12-26 15:53:41,676 INFO [decode.py:655] Decoding started
+2022-12-26 15:53:41,676 INFO [decode.py:661] Device: cuda:0
+2022-12-26 15:53:41,682 INFO [decode.py:671] {'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': '1.23.2', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'efd83642a940dc7db08688cc0791985bed1fafcd', 'k2-git-date': 'Sun Nov 27 19:12:00 2022', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming_zipformer', 'icefall-git-sha1': '3c5ed61-clean', 'icefall-git-date': 'Mon Dec 26 13:01:22 2022', 'icefall-path': '/star-zw/workspace/zipformer/icefall_streaming2', 'k2-path': '/star-zw/workspace/share/k2-last/k2/python/k2/__init__.py', 'lhotse-path': '/star-zw/env/k2_icefall/lib/python3.8/site-packages/lhotse-1.12.0.dev0+git.891bad1.clean-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1216192652-5bcf7587b4-n6q9m', 'IP address': '10.177.74.211'}, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'fast_beam_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 32, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search'), 'suffix': 'epoch-30-avg-9-streaming-chunk-size-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-26 15:53:41,682 INFO [decode.py:673] About to create model
+2022-12-26 15:53:42,327 INFO [zipformer.py:378] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-26 15:53:42,341 INFO [decode.py:744] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2022-12-26 15:53:56,411 INFO [decode.py:778] Number of model parameters: 70369391
+2022-12-26 15:53:56,411 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2022-12-26 15:53:56,421 INFO [asr_datamodule.py:450] About to get test-other cuts
+2022-12-26 15:53:59,270 INFO [decode.py:560] batch 0/?, cuts processed until now is 43
+2022-12-26 15:54:31,199 INFO [decode.py:560] batch 20/?, cuts processed until now is 1430
+2022-12-26 15:54:55,125 INFO [zipformer.py:2453] attn_weights_entropy = tensor([2.6072, 3.4398, 3.2798, 1.3989, 3.5262, 2.7791, 0.6962, 2.0624],
+       device='cuda:0'), covar=tensor([0.1968, 0.0881, 0.1401, 0.4125, 0.0731, 0.0908, 0.4646, 0.1797],
+       device='cuda:0'), in_proj_covar=tensor([0.0147, 0.0144, 0.0158, 0.0122, 0.0147, 0.0112, 0.0142, 0.0113],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0003, 0.0002, 0.0003, 0.0002, 0.0003, 0.0002],
+       device='cuda:0')
+2022-12-26 15:55:04,830 INFO [decode.py:560] batch 40/?, cuts processed until now is 2561
+2022-12-26 15:55:05,057 INFO [zipformer.py:2453] attn_weights_entropy = tensor([1.9880, 1.2823, 1.9022, 1.6360, 2.0849, 2.0807, 1.7661, 1.8778],
+       device='cuda:0'), covar=tensor([0.2520, 0.3568, 0.2734, 0.3188, 0.2412, 0.1061, 0.3954, 0.1427],
+       device='cuda:0'), in_proj_covar=tensor([0.0267, 0.0297, 0.0281, 0.0320, 0.0312, 0.0254, 0.0348, 0.0243],
+       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')
+2022-12-26 15:55:05,901 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-12-26 15:55:05,977 INFO [utils.py:536] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.20% [1685 / 52576, 196 ins, 126 del, 1363 sub ]
+2022-12-26 15:55:06,154 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-12-26 15:55:06,157 INFO [decode.py:605] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.2	best for test-clean
+
+2022-12-26 15:55:08,413 INFO [decode.py:560] batch 0/?, cuts processed until now is 52
+2022-12-26 15:55:19,276 INFO [zipformer.py:2453] attn_weights_entropy = tensor([1.5789, 1.4018, 1.3338, 1.7230, 1.7707, 2.9352, 1.5062, 1.5780],
+       device='cuda:0'), covar=tensor([0.0825, 0.1946, 0.0994, 0.0844, 0.1375, 0.0310, 0.1411, 0.1567],
+       device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0070, 0.0073, 0.0089, 0.0074, 0.0083, 0.0076],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004],
+       device='cuda:0')
+2022-12-26 15:55:38,950 INFO [decode.py:560] batch 20/?, cuts processed until now is 1647
+2022-12-26 15:56:11,524 INFO [decode.py:560] batch 40/?, cuts processed until now is 2870
+2022-12-26 15:56:12,579 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-12-26 15:56:12,658 INFO [utils.py:536] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 8.04% [4210 / 52343, 428 ins, 400 del, 3382 sub ]
+2022-12-26 15:56:12,924 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-12-26 15:56:12,927 INFO [decode.py:605] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	8.04	best for test-other
+
+2022-12-26 15:56:12,928 INFO [decode.py:809] Done!

decoding_results/fast_beam_search/log-decode-epoch-30-avg-9-streaming-chunk-size-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model-2022-12-26-16-03-35

@@ -0,0 +1,45 @@
+2022-12-26 16:03:35,556 INFO [decode.py:655] Decoding started
+2022-12-26 16:03:35,557 INFO [decode.py:661] Device: cuda:0
+2022-12-26 16:03:35,563 INFO [decode.py:671] {'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': '1.23.2', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'efd83642a940dc7db08688cc0791985bed1fafcd', 'k2-git-date': 'Sun Nov 27 19:12:00 2022', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming_zipformer', 'icefall-git-sha1': '82b18bc-clean', 'icefall-git-date': 'Mon Dec 26 16:03:17 2022', 'icefall-path': '/star-zw/workspace/zipformer/icefall_streaming2', 'k2-path': '/star-zw/workspace/share/k2-last/k2/python/k2/__init__.py', 'lhotse-path': '/star-zw/env/k2_icefall/lib/python3.8/site-packages/lhotse-1.12.0.dev0+git.891bad1.clean-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1216192652-5bcf7587b4-n6q9m', 'IP address': '10.177.74.211'}, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'fast_beam_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 64, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search'), 'suffix': 'epoch-30-avg-9-streaming-chunk-size-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-26 16:03:35,563 INFO [decode.py:673] About to create model
+2022-12-26 16:03:36,130 INFO [zipformer.py:378] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-26 16:03:36,143 INFO [decode.py:744] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2022-12-26 16:03:48,140 INFO [decode.py:778] Number of model parameters: 70369391
+2022-12-26 16:03:48,140 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2022-12-26 16:03:48,159 INFO [asr_datamodule.py:450] About to get test-other cuts
+2022-12-26 16:03:50,925 INFO [decode.py:560] batch 0/?, cuts processed until now is 43
+2022-12-26 16:03:58,813 INFO [zipformer.py:2453] attn_weights_entropy = tensor([1.8448, 1.7894, 1.9317, 1.7849, 1.4553, 3.7271, 1.8085, 2.3414],
+       device='cuda:0'), covar=tensor([0.3225, 0.2089, 0.1803, 0.2081, 0.1425, 0.0163, 0.1478, 0.0743],
+       device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0117, 0.0123, 0.0121, 0.0104, 0.0094, 0.0089, 0.0088],
+       device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004],
+       device='cuda:0')
+2022-12-26 16:04:09,638 INFO [zipformer.py:2453] attn_weights_entropy = tensor([2.6905, 2.1925, 1.8600, 2.4678, 1.9859, 2.2498, 2.1110, 2.5181],
+       device='cuda:0'), covar=tensor([0.2180, 0.3849, 0.2003, 0.2866, 0.4265, 0.1033, 0.3543, 0.1013],
+       device='cuda:0'), in_proj_covar=tensor([0.0293, 0.0293, 0.0247, 0.0345, 0.0274, 0.0228, 0.0290, 0.0214],
+       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')
+2022-12-26 16:04:25,948 INFO [decode.py:560] batch 20/?, cuts processed until now is 1430
+2022-12-26 16:04:58,837 INFO [decode.py:560] batch 40/?, cuts processed until now is 2561
+2022-12-26 16:04:59,894 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search/recogs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-12-26 16:04:59,967 INFO [utils.py:536] [test-clean-beam_20.0_max_contexts_8_max_states_64] %WER 3.02% [1590 / 52576, 194 ins, 109 del, 1287 sub ]
+2022-12-26 16:05:00,146 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search/errs-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-12-26 16:05:00,149 INFO [decode.py:605] 
+For test-clean, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	3.02	best for test-clean
+
+2022-12-26 16:05:02,362 INFO [decode.py:560] batch 0/?, cuts processed until now is 52
+2022-12-26 16:05:32,491 INFO [decode.py:560] batch 20/?, cuts processed until now is 1647
+2022-12-26 16:05:58,110 INFO [zipformer.py:2453] attn_weights_entropy = tensor([2.4443, 2.4112, 2.1316, 1.2474, 2.0257, 2.0422, 1.9451, 2.2252],
+       device='cuda:0'), covar=tensor([0.0523, 0.0401, 0.1174, 0.1557, 0.0972, 0.1258, 0.1322, 0.0606],
+       device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0183, 0.0204, 0.0186, 0.0206, 0.0200, 0.0213, 0.0199],
+       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')
+2022-12-26 16:06:06,445 INFO [decode.py:560] batch 40/?, cuts processed until now is 2870
+2022-12-26 16:06:07,632 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search/recogs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-12-26 16:06:07,712 INFO [utils.py:536] [test-other-beam_20.0_max_contexts_8_max_states_64] %WER 7.47% [3910 / 52343, 404 ins, 336 del, 3170 sub ]
+2022-12-26 16:06:07,970 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/fast_beam_search/errs-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt
+2022-12-26 16:06:07,973 INFO [decode.py:605] 
+For test-other, WER of different settings are:
+beam_20.0_max_contexts_8_max_states_64	7.47	best for test-other
+
+2022-12-26 16:06:07,973 INFO [decode.py:809] Done!

decoding_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.2

decoding_results/fast_beam_search/wer-summary-test-clean-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	3.02

decoding_results/fast_beam_search/wer-summary-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-32-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	8.04

decoding_results/fast_beam_search/wer-summary-test-other-beam_20.0_max_contexts_8_max_states_64-epoch-30-avg-9-streaming-chunk-size-64-beam-20.0-max-contexts-8-max-states-64-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_20.0_max_contexts_8_max_states_64	7.47

decoding_results/greedy_search/log-decode-epoch-30-avg-9-streaming-chunk-size-32-context-2-max-sym-per-frame-1-use-averaged-model-2022-12-26-15-46-41

@@ -0,0 +1,31 @@
+2022-12-26 15:46:41,976 INFO [decode.py:655] Decoding started
+2022-12-26 15:46:41,977 INFO [decode.py:661] Device: cuda:0
+2022-12-26 15:46:41,983 INFO [decode.py:671] {'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': '1.23.2', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'efd83642a940dc7db08688cc0791985bed1fafcd', 'k2-git-date': 'Sun Nov 27 19:12:00 2022', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming_zipformer', 'icefall-git-sha1': '3c5ed61-clean', 'icefall-git-date': 'Mon Dec 26 13:01:22 2022', 'icefall-path': '/star-zw/workspace/zipformer/icefall_streaming2', 'k2-path': '/star-zw/workspace/share/k2-last/k2/python/k2/__init__.py', 'lhotse-path': '/star-zw/env/k2_icefall/lib/python3.8/site-packages/lhotse-1.12.0.dev0+git.891bad1.clean-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1216192652-5bcf7587b4-n6q9m', 'IP address': '10.177.74.211'}, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'greedy_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 32, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search'), 'suffix': 'epoch-30-avg-9-streaming-chunk-size-32-context-2-max-sym-per-frame-1-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-26 15:46:41,984 INFO [decode.py:673] About to create model
+2022-12-26 15:46:42,587 INFO [zipformer.py:378] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-26 15:46:42,600 INFO [decode.py:744] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2022-12-26 15:46:55,367 INFO [decode.py:778] Number of model parameters: 70369391
+2022-12-26 15:46:55,368 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2022-12-26 15:46:55,392 INFO [asr_datamodule.py:450] About to get test-other cuts
+2022-12-26 15:46:57,378 INFO [decode.py:560] batch 0/?, cuts processed until now is 43
+2022-12-26 15:47:31,519 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search/recogs-test-clean-greedy_search-epoch-30-avg-9-streaming-chunk-size-32-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-26 15:47:31,595 INFO [utils.py:536] [test-clean-greedy_search] %WER 3.15% [1656 / 52576, 197 ins, 124 del, 1335 sub ]
+2022-12-26 15:47:31,775 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search/errs-test-clean-greedy_search-epoch-30-avg-9-streaming-chunk-size-32-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-26 15:47:31,779 INFO [decode.py:605] 
+For test-clean, WER of different settings are:
+greedy_search	3.15	best for test-clean
+
+2022-12-26 15:47:33,398 INFO [decode.py:560] batch 0/?, cuts processed until now is 52
+2022-12-26 15:47:50,827 INFO [zipformer.py:2453] attn_weights_entropy = tensor([1.1363, 1.1192, 1.1599, 1.0857, 0.7870, 2.1296, 0.8855, 1.2788],
+       device='cuda:0'), covar=tensor([0.2260, 0.1509, 0.1418, 0.1542, 0.1079, 0.0160, 0.1202, 0.0567],
+       device='cuda:0'), in_proj_covar=tensor([0.0131, 0.0117, 0.0123, 0.0121, 0.0104, 0.0094, 0.0089, 0.0088],
+       device='cuda:0'), out_proj_covar=tensor([0.0005, 0.0005, 0.0005, 0.0005, 0.0004, 0.0003, 0.0004, 0.0004],
+       device='cuda:0')
+2022-12-26 15:48:05,354 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search/recogs-test-other-greedy_search-epoch-30-avg-9-streaming-chunk-size-32-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-26 15:48:05,433 INFO [utils.py:536] [test-other-greedy_search] %WER 8.09% [4234 / 52343, 419 ins, 428 del, 3387 sub ]
+2022-12-26 15:48:05,703 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search/errs-test-other-greedy_search-epoch-30-avg-9-streaming-chunk-size-32-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-26 15:48:05,711 INFO [decode.py:605] 
+For test-other, WER of different settings are:
+greedy_search	8.09	best for test-other
+
+2022-12-26 15:48:05,711 INFO [decode.py:809] Done!

decoding_results/greedy_search/log-decode-epoch-30-avg-9-streaming-chunk-size-64-context-2-max-sym-per-frame-1-use-averaged-model-2022-12-26-15-57-02

@@ -0,0 +1,36 @@
+2022-12-26 15:57:02,067 INFO [decode.py:655] Decoding started
+2022-12-26 15:57:02,067 INFO [decode.py:661] Device: cuda:0
+2022-12-26 15:57:02,073 INFO [decode.py:671] {'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': '1.23.2', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'efd83642a940dc7db08688cc0791985bed1fafcd', 'k2-git-date': 'Sun Nov 27 19:12:00 2022', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming_zipformer', 'icefall-git-sha1': '3c5ed61-clean', 'icefall-git-date': 'Mon Dec 26 13:01:22 2022', 'icefall-path': '/star-zw/workspace/zipformer/icefall_streaming2', 'k2-path': '/star-zw/workspace/share/k2-last/k2/python/k2/__init__.py', 'lhotse-path': '/star-zw/env/k2_icefall/lib/python3.8/site-packages/lhotse-1.12.0.dev0+git.891bad1.clean-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1216192652-5bcf7587b4-n6q9m', 'IP address': '10.177.74.211'}, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'greedy_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 64, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search'), 'suffix': 'epoch-30-avg-9-streaming-chunk-size-64-context-2-max-sym-per-frame-1-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-26 15:57:02,074 INFO [decode.py:673] About to create model
+2022-12-26 15:57:02,651 INFO [zipformer.py:378] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-26 15:57:02,665 INFO [decode.py:744] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2022-12-26 15:57:14,195 INFO [decode.py:778] Number of model parameters: 70369391
+2022-12-26 15:57:14,196 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2022-12-26 15:57:14,223 INFO [asr_datamodule.py:450] About to get test-other cuts
+2022-12-26 15:57:16,121 INFO [decode.py:560] batch 0/?, cuts processed until now is 43
+2022-12-26 15:57:39,882 INFO [zipformer.py:2453] attn_weights_entropy = tensor([1.9033, 1.6283, 1.7637, 2.1427, 1.9591, 3.5384, 1.7063, 1.7244],
+       device='cuda:0'), covar=tensor([0.0784, 0.1860, 0.0972, 0.0883, 0.1360, 0.0198, 0.1322, 0.1551],
+       device='cuda:0'), in_proj_covar=tensor([0.0071, 0.0081, 0.0070, 0.0073, 0.0089, 0.0074, 0.0083, 0.0076],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0004, 0.0004, 0.0004, 0.0004, 0.0003, 0.0004, 0.0004],
+       device='cuda:0')
+2022-12-26 15:57:50,436 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search/recogs-test-clean-greedy_search-epoch-30-avg-9-streaming-chunk-size-64-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-26 15:57:50,510 INFO [utils.py:536] [test-clean-greedy_search] %WER 2.97% [1563 / 52576, 182 ins, 118 del, 1263 sub ]
+2022-12-26 15:57:50,688 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search/errs-test-clean-greedy_search-epoch-30-avg-9-streaming-chunk-size-64-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-26 15:57:50,692 INFO [decode.py:605] 
+For test-clean, WER of different settings are:
+greedy_search	2.97	best for test-clean
+
+2022-12-26 15:57:52,364 INFO [decode.py:560] batch 0/?, cuts processed until now is 52
+2022-12-26 15:58:15,082 INFO [zipformer.py:2453] attn_weights_entropy = tensor([1.3437, 1.0758, 1.6689, 2.4451, 1.8342, 2.0346, 0.8291, 1.7668],
+       device='cuda:0'), covar=tensor([0.1387, 0.1287, 0.0960, 0.0457, 0.0726, 0.0757, 0.1515, 0.0821],
+       device='cuda:0'), in_proj_covar=tensor([0.0100, 0.0116, 0.0134, 0.0151, 0.0104, 0.0140, 0.0126, 0.0112],
+       device='cuda:0'), out_proj_covar=tensor([0.0003, 0.0003, 0.0004, 0.0004, 0.0003, 0.0004, 0.0003, 0.0003],
+       device='cuda:0')
+2022-12-26 15:58:24,481 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search/recogs-test-other-greedy_search-epoch-30-avg-9-streaming-chunk-size-64-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-26 15:58:24,558 INFO [utils.py:536] [test-other-greedy_search] %WER 7.50% [3924 / 52343, 386 ins, 372 del, 3166 sub ]
+2022-12-26 15:58:24,817 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/greedy_search/errs-test-other-greedy_search-epoch-30-avg-9-streaming-chunk-size-64-context-2-max-sym-per-frame-1-use-averaged-model.txt
+2022-12-26 15:58:24,820 INFO [decode.py:605] 
+For test-other, WER of different settings are:
+greedy_search	7.5	best for test-other
+
+2022-12-26 15:58:24,820 INFO [decode.py:809] Done!

decoding_results/greedy_search/wer-summary-test-clean-greedy_search-epoch-30-avg-9-streaming-chunk-size-32-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	3.15

decoding_results/greedy_search/wer-summary-test-clean-greedy_search-epoch-30-avg-9-streaming-chunk-size-64-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	2.97

decoding_results/greedy_search/wer-summary-test-other-greedy_search-epoch-30-avg-9-streaming-chunk-size-32-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	8.09

decoding_results/greedy_search/wer-summary-test-other-greedy_search-epoch-30-avg-9-streaming-chunk-size-64-context-2-max-sym-per-frame-1-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+greedy_search	7.5

decoding_results/modified_beam_search/log-decode-epoch-30-avg-9-streaming-chunk-size-32-modified_beam_search-beam-size-4-use-averaged-model-2022-12-26-15-48-13

@@ -0,0 +1,35 @@
+2022-12-26 15:48:13,768 INFO [decode.py:655] Decoding started
+2022-12-26 15:48:13,769 INFO [decode.py:661] Device: cuda:0
+2022-12-26 15:48:13,777 INFO [decode.py:671] {'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': '1.23.2', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'efd83642a940dc7db08688cc0791985bed1fafcd', 'k2-git-date': 'Sun Nov 27 19:12:00 2022', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming_zipformer', 'icefall-git-sha1': '3c5ed61-clean', 'icefall-git-date': 'Mon Dec 26 13:01:22 2022', 'icefall-path': '/star-zw/workspace/zipformer/icefall_streaming2', 'k2-path': '/star-zw/workspace/share/k2-last/k2/python/k2/__init__.py', 'lhotse-path': '/star-zw/env/k2_icefall/lib/python3.8/site-packages/lhotse-1.12.0.dev0+git.891bad1.clean-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1216192652-5bcf7587b4-n6q9m', 'IP address': '10.177.74.211'}, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'modified_beam_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 32, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search'), 'suffix': 'epoch-30-avg-9-streaming-chunk-size-32-modified_beam_search-beam-size-4-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-26 15:48:13,777 INFO [decode.py:673] About to create model
+2022-12-26 15:48:14,356 INFO [zipformer.py:378] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-26 15:48:14,369 INFO [decode.py:744] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2022-12-26 15:48:26,085 INFO [decode.py:778] Number of model parameters: 70369391
+2022-12-26 15:48:26,085 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2022-12-26 15:48:26,097 INFO [asr_datamodule.py:450] About to get test-other cuts
+2022-12-26 15:48:30,965 INFO [decode.py:560] batch 0/?, cuts processed until now is 43
+2022-12-26 15:49:45,893 INFO [decode.py:560] batch 20/?, cuts processed until now is 1430
+2022-12-26 15:50:48,364 INFO [decode.py:560] batch 40/?, cuts processed until now is 2561
+2022-12-26 15:50:50,768 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search/recogs-test-clean-beam_size_4-epoch-30-avg-9-streaming-chunk-size-32-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-26 15:50:50,843 INFO [utils.py:536] [test-clean-beam_size_4] %WER 3.11% [1635 / 52576, 199 ins, 123 del, 1313 sub ]
+2022-12-26 15:50:51,023 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search/errs-test-clean-beam_size_4-epoch-30-avg-9-streaming-chunk-size-32-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-26 15:50:51,026 INFO [decode.py:605] 
+For test-clean, WER of different settings are:
+beam_size_4	3.11	best for test-clean
+
+2022-12-26 15:50:55,556 INFO [decode.py:560] batch 0/?, cuts processed until now is 52
+2022-12-26 15:51:02,903 INFO [zipformer.py:2453] attn_weights_entropy = tensor([1.4207, 1.3964, 1.3638, 1.3553, 1.6503, 1.5348, 1.5299, 1.2176],
+       device='cuda:0'), covar=tensor([0.0333, 0.0229, 0.0501, 0.0422, 0.0376, 0.0399, 0.0272, 0.0292],
+       device='cuda:0'), in_proj_covar=tensor([0.0091, 0.0123, 0.0150, 0.0118, 0.0112, 0.0118, 0.0096, 0.0124],
+       device='cuda:0'), out_proj_covar=tensor([7.2127e-05, 9.7093e-05, 1.2286e-04, 9.3654e-05, 9.0313e-05, 9.0339e-05,
+        7.4492e-05, 9.7424e-05], device='cuda:0')
+2022-12-26 15:52:12,942 INFO [decode.py:560] batch 20/?, cuts processed until now is 1647
+2022-12-26 15:53:27,072 INFO [decode.py:560] batch 40/?, cuts processed until now is 2870
+2022-12-26 15:53:30,052 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search/recogs-test-other-beam_size_4-epoch-30-avg-9-streaming-chunk-size-32-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-26 15:53:30,132 INFO [utils.py:536] [test-other-beam_size_4] %WER 7.93% [4152 / 52343, 431 ins, 368 del, 3353 sub ]
+2022-12-26 15:53:30,324 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search/errs-test-other-beam_size_4-epoch-30-avg-9-streaming-chunk-size-32-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-26 15:53:30,327 INFO [decode.py:605] 
+For test-other, WER of different settings are:
+beam_size_4	7.93	best for test-other
+
+2022-12-26 15:53:30,327 INFO [decode.py:809] Done!

decoding_results/modified_beam_search/log-decode-epoch-30-avg-9-streaming-chunk-size-64-modified_beam_search-beam-size-4-use-averaged-model-2022-12-26-15-58-32

@@ -0,0 +1,40 @@
+2022-12-26 15:58:32,298 INFO [decode.py:655] Decoding started
+2022-12-26 15:58:32,298 INFO [decode.py:661] Device: cuda:0
+2022-12-26 15:58:32,302 INFO [decode.py:671] {'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': '1.23.2', 'k2-build-type': 'Release', 'k2-with-cuda': True, 'k2-git-sha1': 'efd83642a940dc7db08688cc0791985bed1fafcd', 'k2-git-date': 'Sun Nov 27 19:12:00 2022', 'lhotse-version': '1.12.0.dev+git.891bad1.clean', 'torch-version': '1.10.0+cu102', 'torch-cuda-available': True, 'torch-cuda-version': '10.2', 'python-version': '3.8', 'icefall-git-branch': 'streaming_zipformer', 'icefall-git-sha1': '3c5ed61-clean', 'icefall-git-date': 'Mon Dec 26 13:01:22 2022', 'icefall-path': '/star-zw/workspace/zipformer/icefall_streaming2', 'k2-path': '/star-zw/workspace/share/k2-last/k2/python/k2/__init__.py', 'lhotse-path': '/star-zw/env/k2_icefall/lib/python3.8/site-packages/lhotse-1.12.0.dev0+git.891bad1.clean-py3.8.egg/lhotse/__init__.py', 'hostname': 'de-74279-k2-train-2-1216192652-5bcf7587b4-n6q9m', 'IP address': '10.177.74.211'}, 'epoch': 30, 'iter': 0, 'avg': 9, 'use_averaged_model': True, 'exp_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk'), 'bpe_model': 'data/lang_bpe_500/bpe.model', 'lang_dir': PosixPath('data/lang_bpe_500'), 'decoding_method': 'modified_beam_search', 'beam_size': 4, 'beam': 20.0, 'ngram_lm_scale': 0.01, 'max_contexts': 8, 'max_states': 64, 'context_size': 2, 'max_sym_per_frame': 1, 'num_paths': 200, 'nbest_scale': 0.5, 'num_encoder_layers': '2,4,3,2,4', 'feedforward_dims': '1024,1024,2048,2048,1024', 'nhead': '8,8,8,8,8', 'encoder_dims': '384,384,384,384,384', 'attention_dims': '192,192,192,192,192', 'encoder_unmasked_dims': '256,256,256,256,256', 'zipformer_downsampling_factors': '1,2,4,8,2', 'cnn_module_kernels': '31,31,31,31,31', 'decoder_dim': 512, 'joiner_dim': 512, 'short_chunk_size': 50, 'num_left_chunks': 4, 'decode_chunk_len': 64, 'full_libri': True, 'manifest_dir': PosixPath('data/fbank'), 'max_duration': 600, 'bucketing_sampler': True, 'num_buckets': 30, 'concatenate_cuts': False, 'duration_factor': 1.0, 'gap': 1.0, 'on_the_fly_feats': False, 'shuffle': True, 'drop_last': True, 'return_cuts': True, 'num_workers': 2, 'enable_spec_aug': True, 'spec_aug_time_warp_factor': 80, 'enable_musan': True, 'input_strategy': 'PrecomputedFeatures', 'res_dir': PosixPath('pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search'), 'suffix': 'epoch-30-avg-9-streaming-chunk-size-64-modified_beam_search-beam-size-4-use-averaged-model', 'blank_id': 0, 'unk_id': 2, 'vocab_size': 500}
+2022-12-26 15:58:32,303 INFO [decode.py:673] About to create model
+2022-12-26 15:58:32,928 INFO [zipformer.py:378] At encoder stack 4, which has downsampling_factor=2, we will combine the outputs of layers 1 and 3, with downsampling_factors=2 and 8.
+2022-12-26 15:58:32,941 INFO [decode.py:744] Calculating the averaged model over epoch range from 21 (excluded) to 30
+2022-12-26 15:58:45,024 INFO [decode.py:778] Number of model parameters: 70369391
+2022-12-26 15:58:45,024 INFO [asr_datamodule.py:443] About to get test-clean cuts
+2022-12-26 15:58:45,035 INFO [asr_datamodule.py:450] About to get test-other cuts
+2022-12-26 15:58:50,309 INFO [decode.py:560] batch 0/?, cuts processed until now is 43
+2022-12-26 16:00:04,047 INFO [decode.py:560] batch 20/?, cuts processed until now is 1430
+2022-12-26 16:00:38,875 INFO [zipformer.py:2453] attn_weights_entropy = tensor([2.6516, 2.5294, 2.0733, 1.0354, 1.9267, 2.1492, 2.0182, 2.2597],
+       device='cuda:0'), covar=tensor([0.0654, 0.0513, 0.1258, 0.1839, 0.1418, 0.1651, 0.1676, 0.0725],
+       device='cuda:0'), in_proj_covar=tensor([0.0169, 0.0183, 0.0204, 0.0186, 0.0206, 0.0200, 0.0213, 0.0199],
+       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')
+2022-12-26 16:01:05,638 INFO [decode.py:560] batch 40/?, cuts processed until now is 2561
+2022-12-26 16:01:07,958 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search/recogs-test-clean-beam_size_4-epoch-30-avg-9-streaming-chunk-size-64-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-26 16:01:08,031 INFO [utils.py:536] [test-clean-beam_size_4] %WER 2.94% [1546 / 52576, 192 ins, 109 del, 1245 sub ]
+2022-12-26 16:01:08,263 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search/errs-test-clean-beam_size_4-epoch-30-avg-9-streaming-chunk-size-64-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-26 16:01:08,266 INFO [decode.py:605] 
+For test-clean, WER of different settings are:
+beam_size_4	2.94	best for test-clean
+
+2022-12-26 16:01:12,851 INFO [decode.py:560] batch 0/?, cuts processed until now is 52
+2022-12-26 16:02:07,574 INFO [zipformer.py:2453] attn_weights_entropy = tensor([2.2372, 2.1836, 1.7346, 2.6503, 2.4433, 2.1968, 2.9547, 2.3434],
+       device='cuda:0'), covar=tensor([0.0885, 0.1975, 0.2853, 0.1639, 0.1777, 0.0857, 0.0860, 0.1317],
+       device='cuda:0'), in_proj_covar=tensor([0.0177, 0.0211, 0.0252, 0.0288, 0.0237, 0.0190, 0.0202, 0.0204],
+       device='cuda:0'), out_proj_covar=tensor([0.0001, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002, 0.0002],
+       device='cuda:0')
+2022-12-26 16:02:24,954 INFO [decode.py:560] batch 20/?, cuts processed until now is 1647
+2022-12-26 16:03:23,922 INFO [decode.py:560] batch 40/?, cuts processed until now is 2870
+2022-12-26 16:03:26,300 INFO [decode.py:576] The transcripts are stored in pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search/recogs-test-other-beam_size_4-epoch-30-avg-9-streaming-chunk-size-64-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-26 16:03:26,377 INFO [utils.py:536] [test-other-beam_size_4] %WER 7.36% [3855 / 52343, 409 ins, 318 del, 3128 sub ]
+2022-12-26 16:03:26,566 INFO [decode.py:589] Wrote detailed error stats to pruned_transducer_stateless7_streaming/exp-full-dynamic-chunk/modified_beam_search/errs-test-other-beam_size_4-epoch-30-avg-9-streaming-chunk-size-64-modified_beam_search-beam-size-4-use-averaged-model.txt
+2022-12-26 16:03:26,568 INFO [decode.py:605] 
+For test-other, WER of different settings are:
+beam_size_4	7.36	best for test-other
+
+2022-12-26 16:03:26,568 INFO [decode.py:809] Done!

decoding_results/modified_beam_search/wer-summary-test-clean-beam_size_4-epoch-30-avg-9-streaming-chunk-size-32-modified_beam_search-beam-size-4-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_size_4	3.11

decoding_results/modified_beam_search/wer-summary-test-clean-beam_size_4-epoch-30-avg-9-streaming-chunk-size-64-modified_beam_search-beam-size-4-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_size_4	2.94

decoding_results/modified_beam_search/wer-summary-test-other-beam_size_4-epoch-30-avg-9-streaming-chunk-size-32-modified_beam_search-beam-size-4-use-averaged-model.txt

@@ -0,0 +1,2 @@
+settings	WER
+beam_size_4	7.93