Hugging Face's logo

Check
/
vaw2tmp

TensorBoard
Model card Files
xet
 Metrics Community
Check commited on
Commit
600687b
·
1 Parent(s): c8de793

"auto-commit"

Browse files
Files changed (28) hide show
  1. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/config.json +0 -0
  2. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/optimizer.pt +1 -1
  3. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/preprocessor_config.json +0 -0
  4. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55501}/pytorch_model.bin +1 -1
  5. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55501}/rng_state.pth +2 -2
  6. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55501}/scaler.pt +1 -1
  7. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/scheduler.pt +1 -1
  8. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55501}/trainer_state.json +482 -5
  9. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/training_args.bin +0 -0
  10. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/config.json +0 -0
  11. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/optimizer.pt +1 -1
  12. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/preprocessor_config.json +0 -0
  13. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55750}/pytorch_model.bin +1 -1
  14. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55750}/rng_state.pth +2 -2
  15. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55750}/scaler.pt +1 -1
  16. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/scheduler.pt +1 -1
  17. model-bin/finetune/base/{checkpoint-52515 → checkpoint-55750}/trainer_state.json +4122 -6
  18. model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/training_args.bin +0 -0
  19. model-bin/finetune/base/log/1629771721.037489/events.out.tfevents.1629771721.c435e1c5ee04.920.21 +3 -0
  20. model-bin/finetune/base/log/1629772396.2921655/events.out.tfevents.1629772396.c435e1c5ee04.920.23 +3 -0
  21. model-bin/finetune/base/log/1629773061.1068475/events.out.tfevents.1629773061.c435e1c5ee04.920.25 +3 -0
  22. model-bin/finetune/base/log/1629773705.9435754/events.out.tfevents.1629773705.c435e1c5ee04.920.27 +3 -0
  23. model-bin/finetune/base/log/1629774367.3900394/events.out.tfevents.1629774367.c435e1c5ee04.920.29 +3 -0
  24. model-bin/finetune/base/log/events.out.tfevents.1629771721.c435e1c5ee04.920.20 +3 -0
  25. model-bin/finetune/base/log/events.out.tfevents.1629772396.c435e1c5ee04.920.22 +3 -0
  26. model-bin/finetune/base/log/events.out.tfevents.1629773061.c435e1c5ee04.920.24 +3 -0
  27. model-bin/finetune/base/log/events.out.tfevents.1629773705.c435e1c5ee04.920.26 +3 -0
  28. model-bin/finetune/base/log/events.out.tfevents.1629774367.c435e1c5ee04.920.28 +3 -0
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/config.json RENAMED
File without changes
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/optimizer.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:07ad5e8e80074832caa0a98b5c08f323f432edcdb3b2a365d123371f20e8beda
3
  size 722165009
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:417134b490a43dadfd6c863fdd32c93e62f5670589c60de518474bbbcb87192e
3
  size 722165009
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/preprocessor_config.json RENAMED
File without changes
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55501}/pytorch_model.bin RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:f0790fa6fb916d8cbb7e32fc2c62611b3f881e48d0f4439e53ce8ba15b93bfa8
3
  size 377909911
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6666b72c33815713f5ef42633441bdc6ce6b93b09179c37e3c0c9ffe31501f52
3
  size 377909911
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55501}/rng_state.pth RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:2e24db58e87b89cd5a8e4c2bb7f104406fb2f99771af6ab8029b177aa98488cd
3
- size 14567
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4a1406c8db515233304a48cdeea6ce28b26f307346d9edec4a1f735f2edd837
3
+ size 14503
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55501}/scaler.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1ca3c56116107871279dfa4f94323a31be7725479169e3fda63b787a410c07df
3
  size 559
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d90403b234e2899cca64e4c4d1ecd5ac4db6b9c472764daea1d674f115d75b81
3
  size 559
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/scheduler.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:675296120f77ac95e3ef252eb671e40c28127966b4388d2ca4ab32f9c73d4f1a
3
  size 623
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:127e8b3b2266ec02eb34759e17b27d11dfed14d46050a41bc1083ae605156cfe
3
  size 623
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55501}/trainer_state.json RENAMED
@@ -1,8 +1,8 @@
1
  {
2
- "best_metric": 0.18992848189928482,
3
- "best_model_checkpoint": "./model-bin/finetune/base/checkpoint-52515",
4
- "epoch": 441.0,
5
- "global_step": 55127,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
@@ -169503,11 +169503,488 @@
169503
  "eval_steps_per_second": 0.689,
169504
  "eval_wer": 0.19953240301015562,
169505
  "step": 55127
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
169506
  }
169507
  ],
169508
  "max_steps": 625000,
169509
  "num_train_epochs": 5000,
169510
- "total_flos": 1.551324177518058e+20,
169511
  "trial_name": null,
169512
  "trial_params": null
169513
  }
 
1
  {
2
+ "best_metric": 0.18631571186315712,
3
+ "best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
4
+ "epoch": 444.0,
5
+ "global_step": 55501,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
 
169503
  "eval_steps_per_second": 0.689,
169504
  "eval_wer": 0.19953240301015562,
169505
  "step": 55127
169506
+ },
169507
+ {
169508
+ "epoch": 441.02,
169509
+ "learning_rate": 9.13286858974359e-06,
169510
+ "loss": 0.3839,
169511
+ "step": 55130
169512
+ },
169513
+ {
169514
+ "epoch": 441.06,
169515
+ "learning_rate": 9.132788461538462e-06,
169516
+ "loss": 0.3799,
169517
+ "step": 55135
169518
+ },
169519
+ {
169520
+ "epoch": 441.1,
169521
+ "learning_rate": 9.132708333333335e-06,
169522
+ "loss": 0.3318,
169523
+ "step": 55140
169524
+ },
169525
+ {
169526
+ "epoch": 441.14,
169527
+ "learning_rate": 9.132628205128206e-06,
169528
+ "loss": 0.4144,
169529
+ "step": 55145
169530
+ },
169531
+ {
169532
+ "epoch": 441.18,
169533
+ "learning_rate": 9.132548076923078e-06,
169534
+ "loss": 1.2108,
169535
+ "step": 55150
169536
+ },
169537
+ {
169538
+ "epoch": 441.22,
169539
+ "learning_rate": 9.132467948717949e-06,
169540
+ "loss": 1.007,
169541
+ "step": 55155
169542
+ },
169543
+ {
169544
+ "epoch": 441.26,
169545
+ "learning_rate": 9.132387820512822e-06,
169546
+ "loss": 0.341,
169547
+ "step": 55160
169548
+ },
169549
+ {
169550
+ "epoch": 441.3,
169551
+ "learning_rate": 9.132307692307693e-06,
169552
+ "loss": 0.3373,
169553
+ "step": 55165
169554
+ },
169555
+ {
169556
+ "epoch": 441.34,
169557
+ "learning_rate": 9.132227564102565e-06,
169558
+ "loss": 0.3604,
169559
+ "step": 55170
169560
+ },
169561
+ {
169562
+ "epoch": 441.38,
169563
+ "learning_rate": 9.132147435897438e-06,
169564
+ "loss": 0.7739,
169565
+ "step": 55175
169566
+ },
169567
+ {
169568
+ "epoch": 441.42,
169569
+ "learning_rate": 9.132067307692309e-06,
169570
+ "loss": 0.9114,
169571
+ "step": 55180
169572
+ },
169573
+ {
169574
+ "epoch": 441.46,
169575
+ "learning_rate": 9.13198717948718e-06,
169576
+ "loss": 0.3421,
169577
+ "step": 55185
169578
+ },
169579
+ {
169580
+ "epoch": 441.5,
169581
+ "learning_rate": 9.131907051282052e-06,
169582
+ "loss": 0.3511,
169583
+ "step": 55190
169584
+ },
169585
+ {
169586
+ "epoch": 441.54,
169587
+ "learning_rate": 9.131826923076925e-06,
169588
+ "loss": 0.4503,
169589
+ "step": 55195
169590
+ },
169591
+ {
169592
+ "epoch": 441.58,
169593
+ "learning_rate": 9.131746794871795e-06,
169594
+ "loss": 0.6906,
169595
+ "step": 55200
169596
+ },
169597
+ {
169598
+ "epoch": 441.62,
169599
+ "learning_rate": 9.131666666666668e-06,
169600
+ "loss": 1.0657,
169601
+ "step": 55205
169602
+ },
169603
+ {
169604
+ "epoch": 441.66,
169605
+ "learning_rate": 9.131586538461539e-06,
169606
+ "loss": 0.3198,
169607
+ "step": 55210
169608
+ },
169609
+ {
169610
+ "epoch": 441.7,
169611
+ "learning_rate": 9.13150641025641e-06,
169612
+ "loss": 0.4051,
169613
+ "step": 55215
169614
+ },
169615
+ {
169616
+ "epoch": 441.74,
169617
+ "learning_rate": 9.131426282051283e-06,
169618
+ "loss": 0.4729,
169619
+ "step": 55220
169620
+ },
169621
+ {
169622
+ "epoch": 441.78,
169623
+ "learning_rate": 9.131346153846155e-06,
169624
+ "loss": 0.781,
169625
+ "step": 55225
169626
+ },
169627
+ {
169628
+ "epoch": 441.82,
169629
+ "learning_rate": 9.131266025641026e-06,
169630
+ "loss": 0.9563,
169631
+ "step": 55230
169632
+ },
169633
+ {
169634
+ "epoch": 441.86,
169635
+ "learning_rate": 9.131185897435897e-06,
169636
+ "loss": 0.2861,
169637
+ "step": 55235
169638
+ },
169639
+ {
169640
+ "epoch": 441.9,
169641
+ "learning_rate": 9.13110576923077e-06,
169642
+ "loss": 0.3339,
169643
+ "step": 55240
169644
+ },
169645
+ {
169646
+ "epoch": 441.94,
169647
+ "learning_rate": 9.131025641025642e-06,
169648
+ "loss": 0.421,
169649
+ "step": 55245
169650
+ },
169651
+ {
169652
+ "epoch": 441.98,
169653
+ "learning_rate": 9.130945512820513e-06,
169654
+ "loss": 0.8084,
169655
+ "step": 55250
169656
+ },
169657
+ {
169658
+ "epoch": 442.0,
169659
+ "eval_loss": 0.40637868642807007,
169660
+ "eval_runtime": 39.8756,
169661
+ "eval_samples_per_second": 21.066,
169662
+ "eval_steps_per_second": 0.677,
169663
+ "eval_wer": 0.198806860551827,
169664
+ "step": 55252
169665
+ },
169666
+ {
169667
+ "epoch": 445.02,
169668
+ "learning_rate": 9.130865384615385e-06,
169669
+ "loss": 0.3958,
169670
+ "step": 55255
169671
+ },
169672
+ {
169673
+ "epoch": 445.06,
169674
+ "learning_rate": 9.130785256410258e-06,
169675
+ "loss": 0.3301,
169676
+ "step": 55260
169677
+ },
169678
+ {
169679
+ "epoch": 445.1,
169680
+ "learning_rate": 9.130705128205129e-06,
169681
+ "loss": 0.3448,
169682
+ "step": 55265
169683
+ },
169684
+ {
169685
+ "epoch": 445.14,
169686
+ "learning_rate": 9.130625e-06,
169687
+ "loss": 0.4292,
169688
+ "step": 55270
169689
+ },
169690
+ {
169691
+ "epoch": 445.18,
169692
+ "learning_rate": 9.130544871794873e-06,
169693
+ "loss": 0.753,
169694
+ "step": 55275
169695
+ },
169696
+ {
169697
+ "epoch": 445.22,
169698
+ "learning_rate": 9.130464743589745e-06,
169699
+ "loss": 1.0239,
169700
+ "step": 55280
169701
+ },
169702
+ {
169703
+ "epoch": 445.27,
169704
+ "learning_rate": 9.130384615384616e-06,
169705
+ "loss": 0.2834,
169706
+ "step": 55285
169707
+ },
169708
+ {
169709
+ "epoch": 445.31,
169710
+ "learning_rate": 9.130304487179487e-06,
169711
+ "loss": 0.3527,
169712
+ "step": 55290
169713
+ },
169714
+ {
169715
+ "epoch": 445.35,
169716
+ "learning_rate": 9.13022435897436e-06,
169717
+ "loss": 0.4062,
169718
+ "step": 55295
169719
+ },
169720
+ {
169721
+ "epoch": 445.39,
169722
+ "learning_rate": 9.130144230769232e-06,
169723
+ "loss": 0.8633,
169724
+ "step": 55300
169725
+ },
169726
+ {
169727
+ "epoch": 445.43,
169728
+ "learning_rate": 9.130064102564103e-06,
169729
+ "loss": 0.9571,
169730
+ "step": 55305
169731
+ },
169732
+ {
169733
+ "epoch": 445.47,
169734
+ "learning_rate": 9.129983974358975e-06,
169735
+ "loss": 0.3956,
169736
+ "step": 55310
169737
+ },
169738
+ {
169739
+ "epoch": 445.51,
169740
+ "learning_rate": 9.129903846153848e-06,
169741
+ "loss": 0.3443,
169742
+ "step": 55315
169743
+ },
169744
+ {
169745
+ "epoch": 445.55,
169746
+ "learning_rate": 9.129823717948719e-06,
169747
+ "loss": 0.4048,
169748
+ "step": 55320
169749
+ },
169750
+ {
169751
+ "epoch": 445.59,
169752
+ "learning_rate": 9.12974358974359e-06,
169753
+ "loss": 0.7511,
169754
+ "step": 55325
169755
+ },
169756
+ {
169757
+ "epoch": 445.63,
169758
+ "learning_rate": 9.129663461538463e-06,
169759
+ "loss": 0.8738,
169760
+ "step": 55330
169761
+ },
169762
+ {
169763
+ "epoch": 445.67,
169764
+ "learning_rate": 9.129583333333333e-06,
169765
+ "loss": 0.372,
169766
+ "step": 55335
169767
+ },
169768
+ {
169769
+ "epoch": 445.71,
169770
+ "learning_rate": 9.129503205128206e-06,
169771
+ "loss": 0.3997,
169772
+ "step": 55340
169773
+ },
169774
+ {
169775
+ "epoch": 445.75,
169776
+ "learning_rate": 9.129423076923078e-06,
169777
+ "loss": 0.3981,
169778
+ "step": 55345
169779
+ },
169780
+ {
169781
+ "epoch": 445.79,
169782
+ "learning_rate": 9.129342948717949e-06,
169783
+ "loss": 0.7046,
169784
+ "step": 55350
169785
+ },
169786
+ {
169787
+ "epoch": 445.83,
169788
+ "learning_rate": 9.12926282051282e-06,
169789
+ "loss": 0.9529,
169790
+ "step": 55355
169791
+ },
169792
+ {
169793
+ "epoch": 445.87,
169794
+ "learning_rate": 9.129182692307693e-06,
169795
+ "loss": 0.3718,
169796
+ "step": 55360
169797
+ },
169798
+ {
169799
+ "epoch": 445.91,
169800
+ "learning_rate": 9.129102564102565e-06,
169801
+ "loss": 0.353,
169802
+ "step": 55365
169803
+ },
169804
+ {
169805
+ "epoch": 445.95,
169806
+ "learning_rate": 9.129022435897436e-06,
169807
+ "loss": 0.3952,
169808
+ "step": 55370
169809
+ },
169810
+ {
169811
+ "epoch": 445.99,
169812
+ "learning_rate": 9.128942307692309e-06,
169813
+ "loss": 0.9729,
169814
+ "step": 55375
169815
+ },
169816
+ {
169817
+ "epoch": 446.0,
169818
+ "eval_loss": 0.40089958906173706,
169819
+ "eval_runtime": 40.2269,
169820
+ "eval_samples_per_second": 20.882,
169821
+ "eval_steps_per_second": 0.671,
169822
+ "eval_wer": 0.20831232442555644,
169823
+ "step": 55376
169824
+ },
169825
+ {
169826
+ "epoch": 443.03,
169827
+ "learning_rate": 9.12886217948718e-06,
169828
+ "loss": 0.4226,
169829
+ "step": 55380
169830
+ },
169831
+ {
169832
+ "epoch": 443.07,
169833
+ "learning_rate": 9.128782051282052e-06,
169834
+ "loss": 0.3028,
169835
+ "step": 55385
169836
+ },
169837
+ {
169838
+ "epoch": 443.11,
169839
+ "learning_rate": 9.128701923076923e-06,
169840
+ "loss": 0.3304,
169841
+ "step": 55390
169842
+ },
169843
+ {
169844
+ "epoch": 443.15,
169845
+ "learning_rate": 9.128621794871796e-06,
169846
+ "loss": 0.4673,
169847
+ "step": 55395
169848
+ },
169849
+ {
169850
+ "epoch": 443.19,
169851
+ "learning_rate": 9.128541666666668e-06,
169852
+ "loss": 0.9796,
169853
+ "step": 55400
169854
+ },
169855
+ {
169856
+ "epoch": 443.23,
169857
+ "learning_rate": 9.128461538461539e-06,
169858
+ "loss": 0.6241,
169859
+ "step": 55405
169860
+ },
169861
+ {
169862
+ "epoch": 443.27,
169863
+ "learning_rate": 9.12838141025641e-06,
169864
+ "loss": 0.3519,
169865
+ "step": 55410
169866
+ },
169867
+ {
169868
+ "epoch": 443.31,
169869
+ "learning_rate": 9.128301282051283e-06,
169870
+ "loss": 0.2897,
169871
+ "step": 55415
169872
+ },
169873
+ {
169874
+ "epoch": 443.35,
169875
+ "learning_rate": 9.128221153846155e-06,
169876
+ "loss": 0.4566,
169877
+ "step": 55420
169878
+ },
169879
+ {
169880
+ "epoch": 443.39,
169881
+ "learning_rate": 9.128141025641026e-06,
169882
+ "loss": 0.9614,
169883
+ "step": 55425
169884
+ },
169885
+ {
169886
+ "epoch": 443.43,
169887
+ "learning_rate": 9.128060897435899e-06,
169888
+ "loss": 0.7686,
169889
+ "step": 55430
169890
+ },
169891
+ {
169892
+ "epoch": 443.47,
169893
+ "learning_rate": 9.12798076923077e-06,
169894
+ "loss": 0.3547,
169895
+ "step": 55435
169896
+ },
169897
+ {
169898
+ "epoch": 443.51,
169899
+ "learning_rate": 9.127900641025642e-06,
169900
+ "loss": 0.3559,
169901
+ "step": 55440
169902
+ },
169903
+ {
169904
+ "epoch": 443.55,
169905
+ "learning_rate": 9.127820512820513e-06,
169906
+ "loss": 0.5024,
169907
+ "step": 55445
169908
+ },
169909
+ {
169910
+ "epoch": 443.59,
169911
+ "learning_rate": 9.127740384615386e-06,
169912
+ "loss": 0.9951,
169913
+ "step": 55450
169914
+ },
169915
+ {
169916
+ "epoch": 443.63,
169917
+ "learning_rate": 9.127660256410258e-06,
169918
+ "loss": 0.754,
169919
+ "step": 55455
169920
+ },
169921
+ {
169922
+ "epoch": 443.67,
169923
+ "learning_rate": 9.127580128205129e-06,
169924
+ "loss": 0.3864,
169925
+ "step": 55460
169926
+ },
169927
+ {
169928
+ "epoch": 443.71,
169929
+ "learning_rate": 9.1275e-06,
169930
+ "loss": 0.3745,
169931
+ "step": 55465
169932
+ },
169933
+ {
169934
+ "epoch": 443.75,
169935
+ "learning_rate": 9.127419871794873e-06,
169936
+ "loss": 0.4729,
169937
+ "step": 55470
169938
+ },
169939
+ {
169940
+ "epoch": 443.79,
169941
+ "learning_rate": 9.127339743589745e-06,
169942
+ "loss": 1.1605,
169943
+ "step": 55475
169944
+ },
169945
+ {
169946
+ "epoch": 443.83,
169947
+ "learning_rate": 9.127259615384616e-06,
169948
+ "loss": 0.7502,
169949
+ "step": 55480
169950
+ },
169951
+ {
169952
+ "epoch": 443.87,
169953
+ "learning_rate": 9.127179487179489e-06,
169954
+ "loss": 0.3047,
169955
+ "step": 55485
169956
+ },
169957
+ {
169958
+ "epoch": 443.91,
169959
+ "learning_rate": 9.127099358974359e-06,
169960
+ "loss": 0.4334,
169961
+ "step": 55490
169962
+ },
169963
+ {
169964
+ "epoch": 443.95,
169965
+ "learning_rate": 9.127019230769232e-06,
169966
+ "loss": 0.4476,
169967
+ "step": 55495
169968
+ },
169969
+ {
169970
+ "epoch": 443.99,
169971
+ "learning_rate": 9.126939102564103e-06,
169972
+ "loss": 1.1088,
169973
+ "step": 55500
169974
+ },
169975
+ {
169976
+ "epoch": 444.0,
169977
+ "eval_loss": 0.3665480613708496,
169978
+ "eval_runtime": 39.3106,
169979
+ "eval_samples_per_second": 21.368,
169980
+ "eval_steps_per_second": 0.687,
169981
+ "eval_wer": 0.18631571186315712,
169982
+ "step": 55501
169983
  }
169984
  ],
169985
  "max_steps": 625000,
169986
  "num_train_epochs": 5000,
169987
+ "total_flos": 1.5618392760037972e+20,
169988
  "trial_name": null,
169989
  "trial_params": null
169990
  }
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55501}/training_args.bin RENAMED
File without changes
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/config.json RENAMED
File without changes
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/optimizer.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:9a127b791c0d77b579a5e491584452669368364880890f5818cdc61e22e15559
3
  size 722165009
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2ca31c2c300395945cb22ec67409afcfb366052eba132541ae008b64617d5ab2
3
  size 722165009
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/preprocessor_config.json RENAMED
File without changes
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55750}/pytorch_model.bin RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:4ad7bfa19a307825777f0d973e4d3894ebe0ab9f5e474e9337514096826b79b2
3
  size 377909911
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ddd1158ad35ce7fae62714509684411831232f25e68b41babb69369b967ac829
3
  size 377909911
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55750}/rng_state.pth RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:df6b787e2064a641feeeec680cf7d039f7d81c7e51db0009bb53c63dd07dabf9
3
- size 14567
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b820a2f4a911cae7a26f7e1aed58b34ebaaa69de39e21788357c8e9b5ff13d2
3
+ size 14503
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55750}/scaler.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:1a85bd7e9d0b1ebc11f5f5c5604858f3dbf9d33e447f24534a89df3b0a46a7a1
3
  size 559
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b7a273902721759f6ffebc392616ef665d670925e0d8a8d948f2c697a000f2b1
3
  size 559
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/scheduler.pt RENAMED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:e9bca6cf836e457ed4761e4f9061cf21b21bdda6e37eae835e5df9405f49e54a
3
  size 623
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b240be141c87ff668e9478780fc08a2d2b5d6f694d05b7fd787237e4763b992d
3
  size 623
model-bin/finetune/base/{checkpoint-52515 → checkpoint-55750}/trainer_state.json RENAMED
@@ -1,8 +1,8 @@
1
  {
2
- "best_metric": 0.18992848189928482,
3
- "best_model_checkpoint": "./model-bin/finetune/base/checkpoint-52515",
4
- "epoch": 420.0,
5
- "global_step": 52515,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
@@ -166182,11 +166182,4127 @@
166182
  "eval_steps_per_second": 0.7,
166183
  "eval_wer": 0.18992848189928482,
166184
  "step": 52515
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
166185
  }
166186
  ],
166187
- "max_steps": 625000,
166188
  "num_train_epochs": 5000,
166189
- "total_flos": 1.4776861135717535e+20,
166190
  "trial_name": null,
166191
  "trial_params": null
166192
  }
 
1
  {
2
+ "best_metric": 0.18631571186315712,
3
+ "best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
4
+ "epoch": 448.99598393574297,
5
+ "global_step": 55750,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
 
166182
  "eval_steps_per_second": 0.7,
166183
  "eval_wer": 0.18992848189928482,
166184
  "step": 52515
166185
+ },
166186
+ {
166187
+ "epoch": 423.04,
166188
+ "learning_rate": 9.174663461538462e-06,
166189
+ "loss": 0.4337,
166190
+ "step": 52520
166191
+ },
166192
+ {
166193
+ "epoch": 423.08,
166194
+ "learning_rate": 9.174583333333335e-06,
166195
+ "loss": 0.3267,
166196
+ "step": 52525
166197
+ },
166198
+ {
166199
+ "epoch": 423.12,
166200
+ "learning_rate": 9.174503205128205e-06,
166201
+ "loss": 0.4162,
166202
+ "step": 52530
166203
+ },
166204
+ {
166205
+ "epoch": 423.16,
166206
+ "learning_rate": 9.174423076923078e-06,
166207
+ "loss": 0.6651,
166208
+ "step": 52535
166209
+ },
166210
+ {
166211
+ "epoch": 423.2,
166212
+ "learning_rate": 9.17434294871795e-06,
166213
+ "loss": 1.3276,
166214
+ "step": 52540
166215
+ },
166216
+ {
166217
+ "epoch": 423.24,
166218
+ "learning_rate": 9.17426282051282e-06,
166219
+ "loss": 0.3578,
166220
+ "step": 52545
166221
+ },
166222
+ {
166223
+ "epoch": 423.28,
166224
+ "learning_rate": 9.174182692307692e-06,
166225
+ "loss": 0.3429,
166226
+ "step": 52550
166227
+ },
166228
+ {
166229
+ "epoch": 423.32,
166230
+ "learning_rate": 9.174102564102565e-06,
166231
+ "loss": 0.3954,
166232
+ "step": 52555
166233
+ },
166234
+ {
166235
+ "epoch": 423.36,
166236
+ "learning_rate": 9.174022435897436e-06,
166237
+ "loss": 0.5291,
166238
+ "step": 52560
166239
+ },
166240
+ {
166241
+ "epoch": 423.4,
166242
+ "learning_rate": 9.173942307692308e-06,
166243
+ "loss": 1.3878,
166244
+ "step": 52565
166245
+ },
166246
+ {
166247
+ "epoch": 423.44,
166248
+ "learning_rate": 9.17386217948718e-06,
166249
+ "loss": 0.3056,
166250
+ "step": 52570
166251
+ },
166252
+ {
166253
+ "epoch": 423.48,
166254
+ "learning_rate": 9.173782051282052e-06,
166255
+ "loss": 0.4103,
166256
+ "step": 52575
166257
+ },
166258
+ {
166259
+ "epoch": 423.52,
166260
+ "learning_rate": 9.173701923076924e-06,
166261
+ "loss": 0.3949,
166262
+ "step": 52580
166263
+ },
166264
+ {
166265
+ "epoch": 423.56,
166266
+ "learning_rate": 9.173621794871795e-06,
166267
+ "loss": 0.5421,
166268
+ "step": 52585
166269
+ },
166270
+ {
166271
+ "epoch": 423.6,
166272
+ "learning_rate": 9.173541666666668e-06,
166273
+ "loss": 1.2796,
166274
+ "step": 52590
166275
+ },
166276
+ {
166277
+ "epoch": 423.64,
166278
+ "learning_rate": 9.17346153846154e-06,
166279
+ "loss": 0.4487,
166280
+ "step": 52595
166281
+ },
166282
+ {
166283
+ "epoch": 423.68,
166284
+ "learning_rate": 9.17338141025641e-06,
166285
+ "loss": 0.3143,
166286
+ "step": 52600
166287
+ },
166288
+ {
166289
+ "epoch": 423.72,
166290
+ "learning_rate": 9.173301282051282e-06,
166291
+ "loss": 0.3283,
166292
+ "step": 52605
166293
+ },
166294
+ {
166295
+ "epoch": 423.76,
166296
+ "learning_rate": 9.173221153846155e-06,
166297
+ "loss": 0.6825,
166298
+ "step": 52610
166299
+ },
166300
+ {
166301
+ "epoch": 423.8,
166302
+ "learning_rate": 9.173141025641026e-06,
166303
+ "loss": 1.3399,
166304
+ "step": 52615
166305
+ },
166306
+ {
166307
+ "epoch": 423.84,
166308
+ "learning_rate": 9.173060897435898e-06,
166309
+ "loss": 0.4315,
166310
+ "step": 52620
166311
+ },
166312
+ {
166313
+ "epoch": 423.88,
166314
+ "learning_rate": 9.172980769230771e-06,
166315
+ "loss": 0.3595,
166316
+ "step": 52625
166317
+ },
166318
+ {
166319
+ "epoch": 423.92,
166320
+ "learning_rate": 9.172900641025642e-06,
166321
+ "loss": 0.3945,
166322
+ "step": 52630
166323
+ },
166324
+ {
166325
+ "epoch": 423.96,
166326
+ "learning_rate": 9.172820512820514e-06,
166327
+ "loss": 0.5435,
166328
+ "step": 52635
166329
+ },
166330
+ {
166331
+ "epoch": 424.0,
166332
+ "eval_loss": 0.4003154933452606,
166333
+ "eval_runtime": 40.453,
166334
+ "eval_samples_per_second": 20.765,
166335
+ "eval_steps_per_second": 0.667,
166336
+ "eval_wer": 0.20118429709774105,
166337
+ "step": 52639
166338
+ },
166339
+ {
166340
+ "epoch": 421.01,
166341
+ "learning_rate": 9.172740384615385e-06,
166342
+ "loss": 0.3764,
166343
+ "step": 52640
166344
+ },
166345
+ {
166346
+ "epoch": 421.05,
166347
+ "learning_rate": 9.172660256410258e-06,
166348
+ "loss": 0.4723,
166349
+ "step": 52645
166350
+ },
166351
+ {
166352
+ "epoch": 421.09,
166353
+ "learning_rate": 9.17258012820513e-06,
166354
+ "loss": 0.3263,
166355
+ "step": 52650
166356
+ },
166357
+ {
166358
+ "epoch": 421.13,
166359
+ "learning_rate": 9.1725e-06,
166360
+ "loss": 0.3414,
166361
+ "step": 52655
166362
+ },
166363
+ {
166364
+ "epoch": 421.17,
166365
+ "learning_rate": 9.172419871794872e-06,
166366
+ "loss": 0.5096,
166367
+ "step": 52660
166368
+ },
166369
+ {
166370
+ "epoch": 421.21,
166371
+ "learning_rate": 9.172339743589745e-06,
166372
+ "loss": 1.3892,
166373
+ "step": 52665
166374
+ },
166375
+ {
166376
+ "epoch": 421.25,
166377
+ "learning_rate": 9.172259615384616e-06,
166378
+ "loss": 0.3895,
166379
+ "step": 52670
166380
+ },
166381
+ {
166382
+ "epoch": 421.29,
166383
+ "learning_rate": 9.172179487179488e-06,
166384
+ "loss": 0.3716,
166385
+ "step": 52675
166386
+ },
166387
+ {
166388
+ "epoch": 421.33,
166389
+ "learning_rate": 9.172099358974361e-06,
166390
+ "loss": 0.3533,
166391
+ "step": 52680
166392
+ },
166393
+ {
166394
+ "epoch": 421.37,
166395
+ "learning_rate": 9.17201923076923e-06,
166396
+ "loss": 0.6364,
166397
+ "step": 52685
166398
+ },
166399
+ {
166400
+ "epoch": 421.41,
166401
+ "learning_rate": 9.171939102564104e-06,
166402
+ "loss": 1.2984,
166403
+ "step": 52690
166404
+ },
166405
+ {
166406
+ "epoch": 421.45,
166407
+ "learning_rate": 9.171858974358975e-06,
166408
+ "loss": 0.367,
166409
+ "step": 52695
166410
+ },
166411
+ {
166412
+ "epoch": 421.49,
166413
+ "learning_rate": 9.171778846153846e-06,
166414
+ "loss": 0.3319,
166415
+ "step": 52700
166416
+ },
166417
+ {
166418
+ "epoch": 421.53,
166419
+ "learning_rate": 9.171698717948718e-06,
166420
+ "loss": 0.3717,
166421
+ "step": 52705
166422
+ },
166423
+ {
166424
+ "epoch": 421.57,
166425
+ "learning_rate": 9.17161858974359e-06,
166426
+ "loss": 0.5638,
166427
+ "step": 52710
166428
+ },
166429
+ {
166430
+ "epoch": 421.61,
166431
+ "learning_rate": 9.171538461538462e-06,
166432
+ "loss": 1.1557,
166433
+ "step": 52715
166434
+ },
166435
+ {
166436
+ "epoch": 421.65,
166437
+ "learning_rate": 9.171458333333333e-06,
166438
+ "loss": 0.4361,
166439
+ "step": 52720
166440
+ },
166441
+ {
166442
+ "epoch": 421.69,
166443
+ "learning_rate": 9.171378205128206e-06,
166444
+ "loss": 0.3131,
166445
+ "step": 52725
166446
+ },
166447
+ {
166448
+ "epoch": 421.73,
166449
+ "learning_rate": 9.171298076923078e-06,
166450
+ "loss": 0.3745,
166451
+ "step": 52730
166452
+ },
166453
+ {
166454
+ "epoch": 421.77,
166455
+ "learning_rate": 9.17121794871795e-06,
166456
+ "loss": 0.5119,
166457
+ "step": 52735
166458
+ },
166459
+ {
166460
+ "epoch": 421.81,
166461
+ "learning_rate": 9.17113782051282e-06,
166462
+ "loss": 1.1258,
166463
+ "step": 52740
166464
+ },
166465
+ {
166466
+ "epoch": 421.85,
166467
+ "learning_rate": 9.171057692307694e-06,
166468
+ "loss": 0.3384,
166469
+ "step": 52745
166470
+ },
166471
+ {
166472
+ "epoch": 421.89,
166473
+ "learning_rate": 9.170977564102565e-06,
166474
+ "loss": 0.3044,
166475
+ "step": 52750
166476
+ },
166477
+ {
166478
+ "epoch": 421.93,
166479
+ "learning_rate": 9.170897435897436e-06,
166480
+ "loss": 0.4152,
166481
+ "step": 52755
166482
+ },
166483
+ {
166484
+ "epoch": 421.97,
166485
+ "learning_rate": 9.170817307692308e-06,
166486
+ "loss": 0.605,
166487
+ "step": 52760
166488
+ },
166489
+ {
166490
+ "epoch": 422.0,
166491
+ "eval_loss": 0.4416767954826355,
166492
+ "eval_runtime": 38.8014,
166493
+ "eval_samples_per_second": 21.649,
166494
+ "eval_steps_per_second": 0.696,
166495
+ "eval_wer": 0.20573733751680862,
166496
+ "step": 52764
166497
+ },
166498
+ {
166499
+ "epoch": 425.01,
166500
+ "learning_rate": 9.17073717948718e-06,
166501
+ "loss": 0.4205,
166502
+ "step": 52765
166503
+ },
166504
+ {
166505
+ "epoch": 425.05,
166506
+ "learning_rate": 9.170657051282052e-06,
166507
+ "loss": 0.3089,
166508
+ "step": 52770
166509
+ },
166510
+ {
166511
+ "epoch": 425.09,
166512
+ "learning_rate": 9.170576923076923e-06,
166513
+ "loss": 0.3435,
166514
+ "step": 52775
166515
+ },
166516
+ {
166517
+ "epoch": 425.13,
166518
+ "learning_rate": 9.170496794871797e-06,
166519
+ "loss": 0.3688,
166520
+ "step": 52780
166521
+ },
166522
+ {
166523
+ "epoch": 425.17,
166524
+ "learning_rate": 9.170416666666668e-06,
166525
+ "loss": 0.6443,
166526
+ "step": 52785
166527
+ },
166528
+ {
166529
+ "epoch": 425.21,
166530
+ "learning_rate": 9.17033653846154e-06,
166531
+ "loss": 1.2337,
166532
+ "step": 52790
166533
+ },
166534
+ {
166535
+ "epoch": 425.25,
166536
+ "learning_rate": 9.17025641025641e-06,
166537
+ "loss": 0.3467,
166538
+ "step": 52795
166539
+ },
166540
+ {
166541
+ "epoch": 425.29,
166542
+ "learning_rate": 9.170176282051284e-06,
166543
+ "loss": 0.3054,
166544
+ "step": 52800
166545
+ },
166546
+ {
166547
+ "epoch": 425.33,
166548
+ "learning_rate": 9.170096153846153e-06,
166549
+ "loss": 0.3231,
166550
+ "step": 52805
166551
+ },
166552
+ {
166553
+ "epoch": 425.37,
166554
+ "learning_rate": 9.170016025641026e-06,
166555
+ "loss": 0.5788,
166556
+ "step": 52810
166557
+ },
166558
+ {
166559
+ "epoch": 425.41,
166560
+ "learning_rate": 9.1699358974359e-06,
166561
+ "loss": 1.4097,
166562
+ "step": 52815
166563
+ },
166564
+ {
166565
+ "epoch": 425.45,
166566
+ "learning_rate": 9.169855769230769e-06,
166567
+ "loss": 0.3334,
166568
+ "step": 52820
166569
+ },
166570
+ {
166571
+ "epoch": 425.49,
166572
+ "learning_rate": 9.169775641025642e-06,
166573
+ "loss": 0.3256,
166574
+ "step": 52825
166575
+ },
166576
+ {
166577
+ "epoch": 425.53,
166578
+ "learning_rate": 9.169695512820513e-06,
166579
+ "loss": 0.3063,
166580
+ "step": 52830
166581
+ },
166582
+ {
166583
+ "epoch": 425.57,
166584
+ "learning_rate": 9.169615384615387e-06,
166585
+ "loss": 0.6159,
166586
+ "step": 52835
166587
+ },
166588
+ {
166589
+ "epoch": 425.61,
166590
+ "learning_rate": 9.169535256410256e-06,
166591
+ "loss": 1.2831,
166592
+ "step": 52840
166593
+ },
166594
+ {
166595
+ "epoch": 425.65,
166596
+ "learning_rate": 9.16945512820513e-06,
166597
+ "loss": 0.3114,
166598
+ "step": 52845
166599
+ },
166600
+ {
166601
+ "epoch": 425.69,
166602
+ "learning_rate": 9.169375e-06,
166603
+ "loss": 0.3027,
166604
+ "step": 52850
166605
+ },
166606
+ {
166607
+ "epoch": 425.73,
166608
+ "learning_rate": 9.169294871794872e-06,
166609
+ "loss": 0.4015,
166610
+ "step": 52855
166611
+ },
166612
+ {
166613
+ "epoch": 425.77,
166614
+ "learning_rate": 9.169214743589743e-06,
166615
+ "loss": 0.6918,
166616
+ "step": 52860
166617
+ },
166618
+ {
166619
+ "epoch": 425.81,
166620
+ "learning_rate": 9.169134615384616e-06,
166621
+ "loss": 1.3139,
166622
+ "step": 52865
166623
+ },
166624
+ {
166625
+ "epoch": 425.85,
166626
+ "learning_rate": 9.169054487179488e-06,
166627
+ "loss": 0.3231,
166628
+ "step": 52870
166629
+ },
166630
+ {
166631
+ "epoch": 425.89,
166632
+ "learning_rate": 9.168974358974359e-06,
166633
+ "loss": 0.3526,
166634
+ "step": 52875
166635
+ },
166636
+ {
166637
+ "epoch": 425.93,
166638
+ "learning_rate": 9.168894230769232e-06,
166639
+ "loss": 0.3667,
166640
+ "step": 52880
166641
+ },
166642
+ {
166643
+ "epoch": 425.97,
166644
+ "learning_rate": 9.168814102564104e-06,
166645
+ "loss": 0.6799,
166646
+ "step": 52885
166647
+ },
166648
+ {
166649
+ "epoch": 426.0,
166650
+ "eval_loss": 0.4513387680053711,
166651
+ "eval_runtime": 39.7893,
166652
+ "eval_samples_per_second": 21.086,
166653
+ "eval_steps_per_second": 0.679,
166654
+ "eval_wer": 0.20593128390596746,
166655
+ "step": 52888
166656
+ },
166657
+ {
166658
+ "epoch": 423.02,
166659
+ "learning_rate": 9.168733974358975e-06,
166660
+ "loss": 0.5276,
166661
+ "step": 52890
166662
+ },
166663
+ {
166664
+ "epoch": 423.06,
166665
+ "learning_rate": 9.168653846153846e-06,
166666
+ "loss": 0.3482,
166667
+ "step": 52895
166668
+ },
166669
+ {
166670
+ "epoch": 423.1,
166671
+ "learning_rate": 9.16857371794872e-06,
166672
+ "loss": 0.3158,
166673
+ "step": 52900
166674
+ },
166675
+ {
166676
+ "epoch": 423.14,
166677
+ "learning_rate": 9.16849358974359e-06,
166678
+ "loss": 0.3622,
166679
+ "step": 52905
166680
+ },
166681
+ {
166682
+ "epoch": 423.18,
166683
+ "learning_rate": 9.168413461538462e-06,
166684
+ "loss": 0.7214,
166685
+ "step": 52910
166686
+ },
166687
+ {
166688
+ "epoch": 423.22,
166689
+ "learning_rate": 9.168333333333333e-06,
166690
+ "loss": 1.0821,
166691
+ "step": 52915
166692
+ },
166693
+ {
166694
+ "epoch": 423.26,
166695
+ "learning_rate": 9.168253205128206e-06,
166696
+ "loss": 0.3709,
166697
+ "step": 52920
166698
+ },
166699
+ {
166700
+ "epoch": 423.3,
166701
+ "learning_rate": 9.168173076923078e-06,
166702
+ "loss": 0.3363,
166703
+ "step": 52925
166704
+ },
166705
+ {
166706
+ "epoch": 423.34,
166707
+ "learning_rate": 9.168092948717949e-06,
166708
+ "loss": 0.3673,
166709
+ "step": 52930
166710
+ },
166711
+ {
166712
+ "epoch": 423.38,
166713
+ "learning_rate": 9.168012820512822e-06,
166714
+ "loss": 0.7023,
166715
+ "step": 52935
166716
+ },
166717
+ {
166718
+ "epoch": 423.42,
166719
+ "learning_rate": 9.167932692307694e-06,
166720
+ "loss": 1.1365,
166721
+ "step": 52940
166722
+ },
166723
+ {
166724
+ "epoch": 423.46,
166725
+ "learning_rate": 9.167852564102565e-06,
166726
+ "loss": 0.3282,
166727
+ "step": 52945
166728
+ },
166729
+ {
166730
+ "epoch": 423.5,
166731
+ "learning_rate": 9.167772435897436e-06,
166732
+ "loss": 0.3293,
166733
+ "step": 52950
166734
+ },
166735
+ {
166736
+ "epoch": 423.54,
166737
+ "learning_rate": 9.16769230769231e-06,
166738
+ "loss": 0.4172,
166739
+ "step": 52955
166740
+ },
166741
+ {
166742
+ "epoch": 423.58,
166743
+ "learning_rate": 9.167612179487179e-06,
166744
+ "loss": 0.6418,
166745
+ "step": 52960
166746
+ },
166747
+ {
166748
+ "epoch": 423.62,
166749
+ "learning_rate": 9.167532051282052e-06,
166750
+ "loss": 1.0752,
166751
+ "step": 52965
166752
+ },
166753
+ {
166754
+ "epoch": 423.66,
166755
+ "learning_rate": 9.167451923076925e-06,
166756
+ "loss": 0.365,
166757
+ "step": 52970
166758
+ },
166759
+ {
166760
+ "epoch": 423.7,
166761
+ "learning_rate": 9.167371794871795e-06,
166762
+ "loss": 0.3403,
166763
+ "step": 52975
166764
+ },
166765
+ {
166766
+ "epoch": 423.74,
166767
+ "learning_rate": 9.167291666666668e-06,
166768
+ "loss": 0.3715,
166769
+ "step": 52980
166770
+ },
166771
+ {
166772
+ "epoch": 423.78,
166773
+ "learning_rate": 9.16721153846154e-06,
166774
+ "loss": 0.6747,
166775
+ "step": 52985
166776
+ },
166777
+ {
166778
+ "epoch": 423.82,
166779
+ "learning_rate": 9.16713141025641e-06,
166780
+ "loss": 1.2286,
166781
+ "step": 52990
166782
+ },
166783
+ {
166784
+ "epoch": 423.86,
166785
+ "learning_rate": 9.167051282051282e-06,
166786
+ "loss": 0.3437,
166787
+ "step": 52995
166788
+ },
166789
+ {
166790
+ "epoch": 423.9,
166791
+ "learning_rate": 9.166971153846155e-06,
166792
+ "loss": 0.3155,
166793
+ "step": 53000
166794
+ },
166795
+ {
166796
+ "epoch": 423.94,
166797
+ "learning_rate": 9.166891025641026e-06,
166798
+ "loss": 0.3946,
166799
+ "step": 53005
166800
+ },
166801
+ {
166802
+ "epoch": 423.98,
166803
+ "learning_rate": 9.166810897435898e-06,
166804
+ "loss": 0.7243,
166805
+ "step": 53010
166806
+ },
166807
+ {
166808
+ "epoch": 424.0,
166809
+ "eval_loss": 0.37052494287490845,
166810
+ "eval_runtime": 39.814,
166811
+ "eval_samples_per_second": 21.073,
166812
+ "eval_steps_per_second": 0.678,
166813
+ "eval_wer": 0.19977989728539985,
166814
+ "step": 53013
166815
+ },
166816
+ {
166817
+ "epoch": 427.02,
166818
+ "learning_rate": 9.166730769230769e-06,
166819
+ "loss": 0.4042,
166820
+ "step": 53015
166821
+ },
166822
+ {
166823
+ "epoch": 427.06,
166824
+ "learning_rate": 9.166650641025642e-06,
166825
+ "loss": 0.3449,
166826
+ "step": 53020
166827
+ },
166828
+ {
166829
+ "epoch": 427.1,
166830
+ "learning_rate": 9.166570512820513e-06,
166831
+ "loss": 0.3405,
166832
+ "step": 53025
166833
+ },
166834
+ {
166835
+ "epoch": 427.14,
166836
+ "learning_rate": 9.166490384615385e-06,
166837
+ "loss": 0.411,
166838
+ "step": 53030
166839
+ },
166840
+ {
166841
+ "epoch": 427.18,
166842
+ "learning_rate": 9.166410256410258e-06,
166843
+ "loss": 0.5731,
166844
+ "step": 53035
166845
+ },
166846
+ {
166847
+ "epoch": 427.22,
166848
+ "learning_rate": 9.16633012820513e-06,
166849
+ "loss": 1.0737,
166850
+ "step": 53040
166851
+ },
166852
+ {
166853
+ "epoch": 427.26,
166854
+ "learning_rate": 9.16625e-06,
166855
+ "loss": 0.3384,
166856
+ "step": 53045
166857
+ },
166858
+ {
166859
+ "epoch": 427.3,
166860
+ "learning_rate": 9.166169871794872e-06,
166861
+ "loss": 0.3478,
166862
+ "step": 53050
166863
+ },
166864
+ {
166865
+ "epoch": 427.34,
166866
+ "learning_rate": 9.166089743589745e-06,
166867
+ "loss": 0.3982,
166868
+ "step": 53055
166869
+ },
166870
+ {
166871
+ "epoch": 427.38,
166872
+ "learning_rate": 9.166009615384616e-06,
166873
+ "loss": 0.6412,
166874
+ "step": 53060
166875
+ },
166876
+ {
166877
+ "epoch": 427.42,
166878
+ "learning_rate": 9.165929487179488e-06,
166879
+ "loss": 1.1081,
166880
+ "step": 53065
166881
+ },
166882
+ {
166883
+ "epoch": 427.46,
166884
+ "learning_rate": 9.16584935897436e-06,
166885
+ "loss": 0.3373,
166886
+ "step": 53070
166887
+ },
166888
+ {
166889
+ "epoch": 427.5,
166890
+ "learning_rate": 9.165769230769232e-06,
166891
+ "loss": 0.3116,
166892
+ "step": 53075
166893
+ },
166894
+ {
166895
+ "epoch": 427.54,
166896
+ "learning_rate": 9.165689102564103e-06,
166897
+ "loss": 0.482,
166898
+ "step": 53080
166899
+ },
166900
+ {
166901
+ "epoch": 427.58,
166902
+ "learning_rate": 9.165608974358975e-06,
166903
+ "loss": 0.6877,
166904
+ "step": 53085
166905
+ },
166906
+ {
166907
+ "epoch": 427.62,
166908
+ "learning_rate": 9.165528846153848e-06,
166909
+ "loss": 0.9501,
166910
+ "step": 53090
166911
+ },
166912
+ {
166913
+ "epoch": 427.66,
166914
+ "learning_rate": 9.16544871794872e-06,
166915
+ "loss": 0.4583,
166916
+ "step": 53095
166917
+ },
166918
+ {
166919
+ "epoch": 427.7,
166920
+ "learning_rate": 9.16536858974359e-06,
166921
+ "loss": 0.4234,
166922
+ "step": 53100
166923
+ },
166924
+ {
166925
+ "epoch": 427.74,
166926
+ "learning_rate": 9.165288461538462e-06,
166927
+ "loss": 0.4676,
166928
+ "step": 53105
166929
+ },
166930
+ {
166931
+ "epoch": 427.78,
166932
+ "learning_rate": 9.165208333333335e-06,
166933
+ "loss": 0.7411,
166934
+ "step": 53110
166935
+ },
166936
+ {
166937
+ "epoch": 427.82,
166938
+ "learning_rate": 9.165128205128205e-06,
166939
+ "loss": 1.0585,
166940
+ "step": 53115
166941
+ },
166942
+ {
166943
+ "epoch": 427.86,
166944
+ "learning_rate": 9.165048076923078e-06,
166945
+ "loss": 0.3383,
166946
+ "step": 53120
166947
+ },
166948
+ {
166949
+ "epoch": 427.9,
166950
+ "learning_rate": 9.16496794871795e-06,
166951
+ "loss": 0.3084,
166952
+ "step": 53125
166953
+ },
166954
+ {
166955
+ "epoch": 427.94,
166956
+ "learning_rate": 9.16488782051282e-06,
166957
+ "loss": 0.3699,
166958
+ "step": 53130
166959
+ },
166960
+ {
166961
+ "epoch": 427.98,
166962
+ "learning_rate": 9.164807692307693e-06,
166963
+ "loss": 0.7039,
166964
+ "step": 53135
166965
+ },
166966
+ {
166967
+ "epoch": 428.0,
166968
+ "eval_loss": 0.46915048360824585,
166969
+ "eval_runtime": 39.008,
166970
+ "eval_samples_per_second": 21.534,
166971
+ "eval_steps_per_second": 0.692,
166972
+ "eval_wer": 0.20150507781106158,
166973
+ "step": 53137
166974
+ },
166975
+ {
166976
+ "epoch": 428.02,
166977
+ "learning_rate": 9.164727564102565e-06,
166978
+ "loss": 0.4169,
166979
+ "step": 53140
166980
+ },
166981
+ {
166982
+ "epoch": 428.06,
166983
+ "learning_rate": 9.164647435897436e-06,
166984
+ "loss": 0.4001,
166985
+ "step": 53145
166986
+ },
166987
+ {
166988
+ "epoch": 428.1,
166989
+ "learning_rate": 9.164567307692308e-06,
166990
+ "loss": 0.3144,
166991
+ "step": 53150
166992
+ },
166993
+ {
166994
+ "epoch": 428.15,
166995
+ "learning_rate": 9.16448717948718e-06,
166996
+ "loss": 0.375,
166997
+ "step": 53155
166998
+ },
166999
+ {
167000
+ "epoch": 428.19,
167001
+ "learning_rate": 9.164407051282052e-06,
167002
+ "loss": 0.7092,
167003
+ "step": 53160
167004
+ },
167005
+ {
167006
+ "epoch": 428.23,
167007
+ "learning_rate": 9.164326923076923e-06,
167008
+ "loss": 0.8503,
167009
+ "step": 53165
167010
+ },
167011
+ {
167012
+ "epoch": 428.27,
167013
+ "learning_rate": 9.164246794871796e-06,
167014
+ "loss": 0.3192,
167015
+ "step": 53170
167016
+ },
167017
+ {
167018
+ "epoch": 428.31,
167019
+ "learning_rate": 9.164166666666668e-06,
167020
+ "loss": 0.3571,
167021
+ "step": 53175
167022
+ },
167023
+ {
167024
+ "epoch": 428.35,
167025
+ "learning_rate": 9.164086538461539e-06,
167026
+ "loss": 0.3936,
167027
+ "step": 53180
167028
+ },
167029
+ {
167030
+ "epoch": 428.39,
167031
+ "learning_rate": 9.16400641025641e-06,
167032
+ "loss": 0.8202,
167033
+ "step": 53185
167034
+ },
167035
+ {
167036
+ "epoch": 428.43,
167037
+ "learning_rate": 9.163926282051283e-06,
167038
+ "loss": 0.9116,
167039
+ "step": 53190
167040
+ },
167041
+ {
167042
+ "epoch": 428.47,
167043
+ "learning_rate": 9.163846153846155e-06,
167044
+ "loss": 0.301,
167045
+ "step": 53195
167046
+ },
167047
+ {
167048
+ "epoch": 428.51,
167049
+ "learning_rate": 9.163766025641026e-06,
167050
+ "loss": 0.3834,
167051
+ "step": 53200
167052
+ },
167053
+ {
167054
+ "epoch": 428.55,
167055
+ "learning_rate": 9.163685897435898e-06,
167056
+ "loss": 0.3736,
167057
+ "step": 53205
167058
+ },
167059
+ {
167060
+ "epoch": 428.59,
167061
+ "learning_rate": 9.16360576923077e-06,
167062
+ "loss": 0.8737,
167063
+ "step": 53210
167064
+ },
167065
+ {
167066
+ "epoch": 428.63,
167067
+ "learning_rate": 9.163525641025642e-06,
167068
+ "loss": 0.8672,
167069
+ "step": 53215
167070
+ },
167071
+ {
167072
+ "epoch": 428.67,
167073
+ "learning_rate": 9.163445512820513e-06,
167074
+ "loss": 0.3897,
167075
+ "step": 53220
167076
+ },
167077
+ {
167078
+ "epoch": 428.71,
167079
+ "learning_rate": 9.163365384615386e-06,
167080
+ "loss": 0.3164,
167081
+ "step": 53225
167082
+ },
167083
+ {
167084
+ "epoch": 428.75,
167085
+ "learning_rate": 9.163285256410258e-06,
167086
+ "loss": 0.4835,
167087
+ "step": 53230
167088
+ },
167089
+ {
167090
+ "epoch": 428.79,
167091
+ "learning_rate": 9.163205128205129e-06,
167092
+ "loss": 0.8753,
167093
+ "step": 53235
167094
+ },
167095
+ {
167096
+ "epoch": 428.83,
167097
+ "learning_rate": 9.163125e-06,
167098
+ "loss": 0.9728,
167099
+ "step": 53240
167100
+ },
167101
+ {
167102
+ "epoch": 428.87,
167103
+ "learning_rate": 9.163044871794874e-06,
167104
+ "loss": 0.3156,
167105
+ "step": 53245
167106
+ },
167107
+ {
167108
+ "epoch": 428.91,
167109
+ "learning_rate": 9.162964743589743e-06,
167110
+ "loss": 0.3698,
167111
+ "step": 53250
167112
+ },
167113
+ {
167114
+ "epoch": 428.95,
167115
+ "learning_rate": 9.162884615384616e-06,
167116
+ "loss": 0.379,
167117
+ "step": 53255
167118
+ },
167119
+ {
167120
+ "epoch": 428.99,
167121
+ "learning_rate": 9.162804487179488e-06,
167122
+ "loss": 1.0544,
167123
+ "step": 53260
167124
+ },
167125
+ {
167126
+ "epoch": 429.0,
167127
+ "eval_loss": 0.4316280484199524,
167128
+ "eval_runtime": 39.1434,
167129
+ "eval_samples_per_second": 21.46,
167130
+ "eval_steps_per_second": 0.69,
167131
+ "eval_wer": 0.19746457867263237,
167132
+ "step": 53261
167133
+ },
167134
+ {
167135
+ "epoch": 429.03,
167136
+ "learning_rate": 9.162724358974359e-06,
167137
+ "loss": 0.4907,
167138
+ "step": 53265
167139
+ },
167140
+ {
167141
+ "epoch": 429.07,
167142
+ "learning_rate": 9.162644230769232e-06,
167143
+ "loss": 0.413,
167144
+ "step": 53270
167145
+ },
167146
+ {
167147
+ "epoch": 429.11,
167148
+ "learning_rate": 9.162564102564103e-06,
167149
+ "loss": 0.3893,
167150
+ "step": 53275
167151
+ },
167152
+ {
167153
+ "epoch": 429.15,
167154
+ "learning_rate": 9.162483974358975e-06,
167155
+ "loss": 0.4385,
167156
+ "step": 53280
167157
+ },
167158
+ {
167159
+ "epoch": 429.19,
167160
+ "learning_rate": 9.162403846153846e-06,
167161
+ "loss": 1.2454,
167162
+ "step": 53285
167163
+ },
167164
+ {
167165
+ "epoch": 429.23,
167166
+ "learning_rate": 9.162323717948719e-06,
167167
+ "loss": 0.7755,
167168
+ "step": 53290
167169
+ },
167170
+ {
167171
+ "epoch": 429.27,
167172
+ "learning_rate": 9.16224358974359e-06,
167173
+ "loss": 0.3312,
167174
+ "step": 53295
167175
+ },
167176
+ {
167177
+ "epoch": 429.31,
167178
+ "learning_rate": 9.162163461538462e-06,
167179
+ "loss": 0.3771,
167180
+ "step": 53300
167181
+ },
167182
+ {
167183
+ "epoch": 429.35,
167184
+ "learning_rate": 9.162083333333333e-06,
167185
+ "loss": 0.4456,
167186
+ "step": 53305
167187
+ },
167188
+ {
167189
+ "epoch": 429.39,
167190
+ "learning_rate": 9.162003205128206e-06,
167191
+ "loss": 0.9052,
167192
+ "step": 53310
167193
+ },
167194
+ {
167195
+ "epoch": 429.43,
167196
+ "learning_rate": 9.161923076923078e-06,
167197
+ "loss": 0.7603,
167198
+ "step": 53315
167199
+ },
167200
+ {
167201
+ "epoch": 429.47,
167202
+ "learning_rate": 9.161842948717949e-06,
167203
+ "loss": 0.4158,
167204
+ "step": 53320
167205
+ },
167206
+ {
167207
+ "epoch": 429.51,
167208
+ "learning_rate": 9.161762820512822e-06,
167209
+ "loss": 0.3428,
167210
+ "step": 53325
167211
+ },
167212
+ {
167213
+ "epoch": 429.55,
167214
+ "learning_rate": 9.161682692307693e-06,
167215
+ "loss": 0.475,
167216
+ "step": 53330
167217
+ },
167218
+ {
167219
+ "epoch": 429.59,
167220
+ "learning_rate": 9.161602564102565e-06,
167221
+ "loss": 1.054,
167222
+ "step": 53335
167223
+ },
167224
+ {
167225
+ "epoch": 429.63,
167226
+ "learning_rate": 9.161522435897436e-06,
167227
+ "loss": 0.6962,
167228
+ "step": 53340
167229
+ },
167230
+ {
167231
+ "epoch": 429.67,
167232
+ "learning_rate": 9.16144230769231e-06,
167233
+ "loss": 0.3644,
167234
+ "step": 53345
167235
+ },
167236
+ {
167237
+ "epoch": 429.71,
167238
+ "learning_rate": 9.16136217948718e-06,
167239
+ "loss": 0.4176,
167240
+ "step": 53350
167241
+ },
167242
+ {
167243
+ "epoch": 429.76,
167244
+ "learning_rate": 9.161282051282052e-06,
167245
+ "loss": 0.48,
167246
+ "step": 53355
167247
+ },
167248
+ {
167249
+ "epoch": 429.8,
167250
+ "learning_rate": 9.161201923076923e-06,
167251
+ "loss": 0.9505,
167252
+ "step": 53360
167253
+ },
167254
+ {
167255
+ "epoch": 429.84,
167256
+ "learning_rate": 9.161121794871796e-06,
167257
+ "loss": 0.81,
167258
+ "step": 53365
167259
+ },
167260
+ {
167261
+ "epoch": 429.88,
167262
+ "learning_rate": 9.161041666666668e-06,
167263
+ "loss": 0.3829,
167264
+ "step": 53370
167265
+ },
167266
+ {
167267
+ "epoch": 429.92,
167268
+ "learning_rate": 9.160961538461539e-06,
167269
+ "loss": 0.3667,
167270
+ "step": 53375
167271
+ },
167272
+ {
167273
+ "epoch": 429.96,
167274
+ "learning_rate": 9.160881410256412e-06,
167275
+ "loss": 0.4709,
167276
+ "step": 53380
167277
+ },
167278
+ {
167279
+ "epoch": 430.0,
167280
+ "learning_rate": 9.160801282051283e-06,
167281
+ "loss": 1.0048,
167282
+ "step": 53385
167283
+ },
167284
+ {
167285
+ "epoch": 430.0,
167286
+ "eval_loss": 0.4243323802947998,
167287
+ "eval_runtime": 39.8421,
167288
+ "eval_samples_per_second": 21.083,
167289
+ "eval_steps_per_second": 0.678,
167290
+ "eval_wer": 0.20694374414751854,
167291
+ "step": 53385
167292
+ },
167293
+ {
167294
+ "epoch": 430.04,
167295
+ "learning_rate": 9.160721153846155e-06,
167296
+ "loss": 0.3929,
167297
+ "step": 53390
167298
+ },
167299
+ {
167300
+ "epoch": 430.08,
167301
+ "learning_rate": 9.160641025641026e-06,
167302
+ "loss": 0.2955,
167303
+ "step": 53395
167304
+ },
167305
+ {
167306
+ "epoch": 430.12,
167307
+ "learning_rate": 9.1605608974359e-06,
167308
+ "loss": 0.3902,
167309
+ "step": 53400
167310
+ },
167311
+ {
167312
+ "epoch": 430.16,
167313
+ "learning_rate": 9.160480769230769e-06,
167314
+ "loss": 0.4729,
167315
+ "step": 53405
167316
+ },
167317
+ {
167318
+ "epoch": 430.2,
167319
+ "learning_rate": 9.160400641025642e-06,
167320
+ "loss": 1.3877,
167321
+ "step": 53410
167322
+ },
167323
+ {
167324
+ "epoch": 430.24,
167325
+ "learning_rate": 9.160320512820513e-06,
167326
+ "loss": 0.4524,
167327
+ "step": 53415
167328
+ },
167329
+ {
167330
+ "epoch": 430.28,
167331
+ "learning_rate": 9.160240384615385e-06,
167332
+ "loss": 0.3233,
167333
+ "step": 53420
167334
+ },
167335
+ {
167336
+ "epoch": 430.32,
167337
+ "learning_rate": 9.160160256410258e-06,
167338
+ "loss": 0.4162,
167339
+ "step": 53425
167340
+ },
167341
+ {
167342
+ "epoch": 430.36,
167343
+ "learning_rate": 9.160080128205129e-06,
167344
+ "loss": 0.4888,
167345
+ "step": 53430
167346
+ },
167347
+ {
167348
+ "epoch": 430.4,
167349
+ "learning_rate": 9.16e-06,
167350
+ "loss": 1.219,
167351
+ "step": 53435
167352
+ },
167353
+ {
167354
+ "epoch": 430.44,
167355
+ "learning_rate": 9.159919871794872e-06,
167356
+ "loss": 0.3253,
167357
+ "step": 53440
167358
+ },
167359
+ {
167360
+ "epoch": 430.48,
167361
+ "learning_rate": 9.159839743589745e-06,
167362
+ "loss": 0.3146,
167363
+ "step": 53445
167364
+ },
167365
+ {
167366
+ "epoch": 430.52,
167367
+ "learning_rate": 9.159759615384616e-06,
167368
+ "loss": 0.3463,
167369
+ "step": 53450
167370
+ },
167371
+ {
167372
+ "epoch": 430.56,
167373
+ "learning_rate": 9.159679487179488e-06,
167374
+ "loss": 0.4703,
167375
+ "step": 53455
167376
+ },
167377
+ {
167378
+ "epoch": 430.6,
167379
+ "learning_rate": 9.159599358974359e-06,
167380
+ "loss": 1.2574,
167381
+ "step": 53460
167382
+ },
167383
+ {
167384
+ "epoch": 430.64,
167385
+ "learning_rate": 9.159519230769232e-06,
167386
+ "loss": 0.3584,
167387
+ "step": 53465
167388
+ },
167389
+ {
167390
+ "epoch": 430.68,
167391
+ "learning_rate": 9.159439102564103e-06,
167392
+ "loss": 0.3258,
167393
+ "step": 53470
167394
+ },
167395
+ {
167396
+ "epoch": 430.72,
167397
+ "learning_rate": 9.159358974358975e-06,
167398
+ "loss": 0.3546,
167399
+ "step": 53475
167400
+ },
167401
+ {
167402
+ "epoch": 430.76,
167403
+ "learning_rate": 9.159278846153848e-06,
167404
+ "loss": 0.4754,
167405
+ "step": 53480
167406
+ },
167407
+ {
167408
+ "epoch": 430.8,
167409
+ "learning_rate": 9.159198717948719e-06,
167410
+ "loss": 1.2657,
167411
+ "step": 53485
167412
+ },
167413
+ {
167414
+ "epoch": 430.84,
167415
+ "learning_rate": 9.15911858974359e-06,
167416
+ "loss": 0.3547,
167417
+ "step": 53490
167418
+ },
167419
+ {
167420
+ "epoch": 430.88,
167421
+ "learning_rate": 9.159038461538462e-06,
167422
+ "loss": 0.3136,
167423
+ "step": 53495
167424
+ },
167425
+ {
167426
+ "epoch": 430.92,
167427
+ "learning_rate": 9.158958333333335e-06,
167428
+ "loss": 0.397,
167429
+ "step": 53500
167430
+ },
167431
+ {
167432
+ "epoch": 430.96,
167433
+ "learning_rate": 9.158878205128206e-06,
167434
+ "loss": 0.5967,
167435
+ "step": 53505
167436
+ },
167437
+ {
167438
+ "epoch": 431.0,
167439
+ "eval_loss": 0.3771889805793762,
167440
+ "eval_runtime": 39.5555,
167441
+ "eval_samples_per_second": 21.236,
167442
+ "eval_steps_per_second": 0.683,
167443
+ "eval_wer": 0.19051832190518322,
167444
+ "step": 53509
167445
+ },
167446
+ {
167447
+ "epoch": 428.01,
167448
+ "learning_rate": 9.158798076923078e-06,
167449
+ "loss": 0.3878,
167450
+ "step": 53510
167451
+ },
167452
+ {
167453
+ "epoch": 428.05,
167454
+ "learning_rate": 9.158717948717949e-06,
167455
+ "loss": 0.3302,
167456
+ "step": 53515
167457
+ },
167458
+ {
167459
+ "epoch": 428.09,
167460
+ "learning_rate": 9.158637820512822e-06,
167461
+ "loss": 0.3114,
167462
+ "step": 53520
167463
+ },
167464
+ {
167465
+ "epoch": 428.13,
167466
+ "learning_rate": 9.158557692307693e-06,
167467
+ "loss": 0.3538,
167468
+ "step": 53525
167469
+ },
167470
+ {
167471
+ "epoch": 428.17,
167472
+ "learning_rate": 9.158477564102565e-06,
167473
+ "loss": 0.6061,
167474
+ "step": 53530
167475
+ },
167476
+ {
167477
+ "epoch": 428.21,
167478
+ "learning_rate": 9.158397435897438e-06,
167479
+ "loss": 1.3689,
167480
+ "step": 53535
167481
+ },
167482
+ {
167483
+ "epoch": 428.25,
167484
+ "learning_rate": 9.158317307692307e-06,
167485
+ "loss": 0.5001,
167486
+ "step": 53540
167487
+ },
167488
+ {
167489
+ "epoch": 428.29,
167490
+ "learning_rate": 9.15823717948718e-06,
167491
+ "loss": 0.3581,
167492
+ "step": 53545
167493
+ },
167494
+ {
167495
+ "epoch": 428.33,
167496
+ "learning_rate": 9.158157051282052e-06,
167497
+ "loss": 0.335,
167498
+ "step": 53550
167499
+ },
167500
+ {
167501
+ "epoch": 428.37,
167502
+ "learning_rate": 9.158076923076923e-06,
167503
+ "loss": 0.6611,
167504
+ "step": 53555
167505
+ },
167506
+ {
167507
+ "epoch": 428.41,
167508
+ "learning_rate": 9.157996794871795e-06,
167509
+ "loss": 1.2383,
167510
+ "step": 53560
167511
+ },
167512
+ {
167513
+ "epoch": 428.45,
167514
+ "learning_rate": 9.157916666666668e-06,
167515
+ "loss": 0.3176,
167516
+ "step": 53565
167517
+ },
167518
+ {
167519
+ "epoch": 428.49,
167520
+ "learning_rate": 9.157836538461539e-06,
167521
+ "loss": 0.2605,
167522
+ "step": 53570
167523
+ },
167524
+ {
167525
+ "epoch": 428.53,
167526
+ "learning_rate": 9.15775641025641e-06,
167527
+ "loss": 0.3196,
167528
+ "step": 53575
167529
+ },
167530
+ {
167531
+ "epoch": 428.57,
167532
+ "learning_rate": 9.157676282051283e-06,
167533
+ "loss": 0.6539,
167534
+ "step": 53580
167535
+ },
167536
+ {
167537
+ "epoch": 428.61,
167538
+ "learning_rate": 9.157596153846155e-06,
167539
+ "loss": 1.1682,
167540
+ "step": 53585
167541
+ },
167542
+ {
167543
+ "epoch": 428.65,
167544
+ "learning_rate": 9.157516025641026e-06,
167545
+ "loss": 0.338,
167546
+ "step": 53590
167547
+ },
167548
+ {
167549
+ "epoch": 428.69,
167550
+ "learning_rate": 9.157435897435897e-06,
167551
+ "loss": 0.3081,
167552
+ "step": 53595
167553
+ },
167554
+ {
167555
+ "epoch": 428.73,
167556
+ "learning_rate": 9.15735576923077e-06,
167557
+ "loss": 0.3158,
167558
+ "step": 53600
167559
+ },
167560
+ {
167561
+ "epoch": 428.77,
167562
+ "learning_rate": 9.157275641025642e-06,
167563
+ "loss": 0.5656,
167564
+ "step": 53605
167565
+ },
167566
+ {
167567
+ "epoch": 428.81,
167568
+ "learning_rate": 9.157195512820513e-06,
167569
+ "loss": 1.0353,
167570
+ "step": 53610
167571
+ },
167572
+ {
167573
+ "epoch": 428.85,
167574
+ "learning_rate": 9.157115384615385e-06,
167575
+ "loss": 0.3622,
167576
+ "step": 53615
167577
+ },
167578
+ {
167579
+ "epoch": 428.89,
167580
+ "learning_rate": 9.157035256410258e-06,
167581
+ "loss": 0.3066,
167582
+ "step": 53620
167583
+ },
167584
+ {
167585
+ "epoch": 428.93,
167586
+ "learning_rate": 9.156955128205129e-06,
167587
+ "loss": 0.4568,
167588
+ "step": 53625
167589
+ },
167590
+ {
167591
+ "epoch": 428.97,
167592
+ "learning_rate": 9.156875e-06,
167593
+ "loss": 0.5429,
167594
+ "step": 53630
167595
+ },
167596
+ {
167597
+ "epoch": 429.0,
167598
+ "eval_loss": 0.4089108407497406,
167599
+ "eval_runtime": 39.7496,
167600
+ "eval_samples_per_second": 21.132,
167601
+ "eval_steps_per_second": 0.679,
167602
+ "eval_wer": 0.2008918780612618,
167603
+ "step": 53634
167604
+ },
167605
+ {
167606
+ "epoch": 432.01,
167607
+ "learning_rate": 9.156794871794873e-06,
167608
+ "loss": 0.462,
167609
+ "step": 53635
167610
+ },
167611
+ {
167612
+ "epoch": 432.05,
167613
+ "learning_rate": 9.156714743589745e-06,
167614
+ "loss": 0.3044,
167615
+ "step": 53640
167616
+ },
167617
+ {
167618
+ "epoch": 432.09,
167619
+ "learning_rate": 9.156634615384616e-06,
167620
+ "loss": 0.3536,
167621
+ "step": 53645
167622
+ },
167623
+ {
167624
+ "epoch": 432.13,
167625
+ "learning_rate": 9.156554487179487e-06,
167626
+ "loss": 0.3502,
167627
+ "step": 53650
167628
+ },
167629
+ {
167630
+ "epoch": 432.17,
167631
+ "learning_rate": 9.15647435897436e-06,
167632
+ "loss": 0.5653,
167633
+ "step": 53655
167634
+ },
167635
+ {
167636
+ "epoch": 432.21,
167637
+ "learning_rate": 9.156394230769232e-06,
167638
+ "loss": 1.2789,
167639
+ "step": 53660
167640
+ },
167641
+ {
167642
+ "epoch": 432.25,
167643
+ "learning_rate": 9.156314102564103e-06,
167644
+ "loss": 0.3512,
167645
+ "step": 53665
167646
+ },
167647
+ {
167648
+ "epoch": 432.29,
167649
+ "learning_rate": 9.156233974358976e-06,
167650
+ "loss": 0.3391,
167651
+ "step": 53670
167652
+ },
167653
+ {
167654
+ "epoch": 432.33,
167655
+ "learning_rate": 9.156153846153848e-06,
167656
+ "loss": 0.3438,
167657
+ "step": 53675
167658
+ },
167659
+ {
167660
+ "epoch": 432.37,
167661
+ "learning_rate": 9.156073717948719e-06,
167662
+ "loss": 0.5578,
167663
+ "step": 53680
167664
+ },
167665
+ {
167666
+ "epoch": 432.41,
167667
+ "learning_rate": 9.15599358974359e-06,
167668
+ "loss": 1.1817,
167669
+ "step": 53685
167670
+ },
167671
+ {
167672
+ "epoch": 432.45,
167673
+ "learning_rate": 9.155913461538463e-06,
167674
+ "loss": 0.3402,
167675
+ "step": 53690
167676
+ },
167677
+ {
167678
+ "epoch": 432.49,
167679
+ "learning_rate": 9.155833333333333e-06,
167680
+ "loss": 0.2861,
167681
+ "step": 53695
167682
+ },
167683
+ {
167684
+ "epoch": 432.53,
167685
+ "learning_rate": 9.155753205128206e-06,
167686
+ "loss": 0.2979,
167687
+ "step": 53700
167688
+ },
167689
+ {
167690
+ "epoch": 432.57,
167691
+ "learning_rate": 9.155673076923077e-06,
167692
+ "loss": 0.5803,
167693
+ "step": 53705
167694
+ },
167695
+ {
167696
+ "epoch": 432.61,
167697
+ "learning_rate": 9.155592948717949e-06,
167698
+ "loss": 1.1293,
167699
+ "step": 53710
167700
+ },
167701
+ {
167702
+ "epoch": 432.65,
167703
+ "learning_rate": 9.15551282051282e-06,
167704
+ "loss": 0.4187,
167705
+ "step": 53715
167706
+ },
167707
+ {
167708
+ "epoch": 432.69,
167709
+ "learning_rate": 9.155432692307693e-06,
167710
+ "loss": 0.3486,
167711
+ "step": 53720
167712
+ },
167713
+ {
167714
+ "epoch": 432.73,
167715
+ "learning_rate": 9.155352564102565e-06,
167716
+ "loss": 0.4529,
167717
+ "step": 53725
167718
+ },
167719
+ {
167720
+ "epoch": 432.77,
167721
+ "learning_rate": 9.155272435897436e-06,
167722
+ "loss": 0.6183,
167723
+ "step": 53730
167724
+ },
167725
+ {
167726
+ "epoch": 432.81,
167727
+ "learning_rate": 9.155192307692309e-06,
167728
+ "loss": 1.154,
167729
+ "step": 53735
167730
+ },
167731
+ {
167732
+ "epoch": 432.85,
167733
+ "learning_rate": 9.15511217948718e-06,
167734
+ "loss": 0.3461,
167735
+ "step": 53740
167736
+ },
167737
+ {
167738
+ "epoch": 432.9,
167739
+ "learning_rate": 9.155032051282052e-06,
167740
+ "loss": 0.3242,
167741
+ "step": 53745
167742
+ },
167743
+ {
167744
+ "epoch": 432.94,
167745
+ "learning_rate": 9.154951923076923e-06,
167746
+ "loss": 0.4448,
167747
+ "step": 53750
167748
+ },
167749
+ {
167750
+ "epoch": 432.98,
167751
+ "learning_rate": 9.154871794871796e-06,
167752
+ "loss": 0.6214,
167753
+ "step": 53755
167754
+ },
167755
+ {
167756
+ "epoch": 433.0,
167757
+ "eval_loss": 0.45085570216178894,
167758
+ "eval_runtime": 39.3642,
167759
+ "eval_samples_per_second": 21.339,
167760
+ "eval_steps_per_second": 0.686,
167761
+ "eval_wer": 0.2042432392051397,
167762
+ "step": 53758
167763
+ },
167764
+ {
167765
+ "epoch": 430.02,
167766
+ "learning_rate": 9.154791666666667e-06,
167767
+ "loss": 0.3722,
167768
+ "step": 53760
167769
+ },
167770
+ {
167771
+ "epoch": 430.06,
167772
+ "learning_rate": 9.154711538461539e-06,
167773
+ "loss": 0.2811,
167774
+ "step": 53765
167775
+ },
167776
+ {
167777
+ "epoch": 430.1,
167778
+ "learning_rate": 9.154631410256412e-06,
167779
+ "loss": 0.3527,
167780
+ "step": 53770
167781
+ },
167782
+ {
167783
+ "epoch": 430.14,
167784
+ "learning_rate": 9.154551282051283e-06,
167785
+ "loss": 0.4463,
167786
+ "step": 53775
167787
+ },
167788
+ {
167789
+ "epoch": 430.18,
167790
+ "learning_rate": 9.154471153846155e-06,
167791
+ "loss": 0.7272,
167792
+ "step": 53780
167793
+ },
167794
+ {
167795
+ "epoch": 430.22,
167796
+ "learning_rate": 9.154391025641026e-06,
167797
+ "loss": 1.0655,
167798
+ "step": 53785
167799
+ },
167800
+ {
167801
+ "epoch": 430.26,
167802
+ "learning_rate": 9.154310897435899e-06,
167803
+ "loss": 0.3984,
167804
+ "step": 53790
167805
+ },
167806
+ {
167807
+ "epoch": 430.3,
167808
+ "learning_rate": 9.15423076923077e-06,
167809
+ "loss": 0.3087,
167810
+ "step": 53795
167811
+ },
167812
+ {
167813
+ "epoch": 430.34,
167814
+ "learning_rate": 9.154150641025642e-06,
167815
+ "loss": 0.3548,
167816
+ "step": 53800
167817
+ },
167818
+ {
167819
+ "epoch": 430.38,
167820
+ "learning_rate": 9.154070512820513e-06,
167821
+ "loss": 0.6702,
167822
+ "step": 53805
167823
+ },
167824
+ {
167825
+ "epoch": 430.42,
167826
+ "learning_rate": 9.153990384615386e-06,
167827
+ "loss": 1.0987,
167828
+ "step": 53810
167829
+ },
167830
+ {
167831
+ "epoch": 430.46,
167832
+ "learning_rate": 9.153910256410256e-06,
167833
+ "loss": 0.2952,
167834
+ "step": 53815
167835
+ },
167836
+ {
167837
+ "epoch": 430.5,
167838
+ "learning_rate": 9.153830128205129e-06,
167839
+ "loss": 0.3559,
167840
+ "step": 53820
167841
+ },
167842
+ {
167843
+ "epoch": 430.54,
167844
+ "learning_rate": 9.153750000000002e-06,
167845
+ "loss": 0.4175,
167846
+ "step": 53825
167847
+ },
167848
+ {
167849
+ "epoch": 430.58,
167850
+ "learning_rate": 9.153669871794872e-06,
167851
+ "loss": 0.6887,
167852
+ "step": 53830
167853
+ },
167854
+ {
167855
+ "epoch": 430.62,
167856
+ "learning_rate": 9.153589743589745e-06,
167857
+ "loss": 1.087,
167858
+ "step": 53835
167859
+ },
167860
+ {
167861
+ "epoch": 430.66,
167862
+ "learning_rate": 9.153509615384616e-06,
167863
+ "loss": 0.3551,
167864
+ "step": 53840
167865
+ },
167866
+ {
167867
+ "epoch": 430.7,
167868
+ "learning_rate": 9.153429487179487e-06,
167869
+ "loss": 0.4347,
167870
+ "step": 53845
167871
+ },
167872
+ {
167873
+ "epoch": 430.74,
167874
+ "learning_rate": 9.153349358974359e-06,
167875
+ "loss": 0.3289,
167876
+ "step": 53850
167877
+ },
167878
+ {
167879
+ "epoch": 430.78,
167880
+ "learning_rate": 9.153269230769232e-06,
167881
+ "loss": 0.7435,
167882
+ "step": 53855
167883
+ },
167884
+ {
167885
+ "epoch": 430.82,
167886
+ "learning_rate": 9.153189102564103e-06,
167887
+ "loss": 1.0853,
167888
+ "step": 53860
167889
+ },
167890
+ {
167891
+ "epoch": 430.86,
167892
+ "learning_rate": 9.153108974358974e-06,
167893
+ "loss": 0.3564,
167894
+ "step": 53865
167895
+ },
167896
+ {
167897
+ "epoch": 430.9,
167898
+ "learning_rate": 9.153028846153846e-06,
167899
+ "loss": 0.3129,
167900
+ "step": 53870
167901
+ },
167902
+ {
167903
+ "epoch": 430.94,
167904
+ "learning_rate": 9.152948717948719e-06,
167905
+ "loss": 0.392,
167906
+ "step": 53875
167907
+ },
167908
+ {
167909
+ "epoch": 430.98,
167910
+ "learning_rate": 9.15286858974359e-06,
167911
+ "loss": 0.6987,
167912
+ "step": 53880
167913
+ },
167914
+ {
167915
+ "epoch": 431.0,
167916
+ "eval_loss": 0.4261249005794525,
167917
+ "eval_runtime": 40.6123,
167918
+ "eval_samples_per_second": 20.659,
167919
+ "eval_steps_per_second": 0.665,
167920
+ "eval_wer": 0.20484629294755877,
167921
+ "step": 53883
167922
+ },
167923
+ {
167924
+ "epoch": 431.02,
167925
+ "learning_rate": 9.152804487179488e-06,
167926
+ "loss": 0.419,
167927
+ "step": 53885
167928
+ },
167929
+ {
167930
+ "epoch": 431.06,
167931
+ "learning_rate": 9.152724358974359e-06,
167932
+ "loss": 0.3243,
167933
+ "step": 53890
167934
+ },
167935
+ {
167936
+ "epoch": 431.1,
167937
+ "learning_rate": 9.152644230769232e-06,
167938
+ "loss": 0.341,
167939
+ "step": 53895
167940
+ },
167941
+ {
167942
+ "epoch": 431.14,
167943
+ "learning_rate": 9.152564102564103e-06,
167944
+ "loss": 0.4508,
167945
+ "step": 53900
167946
+ },
167947
+ {
167948
+ "epoch": 431.18,
167949
+ "learning_rate": 9.152483974358975e-06,
167950
+ "loss": 0.7833,
167951
+ "step": 53905
167952
+ },
167953
+ {
167954
+ "epoch": 431.22,
167955
+ "learning_rate": 9.152403846153846e-06,
167956
+ "loss": 1.0565,
167957
+ "step": 53910
167958
+ },
167959
+ {
167960
+ "epoch": 431.25,
167961
+ "learning_rate": 9.15232371794872e-06,
167962
+ "loss": 0.3407,
167963
+ "step": 53915
167964
+ },
167965
+ {
167966
+ "epoch": 431.29,
167967
+ "learning_rate": 9.15224358974359e-06,
167968
+ "loss": 0.3398,
167969
+ "step": 53920
167970
+ },
167971
+ {
167972
+ "epoch": 431.33,
167973
+ "learning_rate": 9.152163461538462e-06,
167974
+ "loss": 0.3471,
167975
+ "step": 53925
167976
+ },
167977
+ {
167978
+ "epoch": 431.37,
167979
+ "learning_rate": 9.152083333333333e-06,
167980
+ "loss": 0.7131,
167981
+ "step": 53930
167982
+ },
167983
+ {
167984
+ "epoch": 431.41,
167985
+ "learning_rate": 9.152003205128206e-06,
167986
+ "loss": 0.996,
167987
+ "step": 53935
167988
+ },
167989
+ {
167990
+ "epoch": 431.45,
167991
+ "learning_rate": 9.151923076923078e-06,
167992
+ "loss": 0.3824,
167993
+ "step": 53940
167994
+ },
167995
+ {
167996
+ "epoch": 431.49,
167997
+ "learning_rate": 9.151842948717949e-06,
167998
+ "loss": 0.321,
167999
+ "step": 53945
168000
+ },
168001
+ {
168002
+ "epoch": 431.53,
168003
+ "learning_rate": 9.151762820512822e-06,
168004
+ "loss": 0.3952,
168005
+ "step": 53950
168006
+ },
168007
+ {
168008
+ "epoch": 431.57,
168009
+ "learning_rate": 9.151682692307693e-06,
168010
+ "loss": 0.6883,
168011
+ "step": 53955
168012
+ },
168013
+ {
168014
+ "epoch": 431.61,
168015
+ "learning_rate": 9.151618589743591e-06,
168016
+ "loss": 1.5506,
168017
+ "step": 53960
168018
+ },
168019
+ {
168020
+ "epoch": 431.65,
168021
+ "learning_rate": 9.151538461538462e-06,
168022
+ "loss": 0.3653,
168023
+ "step": 53965
168024
+ },
168025
+ {
168026
+ "epoch": 431.69,
168027
+ "learning_rate": 9.151458333333334e-06,
168028
+ "loss": 0.4062,
168029
+ "step": 53970
168030
+ },
168031
+ {
168032
+ "epoch": 431.73,
168033
+ "learning_rate": 9.151378205128207e-06,
168034
+ "loss": 0.3621,
168035
+ "step": 53975
168036
+ },
168037
+ {
168038
+ "epoch": 431.77,
168039
+ "learning_rate": 9.151298076923078e-06,
168040
+ "loss": 0.6849,
168041
+ "step": 53980
168042
+ },
168043
+ {
168044
+ "epoch": 431.81,
168045
+ "learning_rate": 9.15121794871795e-06,
168046
+ "loss": 1.0669,
168047
+ "step": 53985
168048
+ },
168049
+ {
168050
+ "epoch": 431.85,
168051
+ "learning_rate": 9.15113782051282e-06,
168052
+ "loss": 0.3909,
168053
+ "step": 53990
168054
+ },
168055
+ {
168056
+ "epoch": 431.89,
168057
+ "learning_rate": 9.151057692307694e-06,
168058
+ "loss": 0.3625,
168059
+ "step": 53995
168060
+ },
168061
+ {
168062
+ "epoch": 431.93,
168063
+ "learning_rate": 9.150977564102565e-06,
168064
+ "loss": 0.4017,
168065
+ "step": 54000
168066
+ },
168067
+ {
168068
+ "epoch": 431.97,
168069
+ "learning_rate": 9.150897435897437e-06,
168070
+ "loss": 0.7696,
168071
+ "step": 54005
168072
+ },
168073
+ {
168074
+ "epoch": 432.0,
168075
+ "eval_loss": 0.37287962436676025,
168076
+ "eval_runtime": 38.0521,
168077
+ "eval_samples_per_second": 22.049,
168078
+ "eval_steps_per_second": 0.71,
168079
+ "eval_wer": 0.20122172932877608,
168080
+ "step": 54008
168081
+ },
168082
+ {
168083
+ "epoch": 435.02,
168084
+ "learning_rate": 9.150817307692308e-06,
168085
+ "loss": 0.3971,
168086
+ "step": 54010
168087
+ },
168088
+ {
168089
+ "epoch": 435.06,
168090
+ "learning_rate": 9.150737179487181e-06,
168091
+ "loss": 0.3162,
168092
+ "step": 54015
168093
+ },
168094
+ {
168095
+ "epoch": 435.1,
168096
+ "learning_rate": 9.150657051282052e-06,
168097
+ "loss": 0.3087,
168098
+ "step": 54020
168099
+ },
168100
+ {
168101
+ "epoch": 435.14,
168102
+ "learning_rate": 9.150576923076924e-06,
168103
+ "loss": 0.3867,
168104
+ "step": 54025
168105
+ },
168106
+ {
168107
+ "epoch": 435.18,
168108
+ "learning_rate": 9.150496794871797e-06,
168109
+ "loss": 0.6681,
168110
+ "step": 54030
168111
+ },
168112
+ {
168113
+ "epoch": 435.22,
168114
+ "learning_rate": 9.150416666666666e-06,
168115
+ "loss": 1.0952,
168116
+ "step": 54035
168117
+ },
168118
+ {
168119
+ "epoch": 435.26,
168120
+ "learning_rate": 9.15033653846154e-06,
168121
+ "loss": 0.3676,
168122
+ "step": 54040
168123
+ },
168124
+ {
168125
+ "epoch": 435.3,
168126
+ "learning_rate": 9.15025641025641e-06,
168127
+ "loss": 0.3552,
168128
+ "step": 54045
168129
+ },
168130
+ {
168131
+ "epoch": 435.34,
168132
+ "learning_rate": 9.150176282051282e-06,
168133
+ "loss": 0.3858,
168134
+ "step": 54050
168135
+ },
168136
+ {
168137
+ "epoch": 435.38,
168138
+ "learning_rate": 9.150096153846154e-06,
168139
+ "loss": 0.7447,
168140
+ "step": 54055
168141
+ },
168142
+ {
168143
+ "epoch": 435.42,
168144
+ "learning_rate": 9.150016025641027e-06,
168145
+ "loss": 1.1432,
168146
+ "step": 54060
168147
+ },
168148
+ {
168149
+ "epoch": 435.46,
168150
+ "learning_rate": 9.149935897435898e-06,
168151
+ "loss": 0.2896,
168152
+ "step": 54065
168153
+ },
168154
+ {
168155
+ "epoch": 435.5,
168156
+ "learning_rate": 9.14985576923077e-06,
168157
+ "loss": 0.3781,
168158
+ "step": 54070
168159
+ },
168160
+ {
168161
+ "epoch": 435.54,
168162
+ "learning_rate": 9.149775641025642e-06,
168163
+ "loss": 0.3899,
168164
+ "step": 54075
168165
+ },
168166
+ {
168167
+ "epoch": 435.58,
168168
+ "learning_rate": 9.149695512820514e-06,
168169
+ "loss": 0.6563,
168170
+ "step": 54080
168171
+ },
168172
+ {
168173
+ "epoch": 435.62,
168174
+ "learning_rate": 9.149615384615385e-06,
168175
+ "loss": 1.1815,
168176
+ "step": 54085
168177
+ },
168178
+ {
168179
+ "epoch": 435.66,
168180
+ "learning_rate": 9.149535256410256e-06,
168181
+ "loss": 0.4117,
168182
+ "step": 54090
168183
+ },
168184
+ {
168185
+ "epoch": 435.7,
168186
+ "learning_rate": 9.14945512820513e-06,
168187
+ "loss": 0.2979,
168188
+ "step": 54095
168189
+ },
168190
+ {
168191
+ "epoch": 435.74,
168192
+ "learning_rate": 9.149375e-06,
168193
+ "loss": 0.3699,
168194
+ "step": 54100
168195
+ },
168196
+ {
168197
+ "epoch": 435.78,
168198
+ "learning_rate": 9.149294871794872e-06,
168199
+ "loss": 0.6889,
168200
+ "step": 54105
168201
+ },
168202
+ {
168203
+ "epoch": 435.82,
168204
+ "learning_rate": 9.149214743589744e-06,
168205
+ "loss": 1.0284,
168206
+ "step": 54110
168207
+ },
168208
+ {
168209
+ "epoch": 435.86,
168210
+ "learning_rate": 9.149134615384617e-06,
168211
+ "loss": 0.3884,
168212
+ "step": 54115
168213
+ },
168214
+ {
168215
+ "epoch": 435.9,
168216
+ "learning_rate": 9.149054487179488e-06,
168217
+ "loss": 0.3676,
168218
+ "step": 54120
168219
+ },
168220
+ {
168221
+ "epoch": 435.94,
168222
+ "learning_rate": 9.14897435897436e-06,
168223
+ "loss": 0.3872,
168224
+ "step": 54125
168225
+ },
168226
+ {
168227
+ "epoch": 435.98,
168228
+ "learning_rate": 9.148894230769232e-06,
168229
+ "loss": 0.8216,
168230
+ "step": 54130
168231
+ },
168232
+ {
168233
+ "epoch": 436.0,
168234
+ "eval_loss": 0.44797709584236145,
168235
+ "eval_runtime": 40.441,
168236
+ "eval_samples_per_second": 20.771,
168237
+ "eval_steps_per_second": 0.668,
168238
+ "eval_wer": 0.19947140444901254,
168239
+ "step": 54132
168240
+ },
168241
+ {
168242
+ "epoch": 436.02,
168243
+ "learning_rate": 9.148814102564104e-06,
168244
+ "loss": 0.3841,
168245
+ "step": 54135
168246
+ },
168247
+ {
168248
+ "epoch": 436.06,
168249
+ "learning_rate": 9.148733974358975e-06,
168250
+ "loss": 0.364,
168251
+ "step": 54140
168252
+ },
168253
+ {
168254
+ "epoch": 436.1,
168255
+ "learning_rate": 9.148653846153846e-06,
168256
+ "loss": 0.384,
168257
+ "step": 54145
168258
+ },
168259
+ {
168260
+ "epoch": 436.14,
168261
+ "learning_rate": 9.14857371794872e-06,
168262
+ "loss": 0.374,
168263
+ "step": 54150
168264
+ },
168265
+ {
168266
+ "epoch": 436.18,
168267
+ "learning_rate": 9.14849358974359e-06,
168268
+ "loss": 0.8492,
168269
+ "step": 54155
168270
+ },
168271
+ {
168272
+ "epoch": 436.22,
168273
+ "learning_rate": 9.148413461538462e-06,
168274
+ "loss": 0.9104,
168275
+ "step": 54160
168276
+ },
168277
+ {
168278
+ "epoch": 436.27,
168279
+ "learning_rate": 9.148333333333335e-06,
168280
+ "loss": 0.3081,
168281
+ "step": 54165
168282
+ },
168283
+ {
168284
+ "epoch": 436.31,
168285
+ "learning_rate": 9.148253205128207e-06,
168286
+ "loss": 0.3592,
168287
+ "step": 54170
168288
+ },
168289
+ {
168290
+ "epoch": 436.35,
168291
+ "learning_rate": 9.148173076923078e-06,
168292
+ "loss": 0.4217,
168293
+ "step": 54175
168294
+ },
168295
+ {
168296
+ "epoch": 436.39,
168297
+ "learning_rate": 9.14809294871795e-06,
168298
+ "loss": 0.888,
168299
+ "step": 54180
168300
+ },
168301
+ {
168302
+ "epoch": 436.43,
168303
+ "learning_rate": 9.148012820512822e-06,
168304
+ "loss": 1.0466,
168305
+ "step": 54185
168306
+ },
168307
+ {
168308
+ "epoch": 436.47,
168309
+ "learning_rate": 9.147932692307692e-06,
168310
+ "loss": 0.3215,
168311
+ "step": 54190
168312
+ },
168313
+ {
168314
+ "epoch": 436.51,
168315
+ "learning_rate": 9.147852564102565e-06,
168316
+ "loss": 0.3265,
168317
+ "step": 54195
168318
+ },
168319
+ {
168320
+ "epoch": 436.55,
168321
+ "learning_rate": 9.147772435897436e-06,
168322
+ "loss": 0.4732,
168323
+ "step": 54200
168324
+ },
168325
+ {
168326
+ "epoch": 436.59,
168327
+ "learning_rate": 9.147692307692308e-06,
168328
+ "loss": 0.9318,
168329
+ "step": 54205
168330
+ },
168331
+ {
168332
+ "epoch": 436.63,
168333
+ "learning_rate": 9.14761217948718e-06,
168334
+ "loss": 0.8966,
168335
+ "step": 54210
168336
+ },
168337
+ {
168338
+ "epoch": 436.67,
168339
+ "learning_rate": 9.147532051282052e-06,
168340
+ "loss": 0.3186,
168341
+ "step": 54215
168342
+ },
168343
+ {
168344
+ "epoch": 436.71,
168345
+ "learning_rate": 9.147451923076924e-06,
168346
+ "loss": 0.3307,
168347
+ "step": 54220
168348
+ },
168349
+ {
168350
+ "epoch": 436.75,
168351
+ "learning_rate": 9.147371794871795e-06,
168352
+ "loss": 0.4654,
168353
+ "step": 54225
168354
+ },
168355
+ {
168356
+ "epoch": 436.79,
168357
+ "learning_rate": 9.147291666666668e-06,
168358
+ "loss": 0.7942,
168359
+ "step": 54230
168360
+ },
168361
+ {
168362
+ "epoch": 436.83,
168363
+ "learning_rate": 9.14721153846154e-06,
168364
+ "loss": 0.9858,
168365
+ "step": 54235
168366
+ },
168367
+ {
168368
+ "epoch": 436.87,
168369
+ "learning_rate": 9.14713141025641e-06,
168370
+ "loss": 0.3137,
168371
+ "step": 54240
168372
+ },
168373
+ {
168374
+ "epoch": 436.91,
168375
+ "learning_rate": 9.147051282051282e-06,
168376
+ "loss": 0.3034,
168377
+ "step": 54245
168378
+ },
168379
+ {
168380
+ "epoch": 436.95,
168381
+ "learning_rate": 9.146971153846155e-06,
168382
+ "loss": 0.4034,
168383
+ "step": 54250
168384
+ },
168385
+ {
168386
+ "epoch": 436.99,
168387
+ "learning_rate": 9.146891025641026e-06,
168388
+ "loss": 0.9007,
168389
+ "step": 54255
168390
+ },
168391
+ {
168392
+ "epoch": 437.0,
168393
+ "eval_loss": 0.41687604784965515,
168394
+ "eval_runtime": 39.6761,
168395
+ "eval_samples_per_second": 21.171,
168396
+ "eval_steps_per_second": 0.681,
168397
+ "eval_wer": 0.19979597784902361,
168398
+ "step": 54256
168399
+ },
168400
+ {
168401
+ "epoch": 434.03,
168402
+ "learning_rate": 9.146810897435898e-06,
168403
+ "loss": 0.366,
168404
+ "step": 54260
168405
+ },
168406
+ {
168407
+ "epoch": 434.07,
168408
+ "learning_rate": 9.146730769230771e-06,
168409
+ "loss": 0.3047,
168410
+ "step": 54265
168411
+ },
168412
+ {
168413
+ "epoch": 434.11,
168414
+ "learning_rate": 9.146650641025642e-06,
168415
+ "loss": 0.4099,
168416
+ "step": 54270
168417
+ },
168418
+ {
168419
+ "epoch": 434.15,
168420
+ "learning_rate": 9.146570512820514e-06,
168421
+ "loss": 0.405,
168422
+ "step": 54275
168423
+ },
168424
+ {
168425
+ "epoch": 434.19,
168426
+ "learning_rate": 9.146490384615385e-06,
168427
+ "loss": 1.0881,
168428
+ "step": 54280
168429
+ },
168430
+ {
168431
+ "epoch": 434.23,
168432
+ "learning_rate": 9.146410256410258e-06,
168433
+ "loss": 0.6728,
168434
+ "step": 54285
168435
+ },
168436
+ {
168437
+ "epoch": 434.27,
168438
+ "learning_rate": 9.14633012820513e-06,
168439
+ "loss": 0.3143,
168440
+ "step": 54290
168441
+ },
168442
+ {
168443
+ "epoch": 434.31,
168444
+ "learning_rate": 9.14625e-06,
168445
+ "loss": 0.3771,
168446
+ "step": 54295
168447
+ },
168448
+ {
168449
+ "epoch": 434.35,
168450
+ "learning_rate": 9.146169871794872e-06,
168451
+ "loss": 0.4689,
168452
+ "step": 54300
168453
+ },
168454
+ {
168455
+ "epoch": 434.39,
168456
+ "learning_rate": 9.146089743589745e-06,
168457
+ "loss": 1.0307,
168458
+ "step": 54305
168459
+ },
168460
+ {
168461
+ "epoch": 434.43,
168462
+ "learning_rate": 9.146009615384615e-06,
168463
+ "loss": 0.6574,
168464
+ "step": 54310
168465
+ },
168466
+ {
168467
+ "epoch": 434.47,
168468
+ "learning_rate": 9.145929487179488e-06,
168469
+ "loss": 0.3558,
168470
+ "step": 54315
168471
+ },
168472
+ {
168473
+ "epoch": 434.51,
168474
+ "learning_rate": 9.145849358974361e-06,
168475
+ "loss": 0.3287,
168476
+ "step": 54320
168477
+ },
168478
+ {
168479
+ "epoch": 434.55,
168480
+ "learning_rate": 9.14576923076923e-06,
168481
+ "loss": 0.4354,
168482
+ "step": 54325
168483
+ },
168484
+ {
168485
+ "epoch": 434.59,
168486
+ "learning_rate": 9.145689102564104e-06,
168487
+ "loss": 0.8869,
168488
+ "step": 54330
168489
+ },
168490
+ {
168491
+ "epoch": 434.63,
168492
+ "learning_rate": 9.145608974358975e-06,
168493
+ "loss": 0.9042,
168494
+ "step": 54335
168495
+ },
168496
+ {
168497
+ "epoch": 434.67,
168498
+ "learning_rate": 9.145528846153846e-06,
168499
+ "loss": 0.3911,
168500
+ "step": 54340
168501
+ },
168502
+ {
168503
+ "epoch": 434.71,
168504
+ "learning_rate": 9.145448717948718e-06,
168505
+ "loss": 0.342,
168506
+ "step": 54345
168507
+ },
168508
+ {
168509
+ "epoch": 434.75,
168510
+ "learning_rate": 9.14536858974359e-06,
168511
+ "loss": 0.4601,
168512
+ "step": 54350
168513
+ },
168514
+ {
168515
+ "epoch": 434.79,
168516
+ "learning_rate": 9.145288461538462e-06,
168517
+ "loss": 1.0639,
168518
+ "step": 54355
168519
+ },
168520
+ {
168521
+ "epoch": 434.83,
168522
+ "learning_rate": 9.145208333333333e-06,
168523
+ "loss": 0.7023,
168524
+ "step": 54360
168525
+ },
168526
+ {
168527
+ "epoch": 434.87,
168528
+ "learning_rate": 9.145128205128207e-06,
168529
+ "loss": 0.3901,
168530
+ "step": 54365
168531
+ },
168532
+ {
168533
+ "epoch": 434.91,
168534
+ "learning_rate": 9.145048076923078e-06,
168535
+ "loss": 0.3836,
168536
+ "step": 54370
168537
+ },
168538
+ {
168539
+ "epoch": 434.95,
168540
+ "learning_rate": 9.14496794871795e-06,
168541
+ "loss": 0.4315,
168542
+ "step": 54375
168543
+ },
168544
+ {
168545
+ "epoch": 434.99,
168546
+ "learning_rate": 9.14488782051282e-06,
168547
+ "loss": 1.001,
168548
+ "step": 54380
168549
+ },
168550
+ {
168551
+ "epoch": 435.0,
168552
+ "eval_loss": 0.3676057457923889,
168553
+ "eval_runtime": 40.0739,
168554
+ "eval_samples_per_second": 20.961,
168555
+ "eval_steps_per_second": 0.674,
168556
+ "eval_wer": 0.20047993019197208,
168557
+ "step": 54381
168558
+ },
168559
+ {
168560
+ "epoch": 438.03,
168561
+ "learning_rate": 9.144807692307694e-06,
168562
+ "loss": 0.3355,
168563
+ "step": 54385
168564
+ },
168565
+ {
168566
+ "epoch": 438.07,
168567
+ "learning_rate": 9.144727564102565e-06,
168568
+ "loss": 0.3302,
168569
+ "step": 54390
168570
+ },
168571
+ {
168572
+ "epoch": 438.11,
168573
+ "learning_rate": 9.144647435897436e-06,
168574
+ "loss": 0.3472,
168575
+ "step": 54395
168576
+ },
168577
+ {
168578
+ "epoch": 438.15,
168579
+ "learning_rate": 9.144567307692308e-06,
168580
+ "loss": 0.4144,
168581
+ "step": 54400
168582
+ },
168583
+ {
168584
+ "epoch": 438.19,
168585
+ "learning_rate": 9.14448717948718e-06,
168586
+ "loss": 0.8747,
168587
+ "step": 54405
168588
+ },
168589
+ {
168590
+ "epoch": 438.23,
168591
+ "learning_rate": 9.144407051282052e-06,
168592
+ "loss": 0.6965,
168593
+ "step": 54410
168594
+ },
168595
+ {
168596
+ "epoch": 438.27,
168597
+ "learning_rate": 9.144326923076924e-06,
168598
+ "loss": 0.3358,
168599
+ "step": 54415
168600
+ },
168601
+ {
168602
+ "epoch": 438.31,
168603
+ "learning_rate": 9.144246794871797e-06,
168604
+ "loss": 0.4455,
168605
+ "step": 54420
168606
+ },
168607
+ {
168608
+ "epoch": 438.35,
168609
+ "learning_rate": 9.144166666666668e-06,
168610
+ "loss": 0.4321,
168611
+ "step": 54425
168612
+ },
168613
+ {
168614
+ "epoch": 438.39,
168615
+ "learning_rate": 9.14408653846154e-06,
168616
+ "loss": 1.0387,
168617
+ "step": 54430
168618
+ },
168619
+ {
168620
+ "epoch": 438.43,
168621
+ "learning_rate": 9.14400641025641e-06,
168622
+ "loss": 0.7209,
168623
+ "step": 54435
168624
+ },
168625
+ {
168626
+ "epoch": 438.47,
168627
+ "learning_rate": 9.143926282051284e-06,
168628
+ "loss": 0.3056,
168629
+ "step": 54440
168630
+ },
168631
+ {
168632
+ "epoch": 438.51,
168633
+ "learning_rate": 9.143846153846155e-06,
168634
+ "loss": 0.36,
168635
+ "step": 54445
168636
+ },
168637
+ {
168638
+ "epoch": 438.55,
168639
+ "learning_rate": 9.143766025641026e-06,
168640
+ "loss": 0.5089,
168641
+ "step": 54450
168642
+ },
168643
+ {
168644
+ "epoch": 438.59,
168645
+ "learning_rate": 9.143685897435898e-06,
168646
+ "loss": 1.0719,
168647
+ "step": 54455
168648
+ },
168649
+ {
168650
+ "epoch": 438.63,
168651
+ "learning_rate": 9.14360576923077e-06,
168652
+ "loss": 0.7689,
168653
+ "step": 54460
168654
+ },
168655
+ {
168656
+ "epoch": 438.67,
168657
+ "learning_rate": 9.14352564102564e-06,
168658
+ "loss": 0.3371,
168659
+ "step": 54465
168660
+ },
168661
+ {
168662
+ "epoch": 438.71,
168663
+ "learning_rate": 9.143445512820514e-06,
168664
+ "loss": 0.3606,
168665
+ "step": 54470
168666
+ },
168667
+ {
168668
+ "epoch": 438.76,
168669
+ "learning_rate": 9.143365384615387e-06,
168670
+ "loss": 0.4898,
168671
+ "step": 54475
168672
+ },
168673
+ {
168674
+ "epoch": 438.8,
168675
+ "learning_rate": 9.143285256410256e-06,
168676
+ "loss": 0.8839,
168677
+ "step": 54480
168678
+ },
168679
+ {
168680
+ "epoch": 438.84,
168681
+ "learning_rate": 9.14320512820513e-06,
168682
+ "loss": 0.7543,
168683
+ "step": 54485
168684
+ },
168685
+ {
168686
+ "epoch": 438.88,
168687
+ "learning_rate": 9.143125e-06,
168688
+ "loss": 0.3785,
168689
+ "step": 54490
168690
+ },
168691
+ {
168692
+ "epoch": 438.92,
168693
+ "learning_rate": 9.143044871794872e-06,
168694
+ "loss": 0.3476,
168695
+ "step": 54495
168696
+ },
168697
+ {
168698
+ "epoch": 438.96,
168699
+ "learning_rate": 9.142964743589743e-06,
168700
+ "loss": 0.4472,
168701
+ "step": 54500
168702
+ },
168703
+ {
168704
+ "epoch": 439.0,
168705
+ "learning_rate": 9.142884615384616e-06,
168706
+ "loss": 1.1279,
168707
+ "step": 54505
168708
+ },
168709
+ {
168710
+ "epoch": 439.0,
168711
+ "eval_loss": 0.4378458261489868,
168712
+ "eval_runtime": 39.018,
168713
+ "eval_samples_per_second": 21.529,
168714
+ "eval_steps_per_second": 0.692,
168715
+ "eval_wer": 0.19587552664646316,
168716
+ "step": 54505
168717
+ },
168718
+ {
168719
+ "epoch": 439.04,
168720
+ "learning_rate": 9.142804487179488e-06,
168721
+ "loss": 0.3981,
168722
+ "step": 54510
168723
+ },
168724
+ {
168725
+ "epoch": 439.08,
168726
+ "learning_rate": 9.142724358974359e-06,
168727
+ "loss": 0.3357,
168728
+ "step": 54515
168729
+ },
168730
+ {
168731
+ "epoch": 439.12,
168732
+ "learning_rate": 9.142644230769232e-06,
168733
+ "loss": 0.3933,
168734
+ "step": 54520
168735
+ },
168736
+ {
168737
+ "epoch": 439.16,
168738
+ "learning_rate": 9.142564102564104e-06,
168739
+ "loss": 0.464,
168740
+ "step": 54525
168741
+ },
168742
+ {
168743
+ "epoch": 439.2,
168744
+ "learning_rate": 9.142483974358975e-06,
168745
+ "loss": 1.2545,
168746
+ "step": 54530
168747
+ },
168748
+ {
168749
+ "epoch": 439.24,
168750
+ "learning_rate": 9.142403846153846e-06,
168751
+ "loss": 0.4474,
168752
+ "step": 54535
168753
+ },
168754
+ {
168755
+ "epoch": 439.28,
168756
+ "learning_rate": 9.14232371794872e-06,
168757
+ "loss": 0.2817,
168758
+ "step": 54540
168759
+ },
168760
+ {
168761
+ "epoch": 439.32,
168762
+ "learning_rate": 9.14224358974359e-06,
168763
+ "loss": 0.3687,
168764
+ "step": 54545
168765
+ },
168766
+ {
168767
+ "epoch": 439.36,
168768
+ "learning_rate": 9.142163461538462e-06,
168769
+ "loss": 0.5966,
168770
+ "step": 54550
168771
+ },
168772
+ {
168773
+ "epoch": 439.4,
168774
+ "learning_rate": 9.142083333333333e-06,
168775
+ "loss": 1.4357,
168776
+ "step": 54555
168777
+ },
168778
+ {
168779
+ "epoch": 439.44,
168780
+ "learning_rate": 9.142003205128206e-06,
168781
+ "loss": 0.3903,
168782
+ "step": 54560
168783
+ },
168784
+ {
168785
+ "epoch": 439.48,
168786
+ "learning_rate": 9.141923076923078e-06,
168787
+ "loss": 0.3566,
168788
+ "step": 54565
168789
+ },
168790
+ {
168791
+ "epoch": 439.52,
168792
+ "learning_rate": 9.14184294871795e-06,
168793
+ "loss": 0.4125,
168794
+ "step": 54570
168795
+ },
168796
+ {
168797
+ "epoch": 439.56,
168798
+ "learning_rate": 9.141762820512822e-06,
168799
+ "loss": 0.5697,
168800
+ "step": 54575
168801
+ },
168802
+ {
168803
+ "epoch": 439.6,
168804
+ "learning_rate": 9.141682692307694e-06,
168805
+ "loss": 1.262,
168806
+ "step": 54580
168807
+ },
168808
+ {
168809
+ "epoch": 439.64,
168810
+ "learning_rate": 9.141602564102565e-06,
168811
+ "loss": 0.3247,
168812
+ "step": 54585
168813
+ },
168814
+ {
168815
+ "epoch": 439.68,
168816
+ "learning_rate": 9.141522435897436e-06,
168817
+ "loss": 0.2981,
168818
+ "step": 54590
168819
+ },
168820
+ {
168821
+ "epoch": 439.72,
168822
+ "learning_rate": 9.14144230769231e-06,
168823
+ "loss": 0.4221,
168824
+ "step": 54595
168825
+ },
168826
+ {
168827
+ "epoch": 439.76,
168828
+ "learning_rate": 9.141362179487179e-06,
168829
+ "loss": 0.5187,
168830
+ "step": 54600
168831
+ },
168832
+ {
168833
+ "epoch": 439.8,
168834
+ "learning_rate": 9.141282051282052e-06,
168835
+ "loss": 1.5213,
168836
+ "step": 54605
168837
+ },
168838
+ {
168839
+ "epoch": 439.84,
168840
+ "learning_rate": 9.141201923076923e-06,
168841
+ "loss": 0.4041,
168842
+ "step": 54610
168843
+ },
168844
+ {
168845
+ "epoch": 439.88,
168846
+ "learning_rate": 9.141121794871795e-06,
168847
+ "loss": 0.261,
168848
+ "step": 54615
168849
+ },
168850
+ {
168851
+ "epoch": 439.92,
168852
+ "learning_rate": 9.141041666666668e-06,
168853
+ "loss": 0.4434,
168854
+ "step": 54620
168855
+ },
168856
+ {
168857
+ "epoch": 439.96,
168858
+ "learning_rate": 9.14096153846154e-06,
168859
+ "loss": 0.5753,
168860
+ "step": 54625
168861
+ },
168862
+ {
168863
+ "epoch": 440.0,
168864
+ "eval_loss": 0.4085885286331177,
168865
+ "eval_runtime": 40.9337,
168866
+ "eval_samples_per_second": 20.497,
168867
+ "eval_steps_per_second": 0.66,
168868
+ "eval_wer": 0.2093006224826071,
168869
+ "step": 54629
168870
+ },
168871
+ {
168872
+ "epoch": 440.01,
168873
+ "learning_rate": 9.140881410256412e-06,
168874
+ "loss": 0.418,
168875
+ "step": 54630
168876
+ },
168877
+ {
168878
+ "epoch": 440.05,
168879
+ "learning_rate": 9.140801282051282e-06,
168880
+ "loss": 0.3297,
168881
+ "step": 54635
168882
+ },
168883
+ {
168884
+ "epoch": 440.09,
168885
+ "learning_rate": 9.140721153846155e-06,
168886
+ "loss": 0.3332,
168887
+ "step": 54640
168888
+ },
168889
+ {
168890
+ "epoch": 440.13,
168891
+ "learning_rate": 9.140641025641026e-06,
168892
+ "loss": 0.3913,
168893
+ "step": 54645
168894
+ },
168895
+ {
168896
+ "epoch": 440.17,
168897
+ "learning_rate": 9.140560897435898e-06,
168898
+ "loss": 0.5579,
168899
+ "step": 54650
168900
+ },
168901
+ {
168902
+ "epoch": 440.21,
168903
+ "learning_rate": 9.140480769230769e-06,
168904
+ "loss": 1.2676,
168905
+ "step": 54655
168906
+ },
168907
+ {
168908
+ "epoch": 440.25,
168909
+ "learning_rate": 9.140400641025642e-06,
168910
+ "loss": 0.3456,
168911
+ "step": 54660
168912
+ },
168913
+ {
168914
+ "epoch": 440.29,
168915
+ "learning_rate": 9.140320512820513e-06,
168916
+ "loss": 0.3053,
168917
+ "step": 54665
168918
+ },
168919
+ {
168920
+ "epoch": 440.33,
168921
+ "learning_rate": 9.140240384615385e-06,
168922
+ "loss": 0.3536,
168923
+ "step": 54670
168924
+ },
168925
+ {
168926
+ "epoch": 440.37,
168927
+ "learning_rate": 9.140160256410258e-06,
168928
+ "loss": 0.6171,
168929
+ "step": 54675
168930
+ },
168931
+ {
168932
+ "epoch": 440.41,
168933
+ "learning_rate": 9.14008012820513e-06,
168934
+ "loss": 1.2114,
168935
+ "step": 54680
168936
+ },
168937
+ {
168938
+ "epoch": 440.45,
168939
+ "learning_rate": 9.14e-06,
168940
+ "loss": 0.4063,
168941
+ "step": 54685
168942
+ },
168943
+ {
168944
+ "epoch": 440.49,
168945
+ "learning_rate": 9.139919871794872e-06,
168946
+ "loss": 0.2761,
168947
+ "step": 54690
168948
+ },
168949
+ {
168950
+ "epoch": 440.53,
168951
+ "learning_rate": 9.139839743589745e-06,
168952
+ "loss": 0.3836,
168953
+ "step": 54695
168954
+ },
168955
+ {
168956
+ "epoch": 440.57,
168957
+ "learning_rate": 9.139759615384616e-06,
168958
+ "loss": 0.6551,
168959
+ "step": 54700
168960
+ },
168961
+ {
168962
+ "epoch": 440.61,
168963
+ "learning_rate": 9.139679487179488e-06,
168964
+ "loss": 1.2129,
168965
+ "step": 54705
168966
+ },
168967
+ {
168968
+ "epoch": 440.65,
168969
+ "learning_rate": 9.139599358974359e-06,
168970
+ "loss": 0.4,
168971
+ "step": 54710
168972
+ },
168973
+ {
168974
+ "epoch": 440.69,
168975
+ "learning_rate": 9.139519230769232e-06,
168976
+ "loss": 0.3875,
168977
+ "step": 54715
168978
+ },
168979
+ {
168980
+ "epoch": 440.73,
168981
+ "learning_rate": 9.139439102564103e-06,
168982
+ "loss": 0.3592,
168983
+ "step": 54720
168984
+ },
168985
+ {
168986
+ "epoch": 440.77,
168987
+ "learning_rate": 9.139358974358975e-06,
168988
+ "loss": 0.6077,
168989
+ "step": 54725
168990
+ },
168991
+ {
168992
+ "epoch": 440.81,
168993
+ "learning_rate": 9.139278846153848e-06,
168994
+ "loss": 1.1306,
168995
+ "step": 54730
168996
+ },
168997
+ {
168998
+ "epoch": 440.85,
168999
+ "learning_rate": 9.13919871794872e-06,
169000
+ "loss": 0.4137,
169001
+ "step": 54735
169002
+ },
169003
+ {
169004
+ "epoch": 440.9,
169005
+ "learning_rate": 9.13911858974359e-06,
169006
+ "loss": 0.3306,
169007
+ "step": 54740
169008
+ },
169009
+ {
169010
+ "epoch": 440.94,
169011
+ "learning_rate": 9.139038461538462e-06,
169012
+ "loss": 0.3185,
169013
+ "step": 54745
169014
+ },
169015
+ {
169016
+ "epoch": 440.98,
169017
+ "learning_rate": 9.138958333333335e-06,
169018
+ "loss": 0.5769,
169019
+ "step": 54750
169020
+ },
169021
+ {
169022
+ "epoch": 441.0,
169023
+ "eval_loss": 0.4726785123348236,
169024
+ "eval_runtime": 40.5592,
169025
+ "eval_samples_per_second": 20.686,
169026
+ "eval_steps_per_second": 0.666,
169027
+ "eval_wer": 0.2004497968659315,
169028
+ "step": 54753
169029
+ },
169030
+ {
169031
+ "epoch": 441.02,
169032
+ "learning_rate": 9.138878205128205e-06,
169033
+ "loss": 0.4078,
169034
+ "step": 54755
169035
+ },
169036
+ {
169037
+ "epoch": 441.06,
169038
+ "learning_rate": 9.138798076923078e-06,
169039
+ "loss": 0.3861,
169040
+ "step": 54760
169041
+ },
169042
+ {
169043
+ "epoch": 441.1,
169044
+ "learning_rate": 9.138717948717949e-06,
169045
+ "loss": 0.3216,
169046
+ "step": 54765
169047
+ },
169048
+ {
169049
+ "epoch": 441.14,
169050
+ "learning_rate": 9.13863782051282e-06,
169051
+ "loss": 0.3816,
169052
+ "step": 54770
169053
+ },
169054
+ {
169055
+ "epoch": 441.18,
169056
+ "learning_rate": 9.138557692307694e-06,
169057
+ "loss": 0.5851,
169058
+ "step": 54775
169059
+ },
169060
+ {
169061
+ "epoch": 441.22,
169062
+ "learning_rate": 9.138477564102565e-06,
169063
+ "loss": 0.9708,
169064
+ "step": 54780
169065
+ },
169066
+ {
169067
+ "epoch": 441.26,
169068
+ "learning_rate": 9.138397435897436e-06,
169069
+ "loss": 0.3298,
169070
+ "step": 54785
169071
+ },
169072
+ {
169073
+ "epoch": 441.3,
169074
+ "learning_rate": 9.138317307692308e-06,
169075
+ "loss": 0.3666,
169076
+ "step": 54790
169077
+ },
169078
+ {
169079
+ "epoch": 441.34,
169080
+ "learning_rate": 9.13823717948718e-06,
169081
+ "loss": 0.4135,
169082
+ "step": 54795
169083
+ },
169084
+ {
169085
+ "epoch": 441.38,
169086
+ "learning_rate": 9.138157051282052e-06,
169087
+ "loss": 0.7108,
169088
+ "step": 54800
169089
+ },
169090
+ {
169091
+ "epoch": 441.42,
169092
+ "learning_rate": 9.138076923076923e-06,
169093
+ "loss": 1.1512,
169094
+ "step": 54805
169095
+ },
169096
+ {
169097
+ "epoch": 441.46,
169098
+ "learning_rate": 9.137996794871795e-06,
169099
+ "loss": 0.3209,
169100
+ "step": 54810
169101
+ },
169102
+ {
169103
+ "epoch": 441.5,
169104
+ "learning_rate": 9.137916666666668e-06,
169105
+ "loss": 0.3077,
169106
+ "step": 54815
169107
+ },
169108
+ {
169109
+ "epoch": 441.54,
169110
+ "learning_rate": 9.137836538461539e-06,
169111
+ "loss": 0.363,
169112
+ "step": 54820
169113
+ },
169114
+ {
169115
+ "epoch": 441.58,
169116
+ "learning_rate": 9.13775641025641e-06,
169117
+ "loss": 0.854,
169118
+ "step": 54825
169119
+ },
169120
+ {
169121
+ "epoch": 441.62,
169122
+ "learning_rate": 9.137676282051284e-06,
169123
+ "loss": 1.161,
169124
+ "step": 54830
169125
+ },
169126
+ {
169127
+ "epoch": 441.66,
169128
+ "learning_rate": 9.137596153846155e-06,
169129
+ "loss": 0.2903,
169130
+ "step": 54835
169131
+ },
169132
+ {
169133
+ "epoch": 441.7,
169134
+ "learning_rate": 9.137516025641026e-06,
169135
+ "loss": 0.2767,
169136
+ "step": 54840
169137
+ },
169138
+ {
169139
+ "epoch": 441.74,
169140
+ "learning_rate": 9.137435897435898e-06,
169141
+ "loss": 0.4821,
169142
+ "step": 54845
169143
+ },
169144
+ {
169145
+ "epoch": 441.78,
169146
+ "learning_rate": 9.13735576923077e-06,
169147
+ "loss": 0.6793,
169148
+ "step": 54850
169149
+ },
169150
+ {
169151
+ "epoch": 441.82,
169152
+ "learning_rate": 9.137275641025642e-06,
169153
+ "loss": 1.1459,
169154
+ "step": 54855
169155
+ },
169156
+ {
169157
+ "epoch": 441.86,
169158
+ "learning_rate": 9.137195512820513e-06,
169159
+ "loss": 0.3631,
169160
+ "step": 54860
169161
+ },
169162
+ {
169163
+ "epoch": 441.9,
169164
+ "learning_rate": 9.137115384615385e-06,
169165
+ "loss": 0.3231,
169166
+ "step": 54865
169167
+ },
169168
+ {
169169
+ "epoch": 441.94,
169170
+ "learning_rate": 9.137035256410258e-06,
169171
+ "loss": 0.3946,
169172
+ "step": 54870
169173
+ },
169174
+ {
169175
+ "epoch": 441.98,
169176
+ "learning_rate": 9.136955128205129e-06,
169177
+ "loss": 0.7559,
169178
+ "step": 54875
169179
+ },
169180
+ {
169181
+ "epoch": 442.0,
169182
+ "eval_loss": 0.4403437077999115,
169183
+ "eval_runtime": 38.7472,
169184
+ "eval_samples_per_second": 21.653,
169185
+ "eval_steps_per_second": 0.697,
169186
+ "eval_wer": 0.20528028933092224,
169187
+ "step": 54877
169188
+ },
169189
+ {
169190
+ "epoch": 439.02,
169191
+ "learning_rate": 9.136875e-06,
169192
+ "loss": 0.3878,
169193
+ "step": 54880
169194
+ },
169195
+ {
169196
+ "epoch": 439.06,
169197
+ "learning_rate": 9.136794871794874e-06,
169198
+ "loss": 0.4139,
169199
+ "step": 54885
169200
+ },
169201
+ {
169202
+ "epoch": 439.1,
169203
+ "learning_rate": 9.136714743589745e-06,
169204
+ "loss": 0.3717,
169205
+ "step": 54890
169206
+ },
169207
+ {
169208
+ "epoch": 439.14,
169209
+ "learning_rate": 9.136634615384616e-06,
169210
+ "loss": 0.4635,
169211
+ "step": 54895
169212
+ },
169213
+ {
169214
+ "epoch": 439.18,
169215
+ "learning_rate": 9.136554487179488e-06,
169216
+ "loss": 0.8805,
169217
+ "step": 54900
169218
+ },
169219
+ {
169220
+ "epoch": 439.22,
169221
+ "learning_rate": 9.13647435897436e-06,
169222
+ "loss": 0.7647,
169223
+ "step": 54905
169224
+ },
169225
+ {
169226
+ "epoch": 439.26,
169227
+ "learning_rate": 9.13639423076923e-06,
169228
+ "loss": 0.3908,
169229
+ "step": 54910
169230
+ },
169231
+ {
169232
+ "epoch": 439.3,
169233
+ "learning_rate": 9.136314102564103e-06,
169234
+ "loss": 0.353,
169235
+ "step": 54915
169236
+ },
169237
+ {
169238
+ "epoch": 439.34,
169239
+ "learning_rate": 9.136233974358976e-06,
169240
+ "loss": 0.4116,
169241
+ "step": 54920
169242
+ },
169243
+ {
169244
+ "epoch": 439.38,
169245
+ "learning_rate": 9.136153846153846e-06,
169246
+ "loss": 0.851,
169247
+ "step": 54925
169248
+ },
169249
+ {
169250
+ "epoch": 439.42,
169251
+ "learning_rate": 9.13607371794872e-06,
169252
+ "loss": 0.9931,
169253
+ "step": 54930
169254
+ },
169255
+ {
169256
+ "epoch": 439.46,
169257
+ "learning_rate": 9.13599358974359e-06,
169258
+ "loss": 0.3066,
169259
+ "step": 54935
169260
+ },
169261
+ {
169262
+ "epoch": 439.5,
169263
+ "learning_rate": 9.135913461538462e-06,
169264
+ "loss": 0.2852,
169265
+ "step": 54940
169266
+ },
169267
+ {
169268
+ "epoch": 439.54,
169269
+ "learning_rate": 9.135833333333333e-06,
169270
+ "loss": 0.3729,
169271
+ "step": 54945
169272
+ },
169273
+ {
169274
+ "epoch": 439.58,
169275
+ "learning_rate": 9.135753205128206e-06,
169276
+ "loss": 0.7223,
169277
+ "step": 54950
169278
+ },
169279
+ {
169280
+ "epoch": 439.62,
169281
+ "learning_rate": 9.135673076923078e-06,
169282
+ "loss": 0.8935,
169283
+ "step": 54955
169284
+ },
169285
+ {
169286
+ "epoch": 439.66,
169287
+ "learning_rate": 9.135592948717949e-06,
169288
+ "loss": 0.3228,
169289
+ "step": 54960
169290
+ },
169291
+ {
169292
+ "epoch": 439.7,
169293
+ "learning_rate": 9.13551282051282e-06,
169294
+ "loss": 0.3639,
169295
+ "step": 54965
169296
+ },
169297
+ {
169298
+ "epoch": 439.74,
169299
+ "learning_rate": 9.135432692307693e-06,
169300
+ "loss": 0.3532,
169301
+ "step": 54970
169302
+ },
169303
+ {
169304
+ "epoch": 439.78,
169305
+ "learning_rate": 9.135352564102565e-06,
169306
+ "loss": 0.9238,
169307
+ "step": 54975
169308
+ },
169309
+ {
169310
+ "epoch": 439.82,
169311
+ "learning_rate": 9.135272435897436e-06,
169312
+ "loss": 0.96,
169313
+ "step": 54980
169314
+ },
169315
+ {
169316
+ "epoch": 439.86,
169317
+ "learning_rate": 9.13519230769231e-06,
169318
+ "loss": 0.2877,
169319
+ "step": 54985
169320
+ },
169321
+ {
169322
+ "epoch": 439.9,
169323
+ "learning_rate": 9.13511217948718e-06,
169324
+ "loss": 0.3879,
169325
+ "step": 54990
169326
+ },
169327
+ {
169328
+ "epoch": 439.94,
169329
+ "learning_rate": 9.135032051282052e-06,
169330
+ "loss": 0.4861,
169331
+ "step": 54995
169332
+ },
169333
+ {
169334
+ "epoch": 439.98,
169335
+ "learning_rate": 9.134951923076923e-06,
169336
+ "loss": 0.9325,
169337
+ "step": 55000
169338
+ },
169339
+ {
169340
+ "epoch": 440.0,
169341
+ "eval_loss": 0.3862856924533844,
169342
+ "eval_runtime": 41.5049,
169343
+ "eval_samples_per_second": 20.214,
169344
+ "eval_steps_per_second": 0.651,
169345
+ "eval_wer": 0.19904622157006602,
169346
+ "step": 55002
169347
+ },
169348
+ {
169349
+ "epoch": 440.02,
169350
+ "learning_rate": 9.134871794871796e-06,
169351
+ "loss": 0.3228,
169352
+ "step": 55005
169353
+ },
169354
+ {
169355
+ "epoch": 440.06,
169356
+ "learning_rate": 9.134791666666668e-06,
169357
+ "loss": 0.3744,
169358
+ "step": 55010
169359
+ },
169360
+ {
169361
+ "epoch": 440.1,
169362
+ "learning_rate": 9.134711538461539e-06,
169363
+ "loss": 0.3578,
169364
+ "step": 55015
169365
+ },
169366
+ {
169367
+ "epoch": 440.14,
169368
+ "learning_rate": 9.134631410256412e-06,
169369
+ "loss": 0.4476,
169370
+ "step": 55020
169371
+ },
169372
+ {
169373
+ "epoch": 440.18,
169374
+ "learning_rate": 9.134551282051283e-06,
169375
+ "loss": 0.7147,
169376
+ "step": 55025
169377
+ },
169378
+ {
169379
+ "epoch": 440.22,
169380
+ "learning_rate": 9.134471153846155e-06,
169381
+ "loss": 0.8546,
169382
+ "step": 55030
169383
+ },
169384
+ {
169385
+ "epoch": 440.26,
169386
+ "learning_rate": 9.134391025641026e-06,
169387
+ "loss": 0.3208,
169388
+ "step": 55035
169389
+ },
169390
+ {
169391
+ "epoch": 440.3,
169392
+ "learning_rate": 9.1343108974359e-06,
169393
+ "loss": 0.4333,
169394
+ "step": 55040
169395
+ },
169396
+ {
169397
+ "epoch": 440.34,
169398
+ "learning_rate": 9.134230769230769e-06,
169399
+ "loss": 0.4132,
169400
+ "step": 55045
169401
+ },
169402
+ {
169403
+ "epoch": 440.38,
169404
+ "learning_rate": 9.134150641025642e-06,
169405
+ "loss": 0.7658,
169406
+ "step": 55050
169407
+ },
169408
+ {
169409
+ "epoch": 440.42,
169410
+ "learning_rate": 9.134070512820513e-06,
169411
+ "loss": 0.7724,
169412
+ "step": 55055
169413
+ },
169414
+ {
169415
+ "epoch": 440.46,
169416
+ "learning_rate": 9.133990384615385e-06,
169417
+ "loss": 0.3624,
169418
+ "step": 55060
169419
+ },
169420
+ {
169421
+ "epoch": 440.5,
169422
+ "learning_rate": 9.133910256410256e-06,
169423
+ "loss": 0.3578,
169424
+ "step": 55065
169425
+ },
169426
+ {
169427
+ "epoch": 440.54,
169428
+ "learning_rate": 9.133830128205129e-06,
169429
+ "loss": 0.4768,
169430
+ "step": 55070
169431
+ },
169432
+ {
169433
+ "epoch": 440.58,
169434
+ "learning_rate": 9.13375e-06,
169435
+ "loss": 0.8875,
169436
+ "step": 55075
169437
+ },
169438
+ {
169439
+ "epoch": 440.62,
169440
+ "learning_rate": 9.133669871794872e-06,
169441
+ "loss": 1.0683,
169442
+ "step": 55080
169443
+ },
169444
+ {
169445
+ "epoch": 440.66,
169446
+ "learning_rate": 9.133589743589745e-06,
169447
+ "loss": 0.3035,
169448
+ "step": 55085
169449
+ },
169450
+ {
169451
+ "epoch": 440.7,
169452
+ "learning_rate": 9.133509615384616e-06,
169453
+ "loss": 0.3714,
169454
+ "step": 55090
169455
+ },
169456
+ {
169457
+ "epoch": 440.74,
169458
+ "learning_rate": 9.133429487179488e-06,
169459
+ "loss": 0.4341,
169460
+ "step": 55095
169461
+ },
169462
+ {
169463
+ "epoch": 440.78,
169464
+ "learning_rate": 9.133349358974359e-06,
169465
+ "loss": 0.7635,
169466
+ "step": 55100
169467
+ },
169468
+ {
169469
+ "epoch": 440.82,
169470
+ "learning_rate": 9.133269230769232e-06,
169471
+ "loss": 1.1058,
169472
+ "step": 55105
169473
+ },
169474
+ {
169475
+ "epoch": 440.86,
169476
+ "learning_rate": 9.133189102564103e-06,
169477
+ "loss": 0.2956,
169478
+ "step": 55110
169479
+ },
169480
+ {
169481
+ "epoch": 440.9,
169482
+ "learning_rate": 9.133108974358975e-06,
169483
+ "loss": 0.3866,
169484
+ "step": 55115
169485
+ },
169486
+ {
169487
+ "epoch": 440.94,
169488
+ "learning_rate": 9.133028846153848e-06,
169489
+ "loss": 0.4264,
169490
+ "step": 55120
169491
+ },
169492
+ {
169493
+ "epoch": 440.98,
169494
+ "learning_rate": 9.132948717948719e-06,
169495
+ "loss": 0.8004,
169496
+ "step": 55125
169497
+ },
169498
+ {
169499
+ "epoch": 441.0,
169500
+ "eval_loss": 0.4621056914329529,
169501
+ "eval_runtime": 39.1804,
169502
+ "eval_samples_per_second": 21.439,
169503
+ "eval_steps_per_second": 0.689,
169504
+ "eval_wer": 0.19953240301015562,
169505
+ "step": 55127
169506
+ },
169507
+ {
169508
+ "epoch": 441.02,
169509
+ "learning_rate": 9.13286858974359e-06,
169510
+ "loss": 0.3839,
169511
+ "step": 55130
169512
+ },
169513
+ {
169514
+ "epoch": 441.06,
169515
+ "learning_rate": 9.132788461538462e-06,
169516
+ "loss": 0.3799,
169517
+ "step": 55135
169518
+ },
169519
+ {
169520
+ "epoch": 441.1,
169521
+ "learning_rate": 9.132708333333335e-06,
169522
+ "loss": 0.3318,
169523
+ "step": 55140
169524
+ },
169525
+ {
169526
+ "epoch": 441.14,
169527
+ "learning_rate": 9.132628205128206e-06,
169528
+ "loss": 0.4144,
169529
+ "step": 55145
169530
+ },
169531
+ {
169532
+ "epoch": 441.18,
169533
+ "learning_rate": 9.132548076923078e-06,
169534
+ "loss": 1.2108,
169535
+ "step": 55150
169536
+ },
169537
+ {
169538
+ "epoch": 441.22,
169539
+ "learning_rate": 9.132467948717949e-06,
169540
+ "loss": 1.007,
169541
+ "step": 55155
169542
+ },
169543
+ {
169544
+ "epoch": 441.26,
169545
+ "learning_rate": 9.132387820512822e-06,
169546
+ "loss": 0.341,
169547
+ "step": 55160
169548
+ },
169549
+ {
169550
+ "epoch": 441.3,
169551
+ "learning_rate": 9.132307692307693e-06,
169552
+ "loss": 0.3373,
169553
+ "step": 55165
169554
+ },
169555
+ {
169556
+ "epoch": 441.34,
169557
+ "learning_rate": 9.132227564102565e-06,
169558
+ "loss": 0.3604,
169559
+ "step": 55170
169560
+ },
169561
+ {
169562
+ "epoch": 441.38,
169563
+ "learning_rate": 9.132147435897438e-06,
169564
+ "loss": 0.7739,
169565
+ "step": 55175
169566
+ },
169567
+ {
169568
+ "epoch": 441.42,
169569
+ "learning_rate": 9.132067307692309e-06,
169570
+ "loss": 0.9114,
169571
+ "step": 55180
169572
+ },
169573
+ {
169574
+ "epoch": 441.46,
169575
+ "learning_rate": 9.13198717948718e-06,
169576
+ "loss": 0.3421,
169577
+ "step": 55185
169578
+ },
169579
+ {
169580
+ "epoch": 441.5,
169581
+ "learning_rate": 9.131907051282052e-06,
169582
+ "loss": 0.3511,
169583
+ "step": 55190
169584
+ },
169585
+ {
169586
+ "epoch": 441.54,
169587
+ "learning_rate": 9.131826923076925e-06,
169588
+ "loss": 0.4503,
169589
+ "step": 55195
169590
+ },
169591
+ {
169592
+ "epoch": 441.58,
169593
+ "learning_rate": 9.131746794871795e-06,
169594
+ "loss": 0.6906,
169595
+ "step": 55200
169596
+ },
169597
+ {
169598
+ "epoch": 441.62,
169599
+ "learning_rate": 9.131666666666668e-06,
169600
+ "loss": 1.0657,
169601
+ "step": 55205
169602
+ },
169603
+ {
169604
+ "epoch": 441.66,
169605
+ "learning_rate": 9.131586538461539e-06,
169606
+ "loss": 0.3198,
169607
+ "step": 55210
169608
+ },
169609
+ {
169610
+ "epoch": 441.7,
169611
+ "learning_rate": 9.13150641025641e-06,
169612
+ "loss": 0.4051,
169613
+ "step": 55215
169614
+ },
169615
+ {
169616
+ "epoch": 441.74,
169617
+ "learning_rate": 9.131426282051283e-06,
169618
+ "loss": 0.4729,
169619
+ "step": 55220
169620
+ },
169621
+ {
169622
+ "epoch": 441.78,
169623
+ "learning_rate": 9.131346153846155e-06,
169624
+ "loss": 0.781,
169625
+ "step": 55225
169626
+ },
169627
+ {
169628
+ "epoch": 441.82,
169629
+ "learning_rate": 9.131266025641026e-06,
169630
+ "loss": 0.9563,
169631
+ "step": 55230
169632
+ },
169633
+ {
169634
+ "epoch": 441.86,
169635
+ "learning_rate": 9.131185897435897e-06,
169636
+ "loss": 0.2861,
169637
+ "step": 55235
169638
+ },
169639
+ {
169640
+ "epoch": 441.9,
169641
+ "learning_rate": 9.13110576923077e-06,
169642
+ "loss": 0.3339,
169643
+ "step": 55240
169644
+ },
169645
+ {
169646
+ "epoch": 441.94,
169647
+ "learning_rate": 9.131025641025642e-06,
169648
+ "loss": 0.421,
169649
+ "step": 55245
169650
+ },
169651
+ {
169652
+ "epoch": 441.98,
169653
+ "learning_rate": 9.130945512820513e-06,
169654
+ "loss": 0.8084,
169655
+ "step": 55250
169656
+ },
169657
+ {
169658
+ "epoch": 442.0,
169659
+ "eval_loss": 0.40637868642807007,
169660
+ "eval_runtime": 39.8756,
169661
+ "eval_samples_per_second": 21.066,
169662
+ "eval_steps_per_second": 0.677,
169663
+ "eval_wer": 0.198806860551827,
169664
+ "step": 55252
169665
+ },
169666
+ {
169667
+ "epoch": 445.02,
169668
+ "learning_rate": 9.130865384615385e-06,
169669
+ "loss": 0.3958,
169670
+ "step": 55255
169671
+ },
169672
+ {
169673
+ "epoch": 445.06,
169674
+ "learning_rate": 9.130785256410258e-06,
169675
+ "loss": 0.3301,
169676
+ "step": 55260
169677
+ },
169678
+ {
169679
+ "epoch": 445.1,
169680
+ "learning_rate": 9.130705128205129e-06,
169681
+ "loss": 0.3448,
169682
+ "step": 55265
169683
+ },
169684
+ {
169685
+ "epoch": 445.14,
169686
+ "learning_rate": 9.130625e-06,
169687
+ "loss": 0.4292,
169688
+ "step": 55270
169689
+ },
169690
+ {
169691
+ "epoch": 445.18,
169692
+ "learning_rate": 9.130544871794873e-06,
169693
+ "loss": 0.753,
169694
+ "step": 55275
169695
+ },
169696
+ {
169697
+ "epoch": 445.22,
169698
+ "learning_rate": 9.130464743589745e-06,
169699
+ "loss": 1.0239,
169700
+ "step": 55280
169701
+ },
169702
+ {
169703
+ "epoch": 445.27,
169704
+ "learning_rate": 9.130384615384616e-06,
169705
+ "loss": 0.2834,
169706
+ "step": 55285
169707
+ },
169708
+ {
169709
+ "epoch": 445.31,
169710
+ "learning_rate": 9.130304487179487e-06,
169711
+ "loss": 0.3527,
169712
+ "step": 55290
169713
+ },
169714
+ {
169715
+ "epoch": 445.35,
169716
+ "learning_rate": 9.13022435897436e-06,
169717
+ "loss": 0.4062,
169718
+ "step": 55295
169719
+ },
169720
+ {
169721
+ "epoch": 445.39,
169722
+ "learning_rate": 9.130144230769232e-06,
169723
+ "loss": 0.8633,
169724
+ "step": 55300
169725
+ },
169726
+ {
169727
+ "epoch": 445.43,
169728
+ "learning_rate": 9.130064102564103e-06,
169729
+ "loss": 0.9571,
169730
+ "step": 55305
169731
+ },
169732
+ {
169733
+ "epoch": 445.47,
169734
+ "learning_rate": 9.129983974358975e-06,
169735
+ "loss": 0.3956,
169736
+ "step": 55310
169737
+ },
169738
+ {
169739
+ "epoch": 445.51,
169740
+ "learning_rate": 9.129903846153848e-06,
169741
+ "loss": 0.3443,
169742
+ "step": 55315
169743
+ },
169744
+ {
169745
+ "epoch": 445.55,
169746
+ "learning_rate": 9.129823717948719e-06,
169747
+ "loss": 0.4048,
169748
+ "step": 55320
169749
+ },
169750
+ {
169751
+ "epoch": 445.59,
169752
+ "learning_rate": 9.12974358974359e-06,
169753
+ "loss": 0.7511,
169754
+ "step": 55325
169755
+ },
169756
+ {
169757
+ "epoch": 445.63,
169758
+ "learning_rate": 9.129663461538463e-06,
169759
+ "loss": 0.8738,
169760
+ "step": 55330
169761
+ },
169762
+ {
169763
+ "epoch": 445.67,
169764
+ "learning_rate": 9.129583333333333e-06,
169765
+ "loss": 0.372,
169766
+ "step": 55335
169767
+ },
169768
+ {
169769
+ "epoch": 445.71,
169770
+ "learning_rate": 9.129503205128206e-06,
169771
+ "loss": 0.3997,
169772
+ "step": 55340
169773
+ },
169774
+ {
169775
+ "epoch": 445.75,
169776
+ "learning_rate": 9.129423076923078e-06,
169777
+ "loss": 0.3981,
169778
+ "step": 55345
169779
+ },
169780
+ {
169781
+ "epoch": 445.79,
169782
+ "learning_rate": 9.129342948717949e-06,
169783
+ "loss": 0.7046,
169784
+ "step": 55350
169785
+ },
169786
+ {
169787
+ "epoch": 445.83,
169788
+ "learning_rate": 9.12926282051282e-06,
169789
+ "loss": 0.9529,
169790
+ "step": 55355
169791
+ },
169792
+ {
169793
+ "epoch": 445.87,
169794
+ "learning_rate": 9.129182692307693e-06,
169795
+ "loss": 0.3718,
169796
+ "step": 55360
169797
+ },
169798
+ {
169799
+ "epoch": 445.91,
169800
+ "learning_rate": 9.129102564102565e-06,
169801
+ "loss": 0.353,
169802
+ "step": 55365
169803
+ },
169804
+ {
169805
+ "epoch": 445.95,
169806
+ "learning_rate": 9.129022435897436e-06,
169807
+ "loss": 0.3952,
169808
+ "step": 55370
169809
+ },
169810
+ {
169811
+ "epoch": 445.99,
169812
+ "learning_rate": 9.128942307692309e-06,
169813
+ "loss": 0.9729,
169814
+ "step": 55375
169815
+ },
169816
+ {
169817
+ "epoch": 446.0,
169818
+ "eval_loss": 0.40089958906173706,
169819
+ "eval_runtime": 40.2269,
169820
+ "eval_samples_per_second": 20.882,
169821
+ "eval_steps_per_second": 0.671,
169822
+ "eval_wer": 0.20831232442555644,
169823
+ "step": 55376
169824
+ },
169825
+ {
169826
+ "epoch": 443.03,
169827
+ "learning_rate": 9.12886217948718e-06,
169828
+ "loss": 0.4226,
169829
+ "step": 55380
169830
+ },
169831
+ {
169832
+ "epoch": 443.07,
169833
+ "learning_rate": 9.128782051282052e-06,
169834
+ "loss": 0.3028,
169835
+ "step": 55385
169836
+ },
169837
+ {
169838
+ "epoch": 443.11,
169839
+ "learning_rate": 9.128701923076923e-06,
169840
+ "loss": 0.3304,
169841
+ "step": 55390
169842
+ },
169843
+ {
169844
+ "epoch": 443.15,
169845
+ "learning_rate": 9.128621794871796e-06,
169846
+ "loss": 0.4673,
169847
+ "step": 55395
169848
+ },
169849
+ {
169850
+ "epoch": 443.19,
169851
+ "learning_rate": 9.128541666666668e-06,
169852
+ "loss": 0.9796,
169853
+ "step": 55400
169854
+ },
169855
+ {
169856
+ "epoch": 443.23,
169857
+ "learning_rate": 9.128461538461539e-06,
169858
+ "loss": 0.6241,
169859
+ "step": 55405
169860
+ },
169861
+ {
169862
+ "epoch": 443.27,
169863
+ "learning_rate": 9.12838141025641e-06,
169864
+ "loss": 0.3519,
169865
+ "step": 55410
169866
+ },
169867
+ {
169868
+ "epoch": 443.31,
169869
+ "learning_rate": 9.128301282051283e-06,
169870
+ "loss": 0.2897,
169871
+ "step": 55415
169872
+ },
169873
+ {
169874
+ "epoch": 443.35,
169875
+ "learning_rate": 9.128221153846155e-06,
169876
+ "loss": 0.4566,
169877
+ "step": 55420
169878
+ },
169879
+ {
169880
+ "epoch": 443.39,
169881
+ "learning_rate": 9.128141025641026e-06,
169882
+ "loss": 0.9614,
169883
+ "step": 55425
169884
+ },
169885
+ {
169886
+ "epoch": 443.43,
169887
+ "learning_rate": 9.128060897435899e-06,
169888
+ "loss": 0.7686,
169889
+ "step": 55430
169890
+ },
169891
+ {
169892
+ "epoch": 443.47,
169893
+ "learning_rate": 9.12798076923077e-06,
169894
+ "loss": 0.3547,
169895
+ "step": 55435
169896
+ },
169897
+ {
169898
+ "epoch": 443.51,
169899
+ "learning_rate": 9.127900641025642e-06,
169900
+ "loss": 0.3559,
169901
+ "step": 55440
169902
+ },
169903
+ {
169904
+ "epoch": 443.55,
169905
+ "learning_rate": 9.127820512820513e-06,
169906
+ "loss": 0.5024,
169907
+ "step": 55445
169908
+ },
169909
+ {
169910
+ "epoch": 443.59,
169911
+ "learning_rate": 9.127740384615386e-06,
169912
+ "loss": 0.9951,
169913
+ "step": 55450
169914
+ },
169915
+ {
169916
+ "epoch": 443.63,
169917
+ "learning_rate": 9.127660256410258e-06,
169918
+ "loss": 0.754,
169919
+ "step": 55455
169920
+ },
169921
+ {
169922
+ "epoch": 443.67,
169923
+ "learning_rate": 9.127580128205129e-06,
169924
+ "loss": 0.3864,
169925
+ "step": 55460
169926
+ },
169927
+ {
169928
+ "epoch": 443.71,
169929
+ "learning_rate": 9.1275e-06,
169930
+ "loss": 0.3745,
169931
+ "step": 55465
169932
+ },
169933
+ {
169934
+ "epoch": 443.75,
169935
+ "learning_rate": 9.127419871794873e-06,
169936
+ "loss": 0.4729,
169937
+ "step": 55470
169938
+ },
169939
+ {
169940
+ "epoch": 443.79,
169941
+ "learning_rate": 9.127339743589745e-06,
169942
+ "loss": 1.1605,
169943
+ "step": 55475
169944
+ },
169945
+ {
169946
+ "epoch": 443.83,
169947
+ "learning_rate": 9.127259615384616e-06,
169948
+ "loss": 0.7502,
169949
+ "step": 55480
169950
+ },
169951
+ {
169952
+ "epoch": 443.87,
169953
+ "learning_rate": 9.127179487179489e-06,
169954
+ "loss": 0.3047,
169955
+ "step": 55485
169956
+ },
169957
+ {
169958
+ "epoch": 443.91,
169959
+ "learning_rate": 9.127099358974359e-06,
169960
+ "loss": 0.4334,
169961
+ "step": 55490
169962
+ },
169963
+ {
169964
+ "epoch": 443.95,
169965
+ "learning_rate": 9.127019230769232e-06,
169966
+ "loss": 0.4476,
169967
+ "step": 55495
169968
+ },
169969
+ {
169970
+ "epoch": 443.99,
169971
+ "learning_rate": 9.126939102564103e-06,
169972
+ "loss": 1.1088,
169973
+ "step": 55500
169974
+ },
169975
+ {
169976
+ "epoch": 444.0,
169977
+ "eval_loss": 0.3665480613708496,
169978
+ "eval_runtime": 39.3106,
169979
+ "eval_samples_per_second": 21.368,
169980
+ "eval_steps_per_second": 0.687,
169981
+ "eval_wer": 0.18631571186315712,
169982
+ "step": 55501
169983
+ },
169984
+ {
169985
+ "epoch": 444.03,
169986
+ "learning_rate": 9.126858974358975e-06,
169987
+ "loss": 0.3685,
169988
+ "step": 55505
169989
+ },
169990
+ {
169991
+ "epoch": 444.07,
169992
+ "learning_rate": 9.126778846153846e-06,
169993
+ "loss": 0.3553,
169994
+ "step": 55510
169995
+ },
169996
+ {
169997
+ "epoch": 444.11,
169998
+ "learning_rate": 9.126698717948719e-06,
169999
+ "loss": 0.3282,
170000
+ "step": 55515
170001
+ },
170002
+ {
170003
+ "epoch": 444.15,
170004
+ "learning_rate": 9.12661858974359e-06,
170005
+ "loss": 0.4901,
170006
+ "step": 55520
170007
+ },
170008
+ {
170009
+ "epoch": 444.19,
170010
+ "learning_rate": 9.126538461538462e-06,
170011
+ "loss": 1.001,
170012
+ "step": 55525
170013
+ },
170014
+ {
170015
+ "epoch": 444.23,
170016
+ "learning_rate": 9.126458333333335e-06,
170017
+ "loss": 0.7808,
170018
+ "step": 55530
170019
+ },
170020
+ {
170021
+ "epoch": 444.27,
170022
+ "learning_rate": 9.126378205128206e-06,
170023
+ "loss": 0.3719,
170024
+ "step": 55535
170025
+ },
170026
+ {
170027
+ "epoch": 444.31,
170028
+ "learning_rate": 9.126298076923077e-06,
170029
+ "loss": 0.3072,
170030
+ "step": 55540
170031
+ },
170032
+ {
170033
+ "epoch": 444.35,
170034
+ "learning_rate": 9.126217948717949e-06,
170035
+ "loss": 0.4211,
170036
+ "step": 55545
170037
+ },
170038
+ {
170039
+ "epoch": 444.39,
170040
+ "learning_rate": 9.126137820512822e-06,
170041
+ "loss": 0.8594,
170042
+ "step": 55550
170043
+ },
170044
+ {
170045
+ "epoch": 444.43,
170046
+ "learning_rate": 9.126057692307693e-06,
170047
+ "loss": 0.8003,
170048
+ "step": 55555
170049
+ },
170050
+ {
170051
+ "epoch": 444.47,
170052
+ "learning_rate": 9.125977564102565e-06,
170053
+ "loss": 0.3277,
170054
+ "step": 55560
170055
+ },
170056
+ {
170057
+ "epoch": 444.51,
170058
+ "learning_rate": 9.125897435897436e-06,
170059
+ "loss": 0.3255,
170060
+ "step": 55565
170061
+ },
170062
+ {
170063
+ "epoch": 444.55,
170064
+ "learning_rate": 9.125817307692309e-06,
170065
+ "loss": 0.4195,
170066
+ "step": 55570
170067
+ },
170068
+ {
170069
+ "epoch": 444.59,
170070
+ "learning_rate": 9.12573717948718e-06,
170071
+ "loss": 0.9432,
170072
+ "step": 55575
170073
+ },
170074
+ {
170075
+ "epoch": 444.63,
170076
+ "learning_rate": 9.125657051282052e-06,
170077
+ "loss": 0.6421,
170078
+ "step": 55580
170079
+ },
170080
+ {
170081
+ "epoch": 444.67,
170082
+ "learning_rate": 9.125576923076925e-06,
170083
+ "loss": 0.3007,
170084
+ "step": 55585
170085
+ },
170086
+ {
170087
+ "epoch": 444.71,
170088
+ "learning_rate": 9.125496794871796e-06,
170089
+ "loss": 0.3632,
170090
+ "step": 55590
170091
+ },
170092
+ {
170093
+ "epoch": 444.75,
170094
+ "learning_rate": 9.125416666666667e-06,
170095
+ "loss": 0.4306,
170096
+ "step": 55595
170097
+ },
170098
+ {
170099
+ "epoch": 444.79,
170100
+ "learning_rate": 9.125336538461539e-06,
170101
+ "loss": 1.0631,
170102
+ "step": 55600
170103
+ },
170104
+ {
170105
+ "epoch": 444.83,
170106
+ "learning_rate": 9.125256410256412e-06,
170107
+ "loss": 0.7489,
170108
+ "step": 55605
170109
+ },
170110
+ {
170111
+ "epoch": 444.87,
170112
+ "learning_rate": 9.125176282051282e-06,
170113
+ "loss": 0.3631,
170114
+ "step": 55610
170115
+ },
170116
+ {
170117
+ "epoch": 444.91,
170118
+ "learning_rate": 9.125096153846155e-06,
170119
+ "loss": 0.3395,
170120
+ "step": 55615
170121
+ },
170122
+ {
170123
+ "epoch": 444.95,
170124
+ "learning_rate": 9.125016025641026e-06,
170125
+ "loss": 0.5022,
170126
+ "step": 55620
170127
+ },
170128
+ {
170129
+ "epoch": 444.99,
170130
+ "learning_rate": 9.124935897435897e-06,
170131
+ "loss": 1.0518,
170132
+ "step": 55625
170133
+ },
170134
+ {
170135
+ "epoch": 445.0,
170136
+ "eval_loss": 0.38635382056236267,
170137
+ "eval_runtime": 40.1549,
170138
+ "eval_samples_per_second": 20.919,
170139
+ "eval_steps_per_second": 0.672,
170140
+ "eval_wer": 0.20045324950654286,
170141
+ "step": 55626
170142
+ },
170143
+ {
170144
+ "epoch": 448.03,
170145
+ "learning_rate": 9.12485576923077e-06,
170146
+ "loss": 0.3732,
170147
+ "step": 55630
170148
+ },
170149
+ {
170150
+ "epoch": 448.07,
170151
+ "learning_rate": 9.124775641025642e-06,
170152
+ "loss": 0.3437,
170153
+ "step": 55635
170154
+ },
170155
+ {
170156
+ "epoch": 448.11,
170157
+ "learning_rate": 9.124695512820513e-06,
170158
+ "loss": 0.3662,
170159
+ "step": 55640
170160
+ },
170161
+ {
170162
+ "epoch": 448.15,
170163
+ "learning_rate": 9.124615384615384e-06,
170164
+ "loss": 0.3891,
170165
+ "step": 55645
170166
+ },
170167
+ {
170168
+ "epoch": 448.19,
170169
+ "learning_rate": 9.124535256410257e-06,
170170
+ "loss": 1.0661,
170171
+ "step": 55650
170172
+ },
170173
+ {
170174
+ "epoch": 448.23,
170175
+ "learning_rate": 9.124455128205129e-06,
170176
+ "loss": 0.7778,
170177
+ "step": 55655
170178
+ },
170179
+ {
170180
+ "epoch": 448.27,
170181
+ "learning_rate": 9.124375e-06,
170182
+ "loss": 0.3332,
170183
+ "step": 55660
170184
+ },
170185
+ {
170186
+ "epoch": 448.31,
170187
+ "learning_rate": 9.124294871794872e-06,
170188
+ "loss": 0.3014,
170189
+ "step": 55665
170190
+ },
170191
+ {
170192
+ "epoch": 448.35,
170193
+ "learning_rate": 9.124214743589745e-06,
170194
+ "loss": 0.4644,
170195
+ "step": 55670
170196
+ },
170197
+ {
170198
+ "epoch": 448.39,
170199
+ "learning_rate": 9.124134615384616e-06,
170200
+ "loss": 0.963,
170201
+ "step": 55675
170202
+ },
170203
+ {
170204
+ "epoch": 448.43,
170205
+ "learning_rate": 9.124054487179487e-06,
170206
+ "loss": 0.7479,
170207
+ "step": 55680
170208
+ },
170209
+ {
170210
+ "epoch": 448.47,
170211
+ "learning_rate": 9.12397435897436e-06,
170212
+ "loss": 0.3042,
170213
+ "step": 55685
170214
+ },
170215
+ {
170216
+ "epoch": 448.51,
170217
+ "learning_rate": 9.123894230769232e-06,
170218
+ "loss": 0.3876,
170219
+ "step": 55690
170220
+ },
170221
+ {
170222
+ "epoch": 448.55,
170223
+ "learning_rate": 9.123814102564103e-06,
170224
+ "loss": 0.4285,
170225
+ "step": 55695
170226
+ },
170227
+ {
170228
+ "epoch": 448.59,
170229
+ "learning_rate": 9.123733974358974e-06,
170230
+ "loss": 1.0303,
170231
+ "step": 55700
170232
+ },
170233
+ {
170234
+ "epoch": 448.63,
170235
+ "learning_rate": 9.123653846153848e-06,
170236
+ "loss": 0.856,
170237
+ "step": 55705
170238
+ },
170239
+ {
170240
+ "epoch": 448.67,
170241
+ "learning_rate": 9.123573717948719e-06,
170242
+ "loss": 0.3624,
170243
+ "step": 55710
170244
+ },
170245
+ {
170246
+ "epoch": 448.71,
170247
+ "learning_rate": 9.12349358974359e-06,
170248
+ "loss": 0.3148,
170249
+ "step": 55715
170250
+ },
170251
+ {
170252
+ "epoch": 448.76,
170253
+ "learning_rate": 9.123413461538462e-06,
170254
+ "loss": 0.4128,
170255
+ "step": 55720
170256
+ },
170257
+ {
170258
+ "epoch": 448.8,
170259
+ "learning_rate": 9.123333333333335e-06,
170260
+ "loss": 1.0522,
170261
+ "step": 55725
170262
+ },
170263
+ {
170264
+ "epoch": 448.84,
170265
+ "learning_rate": 9.123253205128206e-06,
170266
+ "loss": 0.7466,
170267
+ "step": 55730
170268
+ },
170269
+ {
170270
+ "epoch": 448.88,
170271
+ "learning_rate": 9.123173076923077e-06,
170272
+ "loss": 0.3298,
170273
+ "step": 55735
170274
+ },
170275
+ {
170276
+ "epoch": 448.92,
170277
+ "learning_rate": 9.12309294871795e-06,
170278
+ "loss": 0.3688,
170279
+ "step": 55740
170280
+ },
170281
+ {
170282
+ "epoch": 448.96,
170283
+ "learning_rate": 9.123012820512822e-06,
170284
+ "loss": 0.4274,
170285
+ "step": 55745
170286
+ },
170287
+ {
170288
+ "epoch": 449.0,
170289
+ "learning_rate": 9.122932692307693e-06,
170290
+ "loss": 1.1088,
170291
+ "step": 55750
170292
+ },
170293
+ {
170294
+ "epoch": 449.0,
170295
+ "eval_loss": 0.4279929995536804,
170296
+ "eval_runtime": 40.0671,
170297
+ "eval_samples_per_second": 20.965,
170298
+ "eval_steps_per_second": 0.674,
170299
+ "eval_wer": 0.2049155834453907,
170300
+ "step": 55750
170301
  }
170302
  ],
170303
+ "max_steps": 620000,
170304
  "num_train_epochs": 5000,
170305
+ "total_flos": 1.5688709822798293e+20,
170306
  "trial_name": null,
170307
  "trial_params": null
170308
  }
model-bin/finetune/base/{checkpoint-55127 → checkpoint-55750}/training_args.bin RENAMED
File without changes
model-bin/finetune/base/log/1629771721.037489/events.out.tfevents.1629771721.c435e1c5ee04.920.21 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90906f14051b6b0309a7b844a6f9c6905fcedaa13e593ff4777f33f2ee486ebd
3
+ size 4194
model-bin/finetune/base/log/1629772396.2921655/events.out.tfevents.1629772396.c435e1c5ee04.920.23 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:676e4f255f6295aa608c141ae082d50397f78c88d5d563ccc784aa5a628676eb
3
+ size 4194
model-bin/finetune/base/log/1629773061.1068475/events.out.tfevents.1629773061.c435e1c5ee04.920.25 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:162b051c558005310e7e93f9b4d465e3ec020b39fbe70fcd26ee5a158056e027
3
+ size 4194
model-bin/finetune/base/log/1629773705.9435754/events.out.tfevents.1629773705.c435e1c5ee04.920.27 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c89f252082b401f2401d989b3da0bbfe8ea566e5e4a25feb16f10ac58322fe3
3
+ size 4194
model-bin/finetune/base/log/1629774367.3900394/events.out.tfevents.1629774367.c435e1c5ee04.920.29 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f1a772dd10313592c6e20434b304f830848a83451451be046d59b0dedc45cec
3
+ size 4194
model-bin/finetune/base/log/events.out.tfevents.1629771721.c435e1c5ee04.920.20 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d95ff99c15077809d0ec4c84e6e1d743dd434fe37fd5dd907a903b2af6d81f0
3
+ size 8622
model-bin/finetune/base/log/events.out.tfevents.1629772396.c435e1c5ee04.920.22 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa8ecfb7fdd996cb40d06b488fcf6bd150d038e16db7caa4ea44f8b7f5848ede
3
+ size 8622
model-bin/finetune/base/log/events.out.tfevents.1629773061.c435e1c5ee04.920.24 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a9bd26e1a63bab60a1a2bf6bfec9aa3cac6867a0613b110dc0c6c284f8548dfd
3
+ size 8622
model-bin/finetune/base/log/events.out.tfevents.1629773705.c435e1c5ee04.920.26 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6d173e1cbe1bd043161f7074db95e6088799a0c94e185e3a1465c0681587300a
3
+ size 8622
model-bin/finetune/base/log/events.out.tfevents.1629774367.c435e1c5ee04.920.28 ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92651a75f053a563953c3104be992a86d121443a96ce960edff8eded4ffd47b8
3
+ size 8622