Commit
·
167ee21
1
Parent(s):
9878e25
End of training
Browse files- all_results.json +16 -0
- eval_results.json +10 -0
- train_results.json +9 -0
- trainer_state.json +1413 -0
all_results.json
ADDED
|
@@ -0,0 +1,16 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 128.0,
|
| 3 |
+
"eval_cer": 0.17881710646041857,
|
| 4 |
+
"eval_loss": 0.5552828907966614,
|
| 5 |
+
"eval_runtime": 52.7828,
|
| 6 |
+
"eval_samples": 542,
|
| 7 |
+
"eval_samples_per_second": 10.269,
|
| 8 |
+
"eval_steps_per_second": 1.288,
|
| 9 |
+
"eval_wer": 0.3466070702966274,
|
| 10 |
+
"total_flos": 1.5995217836155625e+19,
|
| 11 |
+
"train_loss": 1.3990157055003303,
|
| 12 |
+
"train_runtime": 26333.2616,
|
| 13 |
+
"train_samples": 448,
|
| 14 |
+
"train_samples_per_second": 2.552,
|
| 15 |
+
"train_steps_per_second": 0.08
|
| 16 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,10 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 128.0,
|
| 3 |
+
"eval_cer": 0.17881710646041857,
|
| 4 |
+
"eval_loss": 0.5552828907966614,
|
| 5 |
+
"eval_runtime": 52.7828,
|
| 6 |
+
"eval_samples": 542,
|
| 7 |
+
"eval_samples_per_second": 10.269,
|
| 8 |
+
"eval_steps_per_second": 1.288,
|
| 9 |
+
"eval_wer": 0.3466070702966274
|
| 10 |
+
}
|
train_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 128.0,
|
| 3 |
+
"total_flos": 1.5995217836155625e+19,
|
| 4 |
+
"train_loss": 1.3990157055003303,
|
| 5 |
+
"train_runtime": 26333.2616,
|
| 6 |
+
"train_samples": 448,
|
| 7 |
+
"train_samples_per_second": 2.552,
|
| 8 |
+
"train_steps_per_second": 0.08
|
| 9 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,1413 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 0.5552828907966614,
|
| 3 |
+
"best_model_checkpoint": "wav2vec2-large-xlsr-coraa-exp-9/checkpoint-1512",
|
| 4 |
+
"epoch": 128.0,
|
| 5 |
+
"global_step": 1792,
|
| 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": 37.5508,
|
| 14 |
+
"step": 1
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 1.0,
|
| 18 |
+
"eval_cer": 0.9618562329390354,
|
| 19 |
+
"eval_loss": 23.137550354003906,
|
| 20 |
+
"eval_runtime": 52.9452,
|
| 21 |
+
"eval_samples_per_second": 10.237,
|
| 22 |
+
"eval_steps_per_second": 1.284,
|
| 23 |
+
"eval_wer": 1.0,
|
| 24 |
+
"step": 14
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"epoch": 2.0,
|
| 28 |
+
"eval_cer": 0.9618562329390354,
|
| 29 |
+
"eval_loss": 6.50360107421875,
|
| 30 |
+
"eval_runtime": 52.53,
|
| 31 |
+
"eval_samples_per_second": 10.318,
|
| 32 |
+
"eval_steps_per_second": 1.294,
|
| 33 |
+
"eval_wer": 1.0,
|
| 34 |
+
"step": 28
|
| 35 |
+
},
|
| 36 |
+
{
|
| 37 |
+
"epoch": 3.0,
|
| 38 |
+
"eval_cer": 0.9618562329390354,
|
| 39 |
+
"eval_loss": 4.391931056976318,
|
| 40 |
+
"eval_runtime": 52.983,
|
| 41 |
+
"eval_samples_per_second": 10.23,
|
| 42 |
+
"eval_steps_per_second": 1.283,
|
| 43 |
+
"eval_wer": 1.0,
|
| 44 |
+
"step": 42
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 4.0,
|
| 48 |
+
"eval_cer": 0.9618562329390354,
|
| 49 |
+
"eval_loss": 3.944124221801758,
|
| 50 |
+
"eval_runtime": 52.4812,
|
| 51 |
+
"eval_samples_per_second": 10.328,
|
| 52 |
+
"eval_steps_per_second": 1.296,
|
| 53 |
+
"eval_wer": 1.0,
|
| 54 |
+
"step": 56
|
| 55 |
+
},
|
| 56 |
+
{
|
| 57 |
+
"epoch": 5.0,
|
| 58 |
+
"eval_cer": 0.9618562329390354,
|
| 59 |
+
"eval_loss": 3.730649709701538,
|
| 60 |
+
"eval_runtime": 52.6687,
|
| 61 |
+
"eval_samples_per_second": 10.291,
|
| 62 |
+
"eval_steps_per_second": 1.291,
|
| 63 |
+
"eval_wer": 1.0,
|
| 64 |
+
"step": 70
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"epoch": 6.0,
|
| 68 |
+
"eval_cer": 0.9618562329390354,
|
| 69 |
+
"eval_loss": 3.5761733055114746,
|
| 70 |
+
"eval_runtime": 52.4991,
|
| 71 |
+
"eval_samples_per_second": 10.324,
|
| 72 |
+
"eval_steps_per_second": 1.295,
|
| 73 |
+
"eval_wer": 1.0,
|
| 74 |
+
"step": 84
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 7.0,
|
| 78 |
+
"eval_cer": 0.9618562329390354,
|
| 79 |
+
"eval_loss": 3.4128713607788086,
|
| 80 |
+
"eval_runtime": 52.9491,
|
| 81 |
+
"eval_samples_per_second": 10.236,
|
| 82 |
+
"eval_steps_per_second": 1.284,
|
| 83 |
+
"eval_wer": 1.0,
|
| 84 |
+
"step": 98
|
| 85 |
+
},
|
| 86 |
+
{
|
| 87 |
+
"epoch": 7.14,
|
| 88 |
+
"learning_rate": 2.8614285714285714e-05,
|
| 89 |
+
"loss": 8.6902,
|
| 90 |
+
"step": 100
|
| 91 |
+
},
|
| 92 |
+
{
|
| 93 |
+
"epoch": 8.0,
|
| 94 |
+
"eval_cer": 0.9618562329390354,
|
| 95 |
+
"eval_loss": 3.285918951034546,
|
| 96 |
+
"eval_runtime": 52.5221,
|
| 97 |
+
"eval_samples_per_second": 10.319,
|
| 98 |
+
"eval_steps_per_second": 1.295,
|
| 99 |
+
"eval_wer": 1.0,
|
| 100 |
+
"step": 112
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 9.0,
|
| 104 |
+
"eval_cer": 0.9618562329390354,
|
| 105 |
+
"eval_loss": 3.219179391860962,
|
| 106 |
+
"eval_runtime": 53.0718,
|
| 107 |
+
"eval_samples_per_second": 10.213,
|
| 108 |
+
"eval_steps_per_second": 1.281,
|
| 109 |
+
"eval_wer": 1.0,
|
| 110 |
+
"step": 126
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 10.0,
|
| 114 |
+
"eval_cer": 0.9618562329390354,
|
| 115 |
+
"eval_loss": 3.1479437351226807,
|
| 116 |
+
"eval_runtime": 52.485,
|
| 117 |
+
"eval_samples_per_second": 10.327,
|
| 118 |
+
"eval_steps_per_second": 1.296,
|
| 119 |
+
"eval_wer": 1.0,
|
| 120 |
+
"step": 140
|
| 121 |
+
},
|
| 122 |
+
{
|
| 123 |
+
"epoch": 11.0,
|
| 124 |
+
"eval_cer": 0.9618562329390354,
|
| 125 |
+
"eval_loss": 3.106292963027954,
|
| 126 |
+
"eval_runtime": 52.586,
|
| 127 |
+
"eval_samples_per_second": 10.307,
|
| 128 |
+
"eval_steps_per_second": 1.293,
|
| 129 |
+
"eval_wer": 1.0,
|
| 130 |
+
"step": 154
|
| 131 |
+
},
|
| 132 |
+
{
|
| 133 |
+
"epoch": 12.0,
|
| 134 |
+
"eval_cer": 0.9618562329390354,
|
| 135 |
+
"eval_loss": 3.0896878242492676,
|
| 136 |
+
"eval_runtime": 52.4874,
|
| 137 |
+
"eval_samples_per_second": 10.326,
|
| 138 |
+
"eval_steps_per_second": 1.296,
|
| 139 |
+
"eval_wer": 1.0,
|
| 140 |
+
"step": 168
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 13.0,
|
| 144 |
+
"eval_cer": 0.9618562329390354,
|
| 145 |
+
"eval_loss": 3.084855794906616,
|
| 146 |
+
"eval_runtime": 52.6206,
|
| 147 |
+
"eval_samples_per_second": 10.3,
|
| 148 |
+
"eval_steps_per_second": 1.292,
|
| 149 |
+
"eval_wer": 1.0,
|
| 150 |
+
"step": 182
|
| 151 |
+
},
|
| 152 |
+
{
|
| 153 |
+
"epoch": 14.0,
|
| 154 |
+
"eval_cer": 0.9618562329390354,
|
| 155 |
+
"eval_loss": 3.048473596572876,
|
| 156 |
+
"eval_runtime": 52.5487,
|
| 157 |
+
"eval_samples_per_second": 10.314,
|
| 158 |
+
"eval_steps_per_second": 1.294,
|
| 159 |
+
"eval_wer": 1.0,
|
| 160 |
+
"step": 196
|
| 161 |
+
},
|
| 162 |
+
{
|
| 163 |
+
"epoch": 14.29,
|
| 164 |
+
"learning_rate": 2.7185714285714287e-05,
|
| 165 |
+
"loss": 3.059,
|
| 166 |
+
"step": 200
|
| 167 |
+
},
|
| 168 |
+
{
|
| 169 |
+
"epoch": 15.0,
|
| 170 |
+
"eval_cer": 0.9618562329390354,
|
| 171 |
+
"eval_loss": 3.0495524406433105,
|
| 172 |
+
"eval_runtime": 52.5935,
|
| 173 |
+
"eval_samples_per_second": 10.305,
|
| 174 |
+
"eval_steps_per_second": 1.293,
|
| 175 |
+
"eval_wer": 1.0,
|
| 176 |
+
"step": 210
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 16.0,
|
| 180 |
+
"eval_cer": 0.9618562329390354,
|
| 181 |
+
"eval_loss": 3.0509705543518066,
|
| 182 |
+
"eval_runtime": 52.5135,
|
| 183 |
+
"eval_samples_per_second": 10.321,
|
| 184 |
+
"eval_steps_per_second": 1.295,
|
| 185 |
+
"eval_wer": 1.0,
|
| 186 |
+
"step": 224
|
| 187 |
+
},
|
| 188 |
+
{
|
| 189 |
+
"epoch": 17.0,
|
| 190 |
+
"eval_cer": 0.9618562329390354,
|
| 191 |
+
"eval_loss": 3.042844533920288,
|
| 192 |
+
"eval_runtime": 52.5458,
|
| 193 |
+
"eval_samples_per_second": 10.315,
|
| 194 |
+
"eval_steps_per_second": 1.294,
|
| 195 |
+
"eval_wer": 1.0,
|
| 196 |
+
"step": 238
|
| 197 |
+
},
|
| 198 |
+
{
|
| 199 |
+
"epoch": 18.0,
|
| 200 |
+
"eval_cer": 0.9618562329390354,
|
| 201 |
+
"eval_loss": 3.033111572265625,
|
| 202 |
+
"eval_runtime": 52.5883,
|
| 203 |
+
"eval_samples_per_second": 10.306,
|
| 204 |
+
"eval_steps_per_second": 1.293,
|
| 205 |
+
"eval_wer": 1.0,
|
| 206 |
+
"step": 252
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 19.0,
|
| 210 |
+
"eval_cer": 0.9618562329390354,
|
| 211 |
+
"eval_loss": 3.035349130630493,
|
| 212 |
+
"eval_runtime": 52.5405,
|
| 213 |
+
"eval_samples_per_second": 10.316,
|
| 214 |
+
"eval_steps_per_second": 1.294,
|
| 215 |
+
"eval_wer": 1.0,
|
| 216 |
+
"step": 266
|
| 217 |
+
},
|
| 218 |
+
{
|
| 219 |
+
"epoch": 20.0,
|
| 220 |
+
"eval_cer": 0.9618562329390354,
|
| 221 |
+
"eval_loss": 3.0217459201812744,
|
| 222 |
+
"eval_runtime": 52.6413,
|
| 223 |
+
"eval_samples_per_second": 10.296,
|
| 224 |
+
"eval_steps_per_second": 1.292,
|
| 225 |
+
"eval_wer": 1.0,
|
| 226 |
+
"step": 280
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 21.0,
|
| 230 |
+
"eval_cer": 0.9618562329390354,
|
| 231 |
+
"eval_loss": 3.0107431411743164,
|
| 232 |
+
"eval_runtime": 52.5958,
|
| 233 |
+
"eval_samples_per_second": 10.305,
|
| 234 |
+
"eval_steps_per_second": 1.293,
|
| 235 |
+
"eval_wer": 1.0,
|
| 236 |
+
"step": 294
|
| 237 |
+
},
|
| 238 |
+
{
|
| 239 |
+
"epoch": 21.43,
|
| 240 |
+
"learning_rate": 2.5757142857142857e-05,
|
| 241 |
+
"loss": 2.9492,
|
| 242 |
+
"step": 300
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 22.0,
|
| 246 |
+
"eval_cer": 0.9618562329390354,
|
| 247 |
+
"eval_loss": 3.0068299770355225,
|
| 248 |
+
"eval_runtime": 52.8053,
|
| 249 |
+
"eval_samples_per_second": 10.264,
|
| 250 |
+
"eval_steps_per_second": 1.288,
|
| 251 |
+
"eval_wer": 1.0,
|
| 252 |
+
"step": 308
|
| 253 |
+
},
|
| 254 |
+
{
|
| 255 |
+
"epoch": 23.0,
|
| 256 |
+
"eval_cer": 0.9618562329390354,
|
| 257 |
+
"eval_loss": 2.9949989318847656,
|
| 258 |
+
"eval_runtime": 52.5717,
|
| 259 |
+
"eval_samples_per_second": 10.31,
|
| 260 |
+
"eval_steps_per_second": 1.293,
|
| 261 |
+
"eval_wer": 1.0,
|
| 262 |
+
"step": 322
|
| 263 |
+
},
|
| 264 |
+
{
|
| 265 |
+
"epoch": 24.0,
|
| 266 |
+
"eval_cer": 0.9618562329390354,
|
| 267 |
+
"eval_loss": 2.989611864089966,
|
| 268 |
+
"eval_runtime": 53.0076,
|
| 269 |
+
"eval_samples_per_second": 10.225,
|
| 270 |
+
"eval_steps_per_second": 1.283,
|
| 271 |
+
"eval_wer": 1.0,
|
| 272 |
+
"step": 336
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 25.0,
|
| 276 |
+
"eval_cer": 0.9618562329390354,
|
| 277 |
+
"eval_loss": 2.9687271118164062,
|
| 278 |
+
"eval_runtime": 52.6125,
|
| 279 |
+
"eval_samples_per_second": 10.302,
|
| 280 |
+
"eval_steps_per_second": 1.292,
|
| 281 |
+
"eval_wer": 1.0,
|
| 282 |
+
"step": 350
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 26.0,
|
| 286 |
+
"eval_cer": 0.9618562329390354,
|
| 287 |
+
"eval_loss": 2.947374105453491,
|
| 288 |
+
"eval_runtime": 52.5637,
|
| 289 |
+
"eval_samples_per_second": 10.311,
|
| 290 |
+
"eval_steps_per_second": 1.294,
|
| 291 |
+
"eval_wer": 1.0,
|
| 292 |
+
"step": 364
|
| 293 |
+
},
|
| 294 |
+
{
|
| 295 |
+
"epoch": 27.0,
|
| 296 |
+
"eval_cer": 0.9618562329390354,
|
| 297 |
+
"eval_loss": 2.9414024353027344,
|
| 298 |
+
"eval_runtime": 52.648,
|
| 299 |
+
"eval_samples_per_second": 10.295,
|
| 300 |
+
"eval_steps_per_second": 1.292,
|
| 301 |
+
"eval_wer": 1.0,
|
| 302 |
+
"step": 378
|
| 303 |
+
},
|
| 304 |
+
{
|
| 305 |
+
"epoch": 28.0,
|
| 306 |
+
"eval_cer": 0.9618562329390354,
|
| 307 |
+
"eval_loss": 2.84250807762146,
|
| 308 |
+
"eval_runtime": 52.5712,
|
| 309 |
+
"eval_samples_per_second": 10.31,
|
| 310 |
+
"eval_steps_per_second": 1.293,
|
| 311 |
+
"eval_wer": 1.0,
|
| 312 |
+
"step": 392
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 28.57,
|
| 316 |
+
"learning_rate": 2.432857142857143e-05,
|
| 317 |
+
"loss": 2.8892,
|
| 318 |
+
"step": 400
|
| 319 |
+
},
|
| 320 |
+
{
|
| 321 |
+
"epoch": 29.0,
|
| 322 |
+
"eval_cer": 0.9618562329390354,
|
| 323 |
+
"eval_loss": 2.7813332080841064,
|
| 324 |
+
"eval_runtime": 52.64,
|
| 325 |
+
"eval_samples_per_second": 10.296,
|
| 326 |
+
"eval_steps_per_second": 1.292,
|
| 327 |
+
"eval_wer": 1.0,
|
| 328 |
+
"step": 406
|
| 329 |
+
},
|
| 330 |
+
{
|
| 331 |
+
"epoch": 30.0,
|
| 332 |
+
"eval_cer": 0.9618562329390354,
|
| 333 |
+
"eval_loss": 2.7269773483276367,
|
| 334 |
+
"eval_runtime": 52.5989,
|
| 335 |
+
"eval_samples_per_second": 10.304,
|
| 336 |
+
"eval_steps_per_second": 1.293,
|
| 337 |
+
"eval_wer": 1.0,
|
| 338 |
+
"step": 420
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 31.0,
|
| 342 |
+
"eval_cer": 0.9606187443130119,
|
| 343 |
+
"eval_loss": 2.6644980907440186,
|
| 344 |
+
"eval_runtime": 52.6591,
|
| 345 |
+
"eval_samples_per_second": 10.293,
|
| 346 |
+
"eval_steps_per_second": 1.291,
|
| 347 |
+
"eval_wer": 1.0,
|
| 348 |
+
"step": 434
|
| 349 |
+
},
|
| 350 |
+
{
|
| 351 |
+
"epoch": 32.0,
|
| 352 |
+
"eval_cer": 0.9138853503184713,
|
| 353 |
+
"eval_loss": 2.5593061447143555,
|
| 354 |
+
"eval_runtime": 52.5942,
|
| 355 |
+
"eval_samples_per_second": 10.305,
|
| 356 |
+
"eval_steps_per_second": 1.293,
|
| 357 |
+
"eval_wer": 1.0,
|
| 358 |
+
"step": 448
|
| 359 |
+
},
|
| 360 |
+
{
|
| 361 |
+
"epoch": 33.0,
|
| 362 |
+
"eval_cer": 0.7002729754322111,
|
| 363 |
+
"eval_loss": 2.3229565620422363,
|
| 364 |
+
"eval_runtime": 53.1245,
|
| 365 |
+
"eval_samples_per_second": 10.202,
|
| 366 |
+
"eval_steps_per_second": 1.28,
|
| 367 |
+
"eval_wer": 1.0,
|
| 368 |
+
"step": 462
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 34.0,
|
| 372 |
+
"eval_cer": 0.535796178343949,
|
| 373 |
+
"eval_loss": 1.9706292152404785,
|
| 374 |
+
"eval_runtime": 52.7199,
|
| 375 |
+
"eval_samples_per_second": 10.281,
|
| 376 |
+
"eval_steps_per_second": 1.29,
|
| 377 |
+
"eval_wer": 1.0,
|
| 378 |
+
"step": 476
|
| 379 |
+
},
|
| 380 |
+
{
|
| 381 |
+
"epoch": 35.0,
|
| 382 |
+
"eval_cer": 0.45481346678798906,
|
| 383 |
+
"eval_loss": 1.7085232734680176,
|
| 384 |
+
"eval_runtime": 52.7281,
|
| 385 |
+
"eval_samples_per_second": 10.279,
|
| 386 |
+
"eval_steps_per_second": 1.29,
|
| 387 |
+
"eval_wer": 0.9997968305566842,
|
| 388 |
+
"step": 490
|
| 389 |
+
},
|
| 390 |
+
{
|
| 391 |
+
"epoch": 35.71,
|
| 392 |
+
"learning_rate": 2.29e-05,
|
| 393 |
+
"loss": 2.3937,
|
| 394 |
+
"step": 500
|
| 395 |
+
},
|
| 396 |
+
{
|
| 397 |
+
"epoch": 36.0,
|
| 398 |
+
"eval_cer": 0.4063694267515924,
|
| 399 |
+
"eval_loss": 1.449361801147461,
|
| 400 |
+
"eval_runtime": 52.6273,
|
| 401 |
+
"eval_samples_per_second": 10.299,
|
| 402 |
+
"eval_steps_per_second": 1.292,
|
| 403 |
+
"eval_wer": 1.0,
|
| 404 |
+
"step": 504
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 37.0,
|
| 408 |
+
"eval_cer": 0.38471337579617837,
|
| 409 |
+
"eval_loss": 1.286526083946228,
|
| 410 |
+
"eval_runtime": 52.7236,
|
| 411 |
+
"eval_samples_per_second": 10.28,
|
| 412 |
+
"eval_steps_per_second": 1.29,
|
| 413 |
+
"eval_wer": 1.0,
|
| 414 |
+
"step": 518
|
| 415 |
+
},
|
| 416 |
+
{
|
| 417 |
+
"epoch": 38.0,
|
| 418 |
+
"eval_cer": 0.3659326660600546,
|
| 419 |
+
"eval_loss": 1.1509252786636353,
|
| 420 |
+
"eval_runtime": 52.6429,
|
| 421 |
+
"eval_samples_per_second": 10.296,
|
| 422 |
+
"eval_steps_per_second": 1.292,
|
| 423 |
+
"eval_wer": 0.9947175944737912,
|
| 424 |
+
"step": 532
|
| 425 |
+
},
|
| 426 |
+
{
|
| 427 |
+
"epoch": 39.0,
|
| 428 |
+
"eval_cer": 0.3182893539581438,
|
| 429 |
+
"eval_loss": 1.0466662645339966,
|
| 430 |
+
"eval_runtime": 52.7531,
|
| 431 |
+
"eval_samples_per_second": 10.274,
|
| 432 |
+
"eval_steps_per_second": 1.289,
|
| 433 |
+
"eval_wer": 0.903088175538399,
|
| 434 |
+
"step": 546
|
| 435 |
+
},
|
| 436 |
+
{
|
| 437 |
+
"epoch": 40.0,
|
| 438 |
+
"eval_cer": 0.24043676069153777,
|
| 439 |
+
"eval_loss": 0.9831815361976624,
|
| 440 |
+
"eval_runtime": 52.6494,
|
| 441 |
+
"eval_samples_per_second": 10.295,
|
| 442 |
+
"eval_steps_per_second": 1.292,
|
| 443 |
+
"eval_wer": 0.5960991466883381,
|
| 444 |
+
"step": 560
|
| 445 |
+
},
|
| 446 |
+
{
|
| 447 |
+
"epoch": 41.0,
|
| 448 |
+
"eval_cer": 0.22227479526842583,
|
| 449 |
+
"eval_loss": 0.892058789730072,
|
| 450 |
+
"eval_runtime": 53.2461,
|
| 451 |
+
"eval_samples_per_second": 10.179,
|
| 452 |
+
"eval_steps_per_second": 1.277,
|
| 453 |
+
"eval_wer": 0.5048760666395774,
|
| 454 |
+
"step": 574
|
| 455 |
+
},
|
| 456 |
+
{
|
| 457 |
+
"epoch": 42.0,
|
| 458 |
+
"eval_cer": 0.21233848953594175,
|
| 459 |
+
"eval_loss": 0.8306074738502502,
|
| 460 |
+
"eval_runtime": 52.8348,
|
| 461 |
+
"eval_samples_per_second": 10.258,
|
| 462 |
+
"eval_steps_per_second": 1.287,
|
| 463 |
+
"eval_wer": 0.4687119057293783,
|
| 464 |
+
"step": 588
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 42.86,
|
| 468 |
+
"learning_rate": 2.1471428571428574e-05,
|
| 469 |
+
"loss": 1.0877,
|
| 470 |
+
"step": 600
|
| 471 |
+
},
|
| 472 |
+
{
|
| 473 |
+
"epoch": 43.0,
|
| 474 |
+
"eval_cer": 0.20880800727934487,
|
| 475 |
+
"eval_loss": 0.8016981482505798,
|
| 476 |
+
"eval_runtime": 52.6157,
|
| 477 |
+
"eval_samples_per_second": 10.301,
|
| 478 |
+
"eval_steps_per_second": 1.292,
|
| 479 |
+
"eval_wer": 0.45631856968711904,
|
| 480 |
+
"step": 602
|
| 481 |
+
},
|
| 482 |
+
{
|
| 483 |
+
"epoch": 44.0,
|
| 484 |
+
"eval_cer": 0.2045859872611465,
|
| 485 |
+
"eval_loss": 0.7715625166893005,
|
| 486 |
+
"eval_runtime": 52.6399,
|
| 487 |
+
"eval_samples_per_second": 10.296,
|
| 488 |
+
"eval_steps_per_second": 1.292,
|
| 489 |
+
"eval_wer": 0.4404713531084925,
|
| 490 |
+
"step": 616
|
| 491 |
+
},
|
| 492 |
+
{
|
| 493 |
+
"epoch": 45.0,
|
| 494 |
+
"eval_cer": 0.2053867151956324,
|
| 495 |
+
"eval_loss": 0.7693753838539124,
|
| 496 |
+
"eval_runtime": 52.708,
|
| 497 |
+
"eval_samples_per_second": 10.283,
|
| 498 |
+
"eval_steps_per_second": 1.29,
|
| 499 |
+
"eval_wer": 0.4406745225518082,
|
| 500 |
+
"step": 630
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 46.0,
|
| 504 |
+
"eval_cer": 0.20367606915377617,
|
| 505 |
+
"eval_loss": 0.7451383471488953,
|
| 506 |
+
"eval_runtime": 52.7002,
|
| 507 |
+
"eval_samples_per_second": 10.285,
|
| 508 |
+
"eval_steps_per_second": 1.29,
|
| 509 |
+
"eval_wer": 0.4315318976026006,
|
| 510 |
+
"step": 644
|
| 511 |
+
},
|
| 512 |
+
{
|
| 513 |
+
"epoch": 47.0,
|
| 514 |
+
"eval_cer": 0.19956323930846223,
|
| 515 |
+
"eval_loss": 0.7111817598342896,
|
| 516 |
+
"eval_runtime": 53.2448,
|
| 517 |
+
"eval_samples_per_second": 10.179,
|
| 518 |
+
"eval_steps_per_second": 1.277,
|
| 519 |
+
"eval_wer": 0.42503047541649736,
|
| 520 |
+
"step": 658
|
| 521 |
+
},
|
| 522 |
+
{
|
| 523 |
+
"epoch": 48.0,
|
| 524 |
+
"eval_cer": 0.19581437670609644,
|
| 525 |
+
"eval_loss": 0.7007989883422852,
|
| 526 |
+
"eval_runtime": 52.7775,
|
| 527 |
+
"eval_samples_per_second": 10.27,
|
| 528 |
+
"eval_steps_per_second": 1.288,
|
| 529 |
+
"eval_wer": 0.41162129215765947,
|
| 530 |
+
"step": 672
|
| 531 |
+
},
|
| 532 |
+
{
|
| 533 |
+
"epoch": 49.0,
|
| 534 |
+
"eval_cer": 0.19799818016378526,
|
| 535 |
+
"eval_loss": 0.7140166163444519,
|
| 536 |
+
"eval_runtime": 52.6495,
|
| 537 |
+
"eval_samples_per_second": 10.294,
|
| 538 |
+
"eval_steps_per_second": 1.292,
|
| 539 |
+
"eval_wer": 0.40572937830150346,
|
| 540 |
+
"step": 686
|
| 541 |
+
},
|
| 542 |
+
{
|
| 543 |
+
"epoch": 50.0,
|
| 544 |
+
"learning_rate": 2.004285714285714e-05,
|
| 545 |
+
"loss": 0.6292,
|
| 546 |
+
"step": 700
|
| 547 |
+
},
|
| 548 |
+
{
|
| 549 |
+
"epoch": 50.0,
|
| 550 |
+
"eval_cer": 0.19883530482256598,
|
| 551 |
+
"eval_loss": 0.7207656502723694,
|
| 552 |
+
"eval_runtime": 52.6919,
|
| 553 |
+
"eval_samples_per_second": 10.286,
|
| 554 |
+
"eval_steps_per_second": 1.291,
|
| 555 |
+
"eval_wer": 0.41141812271434375,
|
| 556 |
+
"step": 700
|
| 557 |
+
},
|
| 558 |
+
{
|
| 559 |
+
"epoch": 51.0,
|
| 560 |
+
"eval_cer": 0.19366696997270247,
|
| 561 |
+
"eval_loss": 0.6675068140029907,
|
| 562 |
+
"eval_runtime": 52.8778,
|
| 563 |
+
"eval_samples_per_second": 10.25,
|
| 564 |
+
"eval_steps_per_second": 1.286,
|
| 565 |
+
"eval_wer": 0.40329134498171476,
|
| 566 |
+
"step": 714
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 52.0,
|
| 570 |
+
"eval_cer": 0.19384895359417653,
|
| 571 |
+
"eval_loss": 0.6650232672691345,
|
| 572 |
+
"eval_runtime": 52.6583,
|
| 573 |
+
"eval_samples_per_second": 10.293,
|
| 574 |
+
"eval_steps_per_second": 1.291,
|
| 575 |
+
"eval_wer": 0.4014628199918732,
|
| 576 |
+
"step": 728
|
| 577 |
+
},
|
| 578 |
+
{
|
| 579 |
+
"epoch": 53.0,
|
| 580 |
+
"eval_cer": 0.1937761601455869,
|
| 581 |
+
"eval_loss": 0.6550155878067017,
|
| 582 |
+
"eval_runtime": 52.7667,
|
| 583 |
+
"eval_samples_per_second": 10.272,
|
| 584 |
+
"eval_steps_per_second": 1.289,
|
| 585 |
+
"eval_wer": 0.4012596505485575,
|
| 586 |
+
"step": 742
|
| 587 |
+
},
|
| 588 |
+
{
|
| 589 |
+
"epoch": 54.0,
|
| 590 |
+
"eval_cer": 0.19315741583257506,
|
| 591 |
+
"eval_loss": 0.6477207541465759,
|
| 592 |
+
"eval_runtime": 52.6665,
|
| 593 |
+
"eval_samples_per_second": 10.291,
|
| 594 |
+
"eval_steps_per_second": 1.291,
|
| 595 |
+
"eval_wer": 0.3990247866720845,
|
| 596 |
+
"step": 756
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 55.0,
|
| 600 |
+
"eval_cer": 0.19319381255686988,
|
| 601 |
+
"eval_loss": 0.6361698508262634,
|
| 602 |
+
"eval_runtime": 52.6325,
|
| 603 |
+
"eval_samples_per_second": 10.298,
|
| 604 |
+
"eval_steps_per_second": 1.292,
|
| 605 |
+
"eval_wer": 0.39597724502234866,
|
| 606 |
+
"step": 770
|
| 607 |
+
},
|
| 608 |
+
{
|
| 609 |
+
"epoch": 56.0,
|
| 610 |
+
"eval_cer": 0.192975432211101,
|
| 611 |
+
"eval_loss": 0.632267951965332,
|
| 612 |
+
"eval_runtime": 52.7108,
|
| 613 |
+
"eval_samples_per_second": 10.283,
|
| 614 |
+
"eval_steps_per_second": 1.29,
|
| 615 |
+
"eval_wer": 0.39191385615603413,
|
| 616 |
+
"step": 784
|
| 617 |
+
},
|
| 618 |
+
{
|
| 619 |
+
"epoch": 57.0,
|
| 620 |
+
"eval_cer": 0.19206551410373066,
|
| 621 |
+
"eval_loss": 0.626406192779541,
|
| 622 |
+
"eval_runtime": 52.675,
|
| 623 |
+
"eval_samples_per_second": 10.29,
|
| 624 |
+
"eval_steps_per_second": 1.291,
|
| 625 |
+
"eval_wer": 0.38703778951645673,
|
| 626 |
+
"step": 798
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 57.14,
|
| 630 |
+
"learning_rate": 1.8614285714285715e-05,
|
| 631 |
+
"loss": 0.4739,
|
| 632 |
+
"step": 800
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 58.0,
|
| 636 |
+
"eval_cer": 0.19210191082802547,
|
| 637 |
+
"eval_loss": 0.6290065050125122,
|
| 638 |
+
"eval_runtime": 52.661,
|
| 639 |
+
"eval_samples_per_second": 10.292,
|
| 640 |
+
"eval_steps_per_second": 1.291,
|
| 641 |
+
"eval_wer": 0.38724095895977245,
|
| 642 |
+
"step": 812
|
| 643 |
+
},
|
| 644 |
+
{
|
| 645 |
+
"epoch": 59.0,
|
| 646 |
+
"eval_cer": 0.19253867151956325,
|
| 647 |
+
"eval_loss": 0.6206580400466919,
|
| 648 |
+
"eval_runtime": 52.6443,
|
| 649 |
+
"eval_samples_per_second": 10.296,
|
| 650 |
+
"eval_steps_per_second": 1.292,
|
| 651 |
+
"eval_wer": 0.38642828118650957,
|
| 652 |
+
"step": 826
|
| 653 |
+
},
|
| 654 |
+
{
|
| 655 |
+
"epoch": 60.0,
|
| 656 |
+
"eval_cer": 0.19184713375796178,
|
| 657 |
+
"eval_loss": 0.6177613735198975,
|
| 658 |
+
"eval_runtime": 52.6855,
|
| 659 |
+
"eval_samples_per_second": 10.287,
|
| 660 |
+
"eval_steps_per_second": 1.291,
|
| 661 |
+
"eval_wer": 0.38581877285656235,
|
| 662 |
+
"step": 840
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 61.0,
|
| 666 |
+
"eval_cer": 0.19177434030937215,
|
| 667 |
+
"eval_loss": 0.6217214465141296,
|
| 668 |
+
"eval_runtime": 53.1515,
|
| 669 |
+
"eval_samples_per_second": 10.197,
|
| 670 |
+
"eval_steps_per_second": 1.279,
|
| 671 |
+
"eval_wer": 0.3860219422998781,
|
| 672 |
+
"step": 854
|
| 673 |
+
},
|
| 674 |
+
{
|
| 675 |
+
"epoch": 62.0,
|
| 676 |
+
"eval_cer": 0.1899545040946315,
|
| 677 |
+
"eval_loss": 0.6078172922134399,
|
| 678 |
+
"eval_runtime": 52.6994,
|
| 679 |
+
"eval_samples_per_second": 10.285,
|
| 680 |
+
"eval_steps_per_second": 1.29,
|
| 681 |
+
"eval_wer": 0.37992685900040635,
|
| 682 |
+
"step": 868
|
| 683 |
+
},
|
| 684 |
+
{
|
| 685 |
+
"epoch": 63.0,
|
| 686 |
+
"eval_cer": 0.1889353958143767,
|
| 687 |
+
"eval_loss": 0.6072229146957397,
|
| 688 |
+
"eval_runtime": 52.8038,
|
| 689 |
+
"eval_samples_per_second": 10.264,
|
| 690 |
+
"eval_steps_per_second": 1.288,
|
| 691 |
+
"eval_wer": 0.3780983340105648,
|
| 692 |
+
"step": 882
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 64.0,
|
| 696 |
+
"eval_cer": 0.18831665150136487,
|
| 697 |
+
"eval_loss": 0.6067555546760559,
|
| 698 |
+
"eval_runtime": 52.6766,
|
| 699 |
+
"eval_samples_per_second": 10.289,
|
| 700 |
+
"eval_steps_per_second": 1.291,
|
| 701 |
+
"eval_wer": 0.37606663957740755,
|
| 702 |
+
"step": 896
|
| 703 |
+
},
|
| 704 |
+
{
|
| 705 |
+
"epoch": 64.29,
|
| 706 |
+
"learning_rate": 1.7185714285714285e-05,
|
| 707 |
+
"loss": 0.3855,
|
| 708 |
+
"step": 900
|
| 709 |
+
},
|
| 710 |
+
{
|
| 711 |
+
"epoch": 65.0,
|
| 712 |
+
"eval_cer": 0.1870427661510464,
|
| 713 |
+
"eval_loss": 0.5944907665252686,
|
| 714 |
+
"eval_runtime": 52.7345,
|
| 715 |
+
"eval_samples_per_second": 10.278,
|
| 716 |
+
"eval_steps_per_second": 1.289,
|
| 717 |
+
"eval_wer": 0.3748476229175132,
|
| 718 |
+
"step": 910
|
| 719 |
+
},
|
| 720 |
+
{
|
| 721 |
+
"epoch": 66.0,
|
| 722 |
+
"eval_cer": 0.1899545040946315,
|
| 723 |
+
"eval_loss": 0.6194169521331787,
|
| 724 |
+
"eval_runtime": 52.7092,
|
| 725 |
+
"eval_samples_per_second": 10.283,
|
| 726 |
+
"eval_steps_per_second": 1.29,
|
| 727 |
+
"eval_wer": 0.37992685900040635,
|
| 728 |
+
"step": 924
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 67.0,
|
| 732 |
+
"eval_cer": 0.18846223839854412,
|
| 733 |
+
"eval_loss": 0.6044437289237976,
|
| 734 |
+
"eval_runtime": 52.6167,
|
| 735 |
+
"eval_samples_per_second": 10.301,
|
| 736 |
+
"eval_steps_per_second": 1.292,
|
| 737 |
+
"eval_wer": 0.3793173506704592,
|
| 738 |
+
"step": 938
|
| 739 |
+
},
|
| 740 |
+
{
|
| 741 |
+
"epoch": 68.0,
|
| 742 |
+
"eval_cer": 0.18798908098271155,
|
| 743 |
+
"eval_loss": 0.5945860147476196,
|
| 744 |
+
"eval_runtime": 52.6643,
|
| 745 |
+
"eval_samples_per_second": 10.292,
|
| 746 |
+
"eval_steps_per_second": 1.291,
|
| 747 |
+
"eval_wer": 0.37505079236082894,
|
| 748 |
+
"step": 952
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 69.0,
|
| 752 |
+
"eval_cer": 0.18798908098271155,
|
| 753 |
+
"eval_loss": 0.6115556955337524,
|
| 754 |
+
"eval_runtime": 52.6187,
|
| 755 |
+
"eval_samples_per_second": 10.301,
|
| 756 |
+
"eval_steps_per_second": 1.292,
|
| 757 |
+
"eval_wer": 0.37139374238114586,
|
| 758 |
+
"step": 966
|
| 759 |
+
},
|
| 760 |
+
{
|
| 761 |
+
"epoch": 70.0,
|
| 762 |
+
"eval_cer": 0.18609645131938127,
|
| 763 |
+
"eval_loss": 0.5876884460449219,
|
| 764 |
+
"eval_runtime": 52.6386,
|
| 765 |
+
"eval_samples_per_second": 10.297,
|
| 766 |
+
"eval_steps_per_second": 1.292,
|
| 767 |
+
"eval_wer": 0.36793986184477856,
|
| 768 |
+
"step": 980
|
| 769 |
+
},
|
| 770 |
+
{
|
| 771 |
+
"epoch": 71.0,
|
| 772 |
+
"eval_cer": 0.18627843494085533,
|
| 773 |
+
"eval_loss": 0.5861026048660278,
|
| 774 |
+
"eval_runtime": 52.6444,
|
| 775 |
+
"eval_samples_per_second": 10.296,
|
| 776 |
+
"eval_steps_per_second": 1.292,
|
| 777 |
+
"eval_wer": 0.36793986184477856,
|
| 778 |
+
"step": 994
|
| 779 |
+
},
|
| 780 |
+
{
|
| 781 |
+
"epoch": 71.43,
|
| 782 |
+
"learning_rate": 1.5757142857142858e-05,
|
| 783 |
+
"loss": 0.3302,
|
| 784 |
+
"step": 1000
|
| 785 |
+
},
|
| 786 |
+
{
|
| 787 |
+
"epoch": 72.0,
|
| 788 |
+
"eval_cer": 0.18562329390354868,
|
| 789 |
+
"eval_loss": 0.580510139465332,
|
| 790 |
+
"eval_runtime": 52.6419,
|
| 791 |
+
"eval_samples_per_second": 10.296,
|
| 792 |
+
"eval_steps_per_second": 1.292,
|
| 793 |
+
"eval_wer": 0.3685493701747257,
|
| 794 |
+
"step": 1008
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 73.0,
|
| 798 |
+
"eval_cer": 0.1862056414922657,
|
| 799 |
+
"eval_loss": 0.586155354976654,
|
| 800 |
+
"eval_runtime": 52.6421,
|
| 801 |
+
"eval_samples_per_second": 10.296,
|
| 802 |
+
"eval_steps_per_second": 1.292,
|
| 803 |
+
"eval_wer": 0.37139374238114586,
|
| 804 |
+
"step": 1022
|
| 805 |
+
},
|
| 806 |
+
{
|
| 807 |
+
"epoch": 74.0,
|
| 808 |
+
"eval_cer": 0.18656960873521383,
|
| 809 |
+
"eval_loss": 0.5920763611793518,
|
| 810 |
+
"eval_runtime": 52.6765,
|
| 811 |
+
"eval_samples_per_second": 10.289,
|
| 812 |
+
"eval_steps_per_second": 1.291,
|
| 813 |
+
"eval_wer": 0.372003250711093,
|
| 814 |
+
"step": 1036
|
| 815 |
+
},
|
| 816 |
+
{
|
| 817 |
+
"epoch": 75.0,
|
| 818 |
+
"eval_cer": 0.185368516833485,
|
| 819 |
+
"eval_loss": 0.5692393779754639,
|
| 820 |
+
"eval_runtime": 52.7333,
|
| 821 |
+
"eval_samples_per_second": 10.278,
|
| 822 |
+
"eval_steps_per_second": 1.29,
|
| 823 |
+
"eval_wer": 0.36834620073141,
|
| 824 |
+
"step": 1050
|
| 825 |
+
},
|
| 826 |
+
{
|
| 827 |
+
"epoch": 76.0,
|
| 828 |
+
"eval_cer": 0.18777070063694268,
|
| 829 |
+
"eval_loss": 0.5922245979309082,
|
| 830 |
+
"eval_runtime": 52.7512,
|
| 831 |
+
"eval_samples_per_second": 10.275,
|
| 832 |
+
"eval_steps_per_second": 1.289,
|
| 833 |
+
"eval_wer": 0.37017472572125154,
|
| 834 |
+
"step": 1064
|
| 835 |
+
},
|
| 836 |
+
{
|
| 837 |
+
"epoch": 77.0,
|
| 838 |
+
"eval_cer": 0.18828025477707006,
|
| 839 |
+
"eval_loss": 0.6105178594589233,
|
| 840 |
+
"eval_runtime": 52.65,
|
| 841 |
+
"eval_samples_per_second": 10.294,
|
| 842 |
+
"eval_steps_per_second": 1.292,
|
| 843 |
+
"eval_wer": 0.3709874034945144,
|
| 844 |
+
"step": 1078
|
| 845 |
+
},
|
| 846 |
+
{
|
| 847 |
+
"epoch": 78.0,
|
| 848 |
+
"eval_cer": 0.18558689717925386,
|
| 849 |
+
"eval_loss": 0.5873062014579773,
|
| 850 |
+
"eval_runtime": 53.0336,
|
| 851 |
+
"eval_samples_per_second": 10.22,
|
| 852 |
+
"eval_steps_per_second": 1.282,
|
| 853 |
+
"eval_wer": 0.36834620073141,
|
| 854 |
+
"step": 1092
|
| 855 |
+
},
|
| 856 |
+
{
|
| 857 |
+
"epoch": 78.57,
|
| 858 |
+
"learning_rate": 1.4328571428571428e-05,
|
| 859 |
+
"loss": 0.3046,
|
| 860 |
+
"step": 1100
|
| 861 |
+
},
|
| 862 |
+
{
|
| 863 |
+
"epoch": 79.0,
|
| 864 |
+
"eval_cer": 0.18587807097361236,
|
| 865 |
+
"eval_loss": 0.5825892090797424,
|
| 866 |
+
"eval_runtime": 52.7458,
|
| 867 |
+
"eval_samples_per_second": 10.276,
|
| 868 |
+
"eval_steps_per_second": 1.289,
|
| 869 |
+
"eval_wer": 0.3681430312880943,
|
| 870 |
+
"step": 1106
|
| 871 |
+
},
|
| 872 |
+
{
|
| 873 |
+
"epoch": 80.0,
|
| 874 |
+
"eval_cer": 0.18445859872611464,
|
| 875 |
+
"eval_loss": 0.5792315006256104,
|
| 876 |
+
"eval_runtime": 53.1292,
|
| 877 |
+
"eval_samples_per_second": 10.202,
|
| 878 |
+
"eval_steps_per_second": 1.28,
|
| 879 |
+
"eval_wer": 0.3632669646485169,
|
| 880 |
+
"step": 1120
|
| 881 |
+
},
|
| 882 |
+
{
|
| 883 |
+
"epoch": 81.0,
|
| 884 |
+
"eval_cer": 0.18347588717015467,
|
| 885 |
+
"eval_loss": 0.5737511515617371,
|
| 886 |
+
"eval_runtime": 52.6677,
|
| 887 |
+
"eval_samples_per_second": 10.291,
|
| 888 |
+
"eval_steps_per_second": 1.291,
|
| 889 |
+
"eval_wer": 0.3610321007720439,
|
| 890 |
+
"step": 1134
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 82.0,
|
| 894 |
+
"eval_cer": 0.1843130118289354,
|
| 895 |
+
"eval_loss": 0.579399585723877,
|
| 896 |
+
"eval_runtime": 53.0003,
|
| 897 |
+
"eval_samples_per_second": 10.226,
|
| 898 |
+
"eval_steps_per_second": 1.283,
|
| 899 |
+
"eval_wer": 0.36245428687525394,
|
| 900 |
+
"step": 1148
|
| 901 |
+
},
|
| 902 |
+
{
|
| 903 |
+
"epoch": 83.0,
|
| 904 |
+
"eval_cer": 0.18289353958143767,
|
| 905 |
+
"eval_loss": 0.5766276121139526,
|
| 906 |
+
"eval_runtime": 52.6724,
|
| 907 |
+
"eval_samples_per_second": 10.29,
|
| 908 |
+
"eval_steps_per_second": 1.291,
|
| 909 |
+
"eval_wer": 0.3563592035757822,
|
| 910 |
+
"step": 1162
|
| 911 |
+
},
|
| 912 |
+
{
|
| 913 |
+
"epoch": 84.0,
|
| 914 |
+
"eval_cer": 0.1830027297543221,
|
| 915 |
+
"eval_loss": 0.5745117664337158,
|
| 916 |
+
"eval_runtime": 52.7431,
|
| 917 |
+
"eval_samples_per_second": 10.276,
|
| 918 |
+
"eval_steps_per_second": 1.289,
|
| 919 |
+
"eval_wer": 0.35778138967899226,
|
| 920 |
+
"step": 1176
|
| 921 |
+
},
|
| 922 |
+
{
|
| 923 |
+
"epoch": 85.0,
|
| 924 |
+
"eval_cer": 0.18140127388535032,
|
| 925 |
+
"eval_loss": 0.561495840549469,
|
| 926 |
+
"eval_runtime": 52.6016,
|
| 927 |
+
"eval_samples_per_second": 10.304,
|
| 928 |
+
"eval_steps_per_second": 1.293,
|
| 929 |
+
"eval_wer": 0.3555465258025193,
|
| 930 |
+
"step": 1190
|
| 931 |
+
},
|
| 932 |
+
{
|
| 933 |
+
"epoch": 85.71,
|
| 934 |
+
"learning_rate": 1.29e-05,
|
| 935 |
+
"loss": 0.2927,
|
| 936 |
+
"step": 1200
|
| 937 |
+
},
|
| 938 |
+
{
|
| 939 |
+
"epoch": 86.0,
|
| 940 |
+
"eval_cer": 0.18282074613284804,
|
| 941 |
+
"eval_loss": 0.5853651762008667,
|
| 942 |
+
"eval_runtime": 52.6345,
|
| 943 |
+
"eval_samples_per_second": 10.297,
|
| 944 |
+
"eval_steps_per_second": 1.292,
|
| 945 |
+
"eval_wer": 0.36143843965867534,
|
| 946 |
+
"step": 1204
|
| 947 |
+
},
|
| 948 |
+
{
|
| 949 |
+
"epoch": 87.0,
|
| 950 |
+
"eval_cer": 0.1835122838944495,
|
| 951 |
+
"eval_loss": 0.5817971229553223,
|
| 952 |
+
"eval_runtime": 52.7194,
|
| 953 |
+
"eval_samples_per_second": 10.281,
|
| 954 |
+
"eval_steps_per_second": 1.29,
|
| 955 |
+
"eval_wer": 0.36245428687525394,
|
| 956 |
+
"step": 1218
|
| 957 |
+
},
|
| 958 |
+
{
|
| 959 |
+
"epoch": 88.0,
|
| 960 |
+
"eval_cer": 0.18154686078252957,
|
| 961 |
+
"eval_loss": 0.5613023638725281,
|
| 962 |
+
"eval_runtime": 52.6918,
|
| 963 |
+
"eval_samples_per_second": 10.286,
|
| 964 |
+
"eval_steps_per_second": 1.291,
|
| 965 |
+
"eval_wer": 0.35778138967899226,
|
| 966 |
+
"step": 1232
|
| 967 |
+
},
|
| 968 |
+
{
|
| 969 |
+
"epoch": 89.0,
|
| 970 |
+
"eval_cer": 0.18125568698817107,
|
| 971 |
+
"eval_loss": 0.5661023259162903,
|
| 972 |
+
"eval_runtime": 52.6276,
|
| 973 |
+
"eval_samples_per_second": 10.299,
|
| 974 |
+
"eval_steps_per_second": 1.292,
|
| 975 |
+
"eval_wer": 0.3549370174725721,
|
| 976 |
+
"step": 1246
|
| 977 |
+
},
|
| 978 |
+
{
|
| 979 |
+
"epoch": 90.0,
|
| 980 |
+
"eval_cer": 0.18202001819836214,
|
| 981 |
+
"eval_loss": 0.5794751048088074,
|
| 982 |
+
"eval_runtime": 52.7536,
|
| 983 |
+
"eval_samples_per_second": 10.274,
|
| 984 |
+
"eval_steps_per_second": 1.289,
|
| 985 |
+
"eval_wer": 0.36042259244209673,
|
| 986 |
+
"step": 1260
|
| 987 |
+
},
|
| 988 |
+
{
|
| 989 |
+
"epoch": 91.0,
|
| 990 |
+
"eval_cer": 0.18020018198362148,
|
| 991 |
+
"eval_loss": 0.5604133605957031,
|
| 992 |
+
"eval_runtime": 52.6351,
|
| 993 |
+
"eval_samples_per_second": 10.297,
|
| 994 |
+
"eval_steps_per_second": 1.292,
|
| 995 |
+
"eval_wer": 0.35209264526615197,
|
| 996 |
+
"step": 1274
|
| 997 |
+
},
|
| 998 |
+
{
|
| 999 |
+
"epoch": 92.0,
|
| 1000 |
+
"eval_cer": 0.1821656050955414,
|
| 1001 |
+
"eval_loss": 0.5738141536712646,
|
| 1002 |
+
"eval_runtime": 52.6638,
|
| 1003 |
+
"eval_samples_per_second": 10.292,
|
| 1004 |
+
"eval_steps_per_second": 1.291,
|
| 1005 |
+
"eval_wer": 0.35900040633888664,
|
| 1006 |
+
"step": 1288
|
| 1007 |
+
},
|
| 1008 |
+
{
|
| 1009 |
+
"epoch": 92.86,
|
| 1010 |
+
"learning_rate": 1.1471428571428572e-05,
|
| 1011 |
+
"loss": 0.2576,
|
| 1012 |
+
"step": 1300
|
| 1013 |
+
},
|
| 1014 |
+
{
|
| 1015 |
+
"epoch": 93.0,
|
| 1016 |
+
"eval_cer": 0.18140127388535032,
|
| 1017 |
+
"eval_loss": 0.5658465623855591,
|
| 1018 |
+
"eval_runtime": 52.9846,
|
| 1019 |
+
"eval_samples_per_second": 10.229,
|
| 1020 |
+
"eval_steps_per_second": 1.283,
|
| 1021 |
+
"eval_wer": 0.3573750507923608,
|
| 1022 |
+
"step": 1302
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 94.0,
|
| 1026 |
+
"eval_cer": 0.1808189262966333,
|
| 1027 |
+
"eval_loss": 0.5620054006576538,
|
| 1028 |
+
"eval_runtime": 52.617,
|
| 1029 |
+
"eval_samples_per_second": 10.301,
|
| 1030 |
+
"eval_steps_per_second": 1.292,
|
| 1031 |
+
"eval_wer": 0.35107679804957337,
|
| 1032 |
+
"step": 1316
|
| 1033 |
+
},
|
| 1034 |
+
{
|
| 1035 |
+
"epoch": 95.0,
|
| 1036 |
+
"eval_cer": 0.18100090991810738,
|
| 1037 |
+
"eval_loss": 0.5709471106529236,
|
| 1038 |
+
"eval_runtime": 52.6962,
|
| 1039 |
+
"eval_samples_per_second": 10.285,
|
| 1040 |
+
"eval_steps_per_second": 1.29,
|
| 1041 |
+
"eval_wer": 0.35412433969930923,
|
| 1042 |
+
"step": 1330
|
| 1043 |
+
},
|
| 1044 |
+
{
|
| 1045 |
+
"epoch": 96.0,
|
| 1046 |
+
"eval_cer": 0.1799454049135578,
|
| 1047 |
+
"eval_loss": 0.5674740076065063,
|
| 1048 |
+
"eval_runtime": 52.6892,
|
| 1049 |
+
"eval_samples_per_second": 10.287,
|
| 1050 |
+
"eval_steps_per_second": 1.291,
|
| 1051 |
+
"eval_wer": 0.35026412027631043,
|
| 1052 |
+
"step": 1344
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 97.0,
|
| 1056 |
+
"eval_cer": 0.18151046405823476,
|
| 1057 |
+
"eval_loss": 0.5788221955299377,
|
| 1058 |
+
"eval_runtime": 53.0799,
|
| 1059 |
+
"eval_samples_per_second": 10.211,
|
| 1060 |
+
"eval_steps_per_second": 1.281,
|
| 1061 |
+
"eval_wer": 0.3549370174725721,
|
| 1062 |
+
"step": 1358
|
| 1063 |
+
},
|
| 1064 |
+
{
|
| 1065 |
+
"epoch": 98.0,
|
| 1066 |
+
"eval_cer": 0.18096451319381257,
|
| 1067 |
+
"eval_loss": 0.5730317234992981,
|
| 1068 |
+
"eval_runtime": 52.5988,
|
| 1069 |
+
"eval_samples_per_second": 10.304,
|
| 1070 |
+
"eval_steps_per_second": 1.293,
|
| 1071 |
+
"eval_wer": 0.3524989841527834,
|
| 1072 |
+
"step": 1372
|
| 1073 |
+
},
|
| 1074 |
+
{
|
| 1075 |
+
"epoch": 99.0,
|
| 1076 |
+
"eval_cer": 0.1802729754322111,
|
| 1077 |
+
"eval_loss": 0.5693602561950684,
|
| 1078 |
+
"eval_runtime": 52.6503,
|
| 1079 |
+
"eval_samples_per_second": 10.294,
|
| 1080 |
+
"eval_steps_per_second": 1.292,
|
| 1081 |
+
"eval_wer": 0.35107679804957337,
|
| 1082 |
+
"step": 1386
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"epoch": 100.0,
|
| 1086 |
+
"learning_rate": 1.0042857142857144e-05,
|
| 1087 |
+
"loss": 0.2273,
|
| 1088 |
+
"step": 1400
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"epoch": 100.0,
|
| 1092 |
+
"eval_cer": 0.18070973612374885,
|
| 1093 |
+
"eval_loss": 0.5747684240341187,
|
| 1094 |
+
"eval_runtime": 52.7106,
|
| 1095 |
+
"eval_samples_per_second": 10.283,
|
| 1096 |
+
"eval_steps_per_second": 1.29,
|
| 1097 |
+
"eval_wer": 0.35270215359609913,
|
| 1098 |
+
"step": 1400
|
| 1099 |
+
},
|
| 1100 |
+
{
|
| 1101 |
+
"epoch": 101.0,
|
| 1102 |
+
"eval_cer": 0.17965423111919926,
|
| 1103 |
+
"eval_loss": 0.5688263773918152,
|
| 1104 |
+
"eval_runtime": 52.691,
|
| 1105 |
+
"eval_samples_per_second": 10.286,
|
| 1106 |
+
"eval_steps_per_second": 1.291,
|
| 1107 |
+
"eval_wer": 0.3512799674928891,
|
| 1108 |
+
"step": 1414
|
| 1109 |
+
},
|
| 1110 |
+
{
|
| 1111 |
+
"epoch": 102.0,
|
| 1112 |
+
"eval_cer": 0.18045495905368517,
|
| 1113 |
+
"eval_loss": 0.5767450332641602,
|
| 1114 |
+
"eval_runtime": 52.6967,
|
| 1115 |
+
"eval_samples_per_second": 10.285,
|
| 1116 |
+
"eval_steps_per_second": 1.29,
|
| 1117 |
+
"eval_wer": 0.35534335635920355,
|
| 1118 |
+
"step": 1428
|
| 1119 |
+
},
|
| 1120 |
+
{
|
| 1121 |
+
"epoch": 103.0,
|
| 1122 |
+
"eval_cer": 0.18118289353958145,
|
| 1123 |
+
"eval_loss": 0.5758454203605652,
|
| 1124 |
+
"eval_runtime": 52.6575,
|
| 1125 |
+
"eval_samples_per_second": 10.293,
|
| 1126 |
+
"eval_steps_per_second": 1.291,
|
| 1127 |
+
"eval_wer": 0.35290532303941485,
|
| 1128 |
+
"step": 1442
|
| 1129 |
+
},
|
| 1130 |
+
{
|
| 1131 |
+
"epoch": 104.0,
|
| 1132 |
+
"eval_cer": 0.17929026387625113,
|
| 1133 |
+
"eval_loss": 0.5641180872917175,
|
| 1134 |
+
"eval_runtime": 52.7234,
|
| 1135 |
+
"eval_samples_per_second": 10.28,
|
| 1136 |
+
"eval_steps_per_second": 1.29,
|
| 1137 |
+
"eval_wer": 0.35067045916294187,
|
| 1138 |
+
"step": 1456
|
| 1139 |
+
},
|
| 1140 |
+
{
|
| 1141 |
+
"epoch": 105.0,
|
| 1142 |
+
"eval_cer": 0.178926296633303,
|
| 1143 |
+
"eval_loss": 0.5628452897071838,
|
| 1144 |
+
"eval_runtime": 52.9721,
|
| 1145 |
+
"eval_samples_per_second": 10.232,
|
| 1146 |
+
"eval_steps_per_second": 1.284,
|
| 1147 |
+
"eval_wer": 0.34945144250304755,
|
| 1148 |
+
"step": 1470
|
| 1149 |
+
},
|
| 1150 |
+
{
|
| 1151 |
+
"epoch": 106.0,
|
| 1152 |
+
"eval_cer": 0.1788898999090082,
|
| 1153 |
+
"eval_loss": 0.5728613138198853,
|
| 1154 |
+
"eval_runtime": 52.6612,
|
| 1155 |
+
"eval_samples_per_second": 10.292,
|
| 1156 |
+
"eval_steps_per_second": 1.291,
|
| 1157 |
+
"eval_wer": 0.3466070702966274,
|
| 1158 |
+
"step": 1484
|
| 1159 |
+
},
|
| 1160 |
+
{
|
| 1161 |
+
"epoch": 107.0,
|
| 1162 |
+
"eval_cer": 0.17983621474067335,
|
| 1163 |
+
"eval_loss": 0.5722076892852783,
|
| 1164 |
+
"eval_runtime": 52.7987,
|
| 1165 |
+
"eval_samples_per_second": 10.265,
|
| 1166 |
+
"eval_steps_per_second": 1.288,
|
| 1167 |
+
"eval_wer": 0.34965461194636327,
|
| 1168 |
+
"step": 1498
|
| 1169 |
+
},
|
| 1170 |
+
{
|
| 1171 |
+
"epoch": 107.14,
|
| 1172 |
+
"learning_rate": 8.614285714285714e-06,
|
| 1173 |
+
"loss": 0.2181,
|
| 1174 |
+
"step": 1500
|
| 1175 |
+
},
|
| 1176 |
+
{
|
| 1177 |
+
"epoch": 108.0,
|
| 1178 |
+
"eval_cer": 0.17881710646041857,
|
| 1179 |
+
"eval_loss": 0.5552828907966614,
|
| 1180 |
+
"eval_runtime": 52.6793,
|
| 1181 |
+
"eval_samples_per_second": 10.289,
|
| 1182 |
+
"eval_steps_per_second": 1.291,
|
| 1183 |
+
"eval_wer": 0.3466070702966274,
|
| 1184 |
+
"step": 1512
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"epoch": 109.0,
|
| 1188 |
+
"eval_cer": 0.1792174704276615,
|
| 1189 |
+
"eval_loss": 0.5581598877906799,
|
| 1190 |
+
"eval_runtime": 52.8837,
|
| 1191 |
+
"eval_samples_per_second": 10.249,
|
| 1192 |
+
"eval_steps_per_second": 1.286,
|
| 1193 |
+
"eval_wer": 0.3484355952864689,
|
| 1194 |
+
"step": 1526
|
| 1195 |
+
},
|
| 1196 |
+
{
|
| 1197 |
+
"epoch": 110.0,
|
| 1198 |
+
"eval_cer": 0.18020018198362148,
|
| 1199 |
+
"eval_loss": 0.5701535940170288,
|
| 1200 |
+
"eval_runtime": 52.7198,
|
| 1201 |
+
"eval_samples_per_second": 10.281,
|
| 1202 |
+
"eval_steps_per_second": 1.29,
|
| 1203 |
+
"eval_wer": 0.35209264526615197,
|
| 1204 |
+
"step": 1540
|
| 1205 |
+
},
|
| 1206 |
+
{
|
| 1207 |
+
"epoch": 111.0,
|
| 1208 |
+
"eval_cer": 0.17976342129208372,
|
| 1209 |
+
"eval_loss": 0.5691486597061157,
|
| 1210 |
+
"eval_runtime": 52.7934,
|
| 1211 |
+
"eval_samples_per_second": 10.266,
|
| 1212 |
+
"eval_steps_per_second": 1.288,
|
| 1213 |
+
"eval_wer": 0.35046728971962615,
|
| 1214 |
+
"step": 1554
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"epoch": 112.0,
|
| 1218 |
+
"eval_cer": 0.17856232939035488,
|
| 1219 |
+
"eval_loss": 0.560352087020874,
|
| 1220 |
+
"eval_runtime": 52.6568,
|
| 1221 |
+
"eval_samples_per_second": 10.293,
|
| 1222 |
+
"eval_steps_per_second": 1.291,
|
| 1223 |
+
"eval_wer": 0.34701340918325885,
|
| 1224 |
+
"step": 1568
|
| 1225 |
+
},
|
| 1226 |
+
{
|
| 1227 |
+
"epoch": 113.0,
|
| 1228 |
+
"eval_cer": 0.17950864422202,
|
| 1229 |
+
"eval_loss": 0.5661062002182007,
|
| 1230 |
+
"eval_runtime": 52.8124,
|
| 1231 |
+
"eval_samples_per_second": 10.263,
|
| 1232 |
+
"eval_steps_per_second": 1.288,
|
| 1233 |
+
"eval_wer": 0.34823242584315317,
|
| 1234 |
+
"step": 1582
|
| 1235 |
+
},
|
| 1236 |
+
{
|
| 1237 |
+
"epoch": 114.0,
|
| 1238 |
+
"eval_cer": 0.17961783439490445,
|
| 1239 |
+
"eval_loss": 0.5682941675186157,
|
| 1240 |
+
"eval_runtime": 52.7057,
|
| 1241 |
+
"eval_samples_per_second": 10.284,
|
| 1242 |
+
"eval_steps_per_second": 1.29,
|
| 1243 |
+
"eval_wer": 0.35107679804957337,
|
| 1244 |
+
"step": 1596
|
| 1245 |
+
},
|
| 1246 |
+
{
|
| 1247 |
+
"epoch": 114.29,
|
| 1248 |
+
"learning_rate": 7.185714285714286e-06,
|
| 1249 |
+
"loss": 0.2171,
|
| 1250 |
+
"step": 1600
|
| 1251 |
+
},
|
| 1252 |
+
{
|
| 1253 |
+
"epoch": 115.0,
|
| 1254 |
+
"eval_cer": 0.17979981801637854,
|
| 1255 |
+
"eval_loss": 0.573845624923706,
|
| 1256 |
+
"eval_runtime": 52.7754,
|
| 1257 |
+
"eval_samples_per_second": 10.27,
|
| 1258 |
+
"eval_steps_per_second": 1.288,
|
| 1259 |
+
"eval_wer": 0.3508736286062576,
|
| 1260 |
+
"step": 1610
|
| 1261 |
+
},
|
| 1262 |
+
{
|
| 1263 |
+
"epoch": 116.0,
|
| 1264 |
+
"eval_cer": 0.17932666060054595,
|
| 1265 |
+
"eval_loss": 0.5730240941047668,
|
| 1266 |
+
"eval_runtime": 52.7098,
|
| 1267 |
+
"eval_samples_per_second": 10.283,
|
| 1268 |
+
"eval_steps_per_second": 1.29,
|
| 1269 |
+
"eval_wer": 0.34579439252336447,
|
| 1270 |
+
"step": 1624
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"epoch": 117.0,
|
| 1274 |
+
"eval_cer": 0.178926296633303,
|
| 1275 |
+
"eval_loss": 0.5704598426818848,
|
| 1276 |
+
"eval_runtime": 52.6695,
|
| 1277 |
+
"eval_samples_per_second": 10.291,
|
| 1278 |
+
"eval_steps_per_second": 1.291,
|
| 1279 |
+
"eval_wer": 0.34559122308004875,
|
| 1280 |
+
"step": 1638
|
| 1281 |
+
},
|
| 1282 |
+
{
|
| 1283 |
+
"epoch": 118.0,
|
| 1284 |
+
"eval_cer": 0.17961783439490445,
|
| 1285 |
+
"eval_loss": 0.5813525319099426,
|
| 1286 |
+
"eval_runtime": 53.1651,
|
| 1287 |
+
"eval_samples_per_second": 10.195,
|
| 1288 |
+
"eval_steps_per_second": 1.279,
|
| 1289 |
+
"eval_wer": 0.3466070702966274,
|
| 1290 |
+
"step": 1652
|
| 1291 |
+
},
|
| 1292 |
+
{
|
| 1293 |
+
"epoch": 119.0,
|
| 1294 |
+
"eval_cer": 0.17907188353048226,
|
| 1295 |
+
"eval_loss": 0.5714964866638184,
|
| 1296 |
+
"eval_runtime": 52.764,
|
| 1297 |
+
"eval_samples_per_second": 10.272,
|
| 1298 |
+
"eval_steps_per_second": 1.289,
|
| 1299 |
+
"eval_wer": 0.3441690369768387,
|
| 1300 |
+
"step": 1666
|
| 1301 |
+
},
|
| 1302 |
+
{
|
| 1303 |
+
"epoch": 120.0,
|
| 1304 |
+
"eval_cer": 0.17976342129208372,
|
| 1305 |
+
"eval_loss": 0.5720311403274536,
|
| 1306 |
+
"eval_runtime": 53.0408,
|
| 1307 |
+
"eval_samples_per_second": 10.219,
|
| 1308 |
+
"eval_steps_per_second": 1.282,
|
| 1309 |
+
"eval_wer": 0.34701340918325885,
|
| 1310 |
+
"step": 1680
|
| 1311 |
+
},
|
| 1312 |
+
{
|
| 1313 |
+
"epoch": 121.0,
|
| 1314 |
+
"eval_cer": 0.17969062784349407,
|
| 1315 |
+
"eval_loss": 0.5768777132034302,
|
| 1316 |
+
"eval_runtime": 52.7851,
|
| 1317 |
+
"eval_samples_per_second": 10.268,
|
| 1318 |
+
"eval_steps_per_second": 1.288,
|
| 1319 |
+
"eval_wer": 0.34701340918325885,
|
| 1320 |
+
"step": 1694
|
| 1321 |
+
},
|
| 1322 |
+
{
|
| 1323 |
+
"epoch": 121.43,
|
| 1324 |
+
"learning_rate": 5.7571428571428574e-06,
|
| 1325 |
+
"loss": 0.1986,
|
| 1326 |
+
"step": 1700
|
| 1327 |
+
},
|
| 1328 |
+
{
|
| 1329 |
+
"epoch": 122.0,
|
| 1330 |
+
"eval_cer": 0.1791810737033667,
|
| 1331 |
+
"eval_loss": 0.571117639541626,
|
| 1332 |
+
"eval_runtime": 52.9656,
|
| 1333 |
+
"eval_samples_per_second": 10.233,
|
| 1334 |
+
"eval_steps_per_second": 1.284,
|
| 1335 |
+
"eval_wer": 0.3464039008533117,
|
| 1336 |
+
"step": 1708
|
| 1337 |
+
},
|
| 1338 |
+
{
|
| 1339 |
+
"epoch": 123.0,
|
| 1340 |
+
"eval_cer": 0.17903548680618744,
|
| 1341 |
+
"eval_loss": 0.5728168487548828,
|
| 1342 |
+
"eval_runtime": 52.927,
|
| 1343 |
+
"eval_samples_per_second": 10.241,
|
| 1344 |
+
"eval_steps_per_second": 1.285,
|
| 1345 |
+
"eval_wer": 0.3441690369768387,
|
| 1346 |
+
"step": 1722
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 124.0,
|
| 1350 |
+
"eval_cer": 0.17830755232029116,
|
| 1351 |
+
"eval_loss": 0.5667761564254761,
|
| 1352 |
+
"eval_runtime": 52.78,
|
| 1353 |
+
"eval_samples_per_second": 10.269,
|
| 1354 |
+
"eval_steps_per_second": 1.288,
|
| 1355 |
+
"eval_wer": 0.3449817147501016,
|
| 1356 |
+
"step": 1736
|
| 1357 |
+
},
|
| 1358 |
+
{
|
| 1359 |
+
"epoch": 125.0,
|
| 1360 |
+
"eval_cer": 0.17969062784349407,
|
| 1361 |
+
"eval_loss": 0.5855135321617126,
|
| 1362 |
+
"eval_runtime": 52.7911,
|
| 1363 |
+
"eval_samples_per_second": 10.267,
|
| 1364 |
+
"eval_steps_per_second": 1.288,
|
| 1365 |
+
"eval_wer": 0.3484355952864689,
|
| 1366 |
+
"step": 1750
|
| 1367 |
+
},
|
| 1368 |
+
{
|
| 1369 |
+
"epoch": 126.0,
|
| 1370 |
+
"eval_cer": 0.17827115559599635,
|
| 1371 |
+
"eval_loss": 0.5667468905448914,
|
| 1372 |
+
"eval_runtime": 52.8968,
|
| 1373 |
+
"eval_samples_per_second": 10.246,
|
| 1374 |
+
"eval_steps_per_second": 1.286,
|
| 1375 |
+
"eval_wer": 0.3427468508736286,
|
| 1376 |
+
"step": 1764
|
| 1377 |
+
},
|
| 1378 |
+
{
|
| 1379 |
+
"epoch": 127.0,
|
| 1380 |
+
"eval_cer": 0.1788898999090082,
|
| 1381 |
+
"eval_loss": 0.5710840821266174,
|
| 1382 |
+
"eval_runtime": 52.7705,
|
| 1383 |
+
"eval_samples_per_second": 10.271,
|
| 1384 |
+
"eval_steps_per_second": 1.289,
|
| 1385 |
+
"eval_wer": 0.3459975619666802,
|
| 1386 |
+
"step": 1778
|
| 1387 |
+
},
|
| 1388 |
+
{
|
| 1389 |
+
"epoch": 128.0,
|
| 1390 |
+
"eval_cer": 0.1781255686988171,
|
| 1391 |
+
"eval_loss": 0.5681577920913696,
|
| 1392 |
+
"eval_runtime": 52.7919,
|
| 1393 |
+
"eval_samples_per_second": 10.267,
|
| 1394 |
+
"eval_steps_per_second": 1.288,
|
| 1395 |
+
"eval_wer": 0.3443722064201544,
|
| 1396 |
+
"step": 1792
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"epoch": 128.0,
|
| 1400 |
+
"step": 1792,
|
| 1401 |
+
"total_flos": 1.5995217836155625e+19,
|
| 1402 |
+
"train_loss": 1.3990157055003303,
|
| 1403 |
+
"train_runtime": 26333.2616,
|
| 1404 |
+
"train_samples_per_second": 2.552,
|
| 1405 |
+
"train_steps_per_second": 0.08
|
| 1406 |
+
}
|
| 1407 |
+
],
|
| 1408 |
+
"max_steps": 2100,
|
| 1409 |
+
"num_train_epochs": 150,
|
| 1410 |
+
"total_flos": 1.5995217836155625e+19,
|
| 1411 |
+
"trial_name": null,
|
| 1412 |
+
"trial_params": null
|
| 1413 |
+
}
|