SuperKogito commited on
Commit
0349c34
1 Parent(s): 1823c45

End of training

Browse files
Files changed (1) hide show
  1. trainer_state.json +1268 -18
trainer_state.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "best_metric": 18.48141795311607,
3
  "best_model_checkpoint": "whisper-base-nl-3/checkpoint-29000",
4
- "epoch": 3.5599857600569598,
5
- "global_step": 30000,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
@@ -7508,28 +7508,1278 @@
7508
  "step": 30000
7509
  },
7510
  {
7511
- "epoch": 3.56,
7512
- "step": 30000,
7513
- "total_flos": 1.9457998848e+18,
7514
- "train_loss": 0.0,
7515
- "train_runtime": 0.3469,
7516
- "train_samples_per_second": 14412.297,
7517
- "train_steps_per_second": 14412.297
7518
  },
7519
  {
7520
- "epoch": 3.56,
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
7521
  "eval_cer": 9.956974355224656,
7522
- "eval_loss": 0.7926790118217468,
7523
- "eval_runtime": 3112.2219,
7524
- "eval_samples_per_second": 0.696,
7525
- "eval_steps_per_second": 0.348,
7526
  "eval_wer": 31.400474471700036,
7527
- "step": 30000
7528
  }
7529
  ],
7530
- "max_steps": 5000,
7531
- "num_train_epochs": 2,
7532
- "total_flos": 1.9457998848e+18,
7533
  "trial_name": null,
7534
  "trial_params": null
7535
  }
 
1
  {
2
  "best_metric": 18.48141795311607,
3
  "best_model_checkpoint": "whisper-base-nl-3/checkpoint-29000",
4
+ "epoch": 38.16793893129771,
5
+ "global_step": 35000,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
 
7508
  "step": 30000
7509
  },
7510
  {
7511
+ "epoch": 32.74,
7512
+ "learning_rate": 1.446376811594203e-06,
7513
+ "loss": 1.1855,
7514
+ "step": 30025
 
 
 
7515
  },
7516
  {
7517
+ "epoch": 32.77,
7518
+ "learning_rate": 1.4391304347826089e-06,
7519
+ "loss": 1.0754,
7520
+ "step": 30050
7521
+ },
7522
+ {
7523
+ "epoch": 32.8,
7524
+ "learning_rate": 1.4318840579710147e-06,
7525
+ "loss": 0.9609,
7526
+ "step": 30075
7527
+ },
7528
+ {
7529
+ "epoch": 32.82,
7530
+ "learning_rate": 1.4246376811594204e-06,
7531
+ "loss": 0.9363,
7532
+ "step": 30100
7533
+ },
7534
+ {
7535
+ "epoch": 32.85,
7536
+ "learning_rate": 1.4173913043478262e-06,
7537
+ "loss": 0.9573,
7538
+ "step": 30125
7539
+ },
7540
+ {
7541
+ "epoch": 32.88,
7542
+ "learning_rate": 1.4101449275362319e-06,
7543
+ "loss": 0.945,
7544
+ "step": 30150
7545
+ },
7546
+ {
7547
+ "epoch": 32.91,
7548
+ "learning_rate": 1.402898550724638e-06,
7549
+ "loss": 0.8939,
7550
+ "step": 30175
7551
+ },
7552
+ {
7553
+ "epoch": 32.93,
7554
+ "learning_rate": 1.3956521739130436e-06,
7555
+ "loss": 0.9062,
7556
+ "step": 30200
7557
+ },
7558
+ {
7559
+ "epoch": 32.96,
7560
+ "learning_rate": 1.3884057971014494e-06,
7561
+ "loss": 0.9019,
7562
+ "step": 30225
7563
+ },
7564
+ {
7565
+ "epoch": 32.99,
7566
+ "learning_rate": 1.381159420289855e-06,
7567
+ "loss": 0.8691,
7568
+ "step": 30250
7569
+ },
7570
+ {
7571
+ "epoch": 33.02,
7572
+ "learning_rate": 1.3739130434782611e-06,
7573
+ "loss": 0.8198,
7574
+ "step": 30275
7575
+ },
7576
+ {
7577
+ "epoch": 33.04,
7578
+ "learning_rate": 1.3666666666666668e-06,
7579
+ "loss": 0.8642,
7580
+ "step": 30300
7581
+ },
7582
+ {
7583
+ "epoch": 33.07,
7584
+ "learning_rate": 1.3594202898550726e-06,
7585
+ "loss": 0.8705,
7586
+ "step": 30325
7587
+ },
7588
+ {
7589
+ "epoch": 33.1,
7590
+ "learning_rate": 1.3521739130434783e-06,
7591
+ "loss": 0.7469,
7592
+ "step": 30350
7593
+ },
7594
+ {
7595
+ "epoch": 33.12,
7596
+ "learning_rate": 1.3449275362318841e-06,
7597
+ "loss": 0.7704,
7598
+ "step": 30375
7599
+ },
7600
+ {
7601
+ "epoch": 33.15,
7602
+ "learning_rate": 1.3376811594202898e-06,
7603
+ "loss": 0.8126,
7604
+ "step": 30400
7605
+ },
7606
+ {
7607
+ "epoch": 33.18,
7608
+ "learning_rate": 1.3304347826086958e-06,
7609
+ "loss": 0.8152,
7610
+ "step": 30425
7611
+ },
7612
+ {
7613
+ "epoch": 33.21,
7614
+ "learning_rate": 1.3231884057971015e-06,
7615
+ "loss": 0.7214,
7616
+ "step": 30450
7617
+ },
7618
+ {
7619
+ "epoch": 33.23,
7620
+ "learning_rate": 1.3159420289855073e-06,
7621
+ "loss": 0.8131,
7622
+ "step": 30475
7623
+ },
7624
+ {
7625
+ "epoch": 33.26,
7626
+ "learning_rate": 1.308695652173913e-06,
7627
+ "loss": 0.7234,
7628
+ "step": 30500
7629
+ },
7630
+ {
7631
+ "epoch": 33.29,
7632
+ "learning_rate": 1.301449275362319e-06,
7633
+ "loss": 0.7552,
7634
+ "step": 30525
7635
+ },
7636
+ {
7637
+ "epoch": 33.32,
7638
+ "learning_rate": 1.2942028985507247e-06,
7639
+ "loss": 0.7036,
7640
+ "step": 30550
7641
+ },
7642
+ {
7643
+ "epoch": 33.34,
7644
+ "learning_rate": 1.2869565217391305e-06,
7645
+ "loss": 0.7777,
7646
+ "step": 30575
7647
+ },
7648
+ {
7649
+ "epoch": 33.37,
7650
+ "learning_rate": 1.2797101449275362e-06,
7651
+ "loss": 0.6986,
7652
+ "step": 30600
7653
+ },
7654
+ {
7655
+ "epoch": 33.4,
7656
+ "learning_rate": 1.272463768115942e-06,
7657
+ "loss": 0.7156,
7658
+ "step": 30625
7659
+ },
7660
+ {
7661
+ "epoch": 33.42,
7662
+ "learning_rate": 1.265217391304348e-06,
7663
+ "loss": 0.7919,
7664
+ "step": 30650
7665
+ },
7666
+ {
7667
+ "epoch": 33.45,
7668
+ "learning_rate": 1.2579710144927537e-06,
7669
+ "loss": 0.6559,
7670
+ "step": 30675
7671
+ },
7672
+ {
7673
+ "epoch": 33.48,
7674
+ "learning_rate": 1.2507246376811596e-06,
7675
+ "loss": 0.6646,
7676
+ "step": 30700
7677
+ },
7678
+ {
7679
+ "epoch": 33.51,
7680
+ "learning_rate": 1.2434782608695652e-06,
7681
+ "loss": 0.6952,
7682
+ "step": 30725
7683
+ },
7684
+ {
7685
+ "epoch": 33.53,
7686
+ "learning_rate": 1.236231884057971e-06,
7687
+ "loss": 0.7332,
7688
+ "step": 30750
7689
+ },
7690
+ {
7691
+ "epoch": 33.56,
7692
+ "learning_rate": 1.228985507246377e-06,
7693
+ "loss": 0.7673,
7694
+ "step": 30775
7695
+ },
7696
+ {
7697
+ "epoch": 33.59,
7698
+ "learning_rate": 1.2217391304347826e-06,
7699
+ "loss": 0.7079,
7700
+ "step": 30800
7701
+ },
7702
+ {
7703
+ "epoch": 33.62,
7704
+ "learning_rate": 1.2144927536231884e-06,
7705
+ "loss": 0.7293,
7706
+ "step": 30825
7707
+ },
7708
+ {
7709
+ "epoch": 33.64,
7710
+ "learning_rate": 1.2072463768115943e-06,
7711
+ "loss": 0.6716,
7712
+ "step": 30850
7713
+ },
7714
+ {
7715
+ "epoch": 33.67,
7716
+ "learning_rate": 1.2000000000000002e-06,
7717
+ "loss": 0.7037,
7718
+ "step": 30875
7719
+ },
7720
+ {
7721
+ "epoch": 33.7,
7722
+ "learning_rate": 1.1927536231884058e-06,
7723
+ "loss": 0.7235,
7724
+ "step": 30900
7725
+ },
7726
+ {
7727
+ "epoch": 33.72,
7728
+ "learning_rate": 1.1855072463768117e-06,
7729
+ "loss": 0.638,
7730
+ "step": 30925
7731
+ },
7732
+ {
7733
+ "epoch": 33.75,
7734
+ "learning_rate": 1.1782608695652175e-06,
7735
+ "loss": 0.7433,
7736
+ "step": 30950
7737
+ },
7738
+ {
7739
+ "epoch": 33.78,
7740
+ "learning_rate": 1.1710144927536232e-06,
7741
+ "loss": 0.7318,
7742
+ "step": 30975
7743
+ },
7744
+ {
7745
+ "epoch": 33.81,
7746
+ "learning_rate": 1.163768115942029e-06,
7747
+ "loss": 0.6832,
7748
+ "step": 31000
7749
+ },
7750
+ {
7751
+ "epoch": 33.81,
7752
+ "eval_cer": 7.850853989513844,
7753
+ "eval_loss": 0.542524516582489,
7754
+ "eval_runtime": 3197.0599,
7755
+ "eval_samples_per_second": 0.677,
7756
+ "eval_steps_per_second": 0.17,
7757
+ "eval_wer": 24.621636222704176,
7758
+ "step": 31000
7759
+ },
7760
+ {
7761
+ "epoch": 33.83,
7762
+ "learning_rate": 1.1565217391304349e-06,
7763
+ "loss": 0.673,
7764
+ "step": 31025
7765
+ },
7766
+ {
7767
+ "epoch": 33.86,
7768
+ "learning_rate": 1.1492753623188407e-06,
7769
+ "loss": 0.6608,
7770
+ "step": 31050
7771
+ },
7772
+ {
7773
+ "epoch": 33.89,
7774
+ "learning_rate": 1.1420289855072466e-06,
7775
+ "loss": 0.6553,
7776
+ "step": 31075
7777
+ },
7778
+ {
7779
+ "epoch": 33.91,
7780
+ "learning_rate": 1.1347826086956524e-06,
7781
+ "loss": 0.6837,
7782
+ "step": 31100
7783
+ },
7784
+ {
7785
+ "epoch": 33.94,
7786
+ "learning_rate": 1.127536231884058e-06,
7787
+ "loss": 0.6683,
7788
+ "step": 31125
7789
+ },
7790
+ {
7791
+ "epoch": 33.97,
7792
+ "learning_rate": 1.120289855072464e-06,
7793
+ "loss": 0.7043,
7794
+ "step": 31150
7795
+ },
7796
+ {
7797
+ "epoch": 34.0,
7798
+ "learning_rate": 1.1130434782608698e-06,
7799
+ "loss": 0.647,
7800
+ "step": 31175
7801
+ },
7802
+ {
7803
+ "epoch": 34.02,
7804
+ "learning_rate": 1.1057971014492754e-06,
7805
+ "loss": 0.691,
7806
+ "step": 31200
7807
+ },
7808
+ {
7809
+ "epoch": 34.05,
7810
+ "learning_rate": 1.0985507246376813e-06,
7811
+ "loss": 0.6857,
7812
+ "step": 31225
7813
+ },
7814
+ {
7815
+ "epoch": 34.08,
7816
+ "learning_rate": 1.0913043478260871e-06,
7817
+ "loss": 0.5965,
7818
+ "step": 31250
7819
+ },
7820
+ {
7821
+ "epoch": 34.11,
7822
+ "learning_rate": 1.0840579710144928e-06,
7823
+ "loss": 0.6514,
7824
+ "step": 31275
7825
+ },
7826
+ {
7827
+ "epoch": 34.13,
7828
+ "learning_rate": 1.0768115942028986e-06,
7829
+ "loss": 0.5753,
7830
+ "step": 31300
7831
+ },
7832
+ {
7833
+ "epoch": 34.16,
7834
+ "learning_rate": 1.0695652173913045e-06,
7835
+ "loss": 0.6485,
7836
+ "step": 31325
7837
+ },
7838
+ {
7839
+ "epoch": 34.19,
7840
+ "learning_rate": 1.0623188405797103e-06,
7841
+ "loss": 0.6007,
7842
+ "step": 31350
7843
+ },
7844
+ {
7845
+ "epoch": 34.21,
7846
+ "learning_rate": 1.055072463768116e-06,
7847
+ "loss": 0.5861,
7848
+ "step": 31375
7849
+ },
7850
+ {
7851
+ "epoch": 34.24,
7852
+ "learning_rate": 1.0478260869565218e-06,
7853
+ "loss": 0.6009,
7854
+ "step": 31400
7855
+ },
7856
+ {
7857
+ "epoch": 34.27,
7858
+ "learning_rate": 1.0405797101449277e-06,
7859
+ "loss": 0.5646,
7860
+ "step": 31425
7861
+ },
7862
+ {
7863
+ "epoch": 34.3,
7864
+ "learning_rate": 1.0333333333333333e-06,
7865
+ "loss": 0.5896,
7866
+ "step": 31450
7867
+ },
7868
+ {
7869
+ "epoch": 34.32,
7870
+ "learning_rate": 1.0260869565217392e-06,
7871
+ "loss": 0.6017,
7872
+ "step": 31475
7873
+ },
7874
+ {
7875
+ "epoch": 34.35,
7876
+ "learning_rate": 1.018840579710145e-06,
7877
+ "loss": 0.6172,
7878
+ "step": 31500
7879
+ },
7880
+ {
7881
+ "epoch": 34.38,
7882
+ "learning_rate": 1.0115942028985509e-06,
7883
+ "loss": 0.5782,
7884
+ "step": 31525
7885
+ },
7886
+ {
7887
+ "epoch": 34.41,
7888
+ "learning_rate": 1.0043478260869565e-06,
7889
+ "loss": 0.6157,
7890
+ "step": 31550
7891
+ },
7892
+ {
7893
+ "epoch": 34.43,
7894
+ "learning_rate": 9.971014492753624e-07,
7895
+ "loss": 0.6407,
7896
+ "step": 31575
7897
+ },
7898
+ {
7899
+ "epoch": 34.46,
7900
+ "learning_rate": 9.898550724637682e-07,
7901
+ "loss": 0.646,
7902
+ "step": 31600
7903
+ },
7904
+ {
7905
+ "epoch": 34.49,
7906
+ "learning_rate": 9.826086956521739e-07,
7907
+ "loss": 0.5987,
7908
+ "step": 31625
7909
+ },
7910
+ {
7911
+ "epoch": 34.51,
7912
+ "learning_rate": 9.753623188405797e-07,
7913
+ "loss": 0.5978,
7914
+ "step": 31650
7915
+ },
7916
+ {
7917
+ "epoch": 34.54,
7918
+ "learning_rate": 9.681159420289856e-07,
7919
+ "loss": 0.6151,
7920
+ "step": 31675
7921
+ },
7922
+ {
7923
+ "epoch": 34.57,
7924
+ "learning_rate": 9.608695652173912e-07,
7925
+ "loss": 0.6037,
7926
+ "step": 31700
7927
+ },
7928
+ {
7929
+ "epoch": 34.6,
7930
+ "learning_rate": 9.536231884057971e-07,
7931
+ "loss": 0.5898,
7932
+ "step": 31725
7933
+ },
7934
+ {
7935
+ "epoch": 34.62,
7936
+ "learning_rate": 9.463768115942029e-07,
7937
+ "loss": 0.5873,
7938
+ "step": 31750
7939
+ },
7940
+ {
7941
+ "epoch": 34.65,
7942
+ "learning_rate": 9.391304347826087e-07,
7943
+ "loss": 0.6409,
7944
+ "step": 31775
7945
+ },
7946
+ {
7947
+ "epoch": 34.68,
7948
+ "learning_rate": 9.318840579710145e-07,
7949
+ "loss": 0.6163,
7950
+ "step": 31800
7951
+ },
7952
+ {
7953
+ "epoch": 34.71,
7954
+ "learning_rate": 9.246376811594204e-07,
7955
+ "loss": 0.549,
7956
+ "step": 31825
7957
+ },
7958
+ {
7959
+ "epoch": 34.73,
7960
+ "learning_rate": 9.173913043478262e-07,
7961
+ "loss": 0.5967,
7962
+ "step": 31850
7963
+ },
7964
+ {
7965
+ "epoch": 34.76,
7966
+ "learning_rate": 9.10144927536232e-07,
7967
+ "loss": 0.5369,
7968
+ "step": 31875
7969
+ },
7970
+ {
7971
+ "epoch": 34.79,
7972
+ "learning_rate": 9.028985507246379e-07,
7973
+ "loss": 0.5848,
7974
+ "step": 31900
7975
+ },
7976
+ {
7977
+ "epoch": 34.81,
7978
+ "learning_rate": 8.956521739130436e-07,
7979
+ "loss": 0.569,
7980
+ "step": 31925
7981
+ },
7982
+ {
7983
+ "epoch": 34.84,
7984
+ "learning_rate": 8.884057971014493e-07,
7985
+ "loss": 0.5782,
7986
+ "step": 31950
7987
+ },
7988
+ {
7989
+ "epoch": 34.87,
7990
+ "learning_rate": 8.811594202898552e-07,
7991
+ "loss": 0.5425,
7992
+ "step": 31975
7993
+ },
7994
+ {
7995
+ "epoch": 34.9,
7996
+ "learning_rate": 8.73913043478261e-07,
7997
+ "loss": 0.5676,
7998
+ "step": 32000
7999
+ },
8000
+ {
8001
+ "epoch": 34.9,
8002
+ "eval_cer": 7.377586234087609,
8003
+ "eval_loss": 0.5141463875770569,
8004
+ "eval_runtime": 3170.6638,
8005
+ "eval_samples_per_second": 0.683,
8006
+ "eval_steps_per_second": 0.171,
8007
+ "eval_wer": 23.67895647544257,
8008
+ "step": 32000
8009
+ },
8010
+ {
8011
+ "epoch": 34.92,
8012
+ "learning_rate": 8.666666666666668e-07,
8013
+ "loss": 0.5505,
8014
+ "step": 32025
8015
+ },
8016
+ {
8017
+ "epoch": 34.95,
8018
+ "learning_rate": 8.594202898550726e-07,
8019
+ "loss": 0.5515,
8020
+ "step": 32050
8021
+ },
8022
+ {
8023
+ "epoch": 34.98,
8024
+ "learning_rate": 8.521739130434783e-07,
8025
+ "loss": 0.5501,
8026
+ "step": 32075
8027
+ },
8028
+ {
8029
+ "epoch": 35.01,
8030
+ "learning_rate": 8.449275362318842e-07,
8031
+ "loss": 0.5561,
8032
+ "step": 32100
8033
+ },
8034
+ {
8035
+ "epoch": 35.03,
8036
+ "learning_rate": 8.376811594202899e-07,
8037
+ "loss": 0.5846,
8038
+ "step": 32125
8039
+ },
8040
+ {
8041
+ "epoch": 35.06,
8042
+ "learning_rate": 8.304347826086958e-07,
8043
+ "loss": 0.5215,
8044
+ "step": 32150
8045
+ },
8046
+ {
8047
+ "epoch": 35.09,
8048
+ "learning_rate": 8.231884057971015e-07,
8049
+ "loss": 0.5006,
8050
+ "step": 32175
8051
+ },
8052
+ {
8053
+ "epoch": 35.11,
8054
+ "learning_rate": 8.159420289855074e-07,
8055
+ "loss": 0.5434,
8056
+ "step": 32200
8057
+ },
8058
+ {
8059
+ "epoch": 35.14,
8060
+ "learning_rate": 8.086956521739131e-07,
8061
+ "loss": 0.5447,
8062
+ "step": 32225
8063
+ },
8064
+ {
8065
+ "epoch": 35.17,
8066
+ "learning_rate": 8.014492753623189e-07,
8067
+ "loss": 0.5672,
8068
+ "step": 32250
8069
+ },
8070
+ {
8071
+ "epoch": 35.2,
8072
+ "learning_rate": 7.942028985507247e-07,
8073
+ "loss": 0.5617,
8074
+ "step": 32275
8075
+ },
8076
+ {
8077
+ "epoch": 35.22,
8078
+ "learning_rate": 7.869565217391305e-07,
8079
+ "loss": 0.5237,
8080
+ "step": 32300
8081
+ },
8082
+ {
8083
+ "epoch": 35.25,
8084
+ "learning_rate": 7.797101449275363e-07,
8085
+ "loss": 0.6142,
8086
+ "step": 32325
8087
+ },
8088
+ {
8089
+ "epoch": 35.28,
8090
+ "learning_rate": 7.724637681159421e-07,
8091
+ "loss": 0.5075,
8092
+ "step": 32350
8093
+ },
8094
+ {
8095
+ "epoch": 35.31,
8096
+ "learning_rate": 7.652173913043478e-07,
8097
+ "loss": 0.5431,
8098
+ "step": 32375
8099
+ },
8100
+ {
8101
+ "epoch": 35.33,
8102
+ "learning_rate": 7.579710144927537e-07,
8103
+ "loss": 0.4816,
8104
+ "step": 32400
8105
+ },
8106
+ {
8107
+ "epoch": 35.36,
8108
+ "learning_rate": 7.507246376811594e-07,
8109
+ "loss": 0.5244,
8110
+ "step": 32425
8111
+ },
8112
+ {
8113
+ "epoch": 35.39,
8114
+ "learning_rate": 7.434782608695653e-07,
8115
+ "loss": 0.5152,
8116
+ "step": 32450
8117
+ },
8118
+ {
8119
+ "epoch": 35.41,
8120
+ "learning_rate": 7.36231884057971e-07,
8121
+ "loss": 0.5653,
8122
+ "step": 32475
8123
+ },
8124
+ {
8125
+ "epoch": 35.44,
8126
+ "learning_rate": 7.289855072463769e-07,
8127
+ "loss": 0.4795,
8128
+ "step": 32500
8129
+ },
8130
+ {
8131
+ "epoch": 35.47,
8132
+ "learning_rate": 7.217391304347826e-07,
8133
+ "loss": 0.5132,
8134
+ "step": 32525
8135
+ },
8136
+ {
8137
+ "epoch": 35.5,
8138
+ "learning_rate": 7.144927536231884e-07,
8139
+ "loss": 0.5668,
8140
+ "step": 32550
8141
+ },
8142
+ {
8143
+ "epoch": 35.52,
8144
+ "learning_rate": 7.072463768115942e-07,
8145
+ "loss": 0.5509,
8146
+ "step": 32575
8147
+ },
8148
+ {
8149
+ "epoch": 35.55,
8150
+ "learning_rate": 7.000000000000001e-07,
8151
+ "loss": 0.532,
8152
+ "step": 32600
8153
+ },
8154
+ {
8155
+ "epoch": 35.58,
8156
+ "learning_rate": 6.927536231884059e-07,
8157
+ "loss": 0.5321,
8158
+ "step": 32625
8159
+ },
8160
+ {
8161
+ "epoch": 35.61,
8162
+ "learning_rate": 6.855072463768117e-07,
8163
+ "loss": 0.5079,
8164
+ "step": 32650
8165
+ },
8166
+ {
8167
+ "epoch": 35.63,
8168
+ "learning_rate": 6.782608695652175e-07,
8169
+ "loss": 0.5015,
8170
+ "step": 32675
8171
+ },
8172
+ {
8173
+ "epoch": 35.66,
8174
+ "learning_rate": 6.710144927536233e-07,
8175
+ "loss": 0.5693,
8176
+ "step": 32700
8177
+ },
8178
+ {
8179
+ "epoch": 35.69,
8180
+ "learning_rate": 6.637681159420291e-07,
8181
+ "loss": 0.5046,
8182
+ "step": 32725
8183
+ },
8184
+ {
8185
+ "epoch": 35.71,
8186
+ "learning_rate": 6.565217391304349e-07,
8187
+ "loss": 0.5433,
8188
+ "step": 32750
8189
+ },
8190
+ {
8191
+ "epoch": 35.74,
8192
+ "learning_rate": 6.492753623188406e-07,
8193
+ "loss": 0.5056,
8194
+ "step": 32775
8195
+ },
8196
+ {
8197
+ "epoch": 35.77,
8198
+ "learning_rate": 6.420289855072465e-07,
8199
+ "loss": 0.5413,
8200
+ "step": 32800
8201
+ },
8202
+ {
8203
+ "epoch": 35.8,
8204
+ "learning_rate": 6.347826086956522e-07,
8205
+ "loss": 0.5512,
8206
+ "step": 32825
8207
+ },
8208
+ {
8209
+ "epoch": 35.82,
8210
+ "learning_rate": 6.275362318840581e-07,
8211
+ "loss": 0.5224,
8212
+ "step": 32850
8213
+ },
8214
+ {
8215
+ "epoch": 35.85,
8216
+ "learning_rate": 6.202898550724638e-07,
8217
+ "loss": 0.4961,
8218
+ "step": 32875
8219
+ },
8220
+ {
8221
+ "epoch": 35.88,
8222
+ "learning_rate": 6.130434782608696e-07,
8223
+ "loss": 0.5242,
8224
+ "step": 32900
8225
+ },
8226
+ {
8227
+ "epoch": 35.91,
8228
+ "learning_rate": 6.057971014492754e-07,
8229
+ "loss": 0.471,
8230
+ "step": 32925
8231
+ },
8232
+ {
8233
+ "epoch": 35.93,
8234
+ "learning_rate": 5.985507246376812e-07,
8235
+ "loss": 0.547,
8236
+ "step": 32950
8237
+ },
8238
+ {
8239
+ "epoch": 35.96,
8240
+ "learning_rate": 5.91304347826087e-07,
8241
+ "loss": 0.5817,
8242
+ "step": 32975
8243
+ },
8244
+ {
8245
+ "epoch": 35.99,
8246
+ "learning_rate": 5.840579710144928e-07,
8247
+ "loss": 0.4863,
8248
+ "step": 33000
8249
+ },
8250
+ {
8251
+ "epoch": 35.99,
8252
+ "eval_cer": 7.244107809276393,
8253
+ "eval_loss": 0.5002937316894531,
8254
+ "eval_runtime": 3172.2622,
8255
+ "eval_samples_per_second": 0.683,
8256
+ "eval_steps_per_second": 0.171,
8257
+ "eval_wer": 23.0541571080715,
8258
+ "step": 33000
8259
+ },
8260
+ {
8261
+ "epoch": 36.01,
8262
+ "learning_rate": 5.768115942028985e-07,
8263
+ "loss": 0.4949,
8264
+ "step": 33025
8265
+ },
8266
+ {
8267
+ "epoch": 36.04,
8268
+ "learning_rate": 5.695652173913044e-07,
8269
+ "loss": 0.4664,
8270
+ "step": 33050
8271
+ },
8272
+ {
8273
+ "epoch": 36.07,
8274
+ "learning_rate": 5.623188405797101e-07,
8275
+ "loss": 0.4838,
8276
+ "step": 33075
8277
+ },
8278
+ {
8279
+ "epoch": 36.1,
8280
+ "learning_rate": 5.55072463768116e-07,
8281
+ "loss": 0.4896,
8282
+ "step": 33100
8283
+ },
8284
+ {
8285
+ "epoch": 36.12,
8286
+ "learning_rate": 5.478260869565217e-07,
8287
+ "loss": 0.5072,
8288
+ "step": 33125
8289
+ },
8290
+ {
8291
+ "epoch": 36.15,
8292
+ "learning_rate": 5.405797101449276e-07,
8293
+ "loss": 0.4928,
8294
+ "step": 33150
8295
+ },
8296
+ {
8297
+ "epoch": 36.18,
8298
+ "learning_rate": 5.333333333333335e-07,
8299
+ "loss": 0.5154,
8300
+ "step": 33175
8301
+ },
8302
+ {
8303
+ "epoch": 36.21,
8304
+ "learning_rate": 5.260869565217392e-07,
8305
+ "loss": 0.5093,
8306
+ "step": 33200
8307
+ },
8308
+ {
8309
+ "epoch": 36.23,
8310
+ "learning_rate": 5.188405797101449e-07,
8311
+ "loss": 0.4988,
8312
+ "step": 33225
8313
+ },
8314
+ {
8315
+ "epoch": 36.26,
8316
+ "learning_rate": 5.115942028985508e-07,
8317
+ "loss": 0.4689,
8318
+ "step": 33250
8319
+ },
8320
+ {
8321
+ "epoch": 36.29,
8322
+ "learning_rate": 5.043478260869565e-07,
8323
+ "loss": 0.5219,
8324
+ "step": 33275
8325
+ },
8326
+ {
8327
+ "epoch": 36.31,
8328
+ "learning_rate": 4.971014492753624e-07,
8329
+ "loss": 0.4721,
8330
+ "step": 33300
8331
+ },
8332
+ {
8333
+ "epoch": 36.34,
8334
+ "learning_rate": 4.898550724637682e-07,
8335
+ "loss": 0.5071,
8336
+ "step": 33325
8337
+ },
8338
+ {
8339
+ "epoch": 36.37,
8340
+ "learning_rate": 4.82608695652174e-07,
8341
+ "loss": 0.5119,
8342
+ "step": 33350
8343
+ },
8344
+ {
8345
+ "epoch": 36.4,
8346
+ "learning_rate": 4.7536231884057975e-07,
8347
+ "loss": 0.4618,
8348
+ "step": 33375
8349
+ },
8350
+ {
8351
+ "epoch": 36.42,
8352
+ "learning_rate": 4.6811594202898556e-07,
8353
+ "loss": 0.4778,
8354
+ "step": 33400
8355
+ },
8356
+ {
8357
+ "epoch": 36.45,
8358
+ "learning_rate": 4.6086956521739136e-07,
8359
+ "loss": 0.5451,
8360
+ "step": 33425
8361
+ },
8362
+ {
8363
+ "epoch": 36.48,
8364
+ "learning_rate": 4.536231884057971e-07,
8365
+ "loss": 0.4701,
8366
+ "step": 33450
8367
+ },
8368
+ {
8369
+ "epoch": 36.5,
8370
+ "learning_rate": 4.463768115942029e-07,
8371
+ "loss": 0.4444,
8372
+ "step": 33475
8373
+ },
8374
+ {
8375
+ "epoch": 36.53,
8376
+ "learning_rate": 4.391304347826087e-07,
8377
+ "loss": 0.4971,
8378
+ "step": 33500
8379
+ },
8380
+ {
8381
+ "epoch": 36.56,
8382
+ "learning_rate": 4.318840579710145e-07,
8383
+ "loss": 0.5379,
8384
+ "step": 33525
8385
+ },
8386
+ {
8387
+ "epoch": 36.59,
8388
+ "learning_rate": 4.246376811594203e-07,
8389
+ "loss": 0.4883,
8390
+ "step": 33550
8391
+ },
8392
+ {
8393
+ "epoch": 36.61,
8394
+ "learning_rate": 4.1739130434782616e-07,
8395
+ "loss": 0.5602,
8396
+ "step": 33575
8397
+ },
8398
+ {
8399
+ "epoch": 36.64,
8400
+ "learning_rate": 4.1014492753623197e-07,
8401
+ "loss": 0.4486,
8402
+ "step": 33600
8403
+ },
8404
+ {
8405
+ "epoch": 36.67,
8406
+ "learning_rate": 4.028985507246377e-07,
8407
+ "loss": 0.4654,
8408
+ "step": 33625
8409
+ },
8410
+ {
8411
+ "epoch": 36.7,
8412
+ "learning_rate": 3.956521739130435e-07,
8413
+ "loss": 0.4993,
8414
+ "step": 33650
8415
+ },
8416
+ {
8417
+ "epoch": 36.72,
8418
+ "learning_rate": 3.884057971014493e-07,
8419
+ "loss": 0.4552,
8420
+ "step": 33675
8421
+ },
8422
+ {
8423
+ "epoch": 36.75,
8424
+ "learning_rate": 3.811594202898551e-07,
8425
+ "loss": 0.5494,
8426
+ "step": 33700
8427
+ },
8428
+ {
8429
+ "epoch": 36.78,
8430
+ "learning_rate": 3.739130434782609e-07,
8431
+ "loss": 0.5392,
8432
+ "step": 33725
8433
+ },
8434
+ {
8435
+ "epoch": 36.8,
8436
+ "learning_rate": 3.666666666666667e-07,
8437
+ "loss": 0.4993,
8438
+ "step": 33750
8439
+ },
8440
+ {
8441
+ "epoch": 36.83,
8442
+ "learning_rate": 3.5942028985507247e-07,
8443
+ "loss": 0.4896,
8444
+ "step": 33775
8445
+ },
8446
+ {
8447
+ "epoch": 36.86,
8448
+ "learning_rate": 3.5217391304347827e-07,
8449
+ "loss": 0.4804,
8450
+ "step": 33800
8451
+ },
8452
+ {
8453
+ "epoch": 36.89,
8454
+ "learning_rate": 3.4492753623188407e-07,
8455
+ "loss": 0.5006,
8456
+ "step": 33825
8457
+ },
8458
+ {
8459
+ "epoch": 36.91,
8460
+ "learning_rate": 3.3768115942028987e-07,
8461
+ "loss": 0.4965,
8462
+ "step": 33850
8463
+ },
8464
+ {
8465
+ "epoch": 36.94,
8466
+ "learning_rate": 3.304347826086957e-07,
8467
+ "loss": 0.5459,
8468
+ "step": 33875
8469
+ },
8470
+ {
8471
+ "epoch": 36.97,
8472
+ "learning_rate": 3.231884057971015e-07,
8473
+ "loss": 0.4653,
8474
+ "step": 33900
8475
+ },
8476
+ {
8477
+ "epoch": 37.0,
8478
+ "learning_rate": 3.159420289855072e-07,
8479
+ "loss": 0.5432,
8480
+ "step": 33925
8481
+ },
8482
+ {
8483
+ "epoch": 37.02,
8484
+ "learning_rate": 3.086956521739131e-07,
8485
+ "loss": 0.5017,
8486
+ "step": 33950
8487
+ },
8488
+ {
8489
+ "epoch": 37.05,
8490
+ "learning_rate": 3.014492753623189e-07,
8491
+ "loss": 0.473,
8492
+ "step": 33975
8493
+ },
8494
+ {
8495
+ "epoch": 37.08,
8496
+ "learning_rate": 2.942028985507247e-07,
8497
+ "loss": 0.5007,
8498
+ "step": 34000
8499
+ },
8500
+ {
8501
+ "epoch": 37.08,
8502
+ "eval_cer": 7.154500918291419,
8503
+ "eval_loss": 0.4948059916496277,
8504
+ "eval_runtime": 3148.2863,
8505
+ "eval_samples_per_second": 0.688,
8506
+ "eval_steps_per_second": 0.172,
8507
+ "eval_wer": 22.92340335575199,
8508
+ "step": 34000
8509
+ },
8510
+ {
8511
+ "epoch": 37.1,
8512
+ "learning_rate": 2.8695652173913043e-07,
8513
+ "loss": 0.4543,
8514
+ "step": 34025
8515
+ },
8516
+ {
8517
+ "epoch": 37.13,
8518
+ "learning_rate": 2.7971014492753623e-07,
8519
+ "loss": 0.483,
8520
+ "step": 34050
8521
+ },
8522
+ {
8523
+ "epoch": 37.16,
8524
+ "learning_rate": 2.724637681159421e-07,
8525
+ "loss": 0.5062,
8526
+ "step": 34075
8527
+ },
8528
+ {
8529
+ "epoch": 37.19,
8530
+ "learning_rate": 2.6521739130434783e-07,
8531
+ "loss": 0.4614,
8532
+ "step": 34100
8533
+ },
8534
+ {
8535
+ "epoch": 37.21,
8536
+ "learning_rate": 2.5797101449275363e-07,
8537
+ "loss": 0.4805,
8538
+ "step": 34125
8539
+ },
8540
+ {
8541
+ "epoch": 37.24,
8542
+ "learning_rate": 2.5072463768115944e-07,
8543
+ "loss": 0.4723,
8544
+ "step": 34150
8545
+ },
8546
+ {
8547
+ "epoch": 37.27,
8548
+ "learning_rate": 2.4347826086956524e-07,
8549
+ "loss": 0.4788,
8550
+ "step": 34175
8551
+ },
8552
+ {
8553
+ "epoch": 37.3,
8554
+ "learning_rate": 2.3623188405797104e-07,
8555
+ "loss": 0.4736,
8556
+ "step": 34200
8557
+ },
8558
+ {
8559
+ "epoch": 37.32,
8560
+ "learning_rate": 2.289855072463768e-07,
8561
+ "loss": 0.4979,
8562
+ "step": 34225
8563
+ },
8564
+ {
8565
+ "epoch": 37.35,
8566
+ "learning_rate": 2.2173913043478261e-07,
8567
+ "loss": 0.4598,
8568
+ "step": 34250
8569
+ },
8570
+ {
8571
+ "epoch": 37.38,
8572
+ "learning_rate": 2.1449275362318844e-07,
8573
+ "loss": 0.4501,
8574
+ "step": 34275
8575
+ },
8576
+ {
8577
+ "epoch": 37.4,
8578
+ "learning_rate": 2.0724637681159422e-07,
8579
+ "loss": 0.4515,
8580
+ "step": 34300
8581
+ },
8582
+ {
8583
+ "epoch": 37.43,
8584
+ "learning_rate": 2.0000000000000002e-07,
8585
+ "loss": 0.4789,
8586
+ "step": 34325
8587
+ },
8588
+ {
8589
+ "epoch": 37.46,
8590
+ "learning_rate": 1.9275362318840582e-07,
8591
+ "loss": 0.4697,
8592
+ "step": 34350
8593
+ },
8594
+ {
8595
+ "epoch": 37.49,
8596
+ "learning_rate": 1.855072463768116e-07,
8597
+ "loss": 0.468,
8598
+ "step": 34375
8599
+ },
8600
+ {
8601
+ "epoch": 37.51,
8602
+ "learning_rate": 1.782608695652174e-07,
8603
+ "loss": 0.4904,
8604
+ "step": 34400
8605
+ },
8606
+ {
8607
+ "epoch": 37.54,
8608
+ "learning_rate": 1.710144927536232e-07,
8609
+ "loss": 0.4827,
8610
+ "step": 34425
8611
+ },
8612
+ {
8613
+ "epoch": 37.57,
8614
+ "learning_rate": 1.6376811594202902e-07,
8615
+ "loss": 0.4441,
8616
+ "step": 34450
8617
+ },
8618
+ {
8619
+ "epoch": 37.6,
8620
+ "learning_rate": 1.565217391304348e-07,
8621
+ "loss": 0.4665,
8622
+ "step": 34475
8623
+ },
8624
+ {
8625
+ "epoch": 37.62,
8626
+ "learning_rate": 1.492753623188406e-07,
8627
+ "loss": 0.4904,
8628
+ "step": 34500
8629
+ },
8630
+ {
8631
+ "epoch": 37.65,
8632
+ "learning_rate": 1.420289855072464e-07,
8633
+ "loss": 0.4366,
8634
+ "step": 34525
8635
+ },
8636
+ {
8637
+ "epoch": 37.68,
8638
+ "learning_rate": 1.3478260869565218e-07,
8639
+ "loss": 0.5191,
8640
+ "step": 34550
8641
+ },
8642
+ {
8643
+ "epoch": 37.7,
8644
+ "learning_rate": 1.2753623188405798e-07,
8645
+ "loss": 0.4901,
8646
+ "step": 34575
8647
+ },
8648
+ {
8649
+ "epoch": 37.73,
8650
+ "learning_rate": 1.2028985507246378e-07,
8651
+ "loss": 0.5176,
8652
+ "step": 34600
8653
+ },
8654
+ {
8655
+ "epoch": 37.76,
8656
+ "learning_rate": 1.1304347826086958e-07,
8657
+ "loss": 0.4337,
8658
+ "step": 34625
8659
+ },
8660
+ {
8661
+ "epoch": 37.79,
8662
+ "learning_rate": 1.0579710144927537e-07,
8663
+ "loss": 0.4485,
8664
+ "step": 34650
8665
+ },
8666
+ {
8667
+ "epoch": 37.81,
8668
+ "learning_rate": 9.855072463768116e-08,
8669
+ "loss": 0.5057,
8670
+ "step": 34675
8671
+ },
8672
+ {
8673
+ "epoch": 37.84,
8674
+ "learning_rate": 9.130434782608697e-08,
8675
+ "loss": 0.4539,
8676
+ "step": 34700
8677
+ },
8678
+ {
8679
+ "epoch": 37.87,
8680
+ "learning_rate": 8.405797101449276e-08,
8681
+ "loss": 0.4845,
8682
+ "step": 34725
8683
+ },
8684
+ {
8685
+ "epoch": 37.9,
8686
+ "learning_rate": 7.681159420289856e-08,
8687
+ "loss": 0.4679,
8688
+ "step": 34750
8689
+ },
8690
+ {
8691
+ "epoch": 37.92,
8692
+ "learning_rate": 6.956521739130436e-08,
8693
+ "loss": 0.5126,
8694
+ "step": 34775
8695
+ },
8696
+ {
8697
+ "epoch": 37.95,
8698
+ "learning_rate": 6.231884057971015e-08,
8699
+ "loss": 0.5234,
8700
+ "step": 34800
8701
+ },
8702
+ {
8703
+ "epoch": 37.98,
8704
+ "learning_rate": 5.507246376811595e-08,
8705
+ "loss": 0.4573,
8706
+ "step": 34825
8707
+ },
8708
+ {
8709
+ "epoch": 38.0,
8710
+ "learning_rate": 4.782608695652174e-08,
8711
+ "loss": 0.4481,
8712
+ "step": 34850
8713
+ },
8714
+ {
8715
+ "epoch": 38.03,
8716
+ "learning_rate": 4.057971014492754e-08,
8717
+ "loss": 0.4473,
8718
+ "step": 34875
8719
+ },
8720
+ {
8721
+ "epoch": 38.06,
8722
+ "learning_rate": 3.333333333333334e-08,
8723
+ "loss": 0.4978,
8724
+ "step": 34900
8725
+ },
8726
+ {
8727
+ "epoch": 38.09,
8728
+ "learning_rate": 2.6086956521739135e-08,
8729
+ "loss": 0.4456,
8730
+ "step": 34925
8731
+ },
8732
+ {
8733
+ "epoch": 38.11,
8734
+ "learning_rate": 1.884057971014493e-08,
8735
+ "loss": 0.47,
8736
+ "step": 34950
8737
+ },
8738
+ {
8739
+ "epoch": 38.14,
8740
+ "learning_rate": 1.1594202898550725e-08,
8741
+ "loss": 0.4272,
8742
+ "step": 34975
8743
+ },
8744
+ {
8745
+ "epoch": 38.17,
8746
+ "learning_rate": 4.3478260869565225e-09,
8747
+ "loss": 0.4519,
8748
+ "step": 35000
8749
+ },
8750
+ {
8751
+ "epoch": 38.17,
8752
+ "eval_cer": 7.125683342150651,
8753
+ "eval_loss": 0.49221470952033997,
8754
+ "eval_runtime": 3147.7575,
8755
+ "eval_samples_per_second": 0.688,
8756
+ "eval_steps_per_second": 0.172,
8757
+ "eval_wer": 22.824750824061823,
8758
+ "step": 35000
8759
+ },
8760
+ {
8761
+ "epoch": 38.17,
8762
+ "step": 35000,
8763
+ "total_flos": 3.24222148804608e+18,
8764
+ "train_loss": 0.08324751208169119,
8765
+ "train_runtime": 18962.643,
8766
+ "train_samples_per_second": 7.383,
8767
+ "train_steps_per_second": 1.846
8768
+ },
8769
+ {
8770
+ "epoch": 38.17,
8771
  "eval_cer": 9.956974355224656,
8772
+ "eval_loss": 0.7910374999046326,
8773
+ "eval_runtime": 3276.0302,
8774
+ "eval_samples_per_second": 0.661,
8775
+ "eval_steps_per_second": 0.165,
8776
  "eval_wer": 31.400474471700036,
8777
+ "step": 35000
8778
  }
8779
  ],
8780
+ "max_steps": 35000,
8781
+ "num_train_epochs": 39,
8782
+ "total_flos": 3.24222148804608e+18,
8783
  "trial_name": null,
8784
  "trial_params": null
8785
  }