ghofrani commited on
Commit
4219e17
1 Parent(s): 468160b

End of training

Browse files
Files changed (4) hide show
  1. all_results.json +10 -10
  2. eval_results.json +6 -6
  3. train_results.json +5 -5
  4. trainer_state.json +2588 -50
all_results.json CHANGED
@@ -1,14 +1,14 @@
1
  {
2
- "epoch": 123.0,
3
- "eval_loss": 0.34283074736595154,
4
- "eval_runtime": 324.3615,
5
  "eval_samples": 7110,
6
- "eval_samples_per_second": 21.92,
7
- "eval_steps_per_second": 1.372,
8
- "eval_wer": 0.34934675312106844,
9
- "train_loss": 0.7828263307311466,
10
- "train_runtime": 82672.8006,
11
  "train_samples": 19531,
12
- "train_samples_per_second": 29.058,
13
- "train_steps_per_second": 0.226
14
  }
 
1
  {
2
+ "epoch": 150.0,
3
+ "eval_loss": 0.34484201669692993,
4
+ "eval_runtime": 317.3231,
5
  "eval_samples": 7110,
6
+ "eval_samples_per_second": 22.406,
7
+ "eval_steps_per_second": 1.402,
8
+ "eval_wer": 0.347837027000871,
9
+ "train_loss": 0.16461168266179269,
10
+ "train_runtime": 24582.5742,
11
  "train_samples": 19531,
12
+ "train_samples_per_second": 119.176,
13
+ "train_steps_per_second": 0.927
14
  }
eval_results.json CHANGED
@@ -1,9 +1,9 @@
1
  {
2
- "epoch": 123.0,
3
- "eval_loss": 0.34283074736595154,
4
- "eval_runtime": 324.3615,
5
  "eval_samples": 7110,
6
- "eval_samples_per_second": 21.92,
7
- "eval_steps_per_second": 1.372,
8
- "eval_wer": 0.34934675312106844
9
  }
 
1
  {
2
+ "epoch": 150.0,
3
+ "eval_loss": 0.34484201669692993,
4
+ "eval_runtime": 317.3231,
5
  "eval_samples": 7110,
6
+ "eval_samples_per_second": 22.406,
7
+ "eval_steps_per_second": 1.402,
8
+ "eval_wer": 0.347837027000871
9
  }
train_results.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
- "epoch": 123.0,
3
- "train_loss": 0.7828263307311466,
4
- "train_runtime": 82672.8006,
5
  "train_samples": 19531,
6
- "train_samples_per_second": 29.058,
7
- "train_steps_per_second": 0.226
8
  }
 
1
  {
2
+ "epoch": 150.0,
3
+ "train_loss": 0.16461168266179269,
4
+ "train_runtime": 24582.5742,
5
  "train_samples": 19531,
6
+ "train_samples_per_second": 119.176,
7
+ "train_steps_per_second": 0.927
8
  }
trainer_state.json CHANGED
@@ -1,8 +1,8 @@
1
  {
2
  "best_metric": null,
3
  "best_model_checkpoint": null,
4
- "epoch": 122.99509001636662,
5
- "global_step": 18696,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
@@ -11442,131 +11442,2669 @@
11442
  },
11443
  {
11444
  "epoch": 121.77,
11445
- "learning_rate": 6.291675629167564e-07,
11446
- "loss": 0.9699,
11447
  "step": 18510
11448
  },
11449
  {
11450
  "epoch": 121.84,
11451
- "learning_rate": 5.969025596902561e-07,
11452
- "loss": 0.8446,
11453
  "step": 18520
11454
  },
11455
  {
11456
  "epoch": 121.9,
11457
- "learning_rate": 5.646375564637557e-07,
11458
- "loss": 0.8564,
11459
  "step": 18530
11460
  },
11461
  {
11462
  "epoch": 121.97,
11463
- "learning_rate": 5.323725532372553e-07,
11464
- "loss": 0.8883,
11465
  "step": 18540
11466
  },
11467
  {
11468
  "epoch": 122.04,
11469
- "learning_rate": 5.00107550010755e-07,
11470
- "loss": 0.9933,
11471
  "step": 18550
11472
  },
11473
  {
11474
  "epoch": 122.1,
11475
- "learning_rate": 4.678425467842547e-07,
11476
- "loss": 0.8476,
11477
  "step": 18560
11478
  },
11479
  {
11480
  "epoch": 122.17,
11481
- "learning_rate": 4.355775435577544e-07,
11482
- "loss": 0.8747,
11483
  "step": 18570
11484
  },
11485
  {
11486
  "epoch": 122.24,
11487
- "learning_rate": 4.03312540331254e-07,
11488
- "loss": 0.8982,
11489
  "step": 18580
11490
  },
11491
  {
11492
  "epoch": 122.3,
11493
- "learning_rate": 3.710475371047537e-07,
11494
- "loss": 0.8518,
11495
  "step": 18590
11496
  },
11497
  {
11498
  "epoch": 122.37,
11499
- "learning_rate": 3.387825338782534e-07,
11500
- "loss": 0.8722,
11501
  "step": 18600
11502
  },
11503
  {
11504
  "epoch": 122.43,
11505
- "learning_rate": 3.065175306517531e-07,
11506
- "loss": 0.8711,
11507
  "step": 18610
11508
  },
11509
  {
11510
  "epoch": 122.5,
11511
- "learning_rate": 2.7425252742525275e-07,
11512
- "loss": 0.9164,
11513
  "step": 18620
11514
  },
11515
  {
11516
  "epoch": 122.56,
11517
- "learning_rate": 2.4198752419875245e-07,
11518
- "loss": 0.8547,
11519
  "step": 18630
11520
  },
11521
  {
11522
  "epoch": 122.63,
11523
- "learning_rate": 2.0972252097225211e-07,
11524
- "loss": 0.858,
11525
  "step": 18640
11526
  },
11527
  {
11528
  "epoch": 122.69,
11529
- "learning_rate": 1.7745751774575178e-07,
11530
- "loss": 0.8493,
11531
  "step": 18650
11532
  },
11533
  {
11534
  "epoch": 122.76,
11535
- "learning_rate": 1.4519251451925145e-07,
11536
- "loss": 0.9404,
11537
  "step": 18660
11538
  },
11539
  {
11540
  "epoch": 122.82,
11541
- "learning_rate": 1.1292751129275113e-07,
11542
- "loss": 0.8165,
11543
  "step": 18670
11544
  },
11545
  {
11546
  "epoch": 122.89,
11547
- "learning_rate": 8.066250806625082e-08,
11548
- "loss": 0.8654,
11549
  "step": 18680
11550
  },
11551
  {
11552
  "epoch": 122.96,
11553
- "learning_rate": 4.8397504839750484e-08,
11554
- "loss": 0.8826,
11555
  "step": 18690
11556
  },
11557
  {
11558
- "epoch": 123.0,
11559
- "step": 18696,
11560
- "total_flos": 2.5832366503240237e+20,
11561
- "train_loss": 0.7828263307311466,
11562
- "train_runtime": 82672.8006,
11563
- "train_samples_per_second": 29.058,
11564
- "train_steps_per_second": 0.226
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
11565
  }
11566
  ],
11567
- "max_steps": 18696,
11568
- "num_train_epochs": 123,
11569
- "total_flos": 2.5832366503240237e+20,
11570
  "trial_name": null,
11571
  "trial_params": null
11572
  }
 
1
  {
2
  "best_metric": null,
3
  "best_model_checkpoint": null,
4
+ "epoch": 149.99509001636662,
5
+ "global_step": 22800,
6
  "is_hyper_param_search": false,
7
  "is_local_process_zero": true,
8
  "is_world_process_zero": true,
 
11442
  },
11443
  {
11444
  "epoch": 121.77,
11445
+ "learning_rate": 1.1362995594713658e-05,
11446
+ "loss": 0.9705,
11447
  "step": 18510
11448
  },
11449
  {
11450
  "epoch": 121.84,
11451
+ "learning_rate": 1.1336563876651981e-05,
11452
+ "loss": 0.8461,
11453
  "step": 18520
11454
  },
11455
  {
11456
  "epoch": 121.9,
11457
+ "learning_rate": 1.1310132158590309e-05,
11458
+ "loss": 0.8587,
11459
  "step": 18530
11460
  },
11461
  {
11462
  "epoch": 121.97,
11463
+ "learning_rate": 1.1283700440528634e-05,
11464
+ "loss": 0.8934,
11465
  "step": 18540
11466
  },
11467
  {
11468
  "epoch": 122.04,
11469
+ "learning_rate": 1.1257268722466961e-05,
11470
+ "loss": 0.9975,
11471
  "step": 18550
11472
  },
11473
  {
11474
  "epoch": 122.1,
11475
+ "learning_rate": 1.1230837004405287e-05,
11476
+ "loss": 0.8517,
11477
  "step": 18560
11478
  },
11479
  {
11480
  "epoch": 122.17,
11481
+ "learning_rate": 1.1204405286343612e-05,
11482
+ "loss": 0.8799,
11483
  "step": 18570
11484
  },
11485
  {
11486
  "epoch": 122.24,
11487
+ "learning_rate": 1.117797356828194e-05,
11488
+ "loss": 0.9021,
11489
  "step": 18580
11490
  },
11491
  {
11492
  "epoch": 122.3,
11493
+ "learning_rate": 1.1151541850220265e-05,
11494
+ "loss": 0.8569,
11495
  "step": 18590
11496
  },
11497
  {
11498
  "epoch": 122.37,
11499
+ "learning_rate": 1.112511013215859e-05,
11500
+ "loss": 0.8779,
11501
  "step": 18600
11502
  },
11503
  {
11504
  "epoch": 122.43,
11505
+ "learning_rate": 1.1098678414096916e-05,
11506
+ "loss": 0.88,
11507
  "step": 18610
11508
  },
11509
  {
11510
  "epoch": 122.5,
11511
+ "learning_rate": 1.1072246696035243e-05,
11512
+ "loss": 0.9219,
11513
  "step": 18620
11514
  },
11515
  {
11516
  "epoch": 122.56,
11517
+ "learning_rate": 1.1045814977973568e-05,
11518
+ "loss": 0.8598,
11519
  "step": 18630
11520
  },
11521
  {
11522
  "epoch": 122.63,
11523
+ "learning_rate": 1.1019383259911894e-05,
11524
+ "loss": 0.8648,
11525
  "step": 18640
11526
  },
11527
  {
11528
  "epoch": 122.69,
11529
+ "learning_rate": 1.0992951541850221e-05,
11530
+ "loss": 0.8566,
11531
  "step": 18650
11532
  },
11533
  {
11534
  "epoch": 122.76,
11535
+ "learning_rate": 1.0966519823788547e-05,
11536
+ "loss": 0.9491,
11537
  "step": 18660
11538
  },
11539
  {
11540
  "epoch": 122.82,
11541
+ "learning_rate": 1.0940088105726872e-05,
11542
+ "loss": 0.8215,
11543
  "step": 18670
11544
  },
11545
  {
11546
  "epoch": 122.89,
11547
+ "learning_rate": 1.09136563876652e-05,
11548
+ "loss": 0.8726,
11549
  "step": 18680
11550
  },
11551
  {
11552
  "epoch": 122.96,
11553
+ "learning_rate": 1.0887224669603525e-05,
11554
+ "loss": 0.8909,
11555
  "step": 18690
11556
  },
11557
  {
11558
+ "epoch": 123.03,
11559
+ "learning_rate": 1.086079295154185e-05,
11560
+ "loss": 0.9692,
11561
+ "step": 18700
11562
+ },
11563
+ {
11564
+ "epoch": 123.09,
11565
+ "learning_rate": 1.0834361233480176e-05,
11566
+ "loss": 0.85,
11567
+ "step": 18710
11568
+ },
11569
+ {
11570
+ "epoch": 123.16,
11571
+ "learning_rate": 1.0807929515418503e-05,
11572
+ "loss": 0.8889,
11573
+ "step": 18720
11574
+ },
11575
+ {
11576
+ "epoch": 123.22,
11577
+ "learning_rate": 1.0781497797356828e-05,
11578
+ "loss": 0.8586,
11579
+ "step": 18730
11580
+ },
11581
+ {
11582
+ "epoch": 123.29,
11583
+ "learning_rate": 1.0755066079295154e-05,
11584
+ "loss": 0.9315,
11585
+ "step": 18740
11586
+ },
11587
+ {
11588
+ "epoch": 123.35,
11589
+ "learning_rate": 1.0728634361233481e-05,
11590
+ "loss": 0.8433,
11591
+ "step": 18750
11592
+ },
11593
+ {
11594
+ "epoch": 123.42,
11595
+ "learning_rate": 1.0702202643171806e-05,
11596
+ "loss": 0.8922,
11597
+ "step": 18760
11598
+ },
11599
+ {
11600
+ "epoch": 123.48,
11601
+ "learning_rate": 1.0675770925110134e-05,
11602
+ "loss": 0.9134,
11603
+ "step": 18770
11604
+ },
11605
+ {
11606
+ "epoch": 123.55,
11607
+ "learning_rate": 1.0649339207048457e-05,
11608
+ "loss": 0.8996,
11609
+ "step": 18780
11610
+ },
11611
+ {
11612
+ "epoch": 123.62,
11613
+ "learning_rate": 1.0622907488986785e-05,
11614
+ "loss": 0.8622,
11615
+ "step": 18790
11616
+ },
11617
+ {
11618
+ "epoch": 123.68,
11619
+ "learning_rate": 1.0596475770925112e-05,
11620
+ "loss": 0.8691,
11621
+ "step": 18800
11622
+ },
11623
+ {
11624
+ "epoch": 123.75,
11625
+ "learning_rate": 1.0570044052863435e-05,
11626
+ "loss": 0.9627,
11627
+ "step": 18810
11628
+ },
11629
+ {
11630
+ "epoch": 123.81,
11631
+ "learning_rate": 1.0543612334801763e-05,
11632
+ "loss": 0.8812,
11633
+ "step": 18820
11634
+ },
11635
+ {
11636
+ "epoch": 123.88,
11637
+ "learning_rate": 1.0517180616740088e-05,
11638
+ "loss": 0.8543,
11639
+ "step": 18830
11640
+ },
11641
+ {
11642
+ "epoch": 123.94,
11643
+ "learning_rate": 1.0490748898678415e-05,
11644
+ "loss": 0.8661,
11645
+ "step": 18840
11646
+ },
11647
+ {
11648
+ "epoch": 124.01,
11649
+ "learning_rate": 1.0464317180616739e-05,
11650
+ "loss": 1.0126,
11651
+ "step": 18850
11652
+ },
11653
+ {
11654
+ "epoch": 124.08,
11655
+ "learning_rate": 1.0437885462555066e-05,
11656
+ "loss": 0.8392,
11657
+ "step": 18860
11658
+ },
11659
+ {
11660
+ "epoch": 124.14,
11661
+ "learning_rate": 1.0411453744493393e-05,
11662
+ "loss": 0.8753,
11663
+ "step": 18870
11664
+ },
11665
+ {
11666
+ "epoch": 124.21,
11667
+ "learning_rate": 1.0385022026431717e-05,
11668
+ "loss": 0.8561,
11669
+ "step": 18880
11670
+ },
11671
+ {
11672
+ "epoch": 124.27,
11673
+ "learning_rate": 1.0358590308370044e-05,
11674
+ "loss": 0.9043,
11675
+ "step": 18890
11676
+ },
11677
+ {
11678
+ "epoch": 124.34,
11679
+ "learning_rate": 1.033215859030837e-05,
11680
+ "loss": 0.8524,
11681
+ "step": 18900
11682
+ },
11683
+ {
11684
+ "epoch": 124.41,
11685
+ "learning_rate": 1.0305726872246697e-05,
11686
+ "loss": 0.8642,
11687
+ "step": 18910
11688
+ },
11689
+ {
11690
+ "epoch": 124.47,
11691
+ "learning_rate": 1.0279295154185022e-05,
11692
+ "loss": 0.8889,
11693
+ "step": 18920
11694
+ },
11695
+ {
11696
+ "epoch": 124.54,
11697
+ "learning_rate": 1.0252863436123348e-05,
11698
+ "loss": 0.9359,
11699
+ "step": 18930
11700
+ },
11701
+ {
11702
+ "epoch": 124.6,
11703
+ "learning_rate": 1.0226431718061675e-05,
11704
+ "loss": 0.8471,
11705
+ "step": 18940
11706
+ },
11707
+ {
11708
+ "epoch": 124.67,
11709
+ "learning_rate": 1.02e-05,
11710
+ "loss": 0.859,
11711
+ "step": 18950
11712
+ },
11713
+ {
11714
+ "epoch": 124.73,
11715
+ "learning_rate": 1.0173568281938326e-05,
11716
+ "loss": 0.8839,
11717
+ "step": 18960
11718
+ },
11719
+ {
11720
+ "epoch": 124.8,
11721
+ "learning_rate": 1.0147136563876652e-05,
11722
+ "loss": 0.8672,
11723
+ "step": 18970
11724
+ },
11725
+ {
11726
+ "epoch": 124.86,
11727
+ "learning_rate": 1.0120704845814979e-05,
11728
+ "loss": 0.8499,
11729
+ "step": 18980
11730
+ },
11731
+ {
11732
+ "epoch": 124.93,
11733
+ "learning_rate": 1.0094273127753304e-05,
11734
+ "loss": 0.9104,
11735
+ "step": 18990
11736
+ },
11737
+ {
11738
+ "epoch": 125.0,
11739
+ "learning_rate": 1.006784140969163e-05,
11740
+ "loss": 0.9453,
11741
+ "step": 19000
11742
+ },
11743
+ {
11744
+ "epoch": 125.0,
11745
+ "eval_loss": 0.3414785861968994,
11746
+ "eval_runtime": 320.2801,
11747
+ "eval_samples_per_second": 22.199,
11748
+ "eval_steps_per_second": 1.389,
11749
+ "eval_wer": 0.34803058163166556,
11750
+ "step": 19000
11751
+ },
11752
+ {
11753
+ "epoch": 125.07,
11754
+ "learning_rate": 1.0041409691629957e-05,
11755
+ "loss": 0.922,
11756
+ "step": 19010
11757
+ },
11758
+ {
11759
+ "epoch": 125.13,
11760
+ "learning_rate": 1.0014977973568282e-05,
11761
+ "loss": 0.8491,
11762
+ "step": 19020
11763
+ },
11764
+ {
11765
+ "epoch": 125.2,
11766
+ "learning_rate": 9.988546255506608e-06,
11767
+ "loss": 0.8899,
11768
+ "step": 19030
11769
+ },
11770
+ {
11771
+ "epoch": 125.26,
11772
+ "learning_rate": 9.962114537444933e-06,
11773
+ "loss": 0.9163,
11774
+ "step": 19040
11775
+ },
11776
+ {
11777
+ "epoch": 125.33,
11778
+ "learning_rate": 9.93568281938326e-06,
11779
+ "loss": 0.853,
11780
+ "step": 19050
11781
+ },
11782
+ {
11783
+ "epoch": 125.39,
11784
+ "learning_rate": 9.909251101321588e-06,
11785
+ "loss": 0.866,
11786
+ "step": 19060
11787
+ },
11788
+ {
11789
+ "epoch": 125.46,
11790
+ "learning_rate": 9.882819383259911e-06,
11791
+ "loss": 0.8817,
11792
+ "step": 19070
11793
+ },
11794
+ {
11795
+ "epoch": 125.52,
11796
+ "learning_rate": 9.856387665198239e-06,
11797
+ "loss": 0.9513,
11798
+ "step": 19080
11799
+ },
11800
+ {
11801
+ "epoch": 125.59,
11802
+ "learning_rate": 9.829955947136564e-06,
11803
+ "loss": 0.8217,
11804
+ "step": 19090
11805
+ },
11806
+ {
11807
+ "epoch": 125.65,
11808
+ "learning_rate": 9.80352422907489e-06,
11809
+ "loss": 0.8926,
11810
+ "step": 19100
11811
+ },
11812
+ {
11813
+ "epoch": 125.72,
11814
+ "learning_rate": 9.777092511013217e-06,
11815
+ "loss": 0.8642,
11816
+ "step": 19110
11817
+ },
11818
+ {
11819
+ "epoch": 125.79,
11820
+ "learning_rate": 9.750660792951542e-06,
11821
+ "loss": 0.9165,
11822
+ "step": 19120
11823
+ },
11824
+ {
11825
+ "epoch": 125.85,
11826
+ "learning_rate": 9.72422907488987e-06,
11827
+ "loss": 0.8646,
11828
+ "step": 19130
11829
+ },
11830
+ {
11831
+ "epoch": 125.92,
11832
+ "learning_rate": 9.697797356828193e-06,
11833
+ "loss": 0.8568,
11834
+ "step": 19140
11835
+ },
11836
+ {
11837
+ "epoch": 125.98,
11838
+ "learning_rate": 9.67136563876652e-06,
11839
+ "loss": 0.8925,
11840
+ "step": 19150
11841
+ },
11842
+ {
11843
+ "epoch": 126.05,
11844
+ "learning_rate": 9.644933920704846e-06,
11845
+ "loss": 0.9546,
11846
+ "step": 19160
11847
+ },
11848
+ {
11849
+ "epoch": 126.12,
11850
+ "learning_rate": 9.618502202643171e-06,
11851
+ "loss": 0.8381,
11852
+ "step": 19170
11853
+ },
11854
+ {
11855
+ "epoch": 126.18,
11856
+ "learning_rate": 9.592070484581498e-06,
11857
+ "loss": 0.8962,
11858
+ "step": 19180
11859
+ },
11860
+ {
11861
+ "epoch": 126.25,
11862
+ "learning_rate": 9.565638766519824e-06,
11863
+ "loss": 0.9426,
11864
+ "step": 19190
11865
+ },
11866
+ {
11867
+ "epoch": 126.31,
11868
+ "learning_rate": 9.539207048458151e-06,
11869
+ "loss": 0.8578,
11870
+ "step": 19200
11871
+ },
11872
+ {
11873
+ "epoch": 126.38,
11874
+ "learning_rate": 9.512775330396475e-06,
11875
+ "loss": 0.8628,
11876
+ "step": 19210
11877
+ },
11878
+ {
11879
+ "epoch": 126.45,
11880
+ "learning_rate": 9.486343612334802e-06,
11881
+ "loss": 0.8734,
11882
+ "step": 19220
11883
+ },
11884
+ {
11885
+ "epoch": 126.51,
11886
+ "learning_rate": 9.459911894273127e-06,
11887
+ "loss": 0.9106,
11888
+ "step": 19230
11889
+ },
11890
+ {
11891
+ "epoch": 126.58,
11892
+ "learning_rate": 9.433480176211455e-06,
11893
+ "loss": 0.8388,
11894
+ "step": 19240
11895
+ },
11896
+ {
11897
+ "epoch": 126.64,
11898
+ "learning_rate": 9.40704845814978e-06,
11899
+ "loss": 0.8349,
11900
+ "step": 19250
11901
+ },
11902
+ {
11903
+ "epoch": 126.71,
11904
+ "learning_rate": 9.380616740088106e-06,
11905
+ "loss": 0.8748,
11906
+ "step": 19260
11907
+ },
11908
+ {
11909
+ "epoch": 126.77,
11910
+ "learning_rate": 9.354185022026433e-06,
11911
+ "loss": 0.9127,
11912
+ "step": 19270
11913
+ },
11914
+ {
11915
+ "epoch": 126.84,
11916
+ "learning_rate": 9.327753303964757e-06,
11917
+ "loss": 0.8584,
11918
+ "step": 19280
11919
+ },
11920
+ {
11921
+ "epoch": 126.9,
11922
+ "learning_rate": 9.301321585903084e-06,
11923
+ "loss": 0.8903,
11924
+ "step": 19290
11925
+ },
11926
+ {
11927
+ "epoch": 126.97,
11928
+ "learning_rate": 9.274889867841411e-06,
11929
+ "loss": 0.8449,
11930
+ "step": 19300
11931
+ },
11932
+ {
11933
+ "epoch": 127.04,
11934
+ "learning_rate": 9.248458149779736e-06,
11935
+ "loss": 0.9825,
11936
+ "step": 19310
11937
+ },
11938
+ {
11939
+ "epoch": 127.1,
11940
+ "learning_rate": 9.222026431718062e-06,
11941
+ "loss": 0.8495,
11942
+ "step": 19320
11943
+ },
11944
+ {
11945
+ "epoch": 127.17,
11946
+ "learning_rate": 9.195594713656387e-06,
11947
+ "loss": 0.8666,
11948
+ "step": 19330
11949
+ },
11950
+ {
11951
+ "epoch": 127.24,
11952
+ "learning_rate": 9.169162995594714e-06,
11953
+ "loss": 0.8931,
11954
+ "step": 19340
11955
+ },
11956
+ {
11957
+ "epoch": 127.3,
11958
+ "learning_rate": 9.142731277533038e-06,
11959
+ "loss": 0.8946,
11960
+ "step": 19350
11961
+ },
11962
+ {
11963
+ "epoch": 127.37,
11964
+ "learning_rate": 9.116299559471365e-06,
11965
+ "loss": 0.8899,
11966
+ "step": 19360
11967
+ },
11968
+ {
11969
+ "epoch": 127.43,
11970
+ "learning_rate": 9.089867841409693e-06,
11971
+ "loss": 0.8799,
11972
+ "step": 19370
11973
+ },
11974
+ {
11975
+ "epoch": 127.5,
11976
+ "learning_rate": 9.063436123348018e-06,
11977
+ "loss": 0.9319,
11978
+ "step": 19380
11979
+ },
11980
+ {
11981
+ "epoch": 127.56,
11982
+ "learning_rate": 9.037004405286344e-06,
11983
+ "loss": 0.8407,
11984
+ "step": 19390
11985
+ },
11986
+ {
11987
+ "epoch": 127.63,
11988
+ "learning_rate": 9.010572687224669e-06,
11989
+ "loss": 0.8338,
11990
+ "step": 19400
11991
+ },
11992
+ {
11993
+ "epoch": 127.69,
11994
+ "learning_rate": 8.984140969162996e-06,
11995
+ "loss": 0.8834,
11996
+ "step": 19410
11997
+ },
11998
+ {
11999
+ "epoch": 127.76,
12000
+ "learning_rate": 8.957709251101323e-06,
12001
+ "loss": 0.9279,
12002
+ "step": 19420
12003
+ },
12004
+ {
12005
+ "epoch": 127.82,
12006
+ "learning_rate": 8.931277533039647e-06,
12007
+ "loss": 0.8409,
12008
+ "step": 19430
12009
+ },
12010
+ {
12011
+ "epoch": 127.89,
12012
+ "learning_rate": 8.904845814977974e-06,
12013
+ "loss": 0.871,
12014
+ "step": 19440
12015
+ },
12016
+ {
12017
+ "epoch": 127.96,
12018
+ "learning_rate": 8.8784140969163e-06,
12019
+ "loss": 0.8758,
12020
+ "step": 19450
12021
+ },
12022
+ {
12023
+ "epoch": 128.03,
12024
+ "learning_rate": 8.851982378854625e-06,
12025
+ "loss": 0.9881,
12026
+ "step": 19460
12027
+ },
12028
+ {
12029
+ "epoch": 128.09,
12030
+ "learning_rate": 8.82555066079295e-06,
12031
+ "loss": 0.8283,
12032
+ "step": 19470
12033
+ },
12034
+ {
12035
+ "epoch": 128.16,
12036
+ "learning_rate": 8.799118942731278e-06,
12037
+ "loss": 0.8865,
12038
+ "step": 19480
12039
+ },
12040
+ {
12041
+ "epoch": 128.22,
12042
+ "learning_rate": 8.772687224669605e-06,
12043
+ "loss": 0.8691,
12044
+ "step": 19490
12045
+ },
12046
+ {
12047
+ "epoch": 128.29,
12048
+ "learning_rate": 8.746255506607929e-06,
12049
+ "loss": 0.9267,
12050
+ "step": 19500
12051
+ },
12052
+ {
12053
+ "epoch": 128.29,
12054
+ "eval_loss": 0.3477088212966919,
12055
+ "eval_runtime": 318.7167,
12056
+ "eval_samples_per_second": 22.308,
12057
+ "eval_steps_per_second": 1.396,
12058
+ "eval_wer": 0.35029517081196165,
12059
+ "step": 19500
12060
+ },
12061
+ {
12062
+ "epoch": 128.35,
12063
+ "learning_rate": 8.719823788546256e-06,
12064
+ "loss": 0.8531,
12065
+ "step": 19510
12066
+ },
12067
+ {
12068
+ "epoch": 128.42,
12069
+ "learning_rate": 8.693392070484582e-06,
12070
+ "loss": 0.8753,
12071
+ "step": 19520
12072
+ },
12073
+ {
12074
+ "epoch": 128.48,
12075
+ "learning_rate": 8.666960352422909e-06,
12076
+ "loss": 0.8862,
12077
+ "step": 19530
12078
+ },
12079
+ {
12080
+ "epoch": 128.55,
12081
+ "learning_rate": 8.640528634361234e-06,
12082
+ "loss": 0.8788,
12083
+ "step": 19540
12084
+ },
12085
+ {
12086
+ "epoch": 128.62,
12087
+ "learning_rate": 8.61409691629956e-06,
12088
+ "loss": 0.8425,
12089
+ "step": 19550
12090
+ },
12091
+ {
12092
+ "epoch": 128.68,
12093
+ "learning_rate": 8.587665198237887e-06,
12094
+ "loss": 0.8931,
12095
+ "step": 19560
12096
+ },
12097
+ {
12098
+ "epoch": 128.75,
12099
+ "learning_rate": 8.56123348017621e-06,
12100
+ "loss": 0.8952,
12101
+ "step": 19570
12102
+ },
12103
+ {
12104
+ "epoch": 128.81,
12105
+ "learning_rate": 8.534801762114538e-06,
12106
+ "loss": 0.8496,
12107
+ "step": 19580
12108
+ },
12109
+ {
12110
+ "epoch": 128.88,
12111
+ "learning_rate": 8.508370044052863e-06,
12112
+ "loss": 0.8719,
12113
+ "step": 19590
12114
+ },
12115
+ {
12116
+ "epoch": 128.94,
12117
+ "learning_rate": 8.48193832599119e-06,
12118
+ "loss": 0.8714,
12119
+ "step": 19600
12120
+ },
12121
+ {
12122
+ "epoch": 129.01,
12123
+ "learning_rate": 8.455506607929516e-06,
12124
+ "loss": 0.9905,
12125
+ "step": 19610
12126
+ },
12127
+ {
12128
+ "epoch": 129.08,
12129
+ "learning_rate": 8.429074889867841e-06,
12130
+ "loss": 0.8359,
12131
+ "step": 19620
12132
+ },
12133
+ {
12134
+ "epoch": 129.14,
12135
+ "learning_rate": 8.402643171806169e-06,
12136
+ "loss": 0.8713,
12137
+ "step": 19630
12138
+ },
12139
+ {
12140
+ "epoch": 129.21,
12141
+ "learning_rate": 8.376211453744492e-06,
12142
+ "loss": 0.8612,
12143
+ "step": 19640
12144
+ },
12145
+ {
12146
+ "epoch": 129.27,
12147
+ "learning_rate": 8.34977973568282e-06,
12148
+ "loss": 0.9565,
12149
+ "step": 19650
12150
+ },
12151
+ {
12152
+ "epoch": 129.34,
12153
+ "learning_rate": 8.323348017621145e-06,
12154
+ "loss": 0.8521,
12155
+ "step": 19660
12156
+ },
12157
+ {
12158
+ "epoch": 129.41,
12159
+ "learning_rate": 8.296916299559472e-06,
12160
+ "loss": 0.8548,
12161
+ "step": 19670
12162
+ },
12163
+ {
12164
+ "epoch": 129.47,
12165
+ "learning_rate": 8.270484581497798e-06,
12166
+ "loss": 0.861,
12167
+ "step": 19680
12168
+ },
12169
+ {
12170
+ "epoch": 129.54,
12171
+ "learning_rate": 8.244052863436123e-06,
12172
+ "loss": 0.9403,
12173
+ "step": 19690
12174
+ },
12175
+ {
12176
+ "epoch": 129.6,
12177
+ "learning_rate": 8.21762114537445e-06,
12178
+ "loss": 0.8556,
12179
+ "step": 19700
12180
+ },
12181
+ {
12182
+ "epoch": 129.67,
12183
+ "learning_rate": 8.191189427312776e-06,
12184
+ "loss": 0.853,
12185
+ "step": 19710
12186
+ },
12187
+ {
12188
+ "epoch": 129.73,
12189
+ "learning_rate": 8.164757709251101e-06,
12190
+ "loss": 0.8882,
12191
+ "step": 19720
12192
+ },
12193
+ {
12194
+ "epoch": 129.8,
12195
+ "learning_rate": 8.138325991189428e-06,
12196
+ "loss": 0.889,
12197
+ "step": 19730
12198
+ },
12199
+ {
12200
+ "epoch": 129.86,
12201
+ "learning_rate": 8.111894273127754e-06,
12202
+ "loss": 0.8306,
12203
+ "step": 19740
12204
+ },
12205
+ {
12206
+ "epoch": 129.93,
12207
+ "learning_rate": 8.08546255506608e-06,
12208
+ "loss": 0.8692,
12209
+ "step": 19750
12210
+ },
12211
+ {
12212
+ "epoch": 130.0,
12213
+ "learning_rate": 8.059030837004405e-06,
12214
+ "loss": 0.894,
12215
+ "step": 19760
12216
+ },
12217
+ {
12218
+ "epoch": 130.07,
12219
+ "learning_rate": 8.032599118942732e-06,
12220
+ "loss": 0.9069,
12221
+ "step": 19770
12222
+ },
12223
+ {
12224
+ "epoch": 130.13,
12225
+ "learning_rate": 8.006167400881057e-06,
12226
+ "loss": 0.8602,
12227
+ "step": 19780
12228
+ },
12229
+ {
12230
+ "epoch": 130.2,
12231
+ "learning_rate": 7.979735682819383e-06,
12232
+ "loss": 0.8758,
12233
+ "step": 19790
12234
+ },
12235
+ {
12236
+ "epoch": 130.26,
12237
+ "learning_rate": 7.95330396475771e-06,
12238
+ "loss": 0.9363,
12239
+ "step": 19800
12240
+ },
12241
+ {
12242
+ "epoch": 130.33,
12243
+ "learning_rate": 7.926872246696036e-06,
12244
+ "loss": 0.8534,
12245
+ "step": 19810
12246
+ },
12247
+ {
12248
+ "epoch": 130.39,
12249
+ "learning_rate": 7.900440528634361e-06,
12250
+ "loss": 0.8481,
12251
+ "step": 19820
12252
+ },
12253
+ {
12254
+ "epoch": 130.46,
12255
+ "learning_rate": 7.874008810572686e-06,
12256
+ "loss": 0.869,
12257
+ "step": 19830
12258
+ },
12259
+ {
12260
+ "epoch": 130.52,
12261
+ "learning_rate": 7.847577092511014e-06,
12262
+ "loss": 0.9053,
12263
+ "step": 19840
12264
+ },
12265
+ {
12266
+ "epoch": 130.59,
12267
+ "learning_rate": 7.821145374449339e-06,
12268
+ "loss": 0.8458,
12269
+ "step": 19850
12270
+ },
12271
+ {
12272
+ "epoch": 130.65,
12273
+ "learning_rate": 7.794713656387665e-06,
12274
+ "loss": 0.8722,
12275
+ "step": 19860
12276
+ },
12277
+ {
12278
+ "epoch": 130.72,
12279
+ "learning_rate": 7.768281938325992e-06,
12280
+ "loss": 0.8498,
12281
+ "step": 19870
12282
+ },
12283
+ {
12284
+ "epoch": 130.79,
12285
+ "learning_rate": 7.741850220264317e-06,
12286
+ "loss": 0.9224,
12287
+ "step": 19880
12288
+ },
12289
+ {
12290
+ "epoch": 130.85,
12291
+ "learning_rate": 7.715418502202644e-06,
12292
+ "loss": 0.8419,
12293
+ "step": 19890
12294
+ },
12295
+ {
12296
+ "epoch": 130.92,
12297
+ "learning_rate": 7.688986784140968e-06,
12298
+ "loss": 0.8453,
12299
+ "step": 19900
12300
+ },
12301
+ {
12302
+ "epoch": 130.98,
12303
+ "learning_rate": 7.662555066079295e-06,
12304
+ "loss": 0.8761,
12305
+ "step": 19910
12306
+ },
12307
+ {
12308
+ "epoch": 131.05,
12309
+ "learning_rate": 7.636123348017623e-06,
12310
+ "loss": 0.9359,
12311
+ "step": 19920
12312
+ },
12313
+ {
12314
+ "epoch": 131.12,
12315
+ "learning_rate": 7.609691629955946e-06,
12316
+ "loss": 0.8415,
12317
+ "step": 19930
12318
+ },
12319
+ {
12320
+ "epoch": 131.18,
12321
+ "learning_rate": 7.583259911894273e-06,
12322
+ "loss": 0.8569,
12323
+ "step": 19940
12324
+ },
12325
+ {
12326
+ "epoch": 131.25,
12327
+ "learning_rate": 7.5568281938326e-06,
12328
+ "loss": 0.9054,
12329
+ "step": 19950
12330
+ },
12331
+ {
12332
+ "epoch": 131.31,
12333
+ "learning_rate": 7.530396475770926e-06,
12334
+ "loss": 0.8698,
12335
+ "step": 19960
12336
+ },
12337
+ {
12338
+ "epoch": 131.38,
12339
+ "learning_rate": 7.503964757709251e-06,
12340
+ "loss": 0.8543,
12341
+ "step": 19970
12342
+ },
12343
+ {
12344
+ "epoch": 131.45,
12345
+ "learning_rate": 7.477533039647577e-06,
12346
+ "loss": 0.8847,
12347
+ "step": 19980
12348
+ },
12349
+ {
12350
+ "epoch": 131.51,
12351
+ "learning_rate": 7.451101321585903e-06,
12352
+ "loss": 0.9421,
12353
+ "step": 19990
12354
+ },
12355
+ {
12356
+ "epoch": 131.58,
12357
+ "learning_rate": 7.424669603524229e-06,
12358
+ "loss": 0.8315,
12359
+ "step": 20000
12360
+ },
12361
+ {
12362
+ "epoch": 131.58,
12363
+ "eval_loss": 0.3476375639438629,
12364
+ "eval_runtime": 318.7177,
12365
+ "eval_samples_per_second": 22.308,
12366
+ "eval_steps_per_second": 1.396,
12367
+ "eval_wer": 0.3504887254427562,
12368
+ "step": 20000
12369
+ },
12370
+ {
12371
+ "epoch": 131.64,
12372
+ "learning_rate": 7.398237885462555e-06,
12373
+ "loss": 0.8478,
12374
+ "step": 20010
12375
+ },
12376
+ {
12377
+ "epoch": 131.71,
12378
+ "learning_rate": 7.3718061674008815e-06,
12379
+ "loss": 0.8467,
12380
+ "step": 20020
12381
+ },
12382
+ {
12383
+ "epoch": 131.77,
12384
+ "learning_rate": 7.345374449339207e-06,
12385
+ "loss": 0.9429,
12386
+ "step": 20030
12387
+ },
12388
+ {
12389
+ "epoch": 131.84,
12390
+ "learning_rate": 7.318942731277533e-06,
12391
+ "loss": 0.8449,
12392
+ "step": 20040
12393
+ },
12394
+ {
12395
+ "epoch": 131.9,
12396
+ "learning_rate": 7.292511013215859e-06,
12397
+ "loss": 0.8877,
12398
+ "step": 20050
12399
+ },
12400
+ {
12401
+ "epoch": 131.97,
12402
+ "learning_rate": 7.266079295154185e-06,
12403
+ "loss": 0.8775,
12404
+ "step": 20060
12405
+ },
12406
+ {
12407
+ "epoch": 132.04,
12408
+ "learning_rate": 7.2396475770925115e-06,
12409
+ "loss": 0.956,
12410
+ "step": 20070
12411
+ },
12412
+ {
12413
+ "epoch": 132.1,
12414
+ "learning_rate": 7.213215859030838e-06,
12415
+ "loss": 0.8259,
12416
+ "step": 20080
12417
+ },
12418
+ {
12419
+ "epoch": 132.17,
12420
+ "learning_rate": 7.186784140969163e-06,
12421
+ "loss": 0.8629,
12422
+ "step": 20090
12423
+ },
12424
+ {
12425
+ "epoch": 132.24,
12426
+ "learning_rate": 7.160352422907489e-06,
12427
+ "loss": 0.8964,
12428
+ "step": 20100
12429
+ },
12430
+ {
12431
+ "epoch": 132.3,
12432
+ "learning_rate": 7.133920704845815e-06,
12433
+ "loss": 0.9075,
12434
+ "step": 20110
12435
+ },
12436
+ {
12437
+ "epoch": 132.37,
12438
+ "learning_rate": 7.1074889867841405e-06,
12439
+ "loss": 0.8127,
12440
+ "step": 20120
12441
+ },
12442
+ {
12443
+ "epoch": 132.43,
12444
+ "learning_rate": 7.081057268722468e-06,
12445
+ "loss": 0.8789,
12446
+ "step": 20130
12447
+ },
12448
+ {
12449
+ "epoch": 132.5,
12450
+ "learning_rate": 7.054625550660793e-06,
12451
+ "loss": 0.9102,
12452
+ "step": 20140
12453
+ },
12454
+ {
12455
+ "epoch": 132.56,
12456
+ "learning_rate": 7.0281938325991195e-06,
12457
+ "loss": 0.8588,
12458
+ "step": 20150
12459
+ },
12460
+ {
12461
+ "epoch": 132.63,
12462
+ "learning_rate": 7.001762114537445e-06,
12463
+ "loss": 0.8515,
12464
+ "step": 20160
12465
+ },
12466
+ {
12467
+ "epoch": 132.69,
12468
+ "learning_rate": 6.975330396475771e-06,
12469
+ "loss": 0.8963,
12470
+ "step": 20170
12471
+ },
12472
+ {
12473
+ "epoch": 132.76,
12474
+ "learning_rate": 6.948898678414097e-06,
12475
+ "loss": 0.9224,
12476
+ "step": 20180
12477
+ },
12478
+ {
12479
+ "epoch": 132.82,
12480
+ "learning_rate": 6.922466960352423e-06,
12481
+ "loss": 0.8314,
12482
+ "step": 20190
12483
+ },
12484
+ {
12485
+ "epoch": 132.89,
12486
+ "learning_rate": 6.896035242290749e-06,
12487
+ "loss": 0.857,
12488
+ "step": 20200
12489
+ },
12490
+ {
12491
+ "epoch": 132.96,
12492
+ "learning_rate": 6.869603524229075e-06,
12493
+ "loss": 0.8458,
12494
+ "step": 20210
12495
+ },
12496
+ {
12497
+ "epoch": 133.03,
12498
+ "learning_rate": 6.843171806167401e-06,
12499
+ "loss": 0.9865,
12500
+ "step": 20220
12501
+ },
12502
+ {
12503
+ "epoch": 133.09,
12504
+ "learning_rate": 6.816740088105727e-06,
12505
+ "loss": 0.8427,
12506
+ "step": 20230
12507
+ },
12508
+ {
12509
+ "epoch": 133.16,
12510
+ "learning_rate": 6.790308370044053e-06,
12511
+ "loss": 0.8474,
12512
+ "step": 20240
12513
+ },
12514
+ {
12515
+ "epoch": 133.22,
12516
+ "learning_rate": 6.7638766519823785e-06,
12517
+ "loss": 0.8718,
12518
+ "step": 20250
12519
+ },
12520
+ {
12521
+ "epoch": 133.29,
12522
+ "learning_rate": 6.737444933920706e-06,
12523
+ "loss": 0.9156,
12524
+ "step": 20260
12525
+ },
12526
+ {
12527
+ "epoch": 133.35,
12528
+ "learning_rate": 6.711013215859031e-06,
12529
+ "loss": 0.8345,
12530
+ "step": 20270
12531
+ },
12532
+ {
12533
+ "epoch": 133.42,
12534
+ "learning_rate": 6.6845814977973575e-06,
12535
+ "loss": 0.8664,
12536
+ "step": 20280
12537
+ },
12538
+ {
12539
+ "epoch": 133.48,
12540
+ "learning_rate": 6.658149779735683e-06,
12541
+ "loss": 0.9086,
12542
+ "step": 20290
12543
+ },
12544
+ {
12545
+ "epoch": 133.55,
12546
+ "learning_rate": 6.631718061674008e-06,
12547
+ "loss": 0.8815,
12548
+ "step": 20300
12549
+ },
12550
+ {
12551
+ "epoch": 133.62,
12552
+ "learning_rate": 6.605286343612335e-06,
12553
+ "loss": 0.8388,
12554
+ "step": 20310
12555
+ },
12556
+ {
12557
+ "epoch": 133.68,
12558
+ "learning_rate": 6.578854625550661e-06,
12559
+ "loss": 0.8464,
12560
+ "step": 20320
12561
+ },
12562
+ {
12563
+ "epoch": 133.75,
12564
+ "learning_rate": 6.552422907488987e-06,
12565
+ "loss": 0.9118,
12566
+ "step": 20330
12567
+ },
12568
+ {
12569
+ "epoch": 133.81,
12570
+ "learning_rate": 6.525991189427313e-06,
12571
+ "loss": 0.836,
12572
+ "step": 20340
12573
+ },
12574
+ {
12575
+ "epoch": 133.88,
12576
+ "learning_rate": 6.499559471365639e-06,
12577
+ "loss": 0.8364,
12578
+ "step": 20350
12579
+ },
12580
+ {
12581
+ "epoch": 133.94,
12582
+ "learning_rate": 6.473127753303965e-06,
12583
+ "loss": 0.8733,
12584
+ "step": 20360
12585
+ },
12586
+ {
12587
+ "epoch": 134.01,
12588
+ "learning_rate": 6.446696035242291e-06,
12589
+ "loss": 0.9697,
12590
+ "step": 20370
12591
+ },
12592
+ {
12593
+ "epoch": 134.08,
12594
+ "learning_rate": 6.420264317180617e-06,
12595
+ "loss": 0.8447,
12596
+ "step": 20380
12597
+ },
12598
+ {
12599
+ "epoch": 134.14,
12600
+ "learning_rate": 6.393832599118943e-06,
12601
+ "loss": 0.8811,
12602
+ "step": 20390
12603
+ },
12604
+ {
12605
+ "epoch": 134.21,
12606
+ "learning_rate": 6.367400881057269e-06,
12607
+ "loss": 0.8537,
12608
+ "step": 20400
12609
+ },
12610
+ {
12611
+ "epoch": 134.27,
12612
+ "learning_rate": 6.3409691629955946e-06,
12613
+ "loss": 0.9016,
12614
+ "step": 20410
12615
+ },
12616
+ {
12617
+ "epoch": 134.34,
12618
+ "learning_rate": 6.314537444933921e-06,
12619
+ "loss": 0.8262,
12620
+ "step": 20420
12621
+ },
12622
+ {
12623
+ "epoch": 134.41,
12624
+ "learning_rate": 6.288105726872246e-06,
12625
+ "loss": 0.8514,
12626
+ "step": 20430
12627
+ },
12628
+ {
12629
+ "epoch": 134.47,
12630
+ "learning_rate": 6.261674008810573e-06,
12631
+ "loss": 0.8821,
12632
+ "step": 20440
12633
+ },
12634
+ {
12635
+ "epoch": 134.54,
12636
+ "learning_rate": 6.235242290748899e-06,
12637
+ "loss": 0.8956,
12638
+ "step": 20450
12639
+ },
12640
+ {
12641
+ "epoch": 134.6,
12642
+ "learning_rate": 6.208810572687225e-06,
12643
+ "loss": 0.8119,
12644
+ "step": 20460
12645
+ },
12646
+ {
12647
+ "epoch": 134.67,
12648
+ "learning_rate": 6.182378854625551e-06,
12649
+ "loss": 0.8541,
12650
+ "step": 20470
12651
+ },
12652
+ {
12653
+ "epoch": 134.73,
12654
+ "learning_rate": 6.155947136563876e-06,
12655
+ "loss": 0.8941,
12656
+ "step": 20480
12657
+ },
12658
+ {
12659
+ "epoch": 134.8,
12660
+ "learning_rate": 6.129515418502203e-06,
12661
+ "loss": 0.8826,
12662
+ "step": 20490
12663
+ },
12664
+ {
12665
+ "epoch": 134.86,
12666
+ "learning_rate": 6.103083700440528e-06,
12667
+ "loss": 0.8542,
12668
+ "step": 20500
12669
+ },
12670
+ {
12671
+ "epoch": 134.86,
12672
+ "eval_loss": 0.34747639298439026,
12673
+ "eval_runtime": 318.584,
12674
+ "eval_samples_per_second": 22.318,
12675
+ "eval_steps_per_second": 1.397,
12676
+ "eval_wer": 0.3505855027581535,
12677
+ "step": 20500
12678
+ },
12679
+ {
12680
+ "epoch": 134.93,
12681
+ "learning_rate": 6.076651982378855e-06,
12682
+ "loss": 0.8579,
12683
+ "step": 20510
12684
+ },
12685
+ {
12686
+ "epoch": 135.0,
12687
+ "learning_rate": 6.050220264317181e-06,
12688
+ "loss": 0.9107,
12689
+ "step": 20520
12690
+ },
12691
+ {
12692
+ "epoch": 135.07,
12693
+ "learning_rate": 6.023788546255507e-06,
12694
+ "loss": 0.9237,
12695
+ "step": 20530
12696
+ },
12697
+ {
12698
+ "epoch": 135.13,
12699
+ "learning_rate": 5.9973568281938325e-06,
12700
+ "loss": 0.8477,
12701
+ "step": 20540
12702
+ },
12703
+ {
12704
+ "epoch": 135.2,
12705
+ "learning_rate": 5.970925110132159e-06,
12706
+ "loss": 0.8807,
12707
+ "step": 20550
12708
+ },
12709
+ {
12710
+ "epoch": 135.26,
12711
+ "learning_rate": 5.944493392070484e-06,
12712
+ "loss": 0.9002,
12713
+ "step": 20560
12714
+ },
12715
+ {
12716
+ "epoch": 135.33,
12717
+ "learning_rate": 5.918061674008811e-06,
12718
+ "loss": 0.8213,
12719
+ "step": 20570
12720
+ },
12721
+ {
12722
+ "epoch": 135.39,
12723
+ "learning_rate": 5.891629955947137e-06,
12724
+ "loss": 0.8608,
12725
+ "step": 20580
12726
+ },
12727
+ {
12728
+ "epoch": 135.46,
12729
+ "learning_rate": 5.8651982378854624e-06,
12730
+ "loss": 0.8611,
12731
+ "step": 20590
12732
+ },
12733
+ {
12734
+ "epoch": 135.52,
12735
+ "learning_rate": 5.838766519823789e-06,
12736
+ "loss": 0.9184,
12737
+ "step": 20600
12738
+ },
12739
+ {
12740
+ "epoch": 135.59,
12741
+ "learning_rate": 5.812334801762114e-06,
12742
+ "loss": 0.8341,
12743
+ "step": 20610
12744
+ },
12745
+ {
12746
+ "epoch": 135.65,
12747
+ "learning_rate": 5.7859030837004406e-06,
12748
+ "loss": 0.8371,
12749
+ "step": 20620
12750
+ },
12751
+ {
12752
+ "epoch": 135.72,
12753
+ "learning_rate": 5.759471365638767e-06,
12754
+ "loss": 0.866,
12755
+ "step": 20630
12756
+ },
12757
+ {
12758
+ "epoch": 135.79,
12759
+ "learning_rate": 5.733039647577093e-06,
12760
+ "loss": 0.9034,
12761
+ "step": 20640
12762
+ },
12763
+ {
12764
+ "epoch": 135.85,
12765
+ "learning_rate": 5.706607929515419e-06,
12766
+ "loss": 0.85,
12767
+ "step": 20650
12768
+ },
12769
+ {
12770
+ "epoch": 135.92,
12771
+ "learning_rate": 5.680176211453745e-06,
12772
+ "loss": 0.8806,
12773
+ "step": 20660
12774
+ },
12775
+ {
12776
+ "epoch": 135.98,
12777
+ "learning_rate": 5.656387665198238e-06,
12778
+ "loss": 0.8653,
12779
+ "step": 20670
12780
+ },
12781
+ {
12782
+ "epoch": 136.05,
12783
+ "learning_rate": 5.6299559471365644e-06,
12784
+ "loss": 0.9354,
12785
+ "step": 20680
12786
+ },
12787
+ {
12788
+ "epoch": 136.12,
12789
+ "learning_rate": 5.60352422907489e-06,
12790
+ "loss": 0.8336,
12791
+ "step": 20690
12792
+ },
12793
+ {
12794
+ "epoch": 136.18,
12795
+ "learning_rate": 5.577092511013216e-06,
12796
+ "loss": 0.856,
12797
+ "step": 20700
12798
+ },
12799
+ {
12800
+ "epoch": 136.25,
12801
+ "learning_rate": 5.550660792951542e-06,
12802
+ "loss": 0.8877,
12803
+ "step": 20710
12804
+ },
12805
+ {
12806
+ "epoch": 136.31,
12807
+ "learning_rate": 5.524229074889868e-06,
12808
+ "loss": 0.8517,
12809
+ "step": 20720
12810
+ },
12811
+ {
12812
+ "epoch": 136.38,
12813
+ "learning_rate": 5.4977973568281935e-06,
12814
+ "loss": 0.8718,
12815
+ "step": 20730
12816
+ },
12817
+ {
12818
+ "epoch": 136.45,
12819
+ "learning_rate": 5.471365638766521e-06,
12820
+ "loss": 0.8488,
12821
+ "step": 20740
12822
+ },
12823
+ {
12824
+ "epoch": 136.51,
12825
+ "learning_rate": 5.444933920704846e-06,
12826
+ "loss": 0.8924,
12827
+ "step": 20750
12828
+ },
12829
+ {
12830
+ "epoch": 136.58,
12831
+ "learning_rate": 5.418502202643172e-06,
12832
+ "loss": 0.8292,
12833
+ "step": 20760
12834
+ },
12835
+ {
12836
+ "epoch": 136.64,
12837
+ "learning_rate": 5.392070484581498e-06,
12838
+ "loss": 0.8556,
12839
+ "step": 20770
12840
+ },
12841
+ {
12842
+ "epoch": 136.71,
12843
+ "learning_rate": 5.3656387665198234e-06,
12844
+ "loss": 0.8545,
12845
+ "step": 20780
12846
+ },
12847
+ {
12848
+ "epoch": 136.77,
12849
+ "learning_rate": 5.33920704845815e-06,
12850
+ "loss": 0.9007,
12851
+ "step": 20790
12852
+ },
12853
+ {
12854
+ "epoch": 136.84,
12855
+ "learning_rate": 5.312775330396476e-06,
12856
+ "loss": 0.8256,
12857
+ "step": 20800
12858
+ },
12859
+ {
12860
+ "epoch": 136.9,
12861
+ "learning_rate": 5.286343612334802e-06,
12862
+ "loss": 0.8613,
12863
+ "step": 20810
12864
+ },
12865
+ {
12866
+ "epoch": 136.97,
12867
+ "learning_rate": 5.259911894273128e-06,
12868
+ "loss": 0.8568,
12869
+ "step": 20820
12870
+ },
12871
+ {
12872
+ "epoch": 137.04,
12873
+ "learning_rate": 5.233480176211454e-06,
12874
+ "loss": 0.9363,
12875
+ "step": 20830
12876
+ },
12877
+ {
12878
+ "epoch": 137.1,
12879
+ "learning_rate": 5.20704845814978e-06,
12880
+ "loss": 0.8349,
12881
+ "step": 20840
12882
+ },
12883
+ {
12884
+ "epoch": 137.17,
12885
+ "learning_rate": 5.180616740088105e-06,
12886
+ "loss": 0.8585,
12887
+ "step": 20850
12888
+ },
12889
+ {
12890
+ "epoch": 137.24,
12891
+ "learning_rate": 5.154185022026432e-06,
12892
+ "loss": 0.906,
12893
+ "step": 20860
12894
+ },
12895
+ {
12896
+ "epoch": 137.3,
12897
+ "learning_rate": 5.127753303964758e-06,
12898
+ "loss": 0.8841,
12899
+ "step": 20870
12900
+ },
12901
+ {
12902
+ "epoch": 137.37,
12903
+ "learning_rate": 5.101321585903084e-06,
12904
+ "loss": 0.8308,
12905
+ "step": 20880
12906
+ },
12907
+ {
12908
+ "epoch": 137.43,
12909
+ "learning_rate": 5.07488986784141e-06,
12910
+ "loss": 0.868,
12911
+ "step": 20890
12912
+ },
12913
+ {
12914
+ "epoch": 137.5,
12915
+ "learning_rate": 5.048458149779736e-06,
12916
+ "loss": 0.8993,
12917
+ "step": 20900
12918
+ },
12919
+ {
12920
+ "epoch": 137.56,
12921
+ "learning_rate": 5.022026431718061e-06,
12922
+ "loss": 0.8606,
12923
+ "step": 20910
12924
+ },
12925
+ {
12926
+ "epoch": 137.63,
12927
+ "learning_rate": 4.9955947136563886e-06,
12928
+ "loss": 0.8182,
12929
+ "step": 20920
12930
+ },
12931
+ {
12932
+ "epoch": 137.69,
12933
+ "learning_rate": 4.969162995594714e-06,
12934
+ "loss": 0.8791,
12935
+ "step": 20930
12936
+ },
12937
+ {
12938
+ "epoch": 137.76,
12939
+ "learning_rate": 4.9427312775330395e-06,
12940
+ "loss": 0.9186,
12941
+ "step": 20940
12942
+ },
12943
+ {
12944
+ "epoch": 137.82,
12945
+ "learning_rate": 4.916299559471366e-06,
12946
+ "loss": 0.8232,
12947
+ "step": 20950
12948
+ },
12949
+ {
12950
+ "epoch": 137.89,
12951
+ "learning_rate": 4.889867841409691e-06,
12952
+ "loss": 0.8488,
12953
+ "step": 20960
12954
+ },
12955
+ {
12956
+ "epoch": 137.96,
12957
+ "learning_rate": 4.863436123348018e-06,
12958
+ "loss": 0.8555,
12959
+ "step": 20970
12960
+ },
12961
+ {
12962
+ "epoch": 138.03,
12963
+ "learning_rate": 4.837004405286343e-06,
12964
+ "loss": 0.9995,
12965
+ "step": 20980
12966
+ },
12967
+ {
12968
+ "epoch": 138.09,
12969
+ "learning_rate": 4.81057268722467e-06,
12970
+ "loss": 0.8541,
12971
+ "step": 20990
12972
+ },
12973
+ {
12974
+ "epoch": 138.16,
12975
+ "learning_rate": 4.784140969162996e-06,
12976
+ "loss": 0.8478,
12977
+ "step": 21000
12978
+ },
12979
+ {
12980
+ "epoch": 138.16,
12981
+ "eval_loss": 0.34300604462623596,
12982
+ "eval_runtime": 316.2575,
12983
+ "eval_samples_per_second": 22.482,
12984
+ "eval_steps_per_second": 1.407,
12985
+ "eval_wer": 0.3481467144101423,
12986
+ "step": 21000
12987
+ },
12988
+ {
12989
+ "epoch": 138.22,
12990
+ "learning_rate": 4.757709251101322e-06,
12991
+ "loss": 0.8544,
12992
+ "step": 21010
12993
+ },
12994
+ {
12995
+ "epoch": 138.29,
12996
+ "learning_rate": 4.7312775330396475e-06,
12997
+ "loss": 0.8809,
12998
+ "step": 21020
12999
+ },
13000
+ {
13001
+ "epoch": 138.35,
13002
+ "learning_rate": 4.704845814977973e-06,
13003
+ "loss": 0.8406,
13004
+ "step": 21030
13005
+ },
13006
+ {
13007
+ "epoch": 138.42,
13008
+ "learning_rate": 4.678414096916299e-06,
13009
+ "loss": 0.8689,
13010
+ "step": 21040
13011
+ },
13012
+ {
13013
+ "epoch": 138.48,
13014
+ "learning_rate": 4.651982378854626e-06,
13015
+ "loss": 0.8715,
13016
+ "step": 21050
13017
+ },
13018
+ {
13019
+ "epoch": 138.55,
13020
+ "learning_rate": 4.625550660792952e-06,
13021
+ "loss": 0.8716,
13022
+ "step": 21060
13023
+ },
13024
+ {
13025
+ "epoch": 138.62,
13026
+ "learning_rate": 4.5991189427312775e-06,
13027
+ "loss": 0.8309,
13028
+ "step": 21070
13029
+ },
13030
+ {
13031
+ "epoch": 138.68,
13032
+ "learning_rate": 4.572687224669604e-06,
13033
+ "loss": 0.8757,
13034
+ "step": 21080
13035
+ },
13036
+ {
13037
+ "epoch": 138.75,
13038
+ "learning_rate": 4.546255506607929e-06,
13039
+ "loss": 0.9188,
13040
+ "step": 21090
13041
+ },
13042
+ {
13043
+ "epoch": 138.81,
13044
+ "learning_rate": 4.519823788546256e-06,
13045
+ "loss": 0.8356,
13046
+ "step": 21100
13047
+ },
13048
+ {
13049
+ "epoch": 138.88,
13050
+ "learning_rate": 4.493392070484582e-06,
13051
+ "loss": 0.8031,
13052
+ "step": 21110
13053
+ },
13054
+ {
13055
+ "epoch": 138.94,
13056
+ "learning_rate": 4.466960352422908e-06,
13057
+ "loss": 0.8519,
13058
+ "step": 21120
13059
+ },
13060
+ {
13061
+ "epoch": 139.01,
13062
+ "learning_rate": 4.440528634361234e-06,
13063
+ "loss": 0.9894,
13064
+ "step": 21130
13065
+ },
13066
+ {
13067
+ "epoch": 139.08,
13068
+ "learning_rate": 4.414096916299559e-06,
13069
+ "loss": 0.835,
13070
+ "step": 21140
13071
+ },
13072
+ {
13073
+ "epoch": 139.14,
13074
+ "learning_rate": 4.3876651982378855e-06,
13075
+ "loss": 0.8651,
13076
+ "step": 21150
13077
+ },
13078
+ {
13079
+ "epoch": 139.21,
13080
+ "learning_rate": 4.361233480176211e-06,
13081
+ "loss": 0.8675,
13082
+ "step": 21160
13083
+ },
13084
+ {
13085
+ "epoch": 139.27,
13086
+ "learning_rate": 4.334801762114538e-06,
13087
+ "loss": 0.9105,
13088
+ "step": 21170
13089
+ },
13090
+ {
13091
+ "epoch": 139.34,
13092
+ "learning_rate": 4.308370044052864e-06,
13093
+ "loss": 0.8082,
13094
+ "step": 21180
13095
+ },
13096
+ {
13097
+ "epoch": 139.41,
13098
+ "learning_rate": 4.28193832599119e-06,
13099
+ "loss": 0.8535,
13100
+ "step": 21190
13101
+ },
13102
+ {
13103
+ "epoch": 139.47,
13104
+ "learning_rate": 4.255506607929515e-06,
13105
+ "loss": 0.8652,
13106
+ "step": 21200
13107
+ },
13108
+ {
13109
+ "epoch": 139.54,
13110
+ "learning_rate": 4.229074889867842e-06,
13111
+ "loss": 0.9299,
13112
+ "step": 21210
13113
+ },
13114
+ {
13115
+ "epoch": 139.6,
13116
+ "learning_rate": 4.202643171806167e-06,
13117
+ "loss": 0.8279,
13118
+ "step": 21220
13119
+ },
13120
+ {
13121
+ "epoch": 139.67,
13122
+ "learning_rate": 4.1762114537444935e-06,
13123
+ "loss": 0.8695,
13124
+ "step": 21230
13125
+ },
13126
+ {
13127
+ "epoch": 139.73,
13128
+ "learning_rate": 4.14977973568282e-06,
13129
+ "loss": 0.9049,
13130
+ "step": 21240
13131
+ },
13132
+ {
13133
+ "epoch": 139.8,
13134
+ "learning_rate": 4.123348017621145e-06,
13135
+ "loss": 0.8715,
13136
+ "step": 21250
13137
+ },
13138
+ {
13139
+ "epoch": 139.86,
13140
+ "learning_rate": 4.096916299559472e-06,
13141
+ "loss": 0.8472,
13142
+ "step": 21260
13143
+ },
13144
+ {
13145
+ "epoch": 139.93,
13146
+ "learning_rate": 4.070484581497797e-06,
13147
+ "loss": 0.8788,
13148
+ "step": 21270
13149
+ },
13150
+ {
13151
+ "epoch": 140.0,
13152
+ "learning_rate": 4.0440528634361235e-06,
13153
+ "loss": 0.8908,
13154
+ "step": 21280
13155
+ },
13156
+ {
13157
+ "epoch": 140.07,
13158
+ "learning_rate": 4.017621145374449e-06,
13159
+ "loss": 0.9117,
13160
+ "step": 21290
13161
+ },
13162
+ {
13163
+ "epoch": 140.13,
13164
+ "learning_rate": 3.991189427312776e-06,
13165
+ "loss": 0.8214,
13166
+ "step": 21300
13167
+ },
13168
+ {
13169
+ "epoch": 140.2,
13170
+ "learning_rate": 3.964757709251102e-06,
13171
+ "loss": 0.8663,
13172
+ "step": 21310
13173
+ },
13174
+ {
13175
+ "epoch": 140.26,
13176
+ "learning_rate": 3.938325991189427e-06,
13177
+ "loss": 0.9228,
13178
+ "step": 21320
13179
+ },
13180
+ {
13181
+ "epoch": 140.33,
13182
+ "learning_rate": 3.911894273127753e-06,
13183
+ "loss": 0.8202,
13184
+ "step": 21330
13185
+ },
13186
+ {
13187
+ "epoch": 140.39,
13188
+ "learning_rate": 3.885462555066079e-06,
13189
+ "loss": 0.8349,
13190
+ "step": 21340
13191
+ },
13192
+ {
13193
+ "epoch": 140.46,
13194
+ "learning_rate": 3.859030837004405e-06,
13195
+ "loss": 0.8695,
13196
+ "step": 21350
13197
+ },
13198
+ {
13199
+ "epoch": 140.52,
13200
+ "learning_rate": 3.8325991189427315e-06,
13201
+ "loss": 0.9028,
13202
+ "step": 21360
13203
+ },
13204
+ {
13205
+ "epoch": 140.59,
13206
+ "learning_rate": 3.806167400881058e-06,
13207
+ "loss": 0.8398,
13208
+ "step": 21370
13209
+ },
13210
+ {
13211
+ "epoch": 140.65,
13212
+ "learning_rate": 3.7797356828193833e-06,
13213
+ "loss": 0.8379,
13214
+ "step": 21380
13215
+ },
13216
+ {
13217
+ "epoch": 140.72,
13218
+ "learning_rate": 3.7533039647577096e-06,
13219
+ "loss": 0.8491,
13220
+ "step": 21390
13221
+ },
13222
+ {
13223
+ "epoch": 140.79,
13224
+ "learning_rate": 3.7268722466960355e-06,
13225
+ "loss": 0.9013,
13226
+ "step": 21400
13227
+ },
13228
+ {
13229
+ "epoch": 140.85,
13230
+ "learning_rate": 3.7004405286343614e-06,
13231
+ "loss": 0.845,
13232
+ "step": 21410
13233
+ },
13234
+ {
13235
+ "epoch": 140.92,
13236
+ "learning_rate": 3.6740088105726873e-06,
13237
+ "loss": 0.8546,
13238
+ "step": 21420
13239
+ },
13240
+ {
13241
+ "epoch": 140.98,
13242
+ "learning_rate": 3.6475770925110132e-06,
13243
+ "loss": 0.8963,
13244
+ "step": 21430
13245
+ },
13246
+ {
13247
+ "epoch": 141.05,
13248
+ "learning_rate": 3.6211453744493395e-06,
13249
+ "loss": 0.9693,
13250
+ "step": 21440
13251
+ },
13252
+ {
13253
+ "epoch": 141.12,
13254
+ "learning_rate": 3.594713656387665e-06,
13255
+ "loss": 0.8501,
13256
+ "step": 21450
13257
+ },
13258
+ {
13259
+ "epoch": 141.18,
13260
+ "learning_rate": 3.568281938325991e-06,
13261
+ "loss": 0.8597,
13262
+ "step": 21460
13263
+ },
13264
+ {
13265
+ "epoch": 141.25,
13266
+ "learning_rate": 3.5418502202643172e-06,
13267
+ "loss": 0.8948,
13268
+ "step": 21470
13269
+ },
13270
+ {
13271
+ "epoch": 141.31,
13272
+ "learning_rate": 3.515418502202643e-06,
13273
+ "loss": 0.844,
13274
+ "step": 21480
13275
+ },
13276
+ {
13277
+ "epoch": 141.38,
13278
+ "learning_rate": 3.488986784140969e-06,
13279
+ "loss": 0.8161,
13280
+ "step": 21490
13281
+ },
13282
+ {
13283
+ "epoch": 141.45,
13284
+ "learning_rate": 3.4625550660792954e-06,
13285
+ "loss": 0.8643,
13286
+ "step": 21500
13287
+ },
13288
+ {
13289
+ "epoch": 141.45,
13290
+ "eval_loss": 0.3450528085231781,
13291
+ "eval_runtime": 320.1699,
13292
+ "eval_samples_per_second": 22.207,
13293
+ "eval_steps_per_second": 1.39,
13294
+ "eval_wer": 0.3484564018194135,
13295
+ "step": 21500
13296
+ },
13297
+ {
13298
+ "epoch": 141.51,
13299
+ "learning_rate": 3.4361233480176213e-06,
13300
+ "loss": 0.9,
13301
+ "step": 21510
13302
+ },
13303
+ {
13304
+ "epoch": 141.58,
13305
+ "learning_rate": 3.409691629955947e-06,
13306
+ "loss": 0.8424,
13307
+ "step": 21520
13308
+ },
13309
+ {
13310
+ "epoch": 141.64,
13311
+ "learning_rate": 3.3832599118942735e-06,
13312
+ "loss": 0.8523,
13313
+ "step": 21530
13314
+ },
13315
+ {
13316
+ "epoch": 141.71,
13317
+ "learning_rate": 3.3568281938325994e-06,
13318
+ "loss": 0.8772,
13319
+ "step": 21540
13320
+ },
13321
+ {
13322
+ "epoch": 141.77,
13323
+ "learning_rate": 3.330396475770925e-06,
13324
+ "loss": 0.9149,
13325
+ "step": 21550
13326
+ },
13327
+ {
13328
+ "epoch": 141.84,
13329
+ "learning_rate": 3.303964757709251e-06,
13330
+ "loss": 0.8252,
13331
+ "step": 21560
13332
+ },
13333
+ {
13334
+ "epoch": 141.9,
13335
+ "learning_rate": 3.277533039647577e-06,
13336
+ "loss": 0.8308,
13337
+ "step": 21570
13338
+ },
13339
+ {
13340
+ "epoch": 141.97,
13341
+ "learning_rate": 3.251101321585903e-06,
13342
+ "loss": 0.8585,
13343
+ "step": 21580
13344
+ },
13345
+ {
13346
+ "epoch": 142.04,
13347
+ "learning_rate": 3.2246696035242293e-06,
13348
+ "loss": 0.9729,
13349
+ "step": 21590
13350
+ },
13351
+ {
13352
+ "epoch": 142.1,
13353
+ "learning_rate": 3.198237885462555e-06,
13354
+ "loss": 0.8169,
13355
+ "step": 21600
13356
+ },
13357
+ {
13358
+ "epoch": 142.17,
13359
+ "learning_rate": 3.171806167400881e-06,
13360
+ "loss": 0.8536,
13361
+ "step": 21610
13362
+ },
13363
+ {
13364
+ "epoch": 142.24,
13365
+ "learning_rate": 3.1453744493392074e-06,
13366
+ "loss": 0.8764,
13367
+ "step": 21620
13368
+ },
13369
+ {
13370
+ "epoch": 142.3,
13371
+ "learning_rate": 3.1189427312775333e-06,
13372
+ "loss": 0.8731,
13373
+ "step": 21630
13374
+ },
13375
+ {
13376
+ "epoch": 142.37,
13377
+ "learning_rate": 3.092511013215859e-06,
13378
+ "loss": 0.8372,
13379
+ "step": 21640
13380
+ },
13381
+ {
13382
+ "epoch": 142.43,
13383
+ "learning_rate": 3.066079295154185e-06,
13384
+ "loss": 0.8622,
13385
+ "step": 21650
13386
+ },
13387
+ {
13388
+ "epoch": 142.5,
13389
+ "learning_rate": 3.039647577092511e-06,
13390
+ "loss": 0.8759,
13391
+ "step": 21660
13392
+ },
13393
+ {
13394
+ "epoch": 142.56,
13395
+ "learning_rate": 3.013215859030837e-06,
13396
+ "loss": 0.8428,
13397
+ "step": 21670
13398
+ },
13399
+ {
13400
+ "epoch": 142.63,
13401
+ "learning_rate": 2.9867841409691632e-06,
13402
+ "loss": 0.849,
13403
+ "step": 21680
13404
+ },
13405
+ {
13406
+ "epoch": 142.69,
13407
+ "learning_rate": 2.960352422907489e-06,
13408
+ "loss": 0.8815,
13409
+ "step": 21690
13410
+ },
13411
+ {
13412
+ "epoch": 142.76,
13413
+ "learning_rate": 2.933920704845815e-06,
13414
+ "loss": 0.9493,
13415
+ "step": 21700
13416
+ },
13417
+ {
13418
+ "epoch": 142.82,
13419
+ "learning_rate": 2.9074889867841414e-06,
13420
+ "loss": 0.814,
13421
+ "step": 21710
13422
+ },
13423
+ {
13424
+ "epoch": 142.89,
13425
+ "learning_rate": 2.8810572687224673e-06,
13426
+ "loss": 0.8421,
13427
+ "step": 21720
13428
+ },
13429
+ {
13430
+ "epoch": 142.96,
13431
+ "learning_rate": 2.854625550660793e-06,
13432
+ "loss": 0.853,
13433
+ "step": 21730
13434
+ },
13435
+ {
13436
+ "epoch": 143.03,
13437
+ "learning_rate": 2.828193832599119e-06,
13438
+ "loss": 1.0032,
13439
+ "step": 21740
13440
+ },
13441
+ {
13442
+ "epoch": 143.09,
13443
+ "learning_rate": 2.801762114537445e-06,
13444
+ "loss": 0.826,
13445
+ "step": 21750
13446
+ },
13447
+ {
13448
+ "epoch": 143.16,
13449
+ "learning_rate": 2.775330396475771e-06,
13450
+ "loss": 0.8516,
13451
+ "step": 21760
13452
+ },
13453
+ {
13454
+ "epoch": 143.22,
13455
+ "learning_rate": 2.7488986784140968e-06,
13456
+ "loss": 0.8449,
13457
+ "step": 21770
13458
+ },
13459
+ {
13460
+ "epoch": 143.29,
13461
+ "learning_rate": 2.722466960352423e-06,
13462
+ "loss": 0.9136,
13463
+ "step": 21780
13464
+ },
13465
+ {
13466
+ "epoch": 143.35,
13467
+ "learning_rate": 2.696035242290749e-06,
13468
+ "loss": 0.8496,
13469
+ "step": 21790
13470
+ },
13471
+ {
13472
+ "epoch": 143.42,
13473
+ "learning_rate": 2.669603524229075e-06,
13474
+ "loss": 0.8406,
13475
+ "step": 21800
13476
+ },
13477
+ {
13478
+ "epoch": 143.48,
13479
+ "learning_rate": 2.643171806167401e-06,
13480
+ "loss": 0.8884,
13481
+ "step": 21810
13482
+ },
13483
+ {
13484
+ "epoch": 143.55,
13485
+ "learning_rate": 2.616740088105727e-06,
13486
+ "loss": 0.8791,
13487
+ "step": 21820
13488
+ },
13489
+ {
13490
+ "epoch": 143.62,
13491
+ "learning_rate": 2.5903083700440526e-06,
13492
+ "loss": 0.8221,
13493
+ "step": 21830
13494
+ },
13495
+ {
13496
+ "epoch": 143.68,
13497
+ "learning_rate": 2.563876651982379e-06,
13498
+ "loss": 0.8445,
13499
+ "step": 21840
13500
+ },
13501
+ {
13502
+ "epoch": 143.75,
13503
+ "learning_rate": 2.537444933920705e-06,
13504
+ "loss": 0.9059,
13505
+ "step": 21850
13506
+ },
13507
+ {
13508
+ "epoch": 143.81,
13509
+ "learning_rate": 2.5110132158590307e-06,
13510
+ "loss": 0.8484,
13511
+ "step": 21860
13512
+ },
13513
+ {
13514
+ "epoch": 143.88,
13515
+ "learning_rate": 2.484581497797357e-06,
13516
+ "loss": 0.8594,
13517
+ "step": 21870
13518
+ },
13519
+ {
13520
+ "epoch": 143.94,
13521
+ "learning_rate": 2.458149779735683e-06,
13522
+ "loss": 0.8631,
13523
+ "step": 21880
13524
+ },
13525
+ {
13526
+ "epoch": 144.01,
13527
+ "learning_rate": 2.431718061674009e-06,
13528
+ "loss": 0.9621,
13529
+ "step": 21890
13530
+ },
13531
+ {
13532
+ "epoch": 144.08,
13533
+ "learning_rate": 2.405286343612335e-06,
13534
+ "loss": 0.8186,
13535
+ "step": 21900
13536
+ },
13537
+ {
13538
+ "epoch": 144.14,
13539
+ "learning_rate": 2.378854625550661e-06,
13540
+ "loss": 0.8318,
13541
+ "step": 21910
13542
+ },
13543
+ {
13544
+ "epoch": 144.21,
13545
+ "learning_rate": 2.3524229074889865e-06,
13546
+ "loss": 0.8535,
13547
+ "step": 21920
13548
+ },
13549
+ {
13550
+ "epoch": 144.27,
13551
+ "learning_rate": 2.325991189427313e-06,
13552
+ "loss": 0.9028,
13553
+ "step": 21930
13554
+ },
13555
+ {
13556
+ "epoch": 144.34,
13557
+ "learning_rate": 2.2995594713656387e-06,
13558
+ "loss": 0.8314,
13559
+ "step": 21940
13560
+ },
13561
+ {
13562
+ "epoch": 144.41,
13563
+ "learning_rate": 2.2731277533039646e-06,
13564
+ "loss": 0.8417,
13565
+ "step": 21950
13566
+ },
13567
+ {
13568
+ "epoch": 144.47,
13569
+ "learning_rate": 2.249339207048458e-06,
13570
+ "loss": 0.849,
13571
+ "step": 21960
13572
+ },
13573
+ {
13574
+ "epoch": 144.54,
13575
+ "learning_rate": 2.222907488986784e-06,
13576
+ "loss": 0.9065,
13577
+ "step": 21970
13578
+ },
13579
+ {
13580
+ "epoch": 144.6,
13581
+ "learning_rate": 2.1964757709251104e-06,
13582
+ "loss": 0.8296,
13583
+ "step": 21980
13584
+ },
13585
+ {
13586
+ "epoch": 144.67,
13587
+ "learning_rate": 2.1700440528634363e-06,
13588
+ "loss": 0.8638,
13589
+ "step": 21990
13590
+ },
13591
+ {
13592
+ "epoch": 144.73,
13593
+ "learning_rate": 2.143612334801762e-06,
13594
+ "loss": 0.8705,
13595
+ "step": 22000
13596
+ },
13597
+ {
13598
+ "epoch": 144.73,
13599
+ "eval_loss": 0.34444093704223633,
13600
+ "eval_runtime": 318.8078,
13601
+ "eval_samples_per_second": 22.302,
13602
+ "eval_steps_per_second": 1.396,
13603
+ "eval_wer": 0.34743056227620245,
13604
+ "step": 22000
13605
+ },
13606
+ {
13607
+ "epoch": 144.8,
13608
+ "learning_rate": 2.117180616740088e-06,
13609
+ "loss": 0.8493,
13610
+ "step": 22010
13611
+ },
13612
+ {
13613
+ "epoch": 144.86,
13614
+ "learning_rate": 2.090748898678414e-06,
13615
+ "loss": 0.8297,
13616
+ "step": 22020
13617
+ },
13618
+ {
13619
+ "epoch": 144.93,
13620
+ "learning_rate": 2.06431718061674e-06,
13621
+ "loss": 0.8733,
13622
+ "step": 22030
13623
+ },
13624
+ {
13625
+ "epoch": 145.0,
13626
+ "learning_rate": 2.037885462555066e-06,
13627
+ "loss": 0.8968,
13628
+ "step": 22040
13629
+ },
13630
+ {
13631
+ "epoch": 145.07,
13632
+ "learning_rate": 2.011453744493392e-06,
13633
+ "loss": 0.9201,
13634
+ "step": 22050
13635
+ },
13636
+ {
13637
+ "epoch": 145.13,
13638
+ "learning_rate": 1.985022026431718e-06,
13639
+ "loss": 0.8446,
13640
+ "step": 22060
13641
+ },
13642
+ {
13643
+ "epoch": 145.2,
13644
+ "learning_rate": 1.9585903083700443e-06,
13645
+ "loss": 0.849,
13646
+ "step": 22070
13647
+ },
13648
+ {
13649
+ "epoch": 145.26,
13650
+ "learning_rate": 1.93215859030837e-06,
13651
+ "loss": 0.91,
13652
+ "step": 22080
13653
+ },
13654
+ {
13655
+ "epoch": 145.33,
13656
+ "learning_rate": 1.9057268722466963e-06,
13657
+ "loss": 0.8113,
13658
+ "step": 22090
13659
+ },
13660
+ {
13661
+ "epoch": 145.39,
13662
+ "learning_rate": 1.8792951541850218e-06,
13663
+ "loss": 0.8432,
13664
+ "step": 22100
13665
+ },
13666
+ {
13667
+ "epoch": 145.46,
13668
+ "learning_rate": 1.8528634361233481e-06,
13669
+ "loss": 0.8635,
13670
+ "step": 22110
13671
+ },
13672
+ {
13673
+ "epoch": 145.52,
13674
+ "learning_rate": 1.8264317180616742e-06,
13675
+ "loss": 0.8872,
13676
+ "step": 22120
13677
+ },
13678
+ {
13679
+ "epoch": 145.59,
13680
+ "learning_rate": 1.8e-06,
13681
+ "loss": 0.8099,
13682
+ "step": 22130
13683
+ },
13684
+ {
13685
+ "epoch": 145.65,
13686
+ "learning_rate": 1.773568281938326e-06,
13687
+ "loss": 0.862,
13688
+ "step": 22140
13689
+ },
13690
+ {
13691
+ "epoch": 145.72,
13692
+ "learning_rate": 1.7471365638766521e-06,
13693
+ "loss": 0.8395,
13694
+ "step": 22150
13695
+ },
13696
+ {
13697
+ "epoch": 145.79,
13698
+ "learning_rate": 1.720704845814978e-06,
13699
+ "loss": 0.9084,
13700
+ "step": 22160
13701
+ },
13702
+ {
13703
+ "epoch": 145.85,
13704
+ "learning_rate": 1.694273127753304e-06,
13705
+ "loss": 0.8402,
13706
+ "step": 22170
13707
+ },
13708
+ {
13709
+ "epoch": 145.92,
13710
+ "learning_rate": 1.66784140969163e-06,
13711
+ "loss": 0.8627,
13712
+ "step": 22180
13713
+ },
13714
+ {
13715
+ "epoch": 145.98,
13716
+ "learning_rate": 1.641409691629956e-06,
13717
+ "loss": 0.8402,
13718
+ "step": 22190
13719
+ },
13720
+ {
13721
+ "epoch": 146.05,
13722
+ "learning_rate": 1.614977973568282e-06,
13723
+ "loss": 0.9459,
13724
+ "step": 22200
13725
+ },
13726
+ {
13727
+ "epoch": 146.12,
13728
+ "learning_rate": 1.588546255506608e-06,
13729
+ "loss": 0.8286,
13730
+ "step": 22210
13731
+ },
13732
+ {
13733
+ "epoch": 146.18,
13734
+ "learning_rate": 1.5621145374449339e-06,
13735
+ "loss": 0.8557,
13736
+ "step": 22220
13737
+ },
13738
+ {
13739
+ "epoch": 146.25,
13740
+ "learning_rate": 1.53568281938326e-06,
13741
+ "loss": 0.8967,
13742
+ "step": 22230
13743
+ },
13744
+ {
13745
+ "epoch": 146.31,
13746
+ "learning_rate": 1.5092511013215859e-06,
13747
+ "loss": 0.8315,
13748
+ "step": 22240
13749
+ },
13750
+ {
13751
+ "epoch": 146.38,
13752
+ "learning_rate": 1.482819383259912e-06,
13753
+ "loss": 0.8259,
13754
+ "step": 22250
13755
+ },
13756
+ {
13757
+ "epoch": 146.45,
13758
+ "learning_rate": 1.456387665198238e-06,
13759
+ "loss": 0.8549,
13760
+ "step": 22260
13761
+ },
13762
+ {
13763
+ "epoch": 146.51,
13764
+ "learning_rate": 1.4299559471365638e-06,
13765
+ "loss": 0.8958,
13766
+ "step": 22270
13767
+ },
13768
+ {
13769
+ "epoch": 146.58,
13770
+ "learning_rate": 1.4035242290748899e-06,
13771
+ "loss": 0.8303,
13772
+ "step": 22280
13773
+ },
13774
+ {
13775
+ "epoch": 146.64,
13776
+ "learning_rate": 1.377092511013216e-06,
13777
+ "loss": 0.8444,
13778
+ "step": 22290
13779
+ },
13780
+ {
13781
+ "epoch": 146.71,
13782
+ "learning_rate": 1.350660792951542e-06,
13783
+ "loss": 0.8439,
13784
+ "step": 22300
13785
+ },
13786
+ {
13787
+ "epoch": 146.77,
13788
+ "learning_rate": 1.3242290748898678e-06,
13789
+ "loss": 0.9093,
13790
+ "step": 22310
13791
+ },
13792
+ {
13793
+ "epoch": 146.84,
13794
+ "learning_rate": 1.297797356828194e-06,
13795
+ "loss": 0.8584,
13796
+ "step": 22320
13797
+ },
13798
+ {
13799
+ "epoch": 146.9,
13800
+ "learning_rate": 1.2713656387665198e-06,
13801
+ "loss": 0.8541,
13802
+ "step": 22330
13803
+ },
13804
+ {
13805
+ "epoch": 146.97,
13806
+ "learning_rate": 1.244933920704846e-06,
13807
+ "loss": 0.8791,
13808
+ "step": 22340
13809
+ },
13810
+ {
13811
+ "epoch": 147.04,
13812
+ "learning_rate": 1.2185022026431718e-06,
13813
+ "loss": 0.9583,
13814
+ "step": 22350
13815
+ },
13816
+ {
13817
+ "epoch": 147.1,
13818
+ "learning_rate": 1.1920704845814977e-06,
13819
+ "loss": 0.7998,
13820
+ "step": 22360
13821
+ },
13822
+ {
13823
+ "epoch": 147.17,
13824
+ "learning_rate": 1.1656387665198238e-06,
13825
+ "loss": 0.8725,
13826
+ "step": 22370
13827
+ },
13828
+ {
13829
+ "epoch": 147.24,
13830
+ "learning_rate": 1.1392070484581497e-06,
13831
+ "loss": 0.8524,
13832
+ "step": 22380
13833
+ },
13834
+ {
13835
+ "epoch": 147.3,
13836
+ "learning_rate": 1.1127753303964758e-06,
13837
+ "loss": 0.8648,
13838
+ "step": 22390
13839
+ },
13840
+ {
13841
+ "epoch": 147.37,
13842
+ "learning_rate": 1.086343612334802e-06,
13843
+ "loss": 0.8385,
13844
+ "step": 22400
13845
+ },
13846
+ {
13847
+ "epoch": 147.43,
13848
+ "learning_rate": 1.0599118942731276e-06,
13849
+ "loss": 0.8608,
13850
+ "step": 22410
13851
+ },
13852
+ {
13853
+ "epoch": 147.5,
13854
+ "learning_rate": 1.0334801762114537e-06,
13855
+ "loss": 0.8992,
13856
+ "step": 22420
13857
+ },
13858
+ {
13859
+ "epoch": 147.56,
13860
+ "learning_rate": 1.0070484581497799e-06,
13861
+ "loss": 0.8481,
13862
+ "step": 22430
13863
+ },
13864
+ {
13865
+ "epoch": 147.63,
13866
+ "learning_rate": 9.806167400881058e-07,
13867
+ "loss": 0.8521,
13868
+ "step": 22440
13869
+ },
13870
+ {
13871
+ "epoch": 147.69,
13872
+ "learning_rate": 9.541850220264317e-07,
13873
+ "loss": 0.8477,
13874
+ "step": 22450
13875
+ },
13876
+ {
13877
+ "epoch": 147.76,
13878
+ "learning_rate": 9.277533039647578e-07,
13879
+ "loss": 0.9214,
13880
+ "step": 22460
13881
+ },
13882
+ {
13883
+ "epoch": 147.82,
13884
+ "learning_rate": 9.013215859030837e-07,
13885
+ "loss": 0.8459,
13886
+ "step": 22470
13887
+ },
13888
+ {
13889
+ "epoch": 147.89,
13890
+ "learning_rate": 8.748898678414098e-07,
13891
+ "loss": 0.8296,
13892
+ "step": 22480
13893
+ },
13894
+ {
13895
+ "epoch": 147.96,
13896
+ "learning_rate": 8.484581497797357e-07,
13897
+ "loss": 0.8857,
13898
+ "step": 22490
13899
+ },
13900
+ {
13901
+ "epoch": 148.03,
13902
+ "learning_rate": 8.220264317180617e-07,
13903
+ "loss": 0.9869,
13904
+ "step": 22500
13905
+ },
13906
+ {
13907
+ "epoch": 148.03,
13908
+ "eval_loss": 0.3441205620765686,
13909
+ "eval_runtime": 320.0073,
13910
+ "eval_samples_per_second": 22.218,
13911
+ "eval_steps_per_second": 1.391,
13912
+ "eval_wer": 0.3493080421949095,
13913
+ "step": 22500
13914
+ },
13915
+ {
13916
+ "epoch": 148.09,
13917
+ "learning_rate": 7.955947136563877e-07,
13918
+ "loss": 0.8326,
13919
+ "step": 22510
13920
+ },
13921
+ {
13922
+ "epoch": 148.16,
13923
+ "learning_rate": 7.691629955947137e-07,
13924
+ "loss": 0.8538,
13925
+ "step": 22520
13926
+ },
13927
+ {
13928
+ "epoch": 148.22,
13929
+ "learning_rate": 7.427312775330397e-07,
13930
+ "loss": 0.8947,
13931
+ "step": 22530
13932
+ },
13933
+ {
13934
+ "epoch": 148.29,
13935
+ "learning_rate": 7.162995594713656e-07,
13936
+ "loss": 0.9051,
13937
+ "step": 22540
13938
+ },
13939
+ {
13940
+ "epoch": 148.35,
13941
+ "learning_rate": 6.898678414096917e-07,
13942
+ "loss": 0.8197,
13943
+ "step": 22550
13944
+ },
13945
+ {
13946
+ "epoch": 148.42,
13947
+ "learning_rate": 6.634361233480176e-07,
13948
+ "loss": 0.8446,
13949
+ "step": 22560
13950
+ },
13951
+ {
13952
+ "epoch": 148.48,
13953
+ "learning_rate": 6.370044052863436e-07,
13954
+ "loss": 0.9035,
13955
+ "step": 22570
13956
+ },
13957
+ {
13958
+ "epoch": 148.55,
13959
+ "learning_rate": 6.105726872246696e-07,
13960
+ "loss": 0.861,
13961
+ "step": 22580
13962
+ },
13963
+ {
13964
+ "epoch": 148.62,
13965
+ "learning_rate": 5.841409691629956e-07,
13966
+ "loss": 0.8269,
13967
+ "step": 22590
13968
+ },
13969
+ {
13970
+ "epoch": 148.68,
13971
+ "learning_rate": 5.577092511013216e-07,
13972
+ "loss": 0.8597,
13973
+ "step": 22600
13974
+ },
13975
+ {
13976
+ "epoch": 148.75,
13977
+ "learning_rate": 5.312775330396475e-07,
13978
+ "loss": 0.8928,
13979
+ "step": 22610
13980
+ },
13981
+ {
13982
+ "epoch": 148.81,
13983
+ "learning_rate": 5.048458149779736e-07,
13984
+ "loss": 0.8236,
13985
+ "step": 22620
13986
+ },
13987
+ {
13988
+ "epoch": 148.88,
13989
+ "learning_rate": 4.784140969162995e-07,
13990
+ "loss": 0.8326,
13991
+ "step": 22630
13992
+ },
13993
+ {
13994
+ "epoch": 148.94,
13995
+ "learning_rate": 4.519823788546256e-07,
13996
+ "loss": 0.858,
13997
+ "step": 22640
13998
+ },
13999
+ {
14000
+ "epoch": 149.01,
14001
+ "learning_rate": 4.2555066079295154e-07,
14002
+ "loss": 0.9665,
14003
+ "step": 22650
14004
+ },
14005
+ {
14006
+ "epoch": 149.08,
14007
+ "learning_rate": 3.9911894273127755e-07,
14008
+ "loss": 0.8098,
14009
+ "step": 22660
14010
+ },
14011
+ {
14012
+ "epoch": 149.14,
14013
+ "learning_rate": 3.726872246696035e-07,
14014
+ "loss": 0.8435,
14015
+ "step": 22670
14016
+ },
14017
+ {
14018
+ "epoch": 149.21,
14019
+ "learning_rate": 3.462555066079295e-07,
14020
+ "loss": 0.8414,
14021
+ "step": 22680
14022
+ },
14023
+ {
14024
+ "epoch": 149.27,
14025
+ "learning_rate": 3.1982378854625556e-07,
14026
+ "loss": 0.9122,
14027
+ "step": 22690
14028
+ },
14029
+ {
14030
+ "epoch": 149.34,
14031
+ "learning_rate": 2.933920704845815e-07,
14032
+ "loss": 0.8136,
14033
+ "step": 22700
14034
+ },
14035
+ {
14036
+ "epoch": 149.41,
14037
+ "learning_rate": 2.669603524229075e-07,
14038
+ "loss": 0.856,
14039
+ "step": 22710
14040
+ },
14041
+ {
14042
+ "epoch": 149.47,
14043
+ "learning_rate": 2.4052863436123347e-07,
14044
+ "loss": 0.8759,
14045
+ "step": 22720
14046
+ },
14047
+ {
14048
+ "epoch": 149.54,
14049
+ "learning_rate": 2.1409691629955948e-07,
14050
+ "loss": 0.884,
14051
+ "step": 22730
14052
+ },
14053
+ {
14054
+ "epoch": 149.6,
14055
+ "learning_rate": 1.8766519823788548e-07,
14056
+ "loss": 0.8404,
14057
+ "step": 22740
14058
+ },
14059
+ {
14060
+ "epoch": 149.67,
14061
+ "learning_rate": 1.6123348017621146e-07,
14062
+ "loss": 0.8286,
14063
+ "step": 22750
14064
+ },
14065
+ {
14066
+ "epoch": 149.73,
14067
+ "learning_rate": 1.3480176211453744e-07,
14068
+ "loss": 0.8571,
14069
+ "step": 22760
14070
+ },
14071
+ {
14072
+ "epoch": 149.8,
14073
+ "learning_rate": 1.0837004405286344e-07,
14074
+ "loss": 0.8685,
14075
+ "step": 22770
14076
+ },
14077
+ {
14078
+ "epoch": 149.86,
14079
+ "learning_rate": 8.193832599118942e-08,
14080
+ "loss": 0.8315,
14081
+ "step": 22780
14082
+ },
14083
+ {
14084
+ "epoch": 149.93,
14085
+ "learning_rate": 5.550660792951542e-08,
14086
+ "loss": 0.8556,
14087
+ "step": 22790
14088
+ },
14089
+ {
14090
+ "epoch": 150.0,
14091
+ "learning_rate": 2.907488986784141e-08,
14092
+ "loss": 0.8774,
14093
+ "step": 22800
14094
+ },
14095
+ {
14096
+ "epoch": 150.0,
14097
+ "step": 22800,
14098
+ "total_flos": 3.150278594946483e+20,
14099
+ "train_loss": 0.16461168266179269,
14100
+ "train_runtime": 24582.5742,
14101
+ "train_samples_per_second": 119.176,
14102
+ "train_steps_per_second": 0.927
14103
  }
14104
  ],
14105
+ "max_steps": 22800,
14106
+ "num_train_epochs": 150,
14107
+ "total_flos": 3.150278594946483e+20,
14108
  "trial_name": null,
14109
  "trial_params": null
14110
  }