Commit
·
0543857
1
Parent(s):
eaec757
End of training
Browse files- all_results.json +16 -0
- eval_results.json +10 -0
- train_results.json +9 -0
- trainer_state.json +1611 -0
all_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 146.0,
|
| 3 |
+
"eval_cer": 0.17852593266606007,
|
| 4 |
+
"eval_loss": 0.5470190644264221,
|
| 5 |
+
"eval_runtime": 53.7447,
|
| 6 |
+
"eval_samples": 542,
|
| 7 |
+
"eval_samples_per_second": 10.085,
|
| 8 |
+
"eval_steps_per_second": 1.265,
|
| 9 |
+
"eval_wer": 0.34173100365705,
|
| 10 |
+
"total_flos": 1.8244410804942518e+19,
|
| 11 |
+
"train_loss": 1.3513914437676595,
|
| 12 |
+
"train_runtime": 31402.8433,
|
| 13 |
+
"train_samples": 448,
|
| 14 |
+
"train_samples_per_second": 2.14,
|
| 15 |
+
"train_steps_per_second": 0.067
|
| 16 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 146.0,
|
| 3 |
+
"eval_cer": 0.17852593266606007,
|
| 4 |
+
"eval_loss": 0.5470190644264221,
|
| 5 |
+
"eval_runtime": 53.7447,
|
| 6 |
+
"eval_samples": 542,
|
| 7 |
+
"eval_samples_per_second": 10.085,
|
| 8 |
+
"eval_steps_per_second": 1.265,
|
| 9 |
+
"eval_wer": 0.34173100365705
|
| 10 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 146.0,
|
| 3 |
+
"total_flos": 1.8244410804942518e+19,
|
| 4 |
+
"train_loss": 1.3513914437676595,
|
| 5 |
+
"train_runtime": 31402.8433,
|
| 6 |
+
"train_samples": 448,
|
| 7 |
+
"train_samples_per_second": 2.14,
|
| 8 |
+
"train_steps_per_second": 0.067
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1611 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.5470190644264221,
|
| 3 |
+
"best_model_checkpoint": "wav2vec2-large-xlsr-coraa-exp-10/checkpoint-1764",
|
| 4 |
+
"epoch": 146.0,
|
| 5 |
+
"global_step": 2044,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.07,
|
| 12 |
+
"learning_rate": 3e-05,
|
| 13 |
+
"loss": 37.9224,
|
| 14 |
+
"step": 1
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 1.0,
|
| 18 |
+
"eval_cer": 0.9618198362147407,
|
| 19 |
+
"eval_loss": 24.657764434814453,
|
| 20 |
+
"eval_runtime": 53.8844,
|
| 21 |
+
"eval_samples_per_second": 10.059,
|
| 22 |
+
"eval_steps_per_second": 1.262,
|
| 23 |
+
"eval_wer": 1.0,
|
| 24 |
+
"step": 14
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 2.0,
|
| 28 |
+
"eval_cer": 0.9618562329390354,
|
| 29 |
+
"eval_loss": 7.14532470703125,
|
| 30 |
+
"eval_runtime": 53.624,
|
| 31 |
+
"eval_samples_per_second": 10.107,
|
| 32 |
+
"eval_steps_per_second": 1.268,
|
| 33 |
+
"eval_wer": 1.0,
|
| 34 |
+
"step": 28
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 3.0,
|
| 38 |
+
"eval_cer": 0.9618562329390354,
|
| 39 |
+
"eval_loss": 4.43905782699585,
|
| 40 |
+
"eval_runtime": 53.4498,
|
| 41 |
+
"eval_samples_per_second": 10.14,
|
| 42 |
+
"eval_steps_per_second": 1.272,
|
| 43 |
+
"eval_wer": 1.0,
|
| 44 |
+
"step": 42
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 4.0,
|
| 48 |
+
"eval_cer": 0.9618562329390354,
|
| 49 |
+
"eval_loss": 3.90922474861145,
|
| 50 |
+
"eval_runtime": 53.742,
|
| 51 |
+
"eval_samples_per_second": 10.085,
|
| 52 |
+
"eval_steps_per_second": 1.265,
|
| 53 |
+
"eval_wer": 1.0,
|
| 54 |
+
"step": 56
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 5.0,
|
| 58 |
+
"eval_cer": 0.9618562329390354,
|
| 59 |
+
"eval_loss": 3.6834604740142822,
|
| 60 |
+
"eval_runtime": 53.6572,
|
| 61 |
+
"eval_samples_per_second": 10.101,
|
| 62 |
+
"eval_steps_per_second": 1.267,
|
| 63 |
+
"eval_wer": 1.0,
|
| 64 |
+
"step": 70
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"epoch": 6.0,
|
| 68 |
+
"eval_cer": 0.9618562329390354,
|
| 69 |
+
"eval_loss": 3.522313117980957,
|
| 70 |
+
"eval_runtime": 53.7448,
|
| 71 |
+
"eval_samples_per_second": 10.085,
|
| 72 |
+
"eval_steps_per_second": 1.265,
|
| 73 |
+
"eval_wer": 1.0,
|
| 74 |
+
"step": 84
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 7.0,
|
| 78 |
+
"eval_cer": 0.9618562329390354,
|
| 79 |
+
"eval_loss": 3.3716423511505127,
|
| 80 |
+
"eval_runtime": 53.5358,
|
| 81 |
+
"eval_samples_per_second": 10.124,
|
| 82 |
+
"eval_steps_per_second": 1.27,
|
| 83 |
+
"eval_wer": 1.0,
|
| 84 |
+
"step": 98
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 7.14,
|
| 88 |
+
"learning_rate": 2.8614285714285714e-05,
|
| 89 |
+
"loss": 9.0651,
|
| 90 |
+
"step": 100
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"epoch": 8.0,
|
| 94 |
+
"eval_cer": 0.9618562329390354,
|
| 95 |
+
"eval_loss": 3.2723329067230225,
|
| 96 |
+
"eval_runtime": 53.6095,
|
| 97 |
+
"eval_samples_per_second": 10.11,
|
| 98 |
+
"eval_steps_per_second": 1.268,
|
| 99 |
+
"eval_wer": 1.0,
|
| 100 |
+
"step": 112
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 9.0,
|
| 104 |
+
"eval_cer": 0.9618562329390354,
|
| 105 |
+
"eval_loss": 3.1859705448150635,
|
| 106 |
+
"eval_runtime": 53.5878,
|
| 107 |
+
"eval_samples_per_second": 10.114,
|
| 108 |
+
"eval_steps_per_second": 1.269,
|
| 109 |
+
"eval_wer": 1.0,
|
| 110 |
+
"step": 126
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 10.0,
|
| 114 |
+
"eval_cer": 0.9618562329390354,
|
| 115 |
+
"eval_loss": 3.146104574203491,
|
| 116 |
+
"eval_runtime": 53.491,
|
| 117 |
+
"eval_samples_per_second": 10.133,
|
| 118 |
+
"eval_steps_per_second": 1.271,
|
| 119 |
+
"eval_wer": 1.0,
|
| 120 |
+
"step": 140
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"epoch": 11.0,
|
| 124 |
+
"eval_cer": 0.9618562329390354,
|
| 125 |
+
"eval_loss": 3.1367859840393066,
|
| 126 |
+
"eval_runtime": 53.9341,
|
| 127 |
+
"eval_samples_per_second": 10.049,
|
| 128 |
+
"eval_steps_per_second": 1.261,
|
| 129 |
+
"eval_wer": 1.0,
|
| 130 |
+
"step": 154
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 12.0,
|
| 134 |
+
"eval_cer": 0.9618562329390354,
|
| 135 |
+
"eval_loss": 3.0960755348205566,
|
| 136 |
+
"eval_runtime": 53.5674,
|
| 137 |
+
"eval_samples_per_second": 10.118,
|
| 138 |
+
"eval_steps_per_second": 1.269,
|
| 139 |
+
"eval_wer": 1.0,
|
| 140 |
+
"step": 168
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 13.0,
|
| 144 |
+
"eval_cer": 0.9618562329390354,
|
| 145 |
+
"eval_loss": 3.0767099857330322,
|
| 146 |
+
"eval_runtime": 53.6269,
|
| 147 |
+
"eval_samples_per_second": 10.107,
|
| 148 |
+
"eval_steps_per_second": 1.268,
|
| 149 |
+
"eval_wer": 1.0,
|
| 150 |
+
"step": 182
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"epoch": 14.0,
|
| 154 |
+
"eval_cer": 0.9618562329390354,
|
| 155 |
+
"eval_loss": 3.0509135723114014,
|
| 156 |
+
"eval_runtime": 53.9364,
|
| 157 |
+
"eval_samples_per_second": 10.049,
|
| 158 |
+
"eval_steps_per_second": 1.261,
|
| 159 |
+
"eval_wer": 1.0,
|
| 160 |
+
"step": 196
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 14.29,
|
| 164 |
+
"learning_rate": 2.7185714285714287e-05,
|
| 165 |
+
"loss": 3.0601,
|
| 166 |
+
"step": 200
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"epoch": 15.0,
|
| 170 |
+
"eval_cer": 0.9618562329390354,
|
| 171 |
+
"eval_loss": 3.087069511413574,
|
| 172 |
+
"eval_runtime": 53.3522,
|
| 173 |
+
"eval_samples_per_second": 10.159,
|
| 174 |
+
"eval_steps_per_second": 1.275,
|
| 175 |
+
"eval_wer": 1.0,
|
| 176 |
+
"step": 210
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 16.0,
|
| 180 |
+
"eval_cer": 0.9618562329390354,
|
| 181 |
+
"eval_loss": 3.041494131088257,
|
| 182 |
+
"eval_runtime": 53.2391,
|
| 183 |
+
"eval_samples_per_second": 10.18,
|
| 184 |
+
"eval_steps_per_second": 1.277,
|
| 185 |
+
"eval_wer": 1.0,
|
| 186 |
+
"step": 224
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 17.0,
|
| 190 |
+
"eval_cer": 0.9618562329390354,
|
| 191 |
+
"eval_loss": 3.0329771041870117,
|
| 192 |
+
"eval_runtime": 53.3109,
|
| 193 |
+
"eval_samples_per_second": 10.167,
|
| 194 |
+
"eval_steps_per_second": 1.276,
|
| 195 |
+
"eval_wer": 1.0,
|
| 196 |
+
"step": 238
|
| 197 |
+
},
|
| 198 |
+
{
|
| 199 |
+
"epoch": 18.0,
|
| 200 |
+
"eval_cer": 0.9618562329390354,
|
| 201 |
+
"eval_loss": 3.019200086593628,
|
| 202 |
+
"eval_runtime": 53.38,
|
| 203 |
+
"eval_samples_per_second": 10.154,
|
| 204 |
+
"eval_steps_per_second": 1.274,
|
| 205 |
+
"eval_wer": 1.0,
|
| 206 |
+
"step": 252
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 19.0,
|
| 210 |
+
"eval_cer": 0.9618562329390354,
|
| 211 |
+
"eval_loss": 3.026556968688965,
|
| 212 |
+
"eval_runtime": 53.4357,
|
| 213 |
+
"eval_samples_per_second": 10.143,
|
| 214 |
+
"eval_steps_per_second": 1.273,
|
| 215 |
+
"eval_wer": 1.0,
|
| 216 |
+
"step": 266
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 20.0,
|
| 220 |
+
"eval_cer": 0.9618562329390354,
|
| 221 |
+
"eval_loss": 3.024301528930664,
|
| 222 |
+
"eval_runtime": 53.5352,
|
| 223 |
+
"eval_samples_per_second": 10.124,
|
| 224 |
+
"eval_steps_per_second": 1.27,
|
| 225 |
+
"eval_wer": 1.0,
|
| 226 |
+
"step": 280
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 21.0,
|
| 230 |
+
"eval_cer": 0.9618562329390354,
|
| 231 |
+
"eval_loss": 3.0105724334716797,
|
| 232 |
+
"eval_runtime": 53.4874,
|
| 233 |
+
"eval_samples_per_second": 10.133,
|
| 234 |
+
"eval_steps_per_second": 1.271,
|
| 235 |
+
"eval_wer": 1.0,
|
| 236 |
+
"step": 294
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 21.43,
|
| 240 |
+
"learning_rate": 2.5757142857142857e-05,
|
| 241 |
+
"loss": 2.9552,
|
| 242 |
+
"step": 300
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 22.0,
|
| 246 |
+
"eval_cer": 0.9618562329390354,
|
| 247 |
+
"eval_loss": 3.0053164958953857,
|
| 248 |
+
"eval_runtime": 53.5107,
|
| 249 |
+
"eval_samples_per_second": 10.129,
|
| 250 |
+
"eval_steps_per_second": 1.271,
|
| 251 |
+
"eval_wer": 1.0,
|
| 252 |
+
"step": 308
|
| 253 |
+
},
|
| 254 |
+
{
|
| 255 |
+
"epoch": 23.0,
|
| 256 |
+
"eval_cer": 0.9618562329390354,
|
| 257 |
+
"eval_loss": 2.9986302852630615,
|
| 258 |
+
"eval_runtime": 53.679,
|
| 259 |
+
"eval_samples_per_second": 10.097,
|
| 260 |
+
"eval_steps_per_second": 1.267,
|
| 261 |
+
"eval_wer": 1.0,
|
| 262 |
+
"step": 322
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 24.0,
|
| 266 |
+
"eval_cer": 0.9618562329390354,
|
| 267 |
+
"eval_loss": 3.0029869079589844,
|
| 268 |
+
"eval_runtime": 54.0064,
|
| 269 |
+
"eval_samples_per_second": 10.036,
|
| 270 |
+
"eval_steps_per_second": 1.259,
|
| 271 |
+
"eval_wer": 1.0,
|
| 272 |
+
"step": 336
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 25.0,
|
| 276 |
+
"eval_cer": 0.9618562329390354,
|
| 277 |
+
"eval_loss": 2.99497389793396,
|
| 278 |
+
"eval_runtime": 53.5494,
|
| 279 |
+
"eval_samples_per_second": 10.122,
|
| 280 |
+
"eval_steps_per_second": 1.27,
|
| 281 |
+
"eval_wer": 1.0,
|
| 282 |
+
"step": 350
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 26.0,
|
| 286 |
+
"eval_cer": 0.9618562329390354,
|
| 287 |
+
"eval_loss": 3.0058376789093018,
|
| 288 |
+
"eval_runtime": 53.3418,
|
| 289 |
+
"eval_samples_per_second": 10.161,
|
| 290 |
+
"eval_steps_per_second": 1.275,
|
| 291 |
+
"eval_wer": 1.0,
|
| 292 |
+
"step": 364
|
| 293 |
+
},
|
| 294 |
+
{
|
| 295 |
+
"epoch": 27.0,
|
| 296 |
+
"eval_cer": 0.9618562329390354,
|
| 297 |
+
"eval_loss": 2.9943442344665527,
|
| 298 |
+
"eval_runtime": 53.5262,
|
| 299 |
+
"eval_samples_per_second": 10.126,
|
| 300 |
+
"eval_steps_per_second": 1.27,
|
| 301 |
+
"eval_wer": 1.0,
|
| 302 |
+
"step": 378
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 28.0,
|
| 306 |
+
"eval_cer": 0.9618562329390354,
|
| 307 |
+
"eval_loss": 2.9845216274261475,
|
| 308 |
+
"eval_runtime": 53.6167,
|
| 309 |
+
"eval_samples_per_second": 10.109,
|
| 310 |
+
"eval_steps_per_second": 1.268,
|
| 311 |
+
"eval_wer": 1.0,
|
| 312 |
+
"step": 392
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 28.57,
|
| 316 |
+
"learning_rate": 2.432857142857143e-05,
|
| 317 |
+
"loss": 2.9213,
|
| 318 |
+
"step": 400
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"epoch": 29.0,
|
| 322 |
+
"eval_cer": 0.9618562329390354,
|
| 323 |
+
"eval_loss": 2.971311330795288,
|
| 324 |
+
"eval_runtime": 53.7132,
|
| 325 |
+
"eval_samples_per_second": 10.091,
|
| 326 |
+
"eval_steps_per_second": 1.266,
|
| 327 |
+
"eval_wer": 1.0,
|
| 328 |
+
"step": 406
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 30.0,
|
| 332 |
+
"eval_cer": 0.9618562329390354,
|
| 333 |
+
"eval_loss": 2.9484527111053467,
|
| 334 |
+
"eval_runtime": 53.7813,
|
| 335 |
+
"eval_samples_per_second": 10.078,
|
| 336 |
+
"eval_steps_per_second": 1.264,
|
| 337 |
+
"eval_wer": 1.0,
|
| 338 |
+
"step": 420
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 31.0,
|
| 342 |
+
"eval_cer": 0.9618562329390354,
|
| 343 |
+
"eval_loss": 2.941479206085205,
|
| 344 |
+
"eval_runtime": 53.6932,
|
| 345 |
+
"eval_samples_per_second": 10.094,
|
| 346 |
+
"eval_steps_per_second": 1.266,
|
| 347 |
+
"eval_wer": 1.0,
|
| 348 |
+
"step": 434
|
| 349 |
+
},
|
| 350 |
+
{
|
| 351 |
+
"epoch": 32.0,
|
| 352 |
+
"eval_cer": 0.9618562329390354,
|
| 353 |
+
"eval_loss": 2.8912880420684814,
|
| 354 |
+
"eval_runtime": 53.8504,
|
| 355 |
+
"eval_samples_per_second": 10.065,
|
| 356 |
+
"eval_steps_per_second": 1.263,
|
| 357 |
+
"eval_wer": 1.0,
|
| 358 |
+
"step": 448
|
| 359 |
+
},
|
| 360 |
+
{
|
| 361 |
+
"epoch": 33.0,
|
| 362 |
+
"eval_cer": 0.9612010919017289,
|
| 363 |
+
"eval_loss": 2.8057382106781006,
|
| 364 |
+
"eval_runtime": 54.0302,
|
| 365 |
+
"eval_samples_per_second": 10.031,
|
| 366 |
+
"eval_steps_per_second": 1.259,
|
| 367 |
+
"eval_wer": 1.0,
|
| 368 |
+
"step": 462
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 34.0,
|
| 372 |
+
"eval_cer": 0.9598908098271156,
|
| 373 |
+
"eval_loss": 2.698427438735962,
|
| 374 |
+
"eval_runtime": 53.7576,
|
| 375 |
+
"eval_samples_per_second": 10.082,
|
| 376 |
+
"eval_steps_per_second": 1.265,
|
| 377 |
+
"eval_wer": 1.0,
|
| 378 |
+
"step": 476
|
| 379 |
+
},
|
| 380 |
+
{
|
| 381 |
+
"epoch": 35.0,
|
| 382 |
+
"eval_cer": 0.9066787989080982,
|
| 383 |
+
"eval_loss": 2.57851243019104,
|
| 384 |
+
"eval_runtime": 53.7653,
|
| 385 |
+
"eval_samples_per_second": 10.081,
|
| 386 |
+
"eval_steps_per_second": 1.265,
|
| 387 |
+
"eval_wer": 1.0,
|
| 388 |
+
"step": 490
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 35.71,
|
| 392 |
+
"learning_rate": 2.29e-05,
|
| 393 |
+
"loss": 2.7804,
|
| 394 |
+
"step": 500
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"epoch": 36.0,
|
| 398 |
+
"eval_cer": 0.7928662420382165,
|
| 399 |
+
"eval_loss": 2.3545000553131104,
|
| 400 |
+
"eval_runtime": 53.7523,
|
| 401 |
+
"eval_samples_per_second": 10.083,
|
| 402 |
+
"eval_steps_per_second": 1.265,
|
| 403 |
+
"eval_wer": 1.0,
|
| 404 |
+
"step": 504
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 37.0,
|
| 408 |
+
"eval_cer": 0.5933030027297543,
|
| 409 |
+
"eval_loss": 2.0432868003845215,
|
| 410 |
+
"eval_runtime": 53.6932,
|
| 411 |
+
"eval_samples_per_second": 10.094,
|
| 412 |
+
"eval_steps_per_second": 1.266,
|
| 413 |
+
"eval_wer": 1.0,
|
| 414 |
+
"step": 518
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 38.0,
|
| 418 |
+
"eval_cer": 0.4701000909918107,
|
| 419 |
+
"eval_loss": 1.7438023090362549,
|
| 420 |
+
"eval_runtime": 53.7889,
|
| 421 |
+
"eval_samples_per_second": 10.076,
|
| 422 |
+
"eval_steps_per_second": 1.264,
|
| 423 |
+
"eval_wer": 1.0,
|
| 424 |
+
"step": 532
|
| 425 |
+
},
|
| 426 |
+
{
|
| 427 |
+
"epoch": 39.0,
|
| 428 |
+
"eval_cer": 0.4139399454049136,
|
| 429 |
+
"eval_loss": 1.4659055471420288,
|
| 430 |
+
"eval_runtime": 53.7237,
|
| 431 |
+
"eval_samples_per_second": 10.089,
|
| 432 |
+
"eval_steps_per_second": 1.266,
|
| 433 |
+
"eval_wer": 1.0,
|
| 434 |
+
"step": 546
|
| 435 |
+
},
|
| 436 |
+
{
|
| 437 |
+
"epoch": 40.0,
|
| 438 |
+
"eval_cer": 0.3840218380345769,
|
| 439 |
+
"eval_loss": 1.2873483896255493,
|
| 440 |
+
"eval_runtime": 53.8213,
|
| 441 |
+
"eval_samples_per_second": 10.07,
|
| 442 |
+
"eval_steps_per_second": 1.263,
|
| 443 |
+
"eval_wer": 0.9928890694839496,
|
| 444 |
+
"step": 560
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 41.0,
|
| 448 |
+
"eval_cer": 0.3386715195632393,
|
| 449 |
+
"eval_loss": 1.1587954759597778,
|
| 450 |
+
"eval_runtime": 53.6935,
|
| 451 |
+
"eval_samples_per_second": 10.094,
|
| 452 |
+
"eval_steps_per_second": 1.266,
|
| 453 |
+
"eval_wer": 0.9315318976026006,
|
| 454 |
+
"step": 574
|
| 455 |
+
},
|
| 456 |
+
{
|
| 457 |
+
"epoch": 42.0,
|
| 458 |
+
"eval_cer": 0.2705732484076433,
|
| 459 |
+
"eval_loss": 1.016275405883789,
|
| 460 |
+
"eval_runtime": 53.8774,
|
| 461 |
+
"eval_samples_per_second": 10.06,
|
| 462 |
+
"eval_steps_per_second": 1.262,
|
| 463 |
+
"eval_wer": 0.7395367736692402,
|
| 464 |
+
"step": 588
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 42.86,
|
| 468 |
+
"learning_rate": 2.1471428571428574e-05,
|
| 469 |
+
"loss": 1.6517,
|
| 470 |
+
"step": 600
|
| 471 |
+
},
|
| 472 |
+
{
|
| 473 |
+
"epoch": 43.0,
|
| 474 |
+
"eval_cer": 0.22584167424931756,
|
| 475 |
+
"eval_loss": 0.9398883581161499,
|
| 476 |
+
"eval_runtime": 53.8756,
|
| 477 |
+
"eval_samples_per_second": 10.06,
|
| 478 |
+
"eval_steps_per_second": 1.262,
|
| 479 |
+
"eval_wer": 0.5331166192604633,
|
| 480 |
+
"step": 602
|
| 481 |
+
},
|
| 482 |
+
{
|
| 483 |
+
"epoch": 44.0,
|
| 484 |
+
"eval_cer": 0.21670609645131939,
|
| 485 |
+
"eval_loss": 0.91312175989151,
|
| 486 |
+
"eval_runtime": 53.9097,
|
| 487 |
+
"eval_samples_per_second": 10.054,
|
| 488 |
+
"eval_steps_per_second": 1.261,
|
| 489 |
+
"eval_wer": 0.4928890694839496,
|
| 490 |
+
"step": 616
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"epoch": 45.0,
|
| 494 |
+
"eval_cer": 0.2113557779799818,
|
| 495 |
+
"eval_loss": 0.8351526856422424,
|
| 496 |
+
"eval_runtime": 54.2652,
|
| 497 |
+
"eval_samples_per_second": 9.988,
|
| 498 |
+
"eval_steps_per_second": 1.253,
|
| 499 |
+
"eval_wer": 0.477041852905323,
|
| 500 |
+
"step": 630
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 46.0,
|
| 504 |
+
"eval_cer": 0.2083712465878071,
|
| 505 |
+
"eval_loss": 0.8114636540412903,
|
| 506 |
+
"eval_runtime": 53.8109,
|
| 507 |
+
"eval_samples_per_second": 10.072,
|
| 508 |
+
"eval_steps_per_second": 1.264,
|
| 509 |
+
"eval_wer": 0.45550589191385615,
|
| 510 |
+
"step": 644
|
| 511 |
+
},
|
| 512 |
+
{
|
| 513 |
+
"epoch": 47.0,
|
| 514 |
+
"eval_cer": 0.2037852593266606,
|
| 515 |
+
"eval_loss": 0.78495192527771,
|
| 516 |
+
"eval_runtime": 53.8458,
|
| 517 |
+
"eval_samples_per_second": 10.066,
|
| 518 |
+
"eval_steps_per_second": 1.263,
|
| 519 |
+
"eval_wer": 0.4402681836651768,
|
| 520 |
+
"step": 658
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 48.0,
|
| 524 |
+
"eval_cer": 0.2018926296633303,
|
| 525 |
+
"eval_loss": 0.757414698600769,
|
| 526 |
+
"eval_runtime": 53.866,
|
| 527 |
+
"eval_samples_per_second": 10.062,
|
| 528 |
+
"eval_steps_per_second": 1.262,
|
| 529 |
+
"eval_wer": 0.4355952864689151,
|
| 530 |
+
"step": 672
|
| 531 |
+
},
|
| 532 |
+
{
|
| 533 |
+
"epoch": 49.0,
|
| 534 |
+
"eval_cer": 0.1989080982711556,
|
| 535 |
+
"eval_loss": 0.723770022392273,
|
| 536 |
+
"eval_runtime": 54.4837,
|
| 537 |
+
"eval_samples_per_second": 9.948,
|
| 538 |
+
"eval_steps_per_second": 1.248,
|
| 539 |
+
"eval_wer": 0.42482730597318163,
|
| 540 |
+
"step": 686
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"epoch": 50.0,
|
| 544 |
+
"learning_rate": 2.004285714285714e-05,
|
| 545 |
+
"loss": 0.7966,
|
| 546 |
+
"step": 700
|
| 547 |
+
},
|
| 548 |
+
{
|
| 549 |
+
"epoch": 50.0,
|
| 550 |
+
"eval_cer": 0.19603275705186532,
|
| 551 |
+
"eval_loss": 0.7132200002670288,
|
| 552 |
+
"eval_runtime": 53.8591,
|
| 553 |
+
"eval_samples_per_second": 10.063,
|
| 554 |
+
"eval_steps_per_second": 1.263,
|
| 555 |
+
"eval_wer": 0.41304347826086957,
|
| 556 |
+
"step": 700
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"epoch": 51.0,
|
| 560 |
+
"eval_cer": 0.19628753412192904,
|
| 561 |
+
"eval_loss": 0.7053857445716858,
|
| 562 |
+
"eval_runtime": 53.8356,
|
| 563 |
+
"eval_samples_per_second": 10.068,
|
| 564 |
+
"eval_steps_per_second": 1.263,
|
| 565 |
+
"eval_wer": 0.41284030881755385,
|
| 566 |
+
"step": 714
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 52.0,
|
| 570 |
+
"eval_cer": 0.19898089171974523,
|
| 571 |
+
"eval_loss": 0.7119351029396057,
|
| 572 |
+
"eval_runtime": 53.6949,
|
| 573 |
+
"eval_samples_per_second": 10.094,
|
| 574 |
+
"eval_steps_per_second": 1.266,
|
| 575 |
+
"eval_wer": 0.413449817147501,
|
| 576 |
+
"step": 728
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"epoch": 53.0,
|
| 580 |
+
"eval_cer": 0.19450409463148316,
|
| 581 |
+
"eval_loss": 0.6792961955070496,
|
| 582 |
+
"eval_runtime": 53.7823,
|
| 583 |
+
"eval_samples_per_second": 10.078,
|
| 584 |
+
"eval_steps_per_second": 1.264,
|
| 585 |
+
"eval_wer": 0.3990247866720845,
|
| 586 |
+
"step": 742
|
| 587 |
+
},
|
| 588 |
+
{
|
| 589 |
+
"epoch": 54.0,
|
| 590 |
+
"eval_cer": 0.19319381255686988,
|
| 591 |
+
"eval_loss": 0.6717870235443115,
|
| 592 |
+
"eval_runtime": 53.6919,
|
| 593 |
+
"eval_samples_per_second": 10.095,
|
| 594 |
+
"eval_steps_per_second": 1.266,
|
| 595 |
+
"eval_wer": 0.39435188947582284,
|
| 596 |
+
"step": 756
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 55.0,
|
| 600 |
+
"eval_cer": 0.19494085532302094,
|
| 601 |
+
"eval_loss": 0.6717579960823059,
|
| 602 |
+
"eval_runtime": 53.8881,
|
| 603 |
+
"eval_samples_per_second": 10.058,
|
| 604 |
+
"eval_steps_per_second": 1.262,
|
| 605 |
+
"eval_wer": 0.4012596505485575,
|
| 606 |
+
"step": 770
|
| 607 |
+
},
|
| 608 |
+
{
|
| 609 |
+
"epoch": 56.0,
|
| 610 |
+
"eval_cer": 0.19654231119199272,
|
| 611 |
+
"eval_loss": 0.6830848455429077,
|
| 612 |
+
"eval_runtime": 54.2951,
|
| 613 |
+
"eval_samples_per_second": 9.982,
|
| 614 |
+
"eval_steps_per_second": 1.252,
|
| 615 |
+
"eval_wer": 0.3976026005688744,
|
| 616 |
+
"step": 784
|
| 617 |
+
},
|
| 618 |
+
{
|
| 619 |
+
"epoch": 57.0,
|
| 620 |
+
"eval_cer": 0.1916287534121929,
|
| 621 |
+
"eval_loss": 0.6400262117385864,
|
| 622 |
+
"eval_runtime": 53.8725,
|
| 623 |
+
"eval_samples_per_second": 10.061,
|
| 624 |
+
"eval_steps_per_second": 1.262,
|
| 625 |
+
"eval_wer": 0.38703778951645673,
|
| 626 |
+
"step": 798
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 57.14,
|
| 630 |
+
"learning_rate": 1.8614285714285715e-05,
|
| 631 |
+
"loss": 0.5799,
|
| 632 |
+
"step": 800
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 58.0,
|
| 636 |
+
"eval_cer": 0.19060964513193812,
|
| 637 |
+
"eval_loss": 0.642275869846344,
|
| 638 |
+
"eval_runtime": 53.5527,
|
| 639 |
+
"eval_samples_per_second": 10.121,
|
| 640 |
+
"eval_steps_per_second": 1.27,
|
| 641 |
+
"eval_wer": 0.3843965867533523,
|
| 642 |
+
"step": 812
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 59.0,
|
| 646 |
+
"eval_cer": 0.19075523202911737,
|
| 647 |
+
"eval_loss": 0.6394357681274414,
|
| 648 |
+
"eval_runtime": 54.0972,
|
| 649 |
+
"eval_samples_per_second": 10.019,
|
| 650 |
+
"eval_steps_per_second": 1.257,
|
| 651 |
+
"eval_wer": 0.38338073953677365,
|
| 652 |
+
"step": 826
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"epoch": 60.0,
|
| 656 |
+
"eval_cer": 0.1924294813466788,
|
| 657 |
+
"eval_loss": 0.6573936939239502,
|
| 658 |
+
"eval_runtime": 53.9836,
|
| 659 |
+
"eval_samples_per_second": 10.04,
|
| 660 |
+
"eval_steps_per_second": 1.26,
|
| 661 |
+
"eval_wer": 0.37850467289719625,
|
| 662 |
+
"step": 840
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 61.0,
|
| 666 |
+
"eval_cer": 0.19181073703366697,
|
| 667 |
+
"eval_loss": 0.6321007013320923,
|
| 668 |
+
"eval_runtime": 53.881,
|
| 669 |
+
"eval_samples_per_second": 10.059,
|
| 670 |
+
"eval_steps_per_second": 1.262,
|
| 671 |
+
"eval_wer": 0.38155221454693217,
|
| 672 |
+
"step": 854
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 62.0,
|
| 676 |
+
"eval_cer": 0.19126478616924478,
|
| 677 |
+
"eval_loss": 0.63057541847229,
|
| 678 |
+
"eval_runtime": 53.9032,
|
| 679 |
+
"eval_samples_per_second": 10.055,
|
| 680 |
+
"eval_steps_per_second": 1.262,
|
| 681 |
+
"eval_wer": 0.38013002844372207,
|
| 682 |
+
"step": 868
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"epoch": 63.0,
|
| 686 |
+
"eval_cer": 0.1915923566878981,
|
| 687 |
+
"eval_loss": 0.6432803869247437,
|
| 688 |
+
"eval_runtime": 53.9619,
|
| 689 |
+
"eval_samples_per_second": 10.044,
|
| 690 |
+
"eval_steps_per_second": 1.26,
|
| 691 |
+
"eval_wer": 0.37992685900040635,
|
| 692 |
+
"step": 882
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 64.0,
|
| 696 |
+
"eval_cer": 0.18955414012738853,
|
| 697 |
+
"eval_loss": 0.6342417001724243,
|
| 698 |
+
"eval_runtime": 53.7595,
|
| 699 |
+
"eval_samples_per_second": 10.082,
|
| 700 |
+
"eval_steps_per_second": 1.265,
|
| 701 |
+
"eval_wer": 0.38114587566030067,
|
| 702 |
+
"step": 896
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 64.29,
|
| 706 |
+
"learning_rate": 1.7185714285714285e-05,
|
| 707 |
+
"loss": 0.445,
|
| 708 |
+
"step": 900
|
| 709 |
+
},
|
| 710 |
+
{
|
| 711 |
+
"epoch": 65.0,
|
| 712 |
+
"eval_cer": 0.19042766151046406,
|
| 713 |
+
"eval_loss": 0.6211907267570496,
|
| 714 |
+
"eval_runtime": 53.9351,
|
| 715 |
+
"eval_samples_per_second": 10.049,
|
| 716 |
+
"eval_steps_per_second": 1.261,
|
| 717 |
+
"eval_wer": 0.38114587566030067,
|
| 718 |
+
"step": 910
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 66.0,
|
| 722 |
+
"eval_cer": 0.1894813466787989,
|
| 723 |
+
"eval_loss": 0.6164088249206543,
|
| 724 |
+
"eval_runtime": 53.6257,
|
| 725 |
+
"eval_samples_per_second": 10.107,
|
| 726 |
+
"eval_steps_per_second": 1.268,
|
| 727 |
+
"eval_wer": 0.3789110117838277,
|
| 728 |
+
"step": 924
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 67.0,
|
| 732 |
+
"eval_cer": 0.18715195632393083,
|
| 733 |
+
"eval_loss": 0.6005796194076538,
|
| 734 |
+
"eval_runtime": 53.6712,
|
| 735 |
+
"eval_samples_per_second": 10.099,
|
| 736 |
+
"eval_steps_per_second": 1.267,
|
| 737 |
+
"eval_wer": 0.3732222673709874,
|
| 738 |
+
"step": 938
|
| 739 |
+
},
|
| 740 |
+
{
|
| 741 |
+
"epoch": 68.0,
|
| 742 |
+
"eval_cer": 0.18911737943585077,
|
| 743 |
+
"eval_loss": 0.6054388880729675,
|
| 744 |
+
"eval_runtime": 53.6435,
|
| 745 |
+
"eval_samples_per_second": 10.104,
|
| 746 |
+
"eval_steps_per_second": 1.268,
|
| 747 |
+
"eval_wer": 0.3746444534741975,
|
| 748 |
+
"step": 952
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 69.0,
|
| 752 |
+
"eval_cer": 0.18940855323020928,
|
| 753 |
+
"eval_loss": 0.6244754791259766,
|
| 754 |
+
"eval_runtime": 53.8541,
|
| 755 |
+
"eval_samples_per_second": 10.064,
|
| 756 |
+
"eval_steps_per_second": 1.263,
|
| 757 |
+
"eval_wer": 0.3722064201544088,
|
| 758 |
+
"step": 966
|
| 759 |
+
},
|
| 760 |
+
{
|
| 761 |
+
"epoch": 70.0,
|
| 762 |
+
"eval_cer": 0.1878070973612375,
|
| 763 |
+
"eval_loss": 0.6090296506881714,
|
| 764 |
+
"eval_runtime": 53.9168,
|
| 765 |
+
"eval_samples_per_second": 10.053,
|
| 766 |
+
"eval_steps_per_second": 1.261,
|
| 767 |
+
"eval_wer": 0.36875253961804144,
|
| 768 |
+
"step": 980
|
| 769 |
+
},
|
| 770 |
+
{
|
| 771 |
+
"epoch": 71.0,
|
| 772 |
+
"eval_cer": 0.18758871701546861,
|
| 773 |
+
"eval_loss": 0.6072561144828796,
|
| 774 |
+
"eval_runtime": 53.6209,
|
| 775 |
+
"eval_samples_per_second": 10.108,
|
| 776 |
+
"eval_steps_per_second": 1.268,
|
| 777 |
+
"eval_wer": 0.3669240146281999,
|
| 778 |
+
"step": 994
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 71.43,
|
| 782 |
+
"learning_rate": 1.5757142857142858e-05,
|
| 783 |
+
"loss": 0.3746,
|
| 784 |
+
"step": 1000
|
| 785 |
+
},
|
| 786 |
+
{
|
| 787 |
+
"epoch": 72.0,
|
| 788 |
+
"eval_cer": 0.18886260236578709,
|
| 789 |
+
"eval_loss": 0.5989311337471008,
|
| 790 |
+
"eval_runtime": 53.765,
|
| 791 |
+
"eval_samples_per_second": 10.081,
|
| 792 |
+
"eval_steps_per_second": 1.265,
|
| 793 |
+
"eval_wer": 0.3707842340511987,
|
| 794 |
+
"step": 1008
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 73.0,
|
| 798 |
+
"eval_cer": 0.18744313011828936,
|
| 799 |
+
"eval_loss": 0.5967620611190796,
|
| 800 |
+
"eval_runtime": 53.6159,
|
| 801 |
+
"eval_samples_per_second": 10.109,
|
| 802 |
+
"eval_steps_per_second": 1.268,
|
| 803 |
+
"eval_wer": 0.3681430312880943,
|
| 804 |
+
"step": 1022
|
| 805 |
+
},
|
| 806 |
+
{
|
| 807 |
+
"epoch": 74.0,
|
| 808 |
+
"eval_cer": 0.1870427661510464,
|
| 809 |
+
"eval_loss": 0.5945658683776855,
|
| 810 |
+
"eval_runtime": 53.8501,
|
| 811 |
+
"eval_samples_per_second": 10.065,
|
| 812 |
+
"eval_steps_per_second": 1.263,
|
| 813 |
+
"eval_wer": 0.3659081674116213,
|
| 814 |
+
"step": 1036
|
| 815 |
+
},
|
| 816 |
+
{
|
| 817 |
+
"epoch": 75.0,
|
| 818 |
+
"eval_cer": 0.18638762511373977,
|
| 819 |
+
"eval_loss": 0.5873651504516602,
|
| 820 |
+
"eval_runtime": 53.5608,
|
| 821 |
+
"eval_samples_per_second": 10.119,
|
| 822 |
+
"eval_steps_per_second": 1.27,
|
| 823 |
+
"eval_wer": 0.3622511174319382,
|
| 824 |
+
"step": 1050
|
| 825 |
+
},
|
| 826 |
+
{
|
| 827 |
+
"epoch": 76.0,
|
| 828 |
+
"eval_cer": 0.1870427661510464,
|
| 829 |
+
"eval_loss": 0.5927525758743286,
|
| 830 |
+
"eval_runtime": 53.7753,
|
| 831 |
+
"eval_samples_per_second": 10.079,
|
| 832 |
+
"eval_steps_per_second": 1.265,
|
| 833 |
+
"eval_wer": 0.36387647297846404,
|
| 834 |
+
"step": 1064
|
| 835 |
+
},
|
| 836 |
+
{
|
| 837 |
+
"epoch": 77.0,
|
| 838 |
+
"eval_cer": 0.18824385805277524,
|
| 839 |
+
"eval_loss": 0.5889000296592712,
|
| 840 |
+
"eval_runtime": 53.8481,
|
| 841 |
+
"eval_samples_per_second": 10.065,
|
| 842 |
+
"eval_steps_per_second": 1.263,
|
| 843 |
+
"eval_wer": 0.3681430312880943,
|
| 844 |
+
"step": 1078
|
| 845 |
+
},
|
| 846 |
+
{
|
| 847 |
+
"epoch": 78.0,
|
| 848 |
+
"eval_cer": 0.18635122838944496,
|
| 849 |
+
"eval_loss": 0.5722672343254089,
|
| 850 |
+
"eval_runtime": 53.9948,
|
| 851 |
+
"eval_samples_per_second": 10.038,
|
| 852 |
+
"eval_steps_per_second": 1.259,
|
| 853 |
+
"eval_wer": 0.36834620073141,
|
| 854 |
+
"step": 1092
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 78.57,
|
| 858 |
+
"learning_rate": 1.4328571428571428e-05,
|
| 859 |
+
"loss": 0.3543,
|
| 860 |
+
"step": 1100
|
| 861 |
+
},
|
| 862 |
+
{
|
| 863 |
+
"epoch": 79.0,
|
| 864 |
+
"eval_cer": 0.18627843494085533,
|
| 865 |
+
"eval_loss": 0.5927807688713074,
|
| 866 |
+
"eval_runtime": 53.7175,
|
| 867 |
+
"eval_samples_per_second": 10.09,
|
| 868 |
+
"eval_steps_per_second": 1.266,
|
| 869 |
+
"eval_wer": 0.3657049979683056,
|
| 870 |
+
"step": 1106
|
| 871 |
+
},
|
| 872 |
+
{
|
| 873 |
+
"epoch": 80.0,
|
| 874 |
+
"eval_cer": 0.18547770700636942,
|
| 875 |
+
"eval_loss": 0.583193838596344,
|
| 876 |
+
"eval_runtime": 53.7886,
|
| 877 |
+
"eval_samples_per_second": 10.076,
|
| 878 |
+
"eval_steps_per_second": 1.264,
|
| 879 |
+
"eval_wer": 0.36489232019504264,
|
| 880 |
+
"step": 1120
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"epoch": 81.0,
|
| 884 |
+
"eval_cer": 0.18493175614194723,
|
| 885 |
+
"eval_loss": 0.5785495638847351,
|
| 886 |
+
"eval_runtime": 53.8447,
|
| 887 |
+
"eval_samples_per_second": 10.066,
|
| 888 |
+
"eval_steps_per_second": 1.263,
|
| 889 |
+
"eval_wer": 0.3644859813084112,
|
| 890 |
+
"step": 1134
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 82.0,
|
| 894 |
+
"eval_cer": 0.18424021838034577,
|
| 895 |
+
"eval_loss": 0.5877463817596436,
|
| 896 |
+
"eval_runtime": 53.7471,
|
| 897 |
+
"eval_samples_per_second": 10.084,
|
| 898 |
+
"eval_steps_per_second": 1.265,
|
| 899 |
+
"eval_wer": 0.35798455912230803,
|
| 900 |
+
"step": 1148
|
| 901 |
+
},
|
| 902 |
+
{
|
| 903 |
+
"epoch": 83.0,
|
| 904 |
+
"eval_cer": 0.18525932666060055,
|
| 905 |
+
"eval_loss": 0.5869647860527039,
|
| 906 |
+
"eval_runtime": 53.8204,
|
| 907 |
+
"eval_samples_per_second": 10.071,
|
| 908 |
+
"eval_steps_per_second": 1.263,
|
| 909 |
+
"eval_wer": 0.36265745631856966,
|
| 910 |
+
"step": 1162
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"epoch": 84.0,
|
| 914 |
+
"eval_cer": 0.18456778889899908,
|
| 915 |
+
"eval_loss": 0.5738150477409363,
|
| 916 |
+
"eval_runtime": 53.9516,
|
| 917 |
+
"eval_samples_per_second": 10.046,
|
| 918 |
+
"eval_steps_per_second": 1.26,
|
| 919 |
+
"eval_wer": 0.3618447785453068,
|
| 920 |
+
"step": 1176
|
| 921 |
+
},
|
| 922 |
+
{
|
| 923 |
+
"epoch": 85.0,
|
| 924 |
+
"eval_cer": 0.18158325750682439,
|
| 925 |
+
"eval_loss": 0.5640825033187866,
|
| 926 |
+
"eval_runtime": 53.9077,
|
| 927 |
+
"eval_samples_per_second": 10.054,
|
| 928 |
+
"eval_steps_per_second": 1.261,
|
| 929 |
+
"eval_wer": 0.35757822023567654,
|
| 930 |
+
"step": 1190
|
| 931 |
+
},
|
| 932 |
+
{
|
| 933 |
+
"epoch": 85.71,
|
| 934 |
+
"learning_rate": 1.29e-05,
|
| 935 |
+
"loss": 0.3207,
|
| 936 |
+
"step": 1200
|
| 937 |
+
},
|
| 938 |
+
{
|
| 939 |
+
"epoch": 86.0,
|
| 940 |
+
"eval_cer": 0.18209281164695176,
|
| 941 |
+
"eval_loss": 0.5727854371070862,
|
| 942 |
+
"eval_runtime": 54.0413,
|
| 943 |
+
"eval_samples_per_second": 10.029,
|
| 944 |
+
"eval_steps_per_second": 1.258,
|
| 945 |
+
"eval_wer": 0.35656237301909793,
|
| 946 |
+
"step": 1204
|
| 947 |
+
},
|
| 948 |
+
{
|
| 949 |
+
"epoch": 87.0,
|
| 950 |
+
"eval_cer": 0.18169244767970882,
|
| 951 |
+
"eval_loss": 0.5706486105918884,
|
| 952 |
+
"eval_runtime": 54.0703,
|
| 953 |
+
"eval_samples_per_second": 10.024,
|
| 954 |
+
"eval_steps_per_second": 1.258,
|
| 955 |
+
"eval_wer": 0.35595286468915077,
|
| 956 |
+
"step": 1218
|
| 957 |
+
},
|
| 958 |
+
{
|
| 959 |
+
"epoch": 88.0,
|
| 960 |
+
"eval_cer": 0.1813284804367607,
|
| 961 |
+
"eval_loss": 0.5607408881187439,
|
| 962 |
+
"eval_runtime": 54.3555,
|
| 963 |
+
"eval_samples_per_second": 9.971,
|
| 964 |
+
"eval_steps_per_second": 1.251,
|
| 965 |
+
"eval_wer": 0.3569687119057294,
|
| 966 |
+
"step": 1232
|
| 967 |
+
},
|
| 968 |
+
{
|
| 969 |
+
"epoch": 89.0,
|
| 970 |
+
"eval_cer": 0.181656050955414,
|
| 971 |
+
"eval_loss": 0.56435227394104,
|
| 972 |
+
"eval_runtime": 53.7806,
|
| 973 |
+
"eval_samples_per_second": 10.078,
|
| 974 |
+
"eval_steps_per_second": 1.264,
|
| 975 |
+
"eval_wer": 0.35574969524583505,
|
| 976 |
+
"step": 1246
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"epoch": 90.0,
|
| 980 |
+
"eval_cer": 0.1824203821656051,
|
| 981 |
+
"eval_loss": 0.5659990310668945,
|
| 982 |
+
"eval_runtime": 54.3449,
|
| 983 |
+
"eval_samples_per_second": 9.973,
|
| 984 |
+
"eval_steps_per_second": 1.251,
|
| 985 |
+
"eval_wer": 0.35818772856562375,
|
| 986 |
+
"step": 1260
|
| 987 |
+
},
|
| 988 |
+
{
|
| 989 |
+
"epoch": 91.0,
|
| 990 |
+
"eval_cer": 0.18285714285714286,
|
| 991 |
+
"eval_loss": 0.5688397884368896,
|
| 992 |
+
"eval_runtime": 53.7699,
|
| 993 |
+
"eval_samples_per_second": 10.08,
|
| 994 |
+
"eval_steps_per_second": 1.265,
|
| 995 |
+
"eval_wer": 0.35656237301909793,
|
| 996 |
+
"step": 1274
|
| 997 |
+
},
|
| 998 |
+
{
|
| 999 |
+
"epoch": 92.0,
|
| 1000 |
+
"eval_cer": 0.18074613284804367,
|
| 1001 |
+
"eval_loss": 0.563470184803009,
|
| 1002 |
+
"eval_runtime": 54.0176,
|
| 1003 |
+
"eval_samples_per_second": 10.034,
|
| 1004 |
+
"eval_steps_per_second": 1.259,
|
| 1005 |
+
"eval_wer": 0.35412433969930923,
|
| 1006 |
+
"step": 1288
|
| 1007 |
+
},
|
| 1008 |
+
{
|
| 1009 |
+
"epoch": 92.86,
|
| 1010 |
+
"learning_rate": 1.1471428571428572e-05,
|
| 1011 |
+
"loss": 0.2984,
|
| 1012 |
+
"step": 1300
|
| 1013 |
+
},
|
| 1014 |
+
{
|
| 1015 |
+
"epoch": 93.0,
|
| 1016 |
+
"eval_cer": 0.1813648771610555,
|
| 1017 |
+
"eval_loss": 0.5662950277328491,
|
| 1018 |
+
"eval_runtime": 54.0264,
|
| 1019 |
+
"eval_samples_per_second": 10.032,
|
| 1020 |
+
"eval_steps_per_second": 1.259,
|
| 1021 |
+
"eval_wer": 0.35026412027631043,
|
| 1022 |
+
"step": 1302
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 94.0,
|
| 1026 |
+
"eval_cer": 0.18074613284804367,
|
| 1027 |
+
"eval_loss": 0.5514569282531738,
|
| 1028 |
+
"eval_runtime": 53.9373,
|
| 1029 |
+
"eval_samples_per_second": 10.049,
|
| 1030 |
+
"eval_steps_per_second": 1.261,
|
| 1031 |
+
"eval_wer": 0.35432750914262495,
|
| 1032 |
+
"step": 1316
|
| 1033 |
+
},
|
| 1034 |
+
{
|
| 1035 |
+
"epoch": 95.0,
|
| 1036 |
+
"eval_cer": 0.1802729754322111,
|
| 1037 |
+
"eval_loss": 0.5563488602638245,
|
| 1038 |
+
"eval_runtime": 54.0134,
|
| 1039 |
+
"eval_samples_per_second": 10.035,
|
| 1040 |
+
"eval_steps_per_second": 1.259,
|
| 1041 |
+
"eval_wer": 0.35168630637952053,
|
| 1042 |
+
"step": 1330
|
| 1043 |
+
},
|
| 1044 |
+
{
|
| 1045 |
+
"epoch": 96.0,
|
| 1046 |
+
"eval_cer": 0.18089171974522292,
|
| 1047 |
+
"eval_loss": 0.5617594122886658,
|
| 1048 |
+
"eval_runtime": 54.0211,
|
| 1049 |
+
"eval_samples_per_second": 10.033,
|
| 1050 |
+
"eval_steps_per_second": 1.259,
|
| 1051 |
+
"eval_wer": 0.3508736286062576,
|
| 1052 |
+
"step": 1344
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 97.0,
|
| 1056 |
+
"eval_cer": 0.18074613284804367,
|
| 1057 |
+
"eval_loss": 0.5554497838020325,
|
| 1058 |
+
"eval_runtime": 54.2071,
|
| 1059 |
+
"eval_samples_per_second": 9.999,
|
| 1060 |
+
"eval_steps_per_second": 1.254,
|
| 1061 |
+
"eval_wer": 0.35168630637952053,
|
| 1062 |
+
"step": 1358
|
| 1063 |
+
},
|
| 1064 |
+
{
|
| 1065 |
+
"epoch": 98.0,
|
| 1066 |
+
"eval_cer": 0.18092811646951773,
|
| 1067 |
+
"eval_loss": 0.5605758428573608,
|
| 1068 |
+
"eval_runtime": 53.9801,
|
| 1069 |
+
"eval_samples_per_second": 10.041,
|
| 1070 |
+
"eval_steps_per_second": 1.26,
|
| 1071 |
+
"eval_wer": 0.35290532303941485,
|
| 1072 |
+
"step": 1372
|
| 1073 |
+
},
|
| 1074 |
+
{
|
| 1075 |
+
"epoch": 99.0,
|
| 1076 |
+
"eval_cer": 0.1813284804367607,
|
| 1077 |
+
"eval_loss": 0.5596539974212646,
|
| 1078 |
+
"eval_runtime": 53.7711,
|
| 1079 |
+
"eval_samples_per_second": 10.08,
|
| 1080 |
+
"eval_steps_per_second": 1.265,
|
| 1081 |
+
"eval_wer": 0.35107679804957337,
|
| 1082 |
+
"step": 1386
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"epoch": 100.0,
|
| 1086 |
+
"learning_rate": 1.0042857142857144e-05,
|
| 1087 |
+
"loss": 0.2622,
|
| 1088 |
+
"step": 1400
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"epoch": 100.0,
|
| 1092 |
+
"eval_cer": 0.18121929026387626,
|
| 1093 |
+
"eval_loss": 0.5628100037574768,
|
| 1094 |
+
"eval_runtime": 53.9102,
|
| 1095 |
+
"eval_samples_per_second": 10.054,
|
| 1096 |
+
"eval_steps_per_second": 1.261,
|
| 1097 |
+
"eval_wer": 0.35046728971962615,
|
| 1098 |
+
"step": 1400
|
| 1099 |
+
},
|
| 1100 |
+
{
|
| 1101 |
+
"epoch": 101.0,
|
| 1102 |
+
"eval_cer": 0.17990900818926298,
|
| 1103 |
+
"eval_loss": 0.5564116835594177,
|
| 1104 |
+
"eval_runtime": 54.03,
|
| 1105 |
+
"eval_samples_per_second": 10.031,
|
| 1106 |
+
"eval_steps_per_second": 1.259,
|
| 1107 |
+
"eval_wer": 0.34945144250304755,
|
| 1108 |
+
"step": 1414
|
| 1109 |
+
},
|
| 1110 |
+
{
|
| 1111 |
+
"epoch": 102.0,
|
| 1112 |
+
"eval_cer": 0.18114649681528663,
|
| 1113 |
+
"eval_loss": 0.5625892877578735,
|
| 1114 |
+
"eval_runtime": 53.8963,
|
| 1115 |
+
"eval_samples_per_second": 10.056,
|
| 1116 |
+
"eval_steps_per_second": 1.262,
|
| 1117 |
+
"eval_wer": 0.3484355952864689,
|
| 1118 |
+
"step": 1428
|
| 1119 |
+
},
|
| 1120 |
+
{
|
| 1121 |
+
"epoch": 103.0,
|
| 1122 |
+
"eval_cer": 0.1799818016378526,
|
| 1123 |
+
"eval_loss": 0.5555956959724426,
|
| 1124 |
+
"eval_runtime": 54.0019,
|
| 1125 |
+
"eval_samples_per_second": 10.037,
|
| 1126 |
+
"eval_steps_per_second": 1.259,
|
| 1127 |
+
"eval_wer": 0.34701340918325885,
|
| 1128 |
+
"step": 1442
|
| 1129 |
+
},
|
| 1130 |
+
{
|
| 1131 |
+
"epoch": 104.0,
|
| 1132 |
+
"eval_cer": 0.1799454049135578,
|
| 1133 |
+
"eval_loss": 0.5602657198905945,
|
| 1134 |
+
"eval_runtime": 53.864,
|
| 1135 |
+
"eval_samples_per_second": 10.062,
|
| 1136 |
+
"eval_steps_per_second": 1.262,
|
| 1137 |
+
"eval_wer": 0.3464039008533117,
|
| 1138 |
+
"step": 1456
|
| 1139 |
+
},
|
| 1140 |
+
{
|
| 1141 |
+
"epoch": 105.0,
|
| 1142 |
+
"eval_cer": 0.17987261146496816,
|
| 1143 |
+
"eval_loss": 0.5570572018623352,
|
| 1144 |
+
"eval_runtime": 53.8761,
|
| 1145 |
+
"eval_samples_per_second": 10.06,
|
| 1146 |
+
"eval_steps_per_second": 1.262,
|
| 1147 |
+
"eval_wer": 0.34538805363673303,
|
| 1148 |
+
"step": 1470
|
| 1149 |
+
},
|
| 1150 |
+
{
|
| 1151 |
+
"epoch": 106.0,
|
| 1152 |
+
"eval_cer": 0.17990900818926298,
|
| 1153 |
+
"eval_loss": 0.5618259310722351,
|
| 1154 |
+
"eval_runtime": 53.8537,
|
| 1155 |
+
"eval_samples_per_second": 10.064,
|
| 1156 |
+
"eval_steps_per_second": 1.263,
|
| 1157 |
+
"eval_wer": 0.3466070702966274,
|
| 1158 |
+
"step": 1484
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"epoch": 107.0,
|
| 1162 |
+
"eval_cer": 0.17867151956323932,
|
| 1163 |
+
"eval_loss": 0.5518925786018372,
|
| 1164 |
+
"eval_runtime": 54.2003,
|
| 1165 |
+
"eval_samples_per_second": 10.0,
|
| 1166 |
+
"eval_steps_per_second": 1.255,
|
| 1167 |
+
"eval_wer": 0.343965867533523,
|
| 1168 |
+
"step": 1498
|
| 1169 |
+
},
|
| 1170 |
+
{
|
| 1171 |
+
"epoch": 107.14,
|
| 1172 |
+
"learning_rate": 8.614285714285714e-06,
|
| 1173 |
+
"loss": 0.2519,
|
| 1174 |
+
"step": 1500
|
| 1175 |
+
},
|
| 1176 |
+
{
|
| 1177 |
+
"epoch": 108.0,
|
| 1178 |
+
"eval_cer": 0.17903548680618744,
|
| 1179 |
+
"eval_loss": 0.5540814399719238,
|
| 1180 |
+
"eval_runtime": 53.903,
|
| 1181 |
+
"eval_samples_per_second": 10.055,
|
| 1182 |
+
"eval_steps_per_second": 1.262,
|
| 1183 |
+
"eval_wer": 0.343965867533523,
|
| 1184 |
+
"step": 1512
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"epoch": 109.0,
|
| 1188 |
+
"eval_cer": 0.17939945404913557,
|
| 1189 |
+
"eval_loss": 0.557438850402832,
|
| 1190 |
+
"eval_runtime": 54.7258,
|
| 1191 |
+
"eval_samples_per_second": 9.904,
|
| 1192 |
+
"eval_steps_per_second": 1.243,
|
| 1193 |
+
"eval_wer": 0.3464039008533117,
|
| 1194 |
+
"step": 1526
|
| 1195 |
+
},
|
| 1196 |
+
{
|
| 1197 |
+
"epoch": 110.0,
|
| 1198 |
+
"eval_cer": 0.18009099181073704,
|
| 1199 |
+
"eval_loss": 0.5590262413024902,
|
| 1200 |
+
"eval_runtime": 54.0902,
|
| 1201 |
+
"eval_samples_per_second": 10.02,
|
| 1202 |
+
"eval_steps_per_second": 1.257,
|
| 1203 |
+
"eval_wer": 0.34538805363673303,
|
| 1204 |
+
"step": 1540
|
| 1205 |
+
},
|
| 1206 |
+
{
|
| 1207 |
+
"epoch": 111.0,
|
| 1208 |
+
"eval_cer": 0.17958143767060963,
|
| 1209 |
+
"eval_loss": 0.5529592633247375,
|
| 1210 |
+
"eval_runtime": 53.9289,
|
| 1211 |
+
"eval_samples_per_second": 10.05,
|
| 1212 |
+
"eval_steps_per_second": 1.261,
|
| 1213 |
+
"eval_wer": 0.34477854530678587,
|
| 1214 |
+
"step": 1554
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"epoch": 112.0,
|
| 1218 |
+
"eval_cer": 0.1791810737033667,
|
| 1219 |
+
"eval_loss": 0.5500696301460266,
|
| 1220 |
+
"eval_runtime": 53.992,
|
| 1221 |
+
"eval_samples_per_second": 10.039,
|
| 1222 |
+
"eval_steps_per_second": 1.259,
|
| 1223 |
+
"eval_wer": 0.3437626980902072,
|
| 1224 |
+
"step": 1568
|
| 1225 |
+
},
|
| 1226 |
+
{
|
| 1227 |
+
"epoch": 113.0,
|
| 1228 |
+
"eval_cer": 0.17990900818926298,
|
| 1229 |
+
"eval_loss": 0.5595071911811829,
|
| 1230 |
+
"eval_runtime": 54.515,
|
| 1231 |
+
"eval_samples_per_second": 9.942,
|
| 1232 |
+
"eval_steps_per_second": 1.247,
|
| 1233 |
+
"eval_wer": 0.34477854530678587,
|
| 1234 |
+
"step": 1582
|
| 1235 |
+
},
|
| 1236 |
+
{
|
| 1237 |
+
"epoch": 114.0,
|
| 1238 |
+
"eval_cer": 0.18005459508644223,
|
| 1239 |
+
"eval_loss": 0.5535922646522522,
|
| 1240 |
+
"eval_runtime": 53.9067,
|
| 1241 |
+
"eval_samples_per_second": 10.054,
|
| 1242 |
+
"eval_steps_per_second": 1.261,
|
| 1243 |
+
"eval_wer": 0.34457537586347015,
|
| 1244 |
+
"step": 1596
|
| 1245 |
+
},
|
| 1246 |
+
{
|
| 1247 |
+
"epoch": 114.29,
|
| 1248 |
+
"learning_rate": 7.185714285714286e-06,
|
| 1249 |
+
"loss": 0.245,
|
| 1250 |
+
"step": 1600
|
| 1251 |
+
},
|
| 1252 |
+
{
|
| 1253 |
+
"epoch": 115.0,
|
| 1254 |
+
"eval_cer": 0.1788898999090082,
|
| 1255 |
+
"eval_loss": 0.5480047464370728,
|
| 1256 |
+
"eval_runtime": 53.9861,
|
| 1257 |
+
"eval_samples_per_second": 10.04,
|
| 1258 |
+
"eval_steps_per_second": 1.26,
|
| 1259 |
+
"eval_wer": 0.34315318976026005,
|
| 1260 |
+
"step": 1610
|
| 1261 |
+
},
|
| 1262 |
+
{
|
| 1263 |
+
"epoch": 116.0,
|
| 1264 |
+
"eval_cer": 0.17979981801637854,
|
| 1265 |
+
"eval_loss": 0.562286913394928,
|
| 1266 |
+
"eval_runtime": 53.656,
|
| 1267 |
+
"eval_samples_per_second": 10.101,
|
| 1268 |
+
"eval_steps_per_second": 1.267,
|
| 1269 |
+
"eval_wer": 0.34863876472978467,
|
| 1270 |
+
"step": 1624
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"epoch": 117.0,
|
| 1274 |
+
"eval_cer": 0.17903548680618744,
|
| 1275 |
+
"eval_loss": 0.549622118473053,
|
| 1276 |
+
"eval_runtime": 53.5772,
|
| 1277 |
+
"eval_samples_per_second": 10.116,
|
| 1278 |
+
"eval_steps_per_second": 1.269,
|
| 1279 |
+
"eval_wer": 0.3427468508736286,
|
| 1280 |
+
"step": 1638
|
| 1281 |
+
},
|
| 1282 |
+
{
|
| 1283 |
+
"epoch": 118.0,
|
| 1284 |
+
"eval_cer": 0.17885350318471338,
|
| 1285 |
+
"eval_loss": 0.5551720857620239,
|
| 1286 |
+
"eval_runtime": 53.9431,
|
| 1287 |
+
"eval_samples_per_second": 10.048,
|
| 1288 |
+
"eval_steps_per_second": 1.261,
|
| 1289 |
+
"eval_wer": 0.34213734254368144,
|
| 1290 |
+
"step": 1652
|
| 1291 |
+
},
|
| 1292 |
+
{
|
| 1293 |
+
"epoch": 119.0,
|
| 1294 |
+
"eval_cer": 0.17867151956323932,
|
| 1295 |
+
"eval_loss": 0.5557767152786255,
|
| 1296 |
+
"eval_runtime": 54.2974,
|
| 1297 |
+
"eval_samples_per_second": 9.982,
|
| 1298 |
+
"eval_steps_per_second": 1.252,
|
| 1299 |
+
"eval_wer": 0.3437626980902072,
|
| 1300 |
+
"step": 1666
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"epoch": 120.0,
|
| 1304 |
+
"eval_cer": 0.17834394904458598,
|
| 1305 |
+
"eval_loss": 0.552444338798523,
|
| 1306 |
+
"eval_runtime": 53.9105,
|
| 1307 |
+
"eval_samples_per_second": 10.054,
|
| 1308 |
+
"eval_steps_per_second": 1.261,
|
| 1309 |
+
"eval_wer": 0.3425436814303129,
|
| 1310 |
+
"step": 1680
|
| 1311 |
+
},
|
| 1312 |
+
{
|
| 1313 |
+
"epoch": 121.0,
|
| 1314 |
+
"eval_cer": 0.1785987261146497,
|
| 1315 |
+
"eval_loss": 0.5581905841827393,
|
| 1316 |
+
"eval_runtime": 53.9834,
|
| 1317 |
+
"eval_samples_per_second": 10.04,
|
| 1318 |
+
"eval_steps_per_second": 1.26,
|
| 1319 |
+
"eval_wer": 0.34213734254368144,
|
| 1320 |
+
"step": 1694
|
| 1321 |
+
},
|
| 1322 |
+
{
|
| 1323 |
+
"epoch": 121.43,
|
| 1324 |
+
"learning_rate": 5.7571428571428574e-06,
|
| 1325 |
+
"loss": 0.2322,
|
| 1326 |
+
"step": 1700
|
| 1327 |
+
},
|
| 1328 |
+
{
|
| 1329 |
+
"epoch": 122.0,
|
| 1330 |
+
"eval_cer": 0.17856232939035488,
|
| 1331 |
+
"eval_loss": 0.5533866286277771,
|
| 1332 |
+
"eval_runtime": 54.0755,
|
| 1333 |
+
"eval_samples_per_second": 10.023,
|
| 1334 |
+
"eval_steps_per_second": 1.258,
|
| 1335 |
+
"eval_wer": 0.3425436814303129,
|
| 1336 |
+
"step": 1708
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"epoch": 123.0,
|
| 1340 |
+
"eval_cer": 0.18012738853503185,
|
| 1341 |
+
"eval_loss": 0.5595676302909851,
|
| 1342 |
+
"eval_runtime": 53.7843,
|
| 1343 |
+
"eval_samples_per_second": 10.077,
|
| 1344 |
+
"eval_steps_per_second": 1.264,
|
| 1345 |
+
"eval_wer": 0.3464039008533117,
|
| 1346 |
+
"step": 1722
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 124.0,
|
| 1350 |
+
"eval_cer": 0.17899909008189263,
|
| 1351 |
+
"eval_loss": 0.5486122369766235,
|
| 1352 |
+
"eval_runtime": 53.9909,
|
| 1353 |
+
"eval_samples_per_second": 10.039,
|
| 1354 |
+
"eval_steps_per_second": 1.259,
|
| 1355 |
+
"eval_wer": 0.34315318976026005,
|
| 1356 |
+
"step": 1736
|
| 1357 |
+
},
|
| 1358 |
+
{
|
| 1359 |
+
"epoch": 125.0,
|
| 1360 |
+
"eval_cer": 0.1792174704276615,
|
| 1361 |
+
"eval_loss": 0.5580961108207703,
|
| 1362 |
+
"eval_runtime": 53.9973,
|
| 1363 |
+
"eval_samples_per_second": 10.038,
|
| 1364 |
+
"eval_steps_per_second": 1.259,
|
| 1365 |
+
"eval_wer": 0.3425436814303129,
|
| 1366 |
+
"step": 1750
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"epoch": 126.0,
|
| 1370 |
+
"eval_cer": 0.17852593266606007,
|
| 1371 |
+
"eval_loss": 0.5470190644264221,
|
| 1372 |
+
"eval_runtime": 53.6657,
|
| 1373 |
+
"eval_samples_per_second": 10.1,
|
| 1374 |
+
"eval_steps_per_second": 1.267,
|
| 1375 |
+
"eval_wer": 0.34173100365705,
|
| 1376 |
+
"step": 1764
|
| 1377 |
+
},
|
| 1378 |
+
{
|
| 1379 |
+
"epoch": 127.0,
|
| 1380 |
+
"eval_cer": 0.17805277525022747,
|
| 1381 |
+
"eval_loss": 0.5543853640556335,
|
| 1382 |
+
"eval_runtime": 53.9507,
|
| 1383 |
+
"eval_samples_per_second": 10.046,
|
| 1384 |
+
"eval_steps_per_second": 1.26,
|
| 1385 |
+
"eval_wer": 0.3413246647704185,
|
| 1386 |
+
"step": 1778
|
| 1387 |
+
},
|
| 1388 |
+
{
|
| 1389 |
+
"epoch": 128.0,
|
| 1390 |
+
"eval_cer": 0.1780891719745223,
|
| 1391 |
+
"eval_loss": 0.5500943660736084,
|
| 1392 |
+
"eval_runtime": 53.7085,
|
| 1393 |
+
"eval_samples_per_second": 10.092,
|
| 1394 |
+
"eval_steps_per_second": 1.266,
|
| 1395 |
+
"eval_wer": 0.3435595286468915,
|
| 1396 |
+
"step": 1792
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"epoch": 128.57,
|
| 1400 |
+
"learning_rate": 4.328571428571429e-06,
|
| 1401 |
+
"loss": 0.2324,
|
| 1402 |
+
"step": 1800
|
| 1403 |
+
},
|
| 1404 |
+
{
|
| 1405 |
+
"epoch": 129.0,
|
| 1406 |
+
"eval_cer": 0.1781619654231119,
|
| 1407 |
+
"eval_loss": 0.5517907738685608,
|
| 1408 |
+
"eval_runtime": 54.1553,
|
| 1409 |
+
"eval_samples_per_second": 10.008,
|
| 1410 |
+
"eval_steps_per_second": 1.256,
|
| 1411 |
+
"eval_wer": 0.343965867533523,
|
| 1412 |
+
"step": 1806
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 130.0,
|
| 1416 |
+
"eval_cer": 0.1775432211101001,
|
| 1417 |
+
"eval_loss": 0.551121711730957,
|
| 1418 |
+
"eval_runtime": 54.0869,
|
| 1419 |
+
"eval_samples_per_second": 10.021,
|
| 1420 |
+
"eval_steps_per_second": 1.257,
|
| 1421 |
+
"eval_wer": 0.3388866314506298,
|
| 1422 |
+
"step": 1820
|
| 1423 |
+
},
|
| 1424 |
+
{
|
| 1425 |
+
"epoch": 131.0,
|
| 1426 |
+
"eval_cer": 0.17823475887170154,
|
| 1427 |
+
"eval_loss": 0.5584121942520142,
|
| 1428 |
+
"eval_runtime": 54.0126,
|
| 1429 |
+
"eval_samples_per_second": 10.035,
|
| 1430 |
+
"eval_steps_per_second": 1.259,
|
| 1431 |
+
"eval_wer": 0.34173100365705,
|
| 1432 |
+
"step": 1834
|
| 1433 |
+
},
|
| 1434 |
+
{
|
| 1435 |
+
"epoch": 132.0,
|
| 1436 |
+
"eval_cer": 0.17747042766151047,
|
| 1437 |
+
"eval_loss": 0.549301028251648,
|
| 1438 |
+
"eval_runtime": 54.1977,
|
| 1439 |
+
"eval_samples_per_second": 10.0,
|
| 1440 |
+
"eval_steps_per_second": 1.255,
|
| 1441 |
+
"eval_wer": 0.33726127590410404,
|
| 1442 |
+
"step": 1848
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 133.0,
|
| 1446 |
+
"eval_cer": 0.17768880800727935,
|
| 1447 |
+
"eval_loss": 0.5506242513656616,
|
| 1448 |
+
"eval_runtime": 54.3838,
|
| 1449 |
+
"eval_samples_per_second": 9.966,
|
| 1450 |
+
"eval_steps_per_second": 1.25,
|
| 1451 |
+
"eval_wer": 0.339496139780577,
|
| 1452 |
+
"step": 1862
|
| 1453 |
+
},
|
| 1454 |
+
{
|
| 1455 |
+
"epoch": 134.0,
|
| 1456 |
+
"eval_cer": 0.1781619654231119,
|
| 1457 |
+
"eval_loss": 0.5543136596679688,
|
| 1458 |
+
"eval_runtime": 53.8876,
|
| 1459 |
+
"eval_samples_per_second": 10.058,
|
| 1460 |
+
"eval_steps_per_second": 1.262,
|
| 1461 |
+
"eval_wer": 0.34091832588378707,
|
| 1462 |
+
"step": 1876
|
| 1463 |
+
},
|
| 1464 |
+
{
|
| 1465 |
+
"epoch": 135.0,
|
| 1466 |
+
"eval_cer": 0.17805277525022747,
|
| 1467 |
+
"eval_loss": 0.5588898062705994,
|
| 1468 |
+
"eval_runtime": 53.8452,
|
| 1469 |
+
"eval_samples_per_second": 10.066,
|
| 1470 |
+
"eval_steps_per_second": 1.263,
|
| 1471 |
+
"eval_wer": 0.33990247866720846,
|
| 1472 |
+
"step": 1890
|
| 1473 |
+
},
|
| 1474 |
+
{
|
| 1475 |
+
"epoch": 135.71,
|
| 1476 |
+
"learning_rate": 2.9e-06,
|
| 1477 |
+
"loss": 0.2077,
|
| 1478 |
+
"step": 1900
|
| 1479 |
+
},
|
| 1480 |
+
{
|
| 1481 |
+
"epoch": 136.0,
|
| 1482 |
+
"eval_cer": 0.17776160145586897,
|
| 1483 |
+
"eval_loss": 0.5555850267410278,
|
| 1484 |
+
"eval_runtime": 54.3429,
|
| 1485 |
+
"eval_samples_per_second": 9.974,
|
| 1486 |
+
"eval_steps_per_second": 1.251,
|
| 1487 |
+
"eval_wer": 0.3390898008939455,
|
| 1488 |
+
"step": 1904
|
| 1489 |
+
},
|
| 1490 |
+
{
|
| 1491 |
+
"epoch": 137.0,
|
| 1492 |
+
"eval_cer": 0.17779799818016379,
|
| 1493 |
+
"eval_loss": 0.5555258393287659,
|
| 1494 |
+
"eval_runtime": 53.6523,
|
| 1495 |
+
"eval_samples_per_second": 10.102,
|
| 1496 |
+
"eval_steps_per_second": 1.267,
|
| 1497 |
+
"eval_wer": 0.34071515644047135,
|
| 1498 |
+
"step": 1918
|
| 1499 |
+
},
|
| 1500 |
+
{
|
| 1501 |
+
"epoch": 138.0,
|
| 1502 |
+
"eval_cer": 0.17739763421292085,
|
| 1503 |
+
"eval_loss": 0.5501488447189331,
|
| 1504 |
+
"eval_runtime": 53.6556,
|
| 1505 |
+
"eval_samples_per_second": 10.101,
|
| 1506 |
+
"eval_steps_per_second": 1.267,
|
| 1507 |
+
"eval_wer": 0.3390898008939455,
|
| 1508 |
+
"step": 1932
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"epoch": 139.0,
|
| 1512 |
+
"eval_cer": 0.17721565059144678,
|
| 1513 |
+
"eval_loss": 0.5544245839118958,
|
| 1514 |
+
"eval_runtime": 54.0174,
|
| 1515 |
+
"eval_samples_per_second": 10.034,
|
| 1516 |
+
"eval_steps_per_second": 1.259,
|
| 1517 |
+
"eval_wer": 0.33746444534741976,
|
| 1518 |
+
"step": 1946
|
| 1519 |
+
},
|
| 1520 |
+
{
|
| 1521 |
+
"epoch": 140.0,
|
| 1522 |
+
"eval_cer": 0.17732484076433122,
|
| 1523 |
+
"eval_loss": 0.5553618669509888,
|
| 1524 |
+
"eval_runtime": 54.2475,
|
| 1525 |
+
"eval_samples_per_second": 9.991,
|
| 1526 |
+
"eval_steps_per_second": 1.254,
|
| 1527 |
+
"eval_wer": 0.3386834620073141,
|
| 1528 |
+
"step": 1960
|
| 1529 |
+
},
|
| 1530 |
+
{
|
| 1531 |
+
"epoch": 141.0,
|
| 1532 |
+
"eval_cer": 0.17721565059144678,
|
| 1533 |
+
"eval_loss": 0.5504088401794434,
|
| 1534 |
+
"eval_runtime": 54.07,
|
| 1535 |
+
"eval_samples_per_second": 10.024,
|
| 1536 |
+
"eval_steps_per_second": 1.258,
|
| 1537 |
+
"eval_wer": 0.3380739536773669,
|
| 1538 |
+
"step": 1974
|
| 1539 |
+
},
|
| 1540 |
+
{
|
| 1541 |
+
"epoch": 142.0,
|
| 1542 |
+
"eval_cer": 0.17699727024567788,
|
| 1543 |
+
"eval_loss": 0.5484427809715271,
|
| 1544 |
+
"eval_runtime": 54.0621,
|
| 1545 |
+
"eval_samples_per_second": 10.026,
|
| 1546 |
+
"eval_steps_per_second": 1.258,
|
| 1547 |
+
"eval_wer": 0.33827712312068264,
|
| 1548 |
+
"step": 1988
|
| 1549 |
+
},
|
| 1550 |
+
{
|
| 1551 |
+
"epoch": 142.86,
|
| 1552 |
+
"learning_rate": 1.4714285714285716e-06,
|
| 1553 |
+
"loss": 0.2089,
|
| 1554 |
+
"step": 2000
|
| 1555 |
+
},
|
| 1556 |
+
{
|
| 1557 |
+
"epoch": 143.0,
|
| 1558 |
+
"eval_cer": 0.17721565059144678,
|
| 1559 |
+
"eval_loss": 0.5519289374351501,
|
| 1560 |
+
"eval_runtime": 54.2165,
|
| 1561 |
+
"eval_samples_per_second": 9.997,
|
| 1562 |
+
"eval_steps_per_second": 1.254,
|
| 1563 |
+
"eval_wer": 0.33848029256399836,
|
| 1564 |
+
"step": 2002
|
| 1565 |
+
},
|
| 1566 |
+
{
|
| 1567 |
+
"epoch": 144.0,
|
| 1568 |
+
"eval_cer": 0.17721565059144678,
|
| 1569 |
+
"eval_loss": 0.5531746745109558,
|
| 1570 |
+
"eval_runtime": 54.0827,
|
| 1571 |
+
"eval_samples_per_second": 10.022,
|
| 1572 |
+
"eval_steps_per_second": 1.257,
|
| 1573 |
+
"eval_wer": 0.3390898008939455,
|
| 1574 |
+
"step": 2016
|
| 1575 |
+
},
|
| 1576 |
+
{
|
| 1577 |
+
"epoch": 145.0,
|
| 1578 |
+
"eval_cer": 0.17747042766151047,
|
| 1579 |
+
"eval_loss": 0.5529664158821106,
|
| 1580 |
+
"eval_runtime": 54.0306,
|
| 1581 |
+
"eval_samples_per_second": 10.031,
|
| 1582 |
+
"eval_steps_per_second": 1.259,
|
| 1583 |
+
"eval_wer": 0.33969930922389274,
|
| 1584 |
+
"step": 2030
|
| 1585 |
+
},
|
| 1586 |
+
{
|
| 1587 |
+
"epoch": 146.0,
|
| 1588 |
+
"eval_cer": 0.1775432211101001,
|
| 1589 |
+
"eval_loss": 0.5551361441612244,
|
| 1590 |
+
"eval_runtime": 54.0882,
|
| 1591 |
+
"eval_samples_per_second": 10.021,
|
| 1592 |
+
"eval_steps_per_second": 1.257,
|
| 1593 |
+
"eval_wer": 0.33969930922389274,
|
| 1594 |
+
"step": 2044
|
| 1595 |
+
},
|
| 1596 |
+
{
|
| 1597 |
+
"epoch": 146.0,
|
| 1598 |
+
"step": 2044,
|
| 1599 |
+
"total_flos": 1.8244410804942518e+19,
|
| 1600 |
+
"train_loss": 1.3513914437676595,
|
| 1601 |
+
"train_runtime": 31402.8433,
|
| 1602 |
+
"train_samples_per_second": 2.14,
|
| 1603 |
+
"train_steps_per_second": 0.067
|
| 1604 |
+
}
|
| 1605 |
+
],
|
| 1606 |
+
"max_steps": 2100,
|
| 1607 |
+
"num_train_epochs": 150,
|
| 1608 |
+
"total_flos": 1.8244410804942518e+19,
|
| 1609 |
+
"trial_name": null,
|
| 1610 |
+
"trial_params": null
|
| 1611 |
+
}
|