Commit
·
a174c37
1
Parent(s):
b62347a
End of training
Browse files- all_results.json +17 -0
- eval_results.json +11 -0
- train_results.json +9 -0
- trainer_state.json +731 -0
all_results.json
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 44.0,
|
| 3 |
+
"eval_cer": 0.13281812509000143,
|
| 4 |
+
"eval_loss": 0.3162442147731781,
|
| 5 |
+
"eval_per": 0.22600304723209752,
|
| 6 |
+
"eval_runtime": 30.1741,
|
| 7 |
+
"eval_samples": 383,
|
| 8 |
+
"eval_samples_per_second": 12.693,
|
| 9 |
+
"eval_steps_per_second": 1.591,
|
| 10 |
+
"eval_wer": 0.22981208735398678,
|
| 11 |
+
"total_flos": 3.1481023081144828e+19,
|
| 12 |
+
"train_loss": 1.0956278002107298,
|
| 13 |
+
"train_runtime": 26844.9073,
|
| 14 |
+
"train_samples": 2688,
|
| 15 |
+
"train_samples_per_second": 20.026,
|
| 16 |
+
"train_steps_per_second": 0.626
|
| 17 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,11 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 44.0,
|
| 3 |
+
"eval_cer": 0.13281812509000143,
|
| 4 |
+
"eval_loss": 0.3162442147731781,
|
| 5 |
+
"eval_per": 0.22600304723209752,
|
| 6 |
+
"eval_runtime": 30.1741,
|
| 7 |
+
"eval_samples": 383,
|
| 8 |
+
"eval_samples_per_second": 12.693,
|
| 9 |
+
"eval_steps_per_second": 1.591,
|
| 10 |
+
"eval_wer": 0.22981208735398678
|
| 11 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 44.0,
|
| 3 |
+
"total_flos": 3.1481023081144828e+19,
|
| 4 |
+
"train_loss": 1.0956278002107298,
|
| 5 |
+
"train_runtime": 26844.9073,
|
| 6 |
+
"train_samples": 2688,
|
| 7 |
+
"train_samples_per_second": 20.026,
|
| 8 |
+
"train_steps_per_second": 0.626
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,731 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.3162442147731781,
|
| 3 |
+
"best_model_checkpoint": "wav2vec2-large-xlsr-coraa-texts-aug-exp-1/checkpoint-2016",
|
| 4 |
+
"epoch": 44.0,
|
| 5 |
+
"global_step": 3696,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.01,
|
| 12 |
+
"learning_rate": 3e-05,
|
| 13 |
+
"loss": 43.995,
|
| 14 |
+
"step": 1
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 1.0,
|
| 18 |
+
"eval_cer": 0.9670714731435703,
|
| 19 |
+
"eval_loss": 3.3726680278778076,
|
| 20 |
+
"eval_per": 1.0,
|
| 21 |
+
"eval_runtime": 29.95,
|
| 22 |
+
"eval_samples_per_second": 12.788,
|
| 23 |
+
"eval_steps_per_second": 1.603,
|
| 24 |
+
"eval_wer": 1.0,
|
| 25 |
+
"step": 84
|
| 26 |
+
},
|
| 27 |
+
{
|
| 28 |
+
"epoch": 1.19,
|
| 29 |
+
"learning_rate": 2.9826785714285713e-05,
|
| 30 |
+
"loss": 9.1824,
|
| 31 |
+
"step": 100
|
| 32 |
+
},
|
| 33 |
+
{
|
| 34 |
+
"epoch": 2.0,
|
| 35 |
+
"eval_cer": 0.9670714731435703,
|
| 36 |
+
"eval_loss": 2.992058753967285,
|
| 37 |
+
"eval_per": 1.0,
|
| 38 |
+
"eval_runtime": 30.0041,
|
| 39 |
+
"eval_samples_per_second": 12.765,
|
| 40 |
+
"eval_steps_per_second": 1.6,
|
| 41 |
+
"eval_wer": 1.0,
|
| 42 |
+
"step": 168
|
| 43 |
+
},
|
| 44 |
+
{
|
| 45 |
+
"epoch": 2.38,
|
| 46 |
+
"learning_rate": 2.9648214285714286e-05,
|
| 47 |
+
"loss": 3.1225,
|
| 48 |
+
"step": 200
|
| 49 |
+
},
|
| 50 |
+
{
|
| 51 |
+
"epoch": 3.0,
|
| 52 |
+
"eval_cer": 0.9670714731435703,
|
| 53 |
+
"eval_loss": 2.932141065597534,
|
| 54 |
+
"eval_per": 1.0,
|
| 55 |
+
"eval_runtime": 29.8416,
|
| 56 |
+
"eval_samples_per_second": 12.834,
|
| 57 |
+
"eval_steps_per_second": 1.608,
|
| 58 |
+
"eval_wer": 1.0,
|
| 59 |
+
"step": 252
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"epoch": 3.57,
|
| 63 |
+
"learning_rate": 2.9469642857142856e-05,
|
| 64 |
+
"loss": 2.979,
|
| 65 |
+
"step": 300
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 4.0,
|
| 69 |
+
"eval_cer": 0.9670714731435703,
|
| 70 |
+
"eval_loss": 2.9393842220306396,
|
| 71 |
+
"eval_per": 1.0,
|
| 72 |
+
"eval_runtime": 29.8698,
|
| 73 |
+
"eval_samples_per_second": 12.822,
|
| 74 |
+
"eval_steps_per_second": 1.607,
|
| 75 |
+
"eval_wer": 1.0,
|
| 76 |
+
"step": 336
|
| 77 |
+
},
|
| 78 |
+
{
|
| 79 |
+
"epoch": 4.76,
|
| 80 |
+
"learning_rate": 2.929107142857143e-05,
|
| 81 |
+
"loss": 2.9357,
|
| 82 |
+
"step": 400
|
| 83 |
+
},
|
| 84 |
+
{
|
| 85 |
+
"epoch": 5.0,
|
| 86 |
+
"eval_cer": 0.9670714731435703,
|
| 87 |
+
"eval_loss": 2.911760091781616,
|
| 88 |
+
"eval_per": 1.0,
|
| 89 |
+
"eval_runtime": 29.9589,
|
| 90 |
+
"eval_samples_per_second": 12.784,
|
| 91 |
+
"eval_steps_per_second": 1.602,
|
| 92 |
+
"eval_wer": 1.0,
|
| 93 |
+
"step": 420
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 5.95,
|
| 97 |
+
"learning_rate": 2.9112500000000003e-05,
|
| 98 |
+
"loss": 2.9116,
|
| 99 |
+
"step": 500
|
| 100 |
+
},
|
| 101 |
+
{
|
| 102 |
+
"epoch": 6.0,
|
| 103 |
+
"eval_cer": 0.9670714731435703,
|
| 104 |
+
"eval_loss": 2.8400661945343018,
|
| 105 |
+
"eval_per": 1.0,
|
| 106 |
+
"eval_runtime": 29.8579,
|
| 107 |
+
"eval_samples_per_second": 12.827,
|
| 108 |
+
"eval_steps_per_second": 1.608,
|
| 109 |
+
"eval_wer": 1.0,
|
| 110 |
+
"step": 504
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 7.0,
|
| 114 |
+
"eval_cer": 0.7347957567321077,
|
| 115 |
+
"eval_loss": 2.0976722240448,
|
| 116 |
+
"eval_per": 1.0,
|
| 117 |
+
"eval_runtime": 29.9484,
|
| 118 |
+
"eval_samples_per_second": 12.789,
|
| 119 |
+
"eval_steps_per_second": 1.603,
|
| 120 |
+
"eval_wer": 1.0,
|
| 121 |
+
"step": 588
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 7.14,
|
| 125 |
+
"learning_rate": 2.893392857142857e-05,
|
| 126 |
+
"loss": 2.6841,
|
| 127 |
+
"step": 600
|
| 128 |
+
},
|
| 129 |
+
{
|
| 130 |
+
"epoch": 8.0,
|
| 131 |
+
"eval_cer": 0.20928334853357652,
|
| 132 |
+
"eval_loss": 0.8209644556045532,
|
| 133 |
+
"eval_per": 0.5335195530726257,
|
| 134 |
+
"eval_runtime": 30.0806,
|
| 135 |
+
"eval_samples_per_second": 12.732,
|
| 136 |
+
"eval_steps_per_second": 1.596,
|
| 137 |
+
"eval_wer": 0.5520568816658202,
|
| 138 |
+
"step": 672
|
| 139 |
+
},
|
| 140 |
+
{
|
| 141 |
+
"epoch": 8.33,
|
| 142 |
+
"learning_rate": 2.8755357142857143e-05,
|
| 143 |
+
"loss": 1.6358,
|
| 144 |
+
"step": 700
|
| 145 |
+
},
|
| 146 |
+
{
|
| 147 |
+
"epoch": 9.0,
|
| 148 |
+
"eval_cer": 0.16325061200979216,
|
| 149 |
+
"eval_loss": 0.5585940480232239,
|
| 150 |
+
"eval_per": 0.3080243778567801,
|
| 151 |
+
"eval_runtime": 30.0617,
|
| 152 |
+
"eval_samples_per_second": 12.74,
|
| 153 |
+
"eval_steps_per_second": 1.597,
|
| 154 |
+
"eval_wer": 0.3329101066531234,
|
| 155 |
+
"step": 756
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"epoch": 9.52,
|
| 159 |
+
"learning_rate": 2.8576785714285716e-05,
|
| 160 |
+
"loss": 1.1226,
|
| 161 |
+
"step": 800
|
| 162 |
+
},
|
| 163 |
+
{
|
| 164 |
+
"epoch": 10.0,
|
| 165 |
+
"eval_cer": 0.1537464599433591,
|
| 166 |
+
"eval_loss": 0.48173972964286804,
|
| 167 |
+
"eval_per": 0.28339258506856274,
|
| 168 |
+
"eval_runtime": 30.0584,
|
| 169 |
+
"eval_samples_per_second": 12.742,
|
| 170 |
+
"eval_steps_per_second": 1.597,
|
| 171 |
+
"eval_wer": 0.29507364144235654,
|
| 172 |
+
"step": 840
|
| 173 |
+
},
|
| 174 |
+
{
|
| 175 |
+
"epoch": 10.71,
|
| 176 |
+
"learning_rate": 2.8398214285714286e-05,
|
| 177 |
+
"loss": 0.9403,
|
| 178 |
+
"step": 900
|
| 179 |
+
},
|
| 180 |
+
{
|
| 181 |
+
"epoch": 11.0,
|
| 182 |
+
"eval_cer": 0.14817837085393365,
|
| 183 |
+
"eval_loss": 0.4337259531021118,
|
| 184 |
+
"eval_per": 0.2643473844591163,
|
| 185 |
+
"eval_runtime": 30.0682,
|
| 186 |
+
"eval_samples_per_second": 12.738,
|
| 187 |
+
"eval_steps_per_second": 1.596,
|
| 188 |
+
"eval_wer": 0.2755205688166582,
|
| 189 |
+
"step": 924
|
| 190 |
+
},
|
| 191 |
+
{
|
| 192 |
+
"epoch": 11.9,
|
| 193 |
+
"learning_rate": 2.821964285714286e-05,
|
| 194 |
+
"loss": 0.8262,
|
| 195 |
+
"step": 1000
|
| 196 |
+
},
|
| 197 |
+
{
|
| 198 |
+
"epoch": 12.0,
|
| 199 |
+
"eval_cer": 0.1460183362933807,
|
| 200 |
+
"eval_loss": 0.40190622210502625,
|
| 201 |
+
"eval_per": 0.26130015236160486,
|
| 202 |
+
"eval_runtime": 30.0826,
|
| 203 |
+
"eval_samples_per_second": 12.732,
|
| 204 |
+
"eval_steps_per_second": 1.596,
|
| 205 |
+
"eval_wer": 0.27171152869476894,
|
| 206 |
+
"step": 1008
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 13.0,
|
| 210 |
+
"eval_cer": 0.14582633322133154,
|
| 211 |
+
"eval_loss": 0.3872309625148773,
|
| 212 |
+
"eval_per": 0.25901472828847133,
|
| 213 |
+
"eval_runtime": 30.1847,
|
| 214 |
+
"eval_samples_per_second": 12.689,
|
| 215 |
+
"eval_steps_per_second": 1.59,
|
| 216 |
+
"eval_wer": 0.2681564245810056,
|
| 217 |
+
"step": 1092
|
| 218 |
+
},
|
| 219 |
+
{
|
| 220 |
+
"epoch": 13.1,
|
| 221 |
+
"learning_rate": 2.804107142857143e-05,
|
| 222 |
+
"loss": 0.7654,
|
| 223 |
+
"step": 1100
|
| 224 |
+
},
|
| 225 |
+
{
|
| 226 |
+
"epoch": 14.0,
|
| 227 |
+
"eval_cer": 0.14136226179618874,
|
| 228 |
+
"eval_loss": 0.36644813418388367,
|
| 229 |
+
"eval_per": 0.24454037582529203,
|
| 230 |
+
"eval_runtime": 30.0619,
|
| 231 |
+
"eval_samples_per_second": 12.74,
|
| 232 |
+
"eval_steps_per_second": 1.597,
|
| 233 |
+
"eval_wer": 0.2531742001015744,
|
| 234 |
+
"step": 1176
|
| 235 |
+
},
|
| 236 |
+
{
|
| 237 |
+
"epoch": 14.29,
|
| 238 |
+
"learning_rate": 2.78625e-05,
|
| 239 |
+
"loss": 0.6788,
|
| 240 |
+
"step": 1200
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"epoch": 15.0,
|
| 244 |
+
"eval_cer": 0.1390102241635866,
|
| 245 |
+
"eval_loss": 0.3475891649723053,
|
| 246 |
+
"eval_per": 0.2404773996952768,
|
| 247 |
+
"eval_runtime": 30.0631,
|
| 248 |
+
"eval_samples_per_second": 12.74,
|
| 249 |
+
"eval_steps_per_second": 1.597,
|
| 250 |
+
"eval_wer": 0.250126968004063,
|
| 251 |
+
"step": 1260
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 15.48,
|
| 255 |
+
"learning_rate": 2.7683928571428573e-05,
|
| 256 |
+
"loss": 0.647,
|
| 257 |
+
"step": 1300
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"epoch": 16.0,
|
| 261 |
+
"eval_cer": 0.13881822109153746,
|
| 262 |
+
"eval_loss": 0.350846529006958,
|
| 263 |
+
"eval_per": 0.23641442356526154,
|
| 264 |
+
"eval_runtime": 30.6699,
|
| 265 |
+
"eval_samples_per_second": 12.488,
|
| 266 |
+
"eval_steps_per_second": 1.565,
|
| 267 |
+
"eval_wer": 0.2453021838496699,
|
| 268 |
+
"step": 1344
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"epoch": 16.67,
|
| 272 |
+
"learning_rate": 2.7505357142857143e-05,
|
| 273 |
+
"loss": 0.6058,
|
| 274 |
+
"step": 1400
|
| 275 |
+
},
|
| 276 |
+
{
|
| 277 |
+
"epoch": 17.0,
|
| 278 |
+
"eval_cer": 0.13920222723563577,
|
| 279 |
+
"eval_loss": 0.3425397276878357,
|
| 280 |
+
"eval_per": 0.23489080751650585,
|
| 281 |
+
"eval_runtime": 30.0637,
|
| 282 |
+
"eval_samples_per_second": 12.74,
|
| 283 |
+
"eval_steps_per_second": 1.597,
|
| 284 |
+
"eval_wer": 0.244794311833418,
|
| 285 |
+
"step": 1428
|
| 286 |
+
},
|
| 287 |
+
{
|
| 288 |
+
"epoch": 17.86,
|
| 289 |
+
"learning_rate": 2.7326785714285716e-05,
|
| 290 |
+
"loss": 0.5853,
|
| 291 |
+
"step": 1500
|
| 292 |
+
},
|
| 293 |
+
{
|
| 294 |
+
"epoch": 18.0,
|
| 295 |
+
"eval_cer": 0.13512216195459129,
|
| 296 |
+
"eval_loss": 0.3393208980560303,
|
| 297 |
+
"eval_per": 0.2293042153377349,
|
| 298 |
+
"eval_runtime": 30.0796,
|
| 299 |
+
"eval_samples_per_second": 12.733,
|
| 300 |
+
"eval_steps_per_second": 1.596,
|
| 301 |
+
"eval_wer": 0.23768410360589132,
|
| 302 |
+
"step": 1512
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 19.0,
|
| 306 |
+
"eval_cer": 0.13473815581049298,
|
| 307 |
+
"eval_loss": 0.32763150334358215,
|
| 308 |
+
"eval_per": 0.22219400711020823,
|
| 309 |
+
"eval_runtime": 30.0565,
|
| 310 |
+
"eval_samples_per_second": 12.743,
|
| 311 |
+
"eval_steps_per_second": 1.597,
|
| 312 |
+
"eval_wer": 0.2293042153377349,
|
| 313 |
+
"step": 1596
|
| 314 |
+
},
|
| 315 |
+
{
|
| 316 |
+
"epoch": 19.05,
|
| 317 |
+
"learning_rate": 2.7148214285714286e-05,
|
| 318 |
+
"loss": 0.5356,
|
| 319 |
+
"step": 1600
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 20.0,
|
| 323 |
+
"eval_cer": 0.13401814429030864,
|
| 324 |
+
"eval_loss": 0.3274891674518585,
|
| 325 |
+
"eval_per": 0.22752666328085322,
|
| 326 |
+
"eval_runtime": 30.0773,
|
| 327 |
+
"eval_samples_per_second": 12.734,
|
| 328 |
+
"eval_steps_per_second": 1.596,
|
| 329 |
+
"eval_wer": 0.23158963941086846,
|
| 330 |
+
"step": 1680
|
| 331 |
+
},
|
| 332 |
+
{
|
| 333 |
+
"epoch": 20.24,
|
| 334 |
+
"learning_rate": 2.6969642857142856e-05,
|
| 335 |
+
"loss": 0.5323,
|
| 336 |
+
"step": 1700
|
| 337 |
+
},
|
| 338 |
+
{
|
| 339 |
+
"epoch": 21.0,
|
| 340 |
+
"eval_cer": 0.1341621465943455,
|
| 341 |
+
"eval_loss": 0.32451650500297546,
|
| 342 |
+
"eval_per": 0.22651091924834943,
|
| 343 |
+
"eval_runtime": 30.0925,
|
| 344 |
+
"eval_samples_per_second": 12.727,
|
| 345 |
+
"eval_steps_per_second": 1.595,
|
| 346 |
+
"eval_wer": 0.23006602336211274,
|
| 347 |
+
"step": 1764
|
| 348 |
+
},
|
| 349 |
+
{
|
| 350 |
+
"epoch": 21.43,
|
| 351 |
+
"learning_rate": 2.679107142857143e-05,
|
| 352 |
+
"loss": 0.501,
|
| 353 |
+
"step": 1800
|
| 354 |
+
},
|
| 355 |
+
{
|
| 356 |
+
"epoch": 22.0,
|
| 357 |
+
"eval_cer": 0.1333941343061489,
|
| 358 |
+
"eval_loss": 0.325840562582016,
|
| 359 |
+
"eval_per": 0.2196546470289487,
|
| 360 |
+
"eval_runtime": 30.6532,
|
| 361 |
+
"eval_samples_per_second": 12.495,
|
| 362 |
+
"eval_steps_per_second": 1.566,
|
| 363 |
+
"eval_wer": 0.22320975114271205,
|
| 364 |
+
"step": 1848
|
| 365 |
+
},
|
| 366 |
+
{
|
| 367 |
+
"epoch": 22.62,
|
| 368 |
+
"learning_rate": 2.66125e-05,
|
| 369 |
+
"loss": 0.4956,
|
| 370 |
+
"step": 1900
|
| 371 |
+
},
|
| 372 |
+
{
|
| 373 |
+
"epoch": 23.0,
|
| 374 |
+
"eval_cer": 0.13320213123409974,
|
| 375 |
+
"eval_loss": 0.3261657655239105,
|
| 376 |
+
"eval_per": 0.22016251904520062,
|
| 377 |
+
"eval_runtime": 29.9903,
|
| 378 |
+
"eval_samples_per_second": 12.771,
|
| 379 |
+
"eval_steps_per_second": 1.601,
|
| 380 |
+
"eval_wer": 0.22422549517521584,
|
| 381 |
+
"step": 1932
|
| 382 |
+
},
|
| 383 |
+
{
|
| 384 |
+
"epoch": 23.81,
|
| 385 |
+
"learning_rate": 2.6433928571428573e-05,
|
| 386 |
+
"loss": 0.4427,
|
| 387 |
+
"step": 2000
|
| 388 |
+
},
|
| 389 |
+
{
|
| 390 |
+
"epoch": 24.0,
|
| 391 |
+
"eval_cer": 0.13281812509000143,
|
| 392 |
+
"eval_loss": 0.3162442147731781,
|
| 393 |
+
"eval_per": 0.22600304723209752,
|
| 394 |
+
"eval_runtime": 30.0934,
|
| 395 |
+
"eval_samples_per_second": 12.727,
|
| 396 |
+
"eval_steps_per_second": 1.595,
|
| 397 |
+
"eval_wer": 0.22981208735398678,
|
| 398 |
+
"step": 2016
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"epoch": 25.0,
|
| 402 |
+
"learning_rate": 2.6255357142857146e-05,
|
| 403 |
+
"loss": 0.4443,
|
| 404 |
+
"step": 2100
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 25.0,
|
| 408 |
+
"eval_cer": 0.13310612969807517,
|
| 409 |
+
"eval_loss": 0.32252442836761475,
|
| 410 |
+
"eval_per": 0.22524123920771966,
|
| 411 |
+
"eval_runtime": 30.0405,
|
| 412 |
+
"eval_samples_per_second": 12.749,
|
| 413 |
+
"eval_steps_per_second": 1.598,
|
| 414 |
+
"eval_wer": 0.22854240731335704,
|
| 415 |
+
"step": 2100
|
| 416 |
+
},
|
| 417 |
+
{
|
| 418 |
+
"epoch": 26.0,
|
| 419 |
+
"eval_cer": 0.1333461335381366,
|
| 420 |
+
"eval_loss": 0.3246039152145386,
|
| 421 |
+
"eval_per": 0.22270187912646014,
|
| 422 |
+
"eval_runtime": 30.1326,
|
| 423 |
+
"eval_samples_per_second": 12.71,
|
| 424 |
+
"eval_steps_per_second": 1.593,
|
| 425 |
+
"eval_wer": 0.2270187912646013,
|
| 426 |
+
"step": 2184
|
| 427 |
+
},
|
| 428 |
+
{
|
| 429 |
+
"epoch": 26.19,
|
| 430 |
+
"learning_rate": 2.6076785714285713e-05,
|
| 431 |
+
"loss": 0.4234,
|
| 432 |
+
"step": 2200
|
| 433 |
+
},
|
| 434 |
+
{
|
| 435 |
+
"epoch": 27.0,
|
| 436 |
+
"eval_cer": 0.13272212355397686,
|
| 437 |
+
"eval_loss": 0.31707704067230225,
|
| 438 |
+
"eval_per": 0.22143219908583037,
|
| 439 |
+
"eval_runtime": 30.1196,
|
| 440 |
+
"eval_samples_per_second": 12.716,
|
| 441 |
+
"eval_steps_per_second": 1.594,
|
| 442 |
+
"eval_wer": 0.2249873031995937,
|
| 443 |
+
"step": 2268
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 27.38,
|
| 447 |
+
"learning_rate": 2.5898214285714286e-05,
|
| 448 |
+
"loss": 0.4245,
|
| 449 |
+
"step": 2300
|
| 450 |
+
},
|
| 451 |
+
{
|
| 452 |
+
"epoch": 28.0,
|
| 453 |
+
"eval_cer": 0.13440215043440695,
|
| 454 |
+
"eval_loss": 0.3180580735206604,
|
| 455 |
+
"eval_per": 0.2229558151345861,
|
| 456 |
+
"eval_runtime": 30.0474,
|
| 457 |
+
"eval_samples_per_second": 12.747,
|
| 458 |
+
"eval_steps_per_second": 1.597,
|
| 459 |
+
"eval_wer": 0.22651091924834943,
|
| 460 |
+
"step": 2352
|
| 461 |
+
},
|
| 462 |
+
{
|
| 463 |
+
"epoch": 28.57,
|
| 464 |
+
"learning_rate": 2.5719642857142856e-05,
|
| 465 |
+
"loss": 0.3993,
|
| 466 |
+
"step": 2400
|
| 467 |
+
},
|
| 468 |
+
{
|
| 469 |
+
"epoch": 29.0,
|
| 470 |
+
"eval_cer": 0.13310612969807517,
|
| 471 |
+
"eval_loss": 0.3283059597015381,
|
| 472 |
+
"eval_per": 0.22168613509395632,
|
| 473 |
+
"eval_runtime": 30.0868,
|
| 474 |
+
"eval_samples_per_second": 12.73,
|
| 475 |
+
"eval_steps_per_second": 1.595,
|
| 476 |
+
"eval_wer": 0.2249873031995937,
|
| 477 |
+
"step": 2436
|
| 478 |
+
},
|
| 479 |
+
{
|
| 480 |
+
"epoch": 29.76,
|
| 481 |
+
"learning_rate": 2.554107142857143e-05,
|
| 482 |
+
"loss": 0.3769,
|
| 483 |
+
"step": 2500
|
| 484 |
+
},
|
| 485 |
+
{
|
| 486 |
+
"epoch": 30.0,
|
| 487 |
+
"eval_cer": 0.13589017424278788,
|
| 488 |
+
"eval_loss": 0.32919740676879883,
|
| 489 |
+
"eval_per": 0.23235144743524633,
|
| 490 |
+
"eval_runtime": 30.1605,
|
| 491 |
+
"eval_samples_per_second": 12.699,
|
| 492 |
+
"eval_steps_per_second": 1.591,
|
| 493 |
+
"eval_wer": 0.23539867953275775,
|
| 494 |
+
"step": 2520
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 30.95,
|
| 498 |
+
"learning_rate": 2.5362500000000003e-05,
|
| 499 |
+
"loss": 0.3925,
|
| 500 |
+
"step": 2600
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 31.0,
|
| 504 |
+
"eval_cer": 0.13598617577881247,
|
| 505 |
+
"eval_loss": 0.3275951147079468,
|
| 506 |
+
"eval_per": 0.2313357034027425,
|
| 507 |
+
"eval_runtime": 30.0874,
|
| 508 |
+
"eval_samples_per_second": 12.73,
|
| 509 |
+
"eval_steps_per_second": 1.595,
|
| 510 |
+
"eval_wer": 0.23438293550025394,
|
| 511 |
+
"step": 2604
|
| 512 |
+
},
|
| 513 |
+
{
|
| 514 |
+
"epoch": 32.0,
|
| 515 |
+
"eval_cer": 0.13449815197043152,
|
| 516 |
+
"eval_loss": 0.335648775100708,
|
| 517 |
+
"eval_per": 0.22752666328085322,
|
| 518 |
+
"eval_runtime": 30.0908,
|
| 519 |
+
"eval_samples_per_second": 12.728,
|
| 520 |
+
"eval_steps_per_second": 1.595,
|
| 521 |
+
"eval_wer": 0.23006602336211274,
|
| 522 |
+
"step": 2688
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 32.14,
|
| 526 |
+
"learning_rate": 2.518392857142857e-05,
|
| 527 |
+
"loss": 0.3724,
|
| 528 |
+
"step": 2700
|
| 529 |
+
},
|
| 530 |
+
{
|
| 531 |
+
"epoch": 33.0,
|
| 532 |
+
"eval_cer": 0.13565017040272645,
|
| 533 |
+
"eval_loss": 0.3362816274166107,
|
| 534 |
+
"eval_per": 0.2244794311833418,
|
| 535 |
+
"eval_runtime": 30.0472,
|
| 536 |
+
"eval_samples_per_second": 12.747,
|
| 537 |
+
"eval_steps_per_second": 1.597,
|
| 538 |
+
"eval_wer": 0.22727272727272727,
|
| 539 |
+
"step": 2772
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 33.33,
|
| 543 |
+
"learning_rate": 2.5005357142857143e-05,
|
| 544 |
+
"loss": 0.3614,
|
| 545 |
+
"step": 2800
|
| 546 |
+
},
|
| 547 |
+
{
|
| 548 |
+
"epoch": 34.0,
|
| 549 |
+
"eval_cer": 0.1358421734747756,
|
| 550 |
+
"eval_loss": 0.330671101808548,
|
| 551 |
+
"eval_per": 0.22803453529710513,
|
| 552 |
+
"eval_runtime": 29.9742,
|
| 553 |
+
"eval_samples_per_second": 12.778,
|
| 554 |
+
"eval_steps_per_second": 1.601,
|
| 555 |
+
"eval_wer": 0.23108176739461656,
|
| 556 |
+
"step": 2856
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"epoch": 34.52,
|
| 560 |
+
"learning_rate": 2.4826785714285713e-05,
|
| 561 |
+
"loss": 0.3705,
|
| 562 |
+
"step": 2900
|
| 563 |
+
},
|
| 564 |
+
{
|
| 565 |
+
"epoch": 35.0,
|
| 566 |
+
"eval_cer": 0.13560216963471416,
|
| 567 |
+
"eval_loss": 0.32763954997062683,
|
| 568 |
+
"eval_per": 0.22854240731335704,
|
| 569 |
+
"eval_runtime": 29.995,
|
| 570 |
+
"eval_samples_per_second": 12.769,
|
| 571 |
+
"eval_steps_per_second": 1.6,
|
| 572 |
+
"eval_wer": 0.23158963941086846,
|
| 573 |
+
"step": 2940
|
| 574 |
+
},
|
| 575 |
+
{
|
| 576 |
+
"epoch": 35.71,
|
| 577 |
+
"learning_rate": 2.4648214285714287e-05,
|
| 578 |
+
"loss": 0.3376,
|
| 579 |
+
"step": 3000
|
| 580 |
+
},
|
| 581 |
+
{
|
| 582 |
+
"epoch": 36.0,
|
| 583 |
+
"eval_cer": 0.13728219651514426,
|
| 584 |
+
"eval_loss": 0.3309231102466583,
|
| 585 |
+
"eval_per": 0.23362112747587607,
|
| 586 |
+
"eval_runtime": 30.0273,
|
| 587 |
+
"eval_samples_per_second": 12.755,
|
| 588 |
+
"eval_steps_per_second": 1.599,
|
| 589 |
+
"eval_wer": 0.23793803961401727,
|
| 590 |
+
"step": 3024
|
| 591 |
+
},
|
| 592 |
+
{
|
| 593 |
+
"epoch": 36.9,
|
| 594 |
+
"learning_rate": 2.446964285714286e-05,
|
| 595 |
+
"loss": 0.3434,
|
| 596 |
+
"step": 3100
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 37.0,
|
| 600 |
+
"eval_cer": 0.13670618729899678,
|
| 601 |
+
"eval_loss": 0.33398517966270447,
|
| 602 |
+
"eval_per": 0.2254951752158456,
|
| 603 |
+
"eval_runtime": 30.1096,
|
| 604 |
+
"eval_samples_per_second": 12.72,
|
| 605 |
+
"eval_steps_per_second": 1.594,
|
| 606 |
+
"eval_wer": 0.22905027932960895,
|
| 607 |
+
"step": 3108
|
| 608 |
+
},
|
| 609 |
+
{
|
| 610 |
+
"epoch": 38.0,
|
| 611 |
+
"eval_cer": 0.13694619113905823,
|
| 612 |
+
"eval_loss": 0.34280771017074585,
|
| 613 |
+
"eval_per": 0.2293042153377349,
|
| 614 |
+
"eval_runtime": 30.1204,
|
| 615 |
+
"eval_samples_per_second": 12.716,
|
| 616 |
+
"eval_steps_per_second": 1.594,
|
| 617 |
+
"eval_wer": 0.23260538344337228,
|
| 618 |
+
"step": 3192
|
| 619 |
+
},
|
| 620 |
+
{
|
| 621 |
+
"epoch": 38.1,
|
| 622 |
+
"learning_rate": 2.429107142857143e-05,
|
| 623 |
+
"loss": 0.3245,
|
| 624 |
+
"step": 3200
|
| 625 |
+
},
|
| 626 |
+
{
|
| 627 |
+
"epoch": 39.0,
|
| 628 |
+
"eval_cer": 0.13397014352229636,
|
| 629 |
+
"eval_loss": 0.3409918546676636,
|
| 630 |
+
"eval_per": 0.22270187912646014,
|
| 631 |
+
"eval_runtime": 30.0402,
|
| 632 |
+
"eval_samples_per_second": 12.75,
|
| 633 |
+
"eval_steps_per_second": 1.598,
|
| 634 |
+
"eval_wer": 0.22600304723209752,
|
| 635 |
+
"step": 3276
|
| 636 |
+
},
|
| 637 |
+
{
|
| 638 |
+
"epoch": 39.29,
|
| 639 |
+
"learning_rate": 2.41125e-05,
|
| 640 |
+
"loss": 0.3087,
|
| 641 |
+
"step": 3300
|
| 642 |
+
},
|
| 643 |
+
{
|
| 644 |
+
"epoch": 40.0,
|
| 645 |
+
"eval_cer": 0.13478615657850526,
|
| 646 |
+
"eval_loss": 0.35161110758781433,
|
| 647 |
+
"eval_per": 0.22574911122397157,
|
| 648 |
+
"eval_runtime": 30.0824,
|
| 649 |
+
"eval_samples_per_second": 12.732,
|
| 650 |
+
"eval_steps_per_second": 1.596,
|
| 651 |
+
"eval_wer": 0.22854240731335704,
|
| 652 |
+
"step": 3360
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"epoch": 40.48,
|
| 656 |
+
"learning_rate": 2.393392857142857e-05,
|
| 657 |
+
"loss": 0.3025,
|
| 658 |
+
"step": 3400
|
| 659 |
+
},
|
| 660 |
+
{
|
| 661 |
+
"epoch": 41.0,
|
| 662 |
+
"eval_cer": 0.13473815581049298,
|
| 663 |
+
"eval_loss": 0.33956247568130493,
|
| 664 |
+
"eval_per": 0.22371762315896393,
|
| 665 |
+
"eval_runtime": 30.036,
|
| 666 |
+
"eval_samples_per_second": 12.751,
|
| 667 |
+
"eval_steps_per_second": 1.598,
|
| 668 |
+
"eval_wer": 0.22676485525647536,
|
| 669 |
+
"step": 3444
|
| 670 |
+
},
|
| 671 |
+
{
|
| 672 |
+
"epoch": 41.67,
|
| 673 |
+
"learning_rate": 2.3755357142857143e-05,
|
| 674 |
+
"loss": 0.2953,
|
| 675 |
+
"step": 3500
|
| 676 |
+
},
|
| 677 |
+
{
|
| 678 |
+
"epoch": 42.0,
|
| 679 |
+
"eval_cer": 0.1345941535064561,
|
| 680 |
+
"eval_loss": 0.3443167209625244,
|
| 681 |
+
"eval_per": 0.22219400711020823,
|
| 682 |
+
"eval_runtime": 30.1578,
|
| 683 |
+
"eval_samples_per_second": 12.7,
|
| 684 |
+
"eval_steps_per_second": 1.592,
|
| 685 |
+
"eval_wer": 0.2254951752158456,
|
| 686 |
+
"step": 3528
|
| 687 |
+
},
|
| 688 |
+
{
|
| 689 |
+
"epoch": 42.86,
|
| 690 |
+
"learning_rate": 2.3576785714285717e-05,
|
| 691 |
+
"loss": 0.2778,
|
| 692 |
+
"step": 3600
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 43.0,
|
| 696 |
+
"eval_cer": 0.13507416118657897,
|
| 697 |
+
"eval_loss": 0.34061482548713684,
|
| 698 |
+
"eval_per": 0.2229558151345861,
|
| 699 |
+
"eval_runtime": 30.0853,
|
| 700 |
+
"eval_samples_per_second": 12.73,
|
| 701 |
+
"eval_steps_per_second": 1.595,
|
| 702 |
+
"eval_wer": 0.2254951752158456,
|
| 703 |
+
"step": 3612
|
| 704 |
+
},
|
| 705 |
+
{
|
| 706 |
+
"epoch": 44.0,
|
| 707 |
+
"eval_cer": 0.13670618729899678,
|
| 708 |
+
"eval_loss": 0.3456892967224121,
|
| 709 |
+
"eval_per": 0.2308278313864906,
|
| 710 |
+
"eval_runtime": 30.1071,
|
| 711 |
+
"eval_samples_per_second": 12.721,
|
| 712 |
+
"eval_steps_per_second": 1.594,
|
| 713 |
+
"eval_wer": 0.2346368715083799,
|
| 714 |
+
"step": 3696
|
| 715 |
+
},
|
| 716 |
+
{
|
| 717 |
+
"epoch": 44.0,
|
| 718 |
+
"step": 3696,
|
| 719 |
+
"total_flos": 3.1481023081144828e+19,
|
| 720 |
+
"train_loss": 1.0956278002107298,
|
| 721 |
+
"train_runtime": 26844.9073,
|
| 722 |
+
"train_samples_per_second": 20.026,
|
| 723 |
+
"train_steps_per_second": 0.626
|
| 724 |
+
}
|
| 725 |
+
],
|
| 726 |
+
"max_steps": 16800,
|
| 727 |
+
"num_train_epochs": 200,
|
| 728 |
+
"total_flos": 3.1481023081144828e+19,
|
| 729 |
+
"trial_name": null,
|
| 730 |
+
"trial_params": null
|
| 731 |
+
}
|