Commit
•
a40988b
1
Parent(s):
43c0931
End of training
Browse files- all_results.json +14 -0
- eval_results.json +9 -0
- runs/Feb03_09-23-17_ficino/events.out.tfevents.1643992561.ficino.366652.2 +3 -0
- train_results.json +8 -0
- trainer_state.json +2779 -0
all_results.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 20.0,
|
| 3 |
+
"eval_loss": 0.19566693902015686,
|
| 4 |
+
"eval_runtime": 441.5982,
|
| 5 |
+
"eval_samples": 6208,
|
| 6 |
+
"eval_samples_per_second": 14.058,
|
| 7 |
+
"eval_steps_per_second": 0.879,
|
| 8 |
+
"eval_wer": 0.16972918437695592,
|
| 9 |
+
"train_loss": 1.2649466082470566,
|
| 10 |
+
"train_runtime": 115228.5602,
|
| 11 |
+
"train_samples": 56673,
|
| 12 |
+
"train_samples_per_second": 9.837,
|
| 13 |
+
"train_steps_per_second": 0.154
|
| 14 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 20.0,
|
| 3 |
+
"eval_loss": 0.19566693902015686,
|
| 4 |
+
"eval_runtime": 441.5982,
|
| 5 |
+
"eval_samples": 6208,
|
| 6 |
+
"eval_samples_per_second": 14.058,
|
| 7 |
+
"eval_steps_per_second": 0.879,
|
| 8 |
+
"eval_wer": 0.16972918437695592
|
| 9 |
+
}
|
runs/Feb03_09-23-17_ficino/events.out.tfevents.1643992561.ficino.366652.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95127f5c15a6d2e58e0c0d1196729c98d009426e214db10185819378cf36e9fe
|
| 3 |
+
size 364
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 20.0,
|
| 3 |
+
"train_loss": 1.2649466082470566,
|
| 4 |
+
"train_runtime": 115228.5602,
|
| 5 |
+
"train_samples": 56673,
|
| 6 |
+
"train_samples_per_second": 9.837,
|
| 7 |
+
"train_steps_per_second": 0.154
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2779 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 19.999153259949196,
|
| 5 |
+
"global_step": 17700,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.06,
|
| 12 |
+
"learning_rate": 1.8749999999999998e-06,
|
| 13 |
+
"loss": 9.7932,
|
| 14 |
+
"step": 50
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.11,
|
| 18 |
+
"learning_rate": 3.7125e-06,
|
| 19 |
+
"loss": 9.6218,
|
| 20 |
+
"step": 100
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.17,
|
| 24 |
+
"learning_rate": 5.549999999999999e-06,
|
| 25 |
+
"loss": 7.6384,
|
| 26 |
+
"step": 150
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.23,
|
| 30 |
+
"learning_rate": 7.425e-06,
|
| 31 |
+
"loss": 5.5724,
|
| 32 |
+
"step": 200
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.28,
|
| 36 |
+
"learning_rate": 9.299999999999999e-06,
|
| 37 |
+
"loss": 4.4527,
|
| 38 |
+
"step": 250
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.28,
|
| 42 |
+
"eval_loss": 4.014413833618164,
|
| 43 |
+
"eval_runtime": 432.9366,
|
| 44 |
+
"eval_samples_per_second": 14.339,
|
| 45 |
+
"eval_steps_per_second": 0.896,
|
| 46 |
+
"eval_wer": 1.0,
|
| 47 |
+
"step": 250
|
| 48 |
+
},
|
| 49 |
+
{
|
| 50 |
+
"epoch": 0.34,
|
| 51 |
+
"learning_rate": 1.1174999999999999e-05,
|
| 52 |
+
"loss": 3.996,
|
| 53 |
+
"step": 300
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"epoch": 0.4,
|
| 57 |
+
"learning_rate": 1.3049999999999999e-05,
|
| 58 |
+
"loss": 3.6961,
|
| 59 |
+
"step": 350
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"epoch": 0.45,
|
| 63 |
+
"learning_rate": 1.4925e-05,
|
| 64 |
+
"loss": 3.4442,
|
| 65 |
+
"step": 400
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.51,
|
| 69 |
+
"learning_rate": 1.68e-05,
|
| 70 |
+
"loss": 3.3442,
|
| 71 |
+
"step": 450
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"epoch": 0.56,
|
| 75 |
+
"learning_rate": 1.8675e-05,
|
| 76 |
+
"loss": 3.1828,
|
| 77 |
+
"step": 500
|
| 78 |
+
},
|
| 79 |
+
{
|
| 80 |
+
"epoch": 0.56,
|
| 81 |
+
"eval_loss": 3.136876106262207,
|
| 82 |
+
"eval_runtime": 432.521,
|
| 83 |
+
"eval_samples_per_second": 14.353,
|
| 84 |
+
"eval_steps_per_second": 0.897,
|
| 85 |
+
"eval_wer": 1.0,
|
| 86 |
+
"step": 500
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.62,
|
| 90 |
+
"learning_rate": 2.055e-05,
|
| 91 |
+
"loss": 3.1052,
|
| 92 |
+
"step": 550
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 0.68,
|
| 96 |
+
"learning_rate": 2.2424999999999996e-05,
|
| 97 |
+
"loss": 3.0545,
|
| 98 |
+
"step": 600
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 0.73,
|
| 102 |
+
"learning_rate": 2.4299999999999998e-05,
|
| 103 |
+
"loss": 3.0155,
|
| 104 |
+
"step": 650
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 0.79,
|
| 108 |
+
"learning_rate": 2.6174999999999996e-05,
|
| 109 |
+
"loss": 3.0148,
|
| 110 |
+
"step": 700
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 0.85,
|
| 114 |
+
"learning_rate": 2.8049999999999997e-05,
|
| 115 |
+
"loss": 2.9927,
|
| 116 |
+
"step": 750
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 0.85,
|
| 120 |
+
"eval_loss": 3.0182671546936035,
|
| 121 |
+
"eval_runtime": 435.1578,
|
| 122 |
+
"eval_samples_per_second": 14.266,
|
| 123 |
+
"eval_steps_per_second": 0.892,
|
| 124 |
+
"eval_wer": 1.0,
|
| 125 |
+
"step": 750
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"epoch": 0.9,
|
| 129 |
+
"learning_rate": 2.9925e-05,
|
| 130 |
+
"loss": 2.9829,
|
| 131 |
+
"step": 800
|
| 132 |
+
},
|
| 133 |
+
{
|
| 134 |
+
"epoch": 0.96,
|
| 135 |
+
"learning_rate": 3.1799999999999994e-05,
|
| 136 |
+
"loss": 2.9876,
|
| 137 |
+
"step": 850
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.02,
|
| 141 |
+
"learning_rate": 3.3675e-05,
|
| 142 |
+
"loss": 3.0304,
|
| 143 |
+
"step": 900
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"epoch": 1.07,
|
| 147 |
+
"learning_rate": 3.555e-05,
|
| 148 |
+
"loss": 2.9783,
|
| 149 |
+
"step": 950
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 1.13,
|
| 153 |
+
"learning_rate": 3.7424999999999995e-05,
|
| 154 |
+
"loss": 2.9591,
|
| 155 |
+
"step": 1000
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"epoch": 1.13,
|
| 159 |
+
"eval_loss": 2.999102830886841,
|
| 160 |
+
"eval_runtime": 430.4872,
|
| 161 |
+
"eval_samples_per_second": 14.421,
|
| 162 |
+
"eval_steps_per_second": 0.901,
|
| 163 |
+
"eval_wer": 1.0,
|
| 164 |
+
"step": 1000
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 1.19,
|
| 168 |
+
"learning_rate": 3.93e-05,
|
| 169 |
+
"loss": 2.9559,
|
| 170 |
+
"step": 1050
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 1.24,
|
| 174 |
+
"learning_rate": 4.1175e-05,
|
| 175 |
+
"loss": 2.9388,
|
| 176 |
+
"step": 1100
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 1.3,
|
| 180 |
+
"learning_rate": 4.3049999999999996e-05,
|
| 181 |
+
"loss": 2.9321,
|
| 182 |
+
"step": 1150
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 1.36,
|
| 186 |
+
"learning_rate": 4.4924999999999994e-05,
|
| 187 |
+
"loss": 2.9205,
|
| 188 |
+
"step": 1200
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 1.41,
|
| 192 |
+
"learning_rate": 4.68e-05,
|
| 193 |
+
"loss": 2.8989,
|
| 194 |
+
"step": 1250
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 1.41,
|
| 198 |
+
"eval_loss": 2.9000213146209717,
|
| 199 |
+
"eval_runtime": 431.5059,
|
| 200 |
+
"eval_samples_per_second": 14.387,
|
| 201 |
+
"eval_steps_per_second": 0.899,
|
| 202 |
+
"eval_wer": 0.999990658308967,
|
| 203 |
+
"step": 1250
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"epoch": 1.47,
|
| 207 |
+
"learning_rate": 4.8675e-05,
|
| 208 |
+
"loss": 2.8682,
|
| 209 |
+
"step": 1300
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 1.52,
|
| 213 |
+
"learning_rate": 5.055e-05,
|
| 214 |
+
"loss": 2.8476,
|
| 215 |
+
"step": 1350
|
| 216 |
+
},
|
| 217 |
+
{
|
| 218 |
+
"epoch": 1.58,
|
| 219 |
+
"learning_rate": 5.2424999999999994e-05,
|
| 220 |
+
"loss": 2.7956,
|
| 221 |
+
"step": 1400
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 1.64,
|
| 225 |
+
"learning_rate": 5.429999999999999e-05,
|
| 226 |
+
"loss": 2.6754,
|
| 227 |
+
"step": 1450
|
| 228 |
+
},
|
| 229 |
+
{
|
| 230 |
+
"epoch": 1.69,
|
| 231 |
+
"learning_rate": 5.6175e-05,
|
| 232 |
+
"loss": 2.4286,
|
| 233 |
+
"step": 1500
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 1.69,
|
| 237 |
+
"eval_loss": 1.7688498497009277,
|
| 238 |
+
"eval_runtime": 430.3663,
|
| 239 |
+
"eval_samples_per_second": 14.425,
|
| 240 |
+
"eval_steps_per_second": 0.902,
|
| 241 |
+
"eval_wer": 0.9550384410586005,
|
| 242 |
+
"step": 1500
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 1.75,
|
| 246 |
+
"learning_rate": 5.8049999999999995e-05,
|
| 247 |
+
"loss": 2.218,
|
| 248 |
+
"step": 1550
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 1.81,
|
| 252 |
+
"learning_rate": 5.9925e-05,
|
| 253 |
+
"loss": 2.0095,
|
| 254 |
+
"step": 1600
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 1.86,
|
| 258 |
+
"learning_rate": 6.18e-05,
|
| 259 |
+
"loss": 1.8416,
|
| 260 |
+
"step": 1650
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 1.92,
|
| 264 |
+
"learning_rate": 6.367499999999999e-05,
|
| 265 |
+
"loss": 1.7642,
|
| 266 |
+
"step": 1700
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 1.98,
|
| 270 |
+
"learning_rate": 6.555e-05,
|
| 271 |
+
"loss": 1.6765,
|
| 272 |
+
"step": 1750
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 1.98,
|
| 276 |
+
"eval_loss": 0.6841917037963867,
|
| 277 |
+
"eval_runtime": 433.019,
|
| 278 |
+
"eval_samples_per_second": 14.337,
|
| 279 |
+
"eval_steps_per_second": 0.896,
|
| 280 |
+
"eval_wer": 0.48551570805347183,
|
| 281 |
+
"step": 1750
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 2.03,
|
| 285 |
+
"learning_rate": 6.7425e-05,
|
| 286 |
+
"loss": 1.5994,
|
| 287 |
+
"step": 1800
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"epoch": 2.09,
|
| 291 |
+
"learning_rate": 6.93e-05,
|
| 292 |
+
"loss": 1.5522,
|
| 293 |
+
"step": 1850
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 2.15,
|
| 297 |
+
"learning_rate": 7.1175e-05,
|
| 298 |
+
"loss": 1.52,
|
| 299 |
+
"step": 1900
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"epoch": 2.2,
|
| 303 |
+
"learning_rate": 7.304999999999999e-05,
|
| 304 |
+
"loss": 1.5086,
|
| 305 |
+
"step": 1950
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 2.26,
|
| 309 |
+
"learning_rate": 7.492499999999999e-05,
|
| 310 |
+
"loss": 1.4521,
|
| 311 |
+
"step": 2000
|
| 312 |
+
},
|
| 313 |
+
{
|
| 314 |
+
"epoch": 2.26,
|
| 315 |
+
"eval_loss": 0.5096011757850647,
|
| 316 |
+
"eval_runtime": 431.8266,
|
| 317 |
+
"eval_samples_per_second": 14.376,
|
| 318 |
+
"eval_steps_per_second": 0.899,
|
| 319 |
+
"eval_wer": 0.3735835660971349,
|
| 320 |
+
"step": 2000
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 2.32,
|
| 324 |
+
"learning_rate": 7.477070063694266e-05,
|
| 325 |
+
"loss": 1.4457,
|
| 326 |
+
"step": 2050
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 2.37,
|
| 330 |
+
"learning_rate": 7.453184713375795e-05,
|
| 331 |
+
"loss": 1.4276,
|
| 332 |
+
"step": 2100
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 2.43,
|
| 336 |
+
"learning_rate": 7.429299363057323e-05,
|
| 337 |
+
"loss": 1.4028,
|
| 338 |
+
"step": 2150
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 2.49,
|
| 342 |
+
"learning_rate": 7.405414012738853e-05,
|
| 343 |
+
"loss": 1.3887,
|
| 344 |
+
"step": 2200
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 2.54,
|
| 348 |
+
"learning_rate": 7.38152866242038e-05,
|
| 349 |
+
"loss": 1.3589,
|
| 350 |
+
"step": 2250
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 2.54,
|
| 354 |
+
"eval_loss": 0.44788965582847595,
|
| 355 |
+
"eval_runtime": 430.2855,
|
| 356 |
+
"eval_samples_per_second": 14.428,
|
| 357 |
+
"eval_steps_per_second": 0.902,
|
| 358 |
+
"eval_wer": 0.3335450783300793,
|
| 359 |
+
"step": 2250
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 2.6,
|
| 363 |
+
"learning_rate": 7.35764331210191e-05,
|
| 364 |
+
"loss": 1.3935,
|
| 365 |
+
"step": 2300
|
| 366 |
+
},
|
| 367 |
+
{
|
| 368 |
+
"epoch": 2.65,
|
| 369 |
+
"learning_rate": 7.333757961783438e-05,
|
| 370 |
+
"loss": 1.3425,
|
| 371 |
+
"step": 2350
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 2.71,
|
| 375 |
+
"learning_rate": 7.309872611464967e-05,
|
| 376 |
+
"loss": 1.3657,
|
| 377 |
+
"step": 2400
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 2.77,
|
| 381 |
+
"learning_rate": 7.285987261146495e-05,
|
| 382 |
+
"loss": 1.3645,
|
| 383 |
+
"step": 2450
|
| 384 |
+
},
|
| 385 |
+
{
|
| 386 |
+
"epoch": 2.82,
|
| 387 |
+
"learning_rate": 7.262101910828025e-05,
|
| 388 |
+
"loss": 1.3136,
|
| 389 |
+
"step": 2500
|
| 390 |
+
},
|
| 391 |
+
{
|
| 392 |
+
"epoch": 2.82,
|
| 393 |
+
"eval_loss": 0.40564054250717163,
|
| 394 |
+
"eval_runtime": 428.4501,
|
| 395 |
+
"eval_samples_per_second": 14.489,
|
| 396 |
+
"eval_steps_per_second": 0.906,
|
| 397 |
+
"eval_wer": 0.3123020729212402,
|
| 398 |
+
"step": 2500
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"epoch": 2.88,
|
| 402 |
+
"learning_rate": 7.238216560509553e-05,
|
| 403 |
+
"loss": 1.3415,
|
| 404 |
+
"step": 2550
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 2.94,
|
| 408 |
+
"learning_rate": 7.214331210191082e-05,
|
| 409 |
+
"loss": 1.3345,
|
| 410 |
+
"step": 2600
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 2.99,
|
| 414 |
+
"learning_rate": 7.19044585987261e-05,
|
| 415 |
+
"loss": 1.3283,
|
| 416 |
+
"step": 2650
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 3.05,
|
| 420 |
+
"learning_rate": 7.16656050955414e-05,
|
| 421 |
+
"loss": 1.2788,
|
| 422 |
+
"step": 2700
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 3.11,
|
| 426 |
+
"learning_rate": 7.142675159235667e-05,
|
| 427 |
+
"loss": 1.2856,
|
| 428 |
+
"step": 2750
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"epoch": 3.11,
|
| 432 |
+
"eval_loss": 0.38699424266815186,
|
| 433 |
+
"eval_runtime": 430.1514,
|
| 434 |
+
"eval_samples_per_second": 14.432,
|
| 435 |
+
"eval_steps_per_second": 0.902,
|
| 436 |
+
"eval_wer": 0.29870991246835504,
|
| 437 |
+
"step": 2750
|
| 438 |
+
},
|
| 439 |
+
{
|
| 440 |
+
"epoch": 3.16,
|
| 441 |
+
"learning_rate": 7.118789808917197e-05,
|
| 442 |
+
"loss": 1.2817,
|
| 443 |
+
"step": 2800
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 3.22,
|
| 447 |
+
"learning_rate": 7.094904458598725e-05,
|
| 448 |
+
"loss": 1.2502,
|
| 449 |
+
"step": 2850
|
| 450 |
+
},
|
| 451 |
+
{
|
| 452 |
+
"epoch": 3.28,
|
| 453 |
+
"learning_rate": 7.071019108280254e-05,
|
| 454 |
+
"loss": 1.2623,
|
| 455 |
+
"step": 2900
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"epoch": 3.33,
|
| 459 |
+
"learning_rate": 7.047133757961782e-05,
|
| 460 |
+
"loss": 1.2302,
|
| 461 |
+
"step": 2950
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 3.39,
|
| 465 |
+
"learning_rate": 7.023248407643311e-05,
|
| 466 |
+
"loss": 1.2283,
|
| 467 |
+
"step": 3000
|
| 468 |
+
},
|
| 469 |
+
{
|
| 470 |
+
"epoch": 3.39,
|
| 471 |
+
"eval_loss": 0.3645668029785156,
|
| 472 |
+
"eval_runtime": 430.0013,
|
| 473 |
+
"eval_samples_per_second": 14.437,
|
| 474 |
+
"eval_steps_per_second": 0.902,
|
| 475 |
+
"eval_wer": 0.2828290377124067,
|
| 476 |
+
"step": 3000
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 3.45,
|
| 480 |
+
"learning_rate": 6.99936305732484e-05,
|
| 481 |
+
"loss": 1.1993,
|
| 482 |
+
"step": 3050
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 3.5,
|
| 486 |
+
"learning_rate": 6.975477707006369e-05,
|
| 487 |
+
"loss": 1.2627,
|
| 488 |
+
"step": 3100
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 3.56,
|
| 492 |
+
"learning_rate": 6.951592356687897e-05,
|
| 493 |
+
"loss": 1.1969,
|
| 494 |
+
"step": 3150
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 3.62,
|
| 498 |
+
"learning_rate": 6.927707006369426e-05,
|
| 499 |
+
"loss": 1.2054,
|
| 500 |
+
"step": 3200
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 3.67,
|
| 504 |
+
"learning_rate": 6.903821656050954e-05,
|
| 505 |
+
"loss": 1.2053,
|
| 506 |
+
"step": 3250
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 3.67,
|
| 510 |
+
"eval_loss": 0.3499177098274231,
|
| 511 |
+
"eval_runtime": 429.329,
|
| 512 |
+
"eval_samples_per_second": 14.46,
|
| 513 |
+
"eval_steps_per_second": 0.904,
|
| 514 |
+
"eval_wer": 0.2747578166599718,
|
| 515 |
+
"step": 3250
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 3.73,
|
| 519 |
+
"learning_rate": 6.879936305732483e-05,
|
| 520 |
+
"loss": 1.2144,
|
| 521 |
+
"step": 3300
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 3.78,
|
| 525 |
+
"learning_rate": 6.856050955414011e-05,
|
| 526 |
+
"loss": 1.1882,
|
| 527 |
+
"step": 3350
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"epoch": 3.84,
|
| 531 |
+
"learning_rate": 6.832165605095541e-05,
|
| 532 |
+
"loss": 1.1901,
|
| 533 |
+
"step": 3400
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 3.9,
|
| 537 |
+
"learning_rate": 6.808280254777069e-05,
|
| 538 |
+
"loss": 1.2064,
|
| 539 |
+
"step": 3450
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 3.95,
|
| 543 |
+
"learning_rate": 6.784394904458598e-05,
|
| 544 |
+
"loss": 1.2087,
|
| 545 |
+
"step": 3500
|
| 546 |
+
},
|
| 547 |
+
{
|
| 548 |
+
"epoch": 3.95,
|
| 549 |
+
"eval_loss": 0.3345482349395752,
|
| 550 |
+
"eval_runtime": 430.4222,
|
| 551 |
+
"eval_samples_per_second": 14.423,
|
| 552 |
+
"eval_steps_per_second": 0.901,
|
| 553 |
+
"eval_wer": 0.2602781955589601,
|
| 554 |
+
"step": 3500
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 4.01,
|
| 558 |
+
"learning_rate": 6.760509554140126e-05,
|
| 559 |
+
"loss": 1.1945,
|
| 560 |
+
"step": 3550
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 4.07,
|
| 564 |
+
"learning_rate": 6.736624203821655e-05,
|
| 565 |
+
"loss": 1.1674,
|
| 566 |
+
"step": 3600
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 4.12,
|
| 570 |
+
"learning_rate": 6.712738853503183e-05,
|
| 571 |
+
"loss": 1.2197,
|
| 572 |
+
"step": 3650
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 4.18,
|
| 576 |
+
"learning_rate": 6.688853503184713e-05,
|
| 577 |
+
"loss": 1.1832,
|
| 578 |
+
"step": 3700
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 4.24,
|
| 582 |
+
"learning_rate": 6.664968152866241e-05,
|
| 583 |
+
"loss": 1.2002,
|
| 584 |
+
"step": 3750
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 4.24,
|
| 588 |
+
"eval_loss": 0.3320307731628418,
|
| 589 |
+
"eval_runtime": 429.9654,
|
| 590 |
+
"eval_samples_per_second": 14.438,
|
| 591 |
+
"eval_steps_per_second": 0.902,
|
| 592 |
+
"eval_wer": 0.25228170803478844,
|
| 593 |
+
"step": 3750
|
| 594 |
+
},
|
| 595 |
+
{
|
| 596 |
+
"epoch": 4.29,
|
| 597 |
+
"learning_rate": 6.64108280254777e-05,
|
| 598 |
+
"loss": 1.1655,
|
| 599 |
+
"step": 3800
|
| 600 |
+
},
|
| 601 |
+
{
|
| 602 |
+
"epoch": 4.35,
|
| 603 |
+
"learning_rate": 6.617197452229298e-05,
|
| 604 |
+
"loss": 1.1387,
|
| 605 |
+
"step": 3850
|
| 606 |
+
},
|
| 607 |
+
{
|
| 608 |
+
"epoch": 4.41,
|
| 609 |
+
"learning_rate": 6.593312101910828e-05,
|
| 610 |
+
"loss": 1.1344,
|
| 611 |
+
"step": 3900
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 4.46,
|
| 615 |
+
"learning_rate": 6.569426751592356e-05,
|
| 616 |
+
"loss": 1.169,
|
| 617 |
+
"step": 3950
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 4.52,
|
| 621 |
+
"learning_rate": 6.545541401273885e-05,
|
| 622 |
+
"loss": 1.1383,
|
| 623 |
+
"step": 4000
|
| 624 |
+
},
|
| 625 |
+
{
|
| 626 |
+
"epoch": 4.52,
|
| 627 |
+
"eval_loss": 0.31172633171081543,
|
| 628 |
+
"eval_runtime": 428.4618,
|
| 629 |
+
"eval_samples_per_second": 14.489,
|
| 630 |
+
"eval_steps_per_second": 0.906,
|
| 631 |
+
"eval_wer": 0.24393957794239912,
|
| 632 |
+
"step": 4000
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 4.58,
|
| 636 |
+
"learning_rate": 6.521656050955413e-05,
|
| 637 |
+
"loss": 1.1241,
|
| 638 |
+
"step": 4050
|
| 639 |
+
},
|
| 640 |
+
{
|
| 641 |
+
"epoch": 4.63,
|
| 642 |
+
"learning_rate": 6.497770700636942e-05,
|
| 643 |
+
"loss": 1.1505,
|
| 644 |
+
"step": 4100
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 4.69,
|
| 648 |
+
"learning_rate": 6.47388535031847e-05,
|
| 649 |
+
"loss": 1.1309,
|
| 650 |
+
"step": 4150
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 4.75,
|
| 654 |
+
"learning_rate": 6.45e-05,
|
| 655 |
+
"loss": 1.1368,
|
| 656 |
+
"step": 4200
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 4.8,
|
| 660 |
+
"learning_rate": 6.426114649681528e-05,
|
| 661 |
+
"loss": 1.1364,
|
| 662 |
+
"step": 4250
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 4.8,
|
| 666 |
+
"eval_loss": 0.3198467195034027,
|
| 667 |
+
"eval_runtime": 427.239,
|
| 668 |
+
"eval_samples_per_second": 14.531,
|
| 669 |
+
"eval_steps_per_second": 0.908,
|
| 670 |
+
"eval_wer": 0.2382878548674881,
|
| 671 |
+
"step": 4250
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 4.86,
|
| 675 |
+
"learning_rate": 6.402229299363057e-05,
|
| 676 |
+
"loss": 1.1185,
|
| 677 |
+
"step": 4300
|
| 678 |
+
},
|
| 679 |
+
{
|
| 680 |
+
"epoch": 4.91,
|
| 681 |
+
"learning_rate": 6.378343949044585e-05,
|
| 682 |
+
"loss": 1.1214,
|
| 683 |
+
"step": 4350
|
| 684 |
+
},
|
| 685 |
+
{
|
| 686 |
+
"epoch": 4.97,
|
| 687 |
+
"learning_rate": 6.354458598726114e-05,
|
| 688 |
+
"loss": 1.1188,
|
| 689 |
+
"step": 4400
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 5.03,
|
| 693 |
+
"learning_rate": 6.330573248407642e-05,
|
| 694 |
+
"loss": 1.1327,
|
| 695 |
+
"step": 4450
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"epoch": 5.08,
|
| 699 |
+
"learning_rate": 6.306687898089172e-05,
|
| 700 |
+
"loss": 1.158,
|
| 701 |
+
"step": 4500
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"epoch": 5.08,
|
| 705 |
+
"eval_loss": 0.3070796728134155,
|
| 706 |
+
"eval_runtime": 427.2037,
|
| 707 |
+
"eval_samples_per_second": 14.532,
|
| 708 |
+
"eval_steps_per_second": 0.908,
|
| 709 |
+
"eval_wer": 0.23418685250404028,
|
| 710 |
+
"step": 4500
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"epoch": 5.14,
|
| 714 |
+
"learning_rate": 6.2828025477707e-05,
|
| 715 |
+
"loss": 1.1221,
|
| 716 |
+
"step": 4550
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 5.2,
|
| 720 |
+
"learning_rate": 6.258917197452229e-05,
|
| 721 |
+
"loss": 1.1167,
|
| 722 |
+
"step": 4600
|
| 723 |
+
},
|
| 724 |
+
{
|
| 725 |
+
"epoch": 5.25,
|
| 726 |
+
"learning_rate": 6.235031847133757e-05,
|
| 727 |
+
"loss": 1.1067,
|
| 728 |
+
"step": 4650
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 5.31,
|
| 732 |
+
"learning_rate": 6.211146496815286e-05,
|
| 733 |
+
"loss": 1.099,
|
| 734 |
+
"step": 4700
|
| 735 |
+
},
|
| 736 |
+
{
|
| 737 |
+
"epoch": 5.37,
|
| 738 |
+
"learning_rate": 6.187261146496814e-05,
|
| 739 |
+
"loss": 1.108,
|
| 740 |
+
"step": 4750
|
| 741 |
+
},
|
| 742 |
+
{
|
| 743 |
+
"epoch": 5.37,
|
| 744 |
+
"eval_loss": 0.3011206090450287,
|
| 745 |
+
"eval_runtime": 430.4576,
|
| 746 |
+
"eval_samples_per_second": 14.422,
|
| 747 |
+
"eval_steps_per_second": 0.901,
|
| 748 |
+
"eval_wer": 0.23136566181210122,
|
| 749 |
+
"step": 4750
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 5.42,
|
| 753 |
+
"learning_rate": 6.163375796178344e-05,
|
| 754 |
+
"loss": 1.1024,
|
| 755 |
+
"step": 4800
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 5.48,
|
| 759 |
+
"learning_rate": 6.139490445859872e-05,
|
| 760 |
+
"loss": 1.1039,
|
| 761 |
+
"step": 4850
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 5.54,
|
| 765 |
+
"learning_rate": 6.115605095541401e-05,
|
| 766 |
+
"loss": 1.1082,
|
| 767 |
+
"step": 4900
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 5.59,
|
| 771 |
+
"learning_rate": 6.09171974522293e-05,
|
| 772 |
+
"loss": 1.0982,
|
| 773 |
+
"step": 4950
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 5.65,
|
| 777 |
+
"learning_rate": 6.0678343949044583e-05,
|
| 778 |
+
"loss": 1.1025,
|
| 779 |
+
"step": 5000
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"epoch": 5.65,
|
| 783 |
+
"eval_loss": 0.28753861784935,
|
| 784 |
+
"eval_runtime": 431.3779,
|
| 785 |
+
"eval_samples_per_second": 14.391,
|
| 786 |
+
"eval_steps_per_second": 0.899,
|
| 787 |
+
"eval_wer": 0.2289368221435444,
|
| 788 |
+
"step": 5000
|
| 789 |
+
},
|
| 790 |
+
{
|
| 791 |
+
"epoch": 5.71,
|
| 792 |
+
"learning_rate": 6.043949044585987e-05,
|
| 793 |
+
"loss": 1.089,
|
| 794 |
+
"step": 5050
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 5.76,
|
| 798 |
+
"learning_rate": 6.020063694267516e-05,
|
| 799 |
+
"loss": 1.0792,
|
| 800 |
+
"step": 5100
|
| 801 |
+
},
|
| 802 |
+
{
|
| 803 |
+
"epoch": 5.82,
|
| 804 |
+
"learning_rate": 5.9961783439490444e-05,
|
| 805 |
+
"loss": 1.1054,
|
| 806 |
+
"step": 5150
|
| 807 |
+
},
|
| 808 |
+
{
|
| 809 |
+
"epoch": 5.87,
|
| 810 |
+
"learning_rate": 5.972770700636942e-05,
|
| 811 |
+
"loss": 1.078,
|
| 812 |
+
"step": 5200
|
| 813 |
+
},
|
| 814 |
+
{
|
| 815 |
+
"epoch": 5.93,
|
| 816 |
+
"learning_rate": 5.948885350318471e-05,
|
| 817 |
+
"loss": 1.0697,
|
| 818 |
+
"step": 5250
|
| 819 |
+
},
|
| 820 |
+
{
|
| 821 |
+
"epoch": 5.93,
|
| 822 |
+
"eval_loss": 0.29261597990989685,
|
| 823 |
+
"eval_runtime": 429.3286,
|
| 824 |
+
"eval_samples_per_second": 14.46,
|
| 825 |
+
"eval_steps_per_second": 0.904,
|
| 826 |
+
"eval_wer": 0.22559249675376236,
|
| 827 |
+
"step": 5250
|
| 828 |
+
},
|
| 829 |
+
{
|
| 830 |
+
"epoch": 5.99,
|
| 831 |
+
"learning_rate": 5.925e-05,
|
| 832 |
+
"loss": 1.1183,
|
| 833 |
+
"step": 5300
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 6.05,
|
| 837 |
+
"learning_rate": 5.9011146496815284e-05,
|
| 838 |
+
"loss": 1.1614,
|
| 839 |
+
"step": 5350
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 6.1,
|
| 843 |
+
"learning_rate": 5.877229299363057e-05,
|
| 844 |
+
"loss": 1.075,
|
| 845 |
+
"step": 5400
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 6.16,
|
| 849 |
+
"learning_rate": 5.853343949044586e-05,
|
| 850 |
+
"loss": 1.0901,
|
| 851 |
+
"step": 5450
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 6.21,
|
| 855 |
+
"learning_rate": 5.8294585987261144e-05,
|
| 856 |
+
"loss": 1.0904,
|
| 857 |
+
"step": 5500
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 6.21,
|
| 861 |
+
"eval_loss": 0.2695116698741913,
|
| 862 |
+
"eval_runtime": 431.1678,
|
| 863 |
+
"eval_samples_per_second": 14.398,
|
| 864 |
+
"eval_steps_per_second": 0.9,
|
| 865 |
+
"eval_wer": 0.22445281044774726,
|
| 866 |
+
"step": 5500
|
| 867 |
+
},
|
| 868 |
+
{
|
| 869 |
+
"epoch": 6.27,
|
| 870 |
+
"learning_rate": 5.805573248407643e-05,
|
| 871 |
+
"loss": 1.0577,
|
| 872 |
+
"step": 5550
|
| 873 |
+
},
|
| 874 |
+
{
|
| 875 |
+
"epoch": 6.33,
|
| 876 |
+
"learning_rate": 5.781687898089172e-05,
|
| 877 |
+
"loss": 1.0693,
|
| 878 |
+
"step": 5600
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"epoch": 6.38,
|
| 882 |
+
"learning_rate": 5.7578025477707004e-05,
|
| 883 |
+
"loss": 1.0784,
|
| 884 |
+
"step": 5650
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 6.44,
|
| 888 |
+
"learning_rate": 5.733917197452229e-05,
|
| 889 |
+
"loss": 1.0754,
|
| 890 |
+
"step": 5700
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 6.5,
|
| 894 |
+
"learning_rate": 5.710031847133758e-05,
|
| 895 |
+
"loss": 1.0802,
|
| 896 |
+
"step": 5750
|
| 897 |
+
},
|
| 898 |
+
{
|
| 899 |
+
"epoch": 6.5,
|
| 900 |
+
"eval_loss": 0.26020729541778564,
|
| 901 |
+
"eval_runtime": 433.3184,
|
| 902 |
+
"eval_samples_per_second": 14.327,
|
| 903 |
+
"eval_steps_per_second": 0.895,
|
| 904 |
+
"eval_wer": 0.21889450428316534,
|
| 905 |
+
"step": 5750
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 6.55,
|
| 909 |
+
"learning_rate": 5.6861464968152864e-05,
|
| 910 |
+
"loss": 1.0459,
|
| 911 |
+
"step": 5800
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 6.61,
|
| 915 |
+
"learning_rate": 5.662261146496815e-05,
|
| 916 |
+
"loss": 1.0492,
|
| 917 |
+
"step": 5850
|
| 918 |
+
},
|
| 919 |
+
{
|
| 920 |
+
"epoch": 6.67,
|
| 921 |
+
"learning_rate": 5.638375796178344e-05,
|
| 922 |
+
"loss": 1.0526,
|
| 923 |
+
"step": 5900
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"epoch": 6.72,
|
| 927 |
+
"learning_rate": 5.6144904458598724e-05,
|
| 928 |
+
"loss": 1.079,
|
| 929 |
+
"step": 5950
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 6.78,
|
| 933 |
+
"learning_rate": 5.590605095541401e-05,
|
| 934 |
+
"loss": 1.0882,
|
| 935 |
+
"step": 6000
|
| 936 |
+
},
|
| 937 |
+
{
|
| 938 |
+
"epoch": 6.78,
|
| 939 |
+
"eval_loss": 0.2602781653404236,
|
| 940 |
+
"eval_runtime": 434.4762,
|
| 941 |
+
"eval_samples_per_second": 14.288,
|
| 942 |
+
"eval_steps_per_second": 0.893,
|
| 943 |
+
"eval_wer": 0.21684867394695787,
|
| 944 |
+
"step": 6000
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 6.84,
|
| 948 |
+
"learning_rate": 5.56671974522293e-05,
|
| 949 |
+
"loss": 1.0691,
|
| 950 |
+
"step": 6050
|
| 951 |
+
},
|
| 952 |
+
{
|
| 953 |
+
"epoch": 6.89,
|
| 954 |
+
"learning_rate": 5.5428343949044585e-05,
|
| 955 |
+
"loss": 1.0728,
|
| 956 |
+
"step": 6100
|
| 957 |
+
},
|
| 958 |
+
{
|
| 959 |
+
"epoch": 6.95,
|
| 960 |
+
"learning_rate": 5.518949044585987e-05,
|
| 961 |
+
"loss": 1.0308,
|
| 962 |
+
"step": 6150
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 7.01,
|
| 966 |
+
"learning_rate": 5.4955414012738844e-05,
|
| 967 |
+
"loss": 1.0894,
|
| 968 |
+
"step": 6200
|
| 969 |
+
},
|
| 970 |
+
{
|
| 971 |
+
"epoch": 7.06,
|
| 972 |
+
"learning_rate": 5.471656050955413e-05,
|
| 973 |
+
"loss": 1.0881,
|
| 974 |
+
"step": 6250
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 7.06,
|
| 978 |
+
"eval_loss": 0.25403761863708496,
|
| 979 |
+
"eval_runtime": 433.991,
|
| 980 |
+
"eval_samples_per_second": 14.304,
|
| 981 |
+
"eval_steps_per_second": 0.894,
|
| 982 |
+
"eval_wer": 0.2292544396386634,
|
| 983 |
+
"step": 6250
|
| 984 |
+
},
|
| 985 |
+
{
|
| 986 |
+
"epoch": 7.12,
|
| 987 |
+
"learning_rate": 5.447770700636942e-05,
|
| 988 |
+
"loss": 1.0295,
|
| 989 |
+
"step": 6300
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 7.17,
|
| 993 |
+
"learning_rate": 5.4238853503184704e-05,
|
| 994 |
+
"loss": 1.0389,
|
| 995 |
+
"step": 6350
|
| 996 |
+
},
|
| 997 |
+
{
|
| 998 |
+
"epoch": 7.23,
|
| 999 |
+
"learning_rate": 5.399999999999999e-05,
|
| 1000 |
+
"loss": 1.0415,
|
| 1001 |
+
"step": 6400
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 7.29,
|
| 1005 |
+
"learning_rate": 5.376114649681528e-05,
|
| 1006 |
+
"loss": 1.0492,
|
| 1007 |
+
"step": 6450
|
| 1008 |
+
},
|
| 1009 |
+
{
|
| 1010 |
+
"epoch": 7.34,
|
| 1011 |
+
"learning_rate": 5.3522292993630565e-05,
|
| 1012 |
+
"loss": 1.0378,
|
| 1013 |
+
"step": 6500
|
| 1014 |
+
},
|
| 1015 |
+
{
|
| 1016 |
+
"epoch": 7.34,
|
| 1017 |
+
"eval_loss": 0.2614484429359436,
|
| 1018 |
+
"eval_runtime": 432.0675,
|
| 1019 |
+
"eval_samples_per_second": 14.368,
|
| 1020 |
+
"eval_steps_per_second": 0.898,
|
| 1021 |
+
"eval_wer": 0.21932422207067923,
|
| 1022 |
+
"step": 6500
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 7.4,
|
| 1026 |
+
"learning_rate": 5.328343949044585e-05,
|
| 1027 |
+
"loss": 1.0362,
|
| 1028 |
+
"step": 6550
|
| 1029 |
+
},
|
| 1030 |
+
{
|
| 1031 |
+
"epoch": 7.46,
|
| 1032 |
+
"learning_rate": 5.304458598726114e-05,
|
| 1033 |
+
"loss": 1.0444,
|
| 1034 |
+
"step": 6600
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 7.51,
|
| 1038 |
+
"learning_rate": 5.2805732484076425e-05,
|
| 1039 |
+
"loss": 1.0626,
|
| 1040 |
+
"step": 6650
|
| 1041 |
+
},
|
| 1042 |
+
{
|
| 1043 |
+
"epoch": 7.57,
|
| 1044 |
+
"learning_rate": 5.256687898089171e-05,
|
| 1045 |
+
"loss": 1.0307,
|
| 1046 |
+
"step": 6700
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 7.63,
|
| 1050 |
+
"learning_rate": 5.2328025477707e-05,
|
| 1051 |
+
"loss": 1.0397,
|
| 1052 |
+
"step": 6750
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 7.63,
|
| 1056 |
+
"eval_loss": 0.27073222398757935,
|
| 1057 |
+
"eval_runtime": 432.0598,
|
| 1058 |
+
"eval_samples_per_second": 14.368,
|
| 1059 |
+
"eval_steps_per_second": 0.898,
|
| 1060 |
+
"eval_wer": 0.21041224882528237,
|
| 1061 |
+
"step": 6750
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 7.68,
|
| 1065 |
+
"learning_rate": 5.2089171974522285e-05,
|
| 1066 |
+
"loss": 1.0481,
|
| 1067 |
+
"step": 6800
|
| 1068 |
+
},
|
| 1069 |
+
{
|
| 1070 |
+
"epoch": 7.74,
|
| 1071 |
+
"learning_rate": 5.185031847133757e-05,
|
| 1072 |
+
"loss": 1.042,
|
| 1073 |
+
"step": 6850
|
| 1074 |
+
},
|
| 1075 |
+
{
|
| 1076 |
+
"epoch": 7.8,
|
| 1077 |
+
"learning_rate": 5.161146496815286e-05,
|
| 1078 |
+
"loss": 1.0298,
|
| 1079 |
+
"step": 6900
|
| 1080 |
+
},
|
| 1081 |
+
{
|
| 1082 |
+
"epoch": 7.85,
|
| 1083 |
+
"learning_rate": 5.1372611464968145e-05,
|
| 1084 |
+
"loss": 1.0269,
|
| 1085 |
+
"step": 6950
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 7.91,
|
| 1089 |
+
"learning_rate": 5.113375796178343e-05,
|
| 1090 |
+
"loss": 1.0296,
|
| 1091 |
+
"step": 7000
|
| 1092 |
+
},
|
| 1093 |
+
{
|
| 1094 |
+
"epoch": 7.91,
|
| 1095 |
+
"eval_loss": 0.248311385512352,
|
| 1096 |
+
"eval_runtime": 431.8203,
|
| 1097 |
+
"eval_samples_per_second": 14.376,
|
| 1098 |
+
"eval_steps_per_second": 0.899,
|
| 1099 |
+
"eval_wer": 0.2119256027726139,
|
| 1100 |
+
"step": 7000
|
| 1101 |
+
},
|
| 1102 |
+
{
|
| 1103 |
+
"epoch": 7.97,
|
| 1104 |
+
"learning_rate": 5.089490445859872e-05,
|
| 1105 |
+
"loss": 1.0276,
|
| 1106 |
+
"step": 7050
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 8.02,
|
| 1110 |
+
"learning_rate": 5.0656050955414005e-05,
|
| 1111 |
+
"loss": 1.0481,
|
| 1112 |
+
"step": 7100
|
| 1113 |
+
},
|
| 1114 |
+
{
|
| 1115 |
+
"epoch": 8.08,
|
| 1116 |
+
"learning_rate": 5.041719745222929e-05,
|
| 1117 |
+
"loss": 1.006,
|
| 1118 |
+
"step": 7150
|
| 1119 |
+
},
|
| 1120 |
+
{
|
| 1121 |
+
"epoch": 8.14,
|
| 1122 |
+
"learning_rate": 5.017834394904458e-05,
|
| 1123 |
+
"loss": 1.0215,
|
| 1124 |
+
"step": 7200
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 8.19,
|
| 1128 |
+
"learning_rate": 4.9939490445859866e-05,
|
| 1129 |
+
"loss": 1.0249,
|
| 1130 |
+
"step": 7250
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"epoch": 8.19,
|
| 1134 |
+
"eval_loss": 0.24828839302062988,
|
| 1135 |
+
"eval_runtime": 429.7696,
|
| 1136 |
+
"eval_samples_per_second": 14.445,
|
| 1137 |
+
"eval_steps_per_second": 0.903,
|
| 1138 |
+
"eval_wer": 0.20468579222210806,
|
| 1139 |
+
"step": 7250
|
| 1140 |
+
},
|
| 1141 |
+
{
|
| 1142 |
+
"epoch": 8.25,
|
| 1143 |
+
"learning_rate": 4.970063694267515e-05,
|
| 1144 |
+
"loss": 1.0109,
|
| 1145 |
+
"step": 7300
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"epoch": 8.3,
|
| 1149 |
+
"learning_rate": 4.946178343949044e-05,
|
| 1150 |
+
"loss": 1.0154,
|
| 1151 |
+
"step": 7350
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 8.36,
|
| 1155 |
+
"learning_rate": 4.9222929936305726e-05,
|
| 1156 |
+
"loss": 1.0123,
|
| 1157 |
+
"step": 7400
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 8.42,
|
| 1161 |
+
"learning_rate": 4.898407643312101e-05,
|
| 1162 |
+
"loss": 1.0126,
|
| 1163 |
+
"step": 7450
|
| 1164 |
+
},
|
| 1165 |
+
{
|
| 1166 |
+
"epoch": 8.47,
|
| 1167 |
+
"learning_rate": 4.87452229299363e-05,
|
| 1168 |
+
"loss": 1.013,
|
| 1169 |
+
"step": 7500
|
| 1170 |
+
},
|
| 1171 |
+
{
|
| 1172 |
+
"epoch": 8.47,
|
| 1173 |
+
"eval_loss": 0.24869437515735626,
|
| 1174 |
+
"eval_runtime": 430.836,
|
| 1175 |
+
"eval_samples_per_second": 14.409,
|
| 1176 |
+
"eval_steps_per_second": 0.901,
|
| 1177 |
+
"eval_wer": 0.20419068259736378,
|
| 1178 |
+
"step": 7500
|
| 1179 |
+
},
|
| 1180 |
+
{
|
| 1181 |
+
"epoch": 8.53,
|
| 1182 |
+
"learning_rate": 4.8506369426751586e-05,
|
| 1183 |
+
"loss": 1.0077,
|
| 1184 |
+
"step": 7550
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"epoch": 8.59,
|
| 1188 |
+
"learning_rate": 4.826751592356687e-05,
|
| 1189 |
+
"loss": 1.0256,
|
| 1190 |
+
"step": 7600
|
| 1191 |
+
},
|
| 1192 |
+
{
|
| 1193 |
+
"epoch": 8.64,
|
| 1194 |
+
"learning_rate": 4.802866242038216e-05,
|
| 1195 |
+
"loss": 1.0627,
|
| 1196 |
+
"step": 7650
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 8.7,
|
| 1200 |
+
"learning_rate": 4.7789808917197446e-05,
|
| 1201 |
+
"loss": 0.9883,
|
| 1202 |
+
"step": 7700
|
| 1203 |
+
},
|
| 1204 |
+
{
|
| 1205 |
+
"epoch": 8.76,
|
| 1206 |
+
"learning_rate": 4.755095541401273e-05,
|
| 1207 |
+
"loss": 1.0064,
|
| 1208 |
+
"step": 7750
|
| 1209 |
+
},
|
| 1210 |
+
{
|
| 1211 |
+
"epoch": 8.76,
|
| 1212 |
+
"eval_loss": 0.24558775126934052,
|
| 1213 |
+
"eval_runtime": 430.9773,
|
| 1214 |
+
"eval_samples_per_second": 14.404,
|
| 1215 |
+
"eval_steps_per_second": 0.9,
|
| 1216 |
+
"eval_wer": 0.20164974263641205,
|
| 1217 |
+
"step": 7750
|
| 1218 |
+
},
|
| 1219 |
+
{
|
| 1220 |
+
"epoch": 8.81,
|
| 1221 |
+
"learning_rate": 4.731210191082802e-05,
|
| 1222 |
+
"loss": 1.0137,
|
| 1223 |
+
"step": 7800
|
| 1224 |
+
},
|
| 1225 |
+
{
|
| 1226 |
+
"epoch": 8.87,
|
| 1227 |
+
"learning_rate": 4.7073248407643306e-05,
|
| 1228 |
+
"loss": 1.0178,
|
| 1229 |
+
"step": 7850
|
| 1230 |
+
},
|
| 1231 |
+
{
|
| 1232 |
+
"epoch": 8.93,
|
| 1233 |
+
"learning_rate": 4.683439490445859e-05,
|
| 1234 |
+
"loss": 1.0035,
|
| 1235 |
+
"step": 7900
|
| 1236 |
+
},
|
| 1237 |
+
{
|
| 1238 |
+
"epoch": 8.98,
|
| 1239 |
+
"learning_rate": 4.659554140127388e-05,
|
| 1240 |
+
"loss": 1.0457,
|
| 1241 |
+
"step": 7950
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 9.04,
|
| 1245 |
+
"learning_rate": 4.6356687898089167e-05,
|
| 1246 |
+
"loss": 1.0668,
|
| 1247 |
+
"step": 8000
|
| 1248 |
+
},
|
| 1249 |
+
{
|
| 1250 |
+
"epoch": 9.04,
|
| 1251 |
+
"eval_loss": 0.2397284209728241,
|
| 1252 |
+
"eval_runtime": 430.6925,
|
| 1253 |
+
"eval_samples_per_second": 14.414,
|
| 1254 |
+
"eval_steps_per_second": 0.901,
|
| 1255 |
+
"eval_wer": 0.19949181200780966,
|
| 1256 |
+
"step": 8000
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 9.1,
|
| 1260 |
+
"learning_rate": 4.611783439490445e-05,
|
| 1261 |
+
"loss": 1.0054,
|
| 1262 |
+
"step": 8050
|
| 1263 |
+
},
|
| 1264 |
+
{
|
| 1265 |
+
"epoch": 9.15,
|
| 1266 |
+
"learning_rate": 4.587898089171974e-05,
|
| 1267 |
+
"loss": 1.0224,
|
| 1268 |
+
"step": 8100
|
| 1269 |
+
},
|
| 1270 |
+
{
|
| 1271 |
+
"epoch": 9.21,
|
| 1272 |
+
"learning_rate": 4.564012738853503e-05,
|
| 1273 |
+
"loss": 1.0019,
|
| 1274 |
+
"step": 8150
|
| 1275 |
+
},
|
| 1276 |
+
{
|
| 1277 |
+
"epoch": 9.27,
|
| 1278 |
+
"learning_rate": 4.5401273885350314e-05,
|
| 1279 |
+
"loss": 1.0033,
|
| 1280 |
+
"step": 8200
|
| 1281 |
+
},
|
| 1282 |
+
{
|
| 1283 |
+
"epoch": 9.32,
|
| 1284 |
+
"learning_rate": 4.51624203821656e-05,
|
| 1285 |
+
"loss": 1.0129,
|
| 1286 |
+
"step": 8250
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"epoch": 9.32,
|
| 1290 |
+
"eval_loss": 0.23742474615573883,
|
| 1291 |
+
"eval_runtime": 432.9935,
|
| 1292 |
+
"eval_samples_per_second": 14.337,
|
| 1293 |
+
"eval_steps_per_second": 0.896,
|
| 1294 |
+
"eval_wer": 0.19942642017057927,
|
| 1295 |
+
"step": 8250
|
| 1296 |
+
},
|
| 1297 |
+
{
|
| 1298 |
+
"epoch": 9.38,
|
| 1299 |
+
"learning_rate": 4.492356687898089e-05,
|
| 1300 |
+
"loss": 0.9864,
|
| 1301 |
+
"step": 8300
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 9.43,
|
| 1305 |
+
"learning_rate": 4.4689490445859874e-05,
|
| 1306 |
+
"loss": 1.0021,
|
| 1307 |
+
"step": 8350
|
| 1308 |
+
},
|
| 1309 |
+
{
|
| 1310 |
+
"epoch": 9.49,
|
| 1311 |
+
"learning_rate": 4.445063694267516e-05,
|
| 1312 |
+
"loss": 1.0073,
|
| 1313 |
+
"step": 8400
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 9.55,
|
| 1317 |
+
"learning_rate": 4.421178343949045e-05,
|
| 1318 |
+
"loss": 0.9999,
|
| 1319 |
+
"step": 8450
|
| 1320 |
+
},
|
| 1321 |
+
{
|
| 1322 |
+
"epoch": 9.6,
|
| 1323 |
+
"learning_rate": 4.3972929936305734e-05,
|
| 1324 |
+
"loss": 1.0164,
|
| 1325 |
+
"step": 8500
|
| 1326 |
+
},
|
| 1327 |
+
{
|
| 1328 |
+
"epoch": 9.6,
|
| 1329 |
+
"eval_loss": 0.2206413298845291,
|
| 1330 |
+
"eval_runtime": 431.5354,
|
| 1331 |
+
"eval_samples_per_second": 14.386,
|
| 1332 |
+
"eval_steps_per_second": 0.899,
|
| 1333 |
+
"eval_wer": 0.19915551113062488,
|
| 1334 |
+
"step": 8500
|
| 1335 |
+
},
|
| 1336 |
+
{
|
| 1337 |
+
"epoch": 9.66,
|
| 1338 |
+
"learning_rate": 4.373407643312102e-05,
|
| 1339 |
+
"loss": 0.9956,
|
| 1340 |
+
"step": 8550
|
| 1341 |
+
},
|
| 1342 |
+
{
|
| 1343 |
+
"epoch": 9.72,
|
| 1344 |
+
"learning_rate": 4.349522292993631e-05,
|
| 1345 |
+
"loss": 0.9662,
|
| 1346 |
+
"step": 8600
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 9.77,
|
| 1350 |
+
"learning_rate": 4.3256369426751594e-05,
|
| 1351 |
+
"loss": 0.9781,
|
| 1352 |
+
"step": 8650
|
| 1353 |
+
},
|
| 1354 |
+
{
|
| 1355 |
+
"epoch": 9.83,
|
| 1356 |
+
"learning_rate": 4.301751592356688e-05,
|
| 1357 |
+
"loss": 0.9863,
|
| 1358 |
+
"step": 8700
|
| 1359 |
+
},
|
| 1360 |
+
{
|
| 1361 |
+
"epoch": 9.89,
|
| 1362 |
+
"learning_rate": 4.277866242038217e-05,
|
| 1363 |
+
"loss": 0.975,
|
| 1364 |
+
"step": 8750
|
| 1365 |
+
},
|
| 1366 |
+
{
|
| 1367 |
+
"epoch": 9.89,
|
| 1368 |
+
"eval_loss": 0.22473624348640442,
|
| 1369 |
+
"eval_runtime": 432.1534,
|
| 1370 |
+
"eval_samples_per_second": 14.365,
|
| 1371 |
+
"eval_steps_per_second": 0.898,
|
| 1372 |
+
"eval_wer": 0.19731519799714145,
|
| 1373 |
+
"step": 8750
|
| 1374 |
+
},
|
| 1375 |
+
{
|
| 1376 |
+
"epoch": 9.94,
|
| 1377 |
+
"learning_rate": 4.2539808917197454e-05,
|
| 1378 |
+
"loss": 0.9931,
|
| 1379 |
+
"step": 8800
|
| 1380 |
+
},
|
| 1381 |
+
{
|
| 1382 |
+
"epoch": 10.0,
|
| 1383 |
+
"learning_rate": 4.230095541401274e-05,
|
| 1384 |
+
"loss": 1.0101,
|
| 1385 |
+
"step": 8850
|
| 1386 |
+
},
|
| 1387 |
+
{
|
| 1388 |
+
"epoch": 10.06,
|
| 1389 |
+
"learning_rate": 4.206210191082803e-05,
|
| 1390 |
+
"loss": 1.0034,
|
| 1391 |
+
"step": 8900
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"epoch": 10.11,
|
| 1395 |
+
"learning_rate": 4.1823248407643314e-05,
|
| 1396 |
+
"loss": 1.0018,
|
| 1397 |
+
"step": 8950
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 10.17,
|
| 1401 |
+
"learning_rate": 4.15843949044586e-05,
|
| 1402 |
+
"loss": 0.9849,
|
| 1403 |
+
"step": 9000
|
| 1404 |
+
},
|
| 1405 |
+
{
|
| 1406 |
+
"epoch": 10.17,
|
| 1407 |
+
"eval_loss": 0.23245184123516083,
|
| 1408 |
+
"eval_runtime": 431.4778,
|
| 1409 |
+
"eval_samples_per_second": 14.388,
|
| 1410 |
+
"eval_steps_per_second": 0.899,
|
| 1411 |
+
"eval_wer": 0.19526002596990108,
|
| 1412 |
+
"step": 9000
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 10.23,
|
| 1416 |
+
"learning_rate": 4.134554140127389e-05,
|
| 1417 |
+
"loss": 0.9953,
|
| 1418 |
+
"step": 9050
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 10.28,
|
| 1422 |
+
"learning_rate": 4.1106687898089175e-05,
|
| 1423 |
+
"loss": 0.9639,
|
| 1424 |
+
"step": 9100
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 10.34,
|
| 1428 |
+
"learning_rate": 4.086783439490446e-05,
|
| 1429 |
+
"loss": 0.9862,
|
| 1430 |
+
"step": 9150
|
| 1431 |
+
},
|
| 1432 |
+
{
|
| 1433 |
+
"epoch": 10.4,
|
| 1434 |
+
"learning_rate": 4.062898089171975e-05,
|
| 1435 |
+
"loss": 1.0222,
|
| 1436 |
+
"step": 9200
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 10.45,
|
| 1440 |
+
"learning_rate": 4.0390127388535035e-05,
|
| 1441 |
+
"loss": 0.9826,
|
| 1442 |
+
"step": 9250
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 10.45,
|
| 1446 |
+
"eval_loss": 0.2301308959722519,
|
| 1447 |
+
"eval_runtime": 432.6762,
|
| 1448 |
+
"eval_samples_per_second": 14.348,
|
| 1449 |
+
"eval_steps_per_second": 0.897,
|
| 1450 |
+
"eval_wer": 0.1933730043812531,
|
| 1451 |
+
"step": 9250
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"epoch": 10.51,
|
| 1455 |
+
"learning_rate": 4.015127388535032e-05,
|
| 1456 |
+
"loss": 0.9867,
|
| 1457 |
+
"step": 9300
|
| 1458 |
+
},
|
| 1459 |
+
{
|
| 1460 |
+
"epoch": 10.56,
|
| 1461 |
+
"learning_rate": 3.991242038216561e-05,
|
| 1462 |
+
"loss": 0.9687,
|
| 1463 |
+
"step": 9350
|
| 1464 |
+
},
|
| 1465 |
+
{
|
| 1466 |
+
"epoch": 10.62,
|
| 1467 |
+
"learning_rate": 3.9673566878980895e-05,
|
| 1468 |
+
"loss": 0.9715,
|
| 1469 |
+
"step": 9400
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 10.68,
|
| 1473 |
+
"learning_rate": 3.943471337579618e-05,
|
| 1474 |
+
"loss": 0.9914,
|
| 1475 |
+
"step": 9450
|
| 1476 |
+
},
|
| 1477 |
+
{
|
| 1478 |
+
"epoch": 10.73,
|
| 1479 |
+
"learning_rate": 3.919585987261147e-05,
|
| 1480 |
+
"loss": 0.9835,
|
| 1481 |
+
"step": 9500
|
| 1482 |
+
},
|
| 1483 |
+
{
|
| 1484 |
+
"epoch": 10.73,
|
| 1485 |
+
"eval_loss": 0.2191852629184723,
|
| 1486 |
+
"eval_runtime": 439.1976,
|
| 1487 |
+
"eval_samples_per_second": 14.135,
|
| 1488 |
+
"eval_steps_per_second": 0.883,
|
| 1489 |
+
"eval_wer": 0.19420441488318216,
|
| 1490 |
+
"step": 9500
|
| 1491 |
+
},
|
| 1492 |
+
{
|
| 1493 |
+
"epoch": 10.79,
|
| 1494 |
+
"learning_rate": 3.8957006369426755e-05,
|
| 1495 |
+
"loss": 0.9652,
|
| 1496 |
+
"step": 9550
|
| 1497 |
+
},
|
| 1498 |
+
{
|
| 1499 |
+
"epoch": 10.85,
|
| 1500 |
+
"learning_rate": 3.871815286624204e-05,
|
| 1501 |
+
"loss": 0.9614,
|
| 1502 |
+
"step": 9600
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 10.9,
|
| 1506 |
+
"learning_rate": 3.847929936305733e-05,
|
| 1507 |
+
"loss": 0.97,
|
| 1508 |
+
"step": 9650
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"epoch": 10.96,
|
| 1512 |
+
"learning_rate": 3.8240445859872615e-05,
|
| 1513 |
+
"loss": 0.9764,
|
| 1514 |
+
"step": 9700
|
| 1515 |
+
},
|
| 1516 |
+
{
|
| 1517 |
+
"epoch": 11.02,
|
| 1518 |
+
"learning_rate": 3.80015923566879e-05,
|
| 1519 |
+
"loss": 0.9676,
|
| 1520 |
+
"step": 9750
|
| 1521 |
+
},
|
| 1522 |
+
{
|
| 1523 |
+
"epoch": 11.02,
|
| 1524 |
+
"eval_loss": 0.2265927493572235,
|
| 1525 |
+
"eval_runtime": 430.7748,
|
| 1526 |
+
"eval_samples_per_second": 14.411,
|
| 1527 |
+
"eval_steps_per_second": 0.901,
|
| 1528 |
+
"eval_wer": 0.19133651573607854,
|
| 1529 |
+
"step": 9750
|
| 1530 |
+
},
|
| 1531 |
+
{
|
| 1532 |
+
"epoch": 11.07,
|
| 1533 |
+
"learning_rate": 3.776273885350319e-05,
|
| 1534 |
+
"loss": 0.9609,
|
| 1535 |
+
"step": 9800
|
| 1536 |
+
},
|
| 1537 |
+
{
|
| 1538 |
+
"epoch": 11.13,
|
| 1539 |
+
"learning_rate": 3.7523885350318475e-05,
|
| 1540 |
+
"loss": 0.9721,
|
| 1541 |
+
"step": 9850
|
| 1542 |
+
},
|
| 1543 |
+
{
|
| 1544 |
+
"epoch": 11.19,
|
| 1545 |
+
"learning_rate": 3.7285031847133755e-05,
|
| 1546 |
+
"loss": 0.9669,
|
| 1547 |
+
"step": 9900
|
| 1548 |
+
},
|
| 1549 |
+
{
|
| 1550 |
+
"epoch": 11.24,
|
| 1551 |
+
"learning_rate": 3.704617834394904e-05,
|
| 1552 |
+
"loss": 0.9643,
|
| 1553 |
+
"step": 9950
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 11.3,
|
| 1557 |
+
"learning_rate": 3.680732484076433e-05,
|
| 1558 |
+
"loss": 0.9627,
|
| 1559 |
+
"step": 10000
|
| 1560 |
+
},
|
| 1561 |
+
{
|
| 1562 |
+
"epoch": 11.3,
|
| 1563 |
+
"eval_loss": 0.2193416953086853,
|
| 1564 |
+
"eval_runtime": 432.6083,
|
| 1565 |
+
"eval_samples_per_second": 14.35,
|
| 1566 |
+
"eval_steps_per_second": 0.897,
|
| 1567 |
+
"eval_wer": 0.19205582594561268,
|
| 1568 |
+
"step": 10000
|
| 1569 |
+
},
|
| 1570 |
+
{
|
| 1571 |
+
"epoch": 11.36,
|
| 1572 |
+
"learning_rate": 3.6568471337579616e-05,
|
| 1573 |
+
"loss": 1.0179,
|
| 1574 |
+
"step": 10050
|
| 1575 |
+
},
|
| 1576 |
+
{
|
| 1577 |
+
"epoch": 11.41,
|
| 1578 |
+
"learning_rate": 3.63296178343949e-05,
|
| 1579 |
+
"loss": 0.9575,
|
| 1580 |
+
"step": 10100
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 11.47,
|
| 1584 |
+
"learning_rate": 3.609076433121019e-05,
|
| 1585 |
+
"loss": 0.98,
|
| 1586 |
+
"step": 10150
|
| 1587 |
+
},
|
| 1588 |
+
{
|
| 1589 |
+
"epoch": 11.52,
|
| 1590 |
+
"learning_rate": 3.5851910828025476e-05,
|
| 1591 |
+
"loss": 0.9542,
|
| 1592 |
+
"step": 10200
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 11.58,
|
| 1596 |
+
"learning_rate": 3.561305732484076e-05,
|
| 1597 |
+
"loss": 0.976,
|
| 1598 |
+
"step": 10250
|
| 1599 |
+
},
|
| 1600 |
+
{
|
| 1601 |
+
"epoch": 11.58,
|
| 1602 |
+
"eval_loss": 0.23090308904647827,
|
| 1603 |
+
"eval_runtime": 432.1501,
|
| 1604 |
+
"eval_samples_per_second": 14.365,
|
| 1605 |
+
"eval_steps_per_second": 0.898,
|
| 1606 |
+
"eval_wer": 0.1881790241669547,
|
| 1607 |
+
"step": 10250
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"epoch": 11.64,
|
| 1611 |
+
"learning_rate": 3.537420382165605e-05,
|
| 1612 |
+
"loss": 0.972,
|
| 1613 |
+
"step": 10300
|
| 1614 |
+
},
|
| 1615 |
+
{
|
| 1616 |
+
"epoch": 11.69,
|
| 1617 |
+
"learning_rate": 3.5135350318471336e-05,
|
| 1618 |
+
"loss": 0.9634,
|
| 1619 |
+
"step": 10350
|
| 1620 |
+
},
|
| 1621 |
+
{
|
| 1622 |
+
"epoch": 11.75,
|
| 1623 |
+
"learning_rate": 3.489649681528662e-05,
|
| 1624 |
+
"loss": 0.9682,
|
| 1625 |
+
"step": 10400
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 11.81,
|
| 1629 |
+
"learning_rate": 3.465764331210191e-05,
|
| 1630 |
+
"loss": 0.9638,
|
| 1631 |
+
"step": 10450
|
| 1632 |
+
},
|
| 1633 |
+
{
|
| 1634 |
+
"epoch": 11.86,
|
| 1635 |
+
"learning_rate": 3.4418789808917196e-05,
|
| 1636 |
+
"loss": 0.969,
|
| 1637 |
+
"step": 10500
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 11.86,
|
| 1641 |
+
"eval_loss": 0.2268366813659668,
|
| 1642 |
+
"eval_runtime": 433.2795,
|
| 1643 |
+
"eval_samples_per_second": 14.328,
|
| 1644 |
+
"eval_steps_per_second": 0.895,
|
| 1645 |
+
"eval_wer": 0.18859005857240277,
|
| 1646 |
+
"step": 10500
|
| 1647 |
+
},
|
| 1648 |
+
{
|
| 1649 |
+
"epoch": 11.92,
|
| 1650 |
+
"learning_rate": 3.417993630573248e-05,
|
| 1651 |
+
"loss": 0.9698,
|
| 1652 |
+
"step": 10550
|
| 1653 |
+
},
|
| 1654 |
+
{
|
| 1655 |
+
"epoch": 11.98,
|
| 1656 |
+
"learning_rate": 3.394108280254777e-05,
|
| 1657 |
+
"loss": 0.9369,
|
| 1658 |
+
"step": 10600
|
| 1659 |
+
},
|
| 1660 |
+
{
|
| 1661 |
+
"epoch": 12.03,
|
| 1662 |
+
"learning_rate": 3.3702229299363056e-05,
|
| 1663 |
+
"loss": 0.9699,
|
| 1664 |
+
"step": 10650
|
| 1665 |
+
},
|
| 1666 |
+
{
|
| 1667 |
+
"epoch": 12.09,
|
| 1668 |
+
"learning_rate": 3.346337579617834e-05,
|
| 1669 |
+
"loss": 1.0013,
|
| 1670 |
+
"step": 10700
|
| 1671 |
+
},
|
| 1672 |
+
{
|
| 1673 |
+
"epoch": 12.15,
|
| 1674 |
+
"learning_rate": 3.322929936305732e-05,
|
| 1675 |
+
"loss": 0.9611,
|
| 1676 |
+
"step": 10750
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"epoch": 12.15,
|
| 1680 |
+
"eval_loss": 0.2322191596031189,
|
| 1681 |
+
"eval_runtime": 429.3587,
|
| 1682 |
+
"eval_samples_per_second": 14.459,
|
| 1683 |
+
"eval_steps_per_second": 0.904,
|
| 1684 |
+
"eval_wer": 0.18626397750520798,
|
| 1685 |
+
"step": 10750
|
| 1686 |
+
},
|
| 1687 |
+
{
|
| 1688 |
+
"epoch": 12.2,
|
| 1689 |
+
"learning_rate": 3.299044585987261e-05,
|
| 1690 |
+
"loss": 0.9418,
|
| 1691 |
+
"step": 10800
|
| 1692 |
+
},
|
| 1693 |
+
{
|
| 1694 |
+
"epoch": 12.26,
|
| 1695 |
+
"learning_rate": 3.2751592356687896e-05,
|
| 1696 |
+
"loss": 0.9582,
|
| 1697 |
+
"step": 10850
|
| 1698 |
+
},
|
| 1699 |
+
{
|
| 1700 |
+
"epoch": 12.32,
|
| 1701 |
+
"learning_rate": 3.251273885350318e-05,
|
| 1702 |
+
"loss": 0.945,
|
| 1703 |
+
"step": 10900
|
| 1704 |
+
},
|
| 1705 |
+
{
|
| 1706 |
+
"epoch": 12.37,
|
| 1707 |
+
"learning_rate": 3.227388535031847e-05,
|
| 1708 |
+
"loss": 0.9386,
|
| 1709 |
+
"step": 10950
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 12.43,
|
| 1713 |
+
"learning_rate": 3.2035031847133757e-05,
|
| 1714 |
+
"loss": 0.9397,
|
| 1715 |
+
"step": 11000
|
| 1716 |
+
},
|
| 1717 |
+
{
|
| 1718 |
+
"epoch": 12.43,
|
| 1719 |
+
"eval_loss": 0.21969455480575562,
|
| 1720 |
+
"eval_runtime": 432.2628,
|
| 1721 |
+
"eval_samples_per_second": 14.362,
|
| 1722 |
+
"eval_steps_per_second": 0.898,
|
| 1723 |
+
"eval_wer": 0.1843676142255271,
|
| 1724 |
+
"step": 11000
|
| 1725 |
+
},
|
| 1726 |
+
{
|
| 1727 |
+
"epoch": 12.49,
|
| 1728 |
+
"learning_rate": 3.179617834394904e-05,
|
| 1729 |
+
"loss": 0.9594,
|
| 1730 |
+
"step": 11050
|
| 1731 |
+
},
|
| 1732 |
+
{
|
| 1733 |
+
"epoch": 12.54,
|
| 1734 |
+
"learning_rate": 3.155732484076433e-05,
|
| 1735 |
+
"loss": 0.9467,
|
| 1736 |
+
"step": 11100
|
| 1737 |
+
},
|
| 1738 |
+
{
|
| 1739 |
+
"epoch": 12.6,
|
| 1740 |
+
"learning_rate": 3.131847133757962e-05,
|
| 1741 |
+
"loss": 0.9609,
|
| 1742 |
+
"step": 11150
|
| 1743 |
+
},
|
| 1744 |
+
{
|
| 1745 |
+
"epoch": 12.65,
|
| 1746 |
+
"learning_rate": 3.1079617834394904e-05,
|
| 1747 |
+
"loss": 0.9446,
|
| 1748 |
+
"step": 11200
|
| 1749 |
+
},
|
| 1750 |
+
{
|
| 1751 |
+
"epoch": 12.71,
|
| 1752 |
+
"learning_rate": 3.084076433121019e-05,
|
| 1753 |
+
"loss": 0.9601,
|
| 1754 |
+
"step": 11250
|
| 1755 |
+
},
|
| 1756 |
+
{
|
| 1757 |
+
"epoch": 12.71,
|
| 1758 |
+
"eval_loss": 0.22107979655265808,
|
| 1759 |
+
"eval_runtime": 432.9535,
|
| 1760 |
+
"eval_samples_per_second": 14.339,
|
| 1761 |
+
"eval_steps_per_second": 0.896,
|
| 1762 |
+
"eval_wer": 0.18711407138920289,
|
| 1763 |
+
"step": 11250
|
| 1764 |
+
},
|
| 1765 |
+
{
|
| 1766 |
+
"epoch": 12.77,
|
| 1767 |
+
"learning_rate": 3.060191082802548e-05,
|
| 1768 |
+
"loss": 0.9497,
|
| 1769 |
+
"step": 11300
|
| 1770 |
+
},
|
| 1771 |
+
{
|
| 1772 |
+
"epoch": 12.82,
|
| 1773 |
+
"learning_rate": 3.036305732484076e-05,
|
| 1774 |
+
"loss": 0.939,
|
| 1775 |
+
"step": 11350
|
| 1776 |
+
},
|
| 1777 |
+
{
|
| 1778 |
+
"epoch": 12.88,
|
| 1779 |
+
"learning_rate": 3.0124203821656047e-05,
|
| 1780 |
+
"loss": 0.9462,
|
| 1781 |
+
"step": 11400
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 12.94,
|
| 1785 |
+
"learning_rate": 2.9885350318471334e-05,
|
| 1786 |
+
"loss": 0.9243,
|
| 1787 |
+
"step": 11450
|
| 1788 |
+
},
|
| 1789 |
+
{
|
| 1790 |
+
"epoch": 12.99,
|
| 1791 |
+
"learning_rate": 2.964649681528662e-05,
|
| 1792 |
+
"loss": 0.9718,
|
| 1793 |
+
"step": 11500
|
| 1794 |
+
},
|
| 1795 |
+
{
|
| 1796 |
+
"epoch": 12.99,
|
| 1797 |
+
"eval_loss": 0.20792651176452637,
|
| 1798 |
+
"eval_runtime": 429.7801,
|
| 1799 |
+
"eval_samples_per_second": 14.445,
|
| 1800 |
+
"eval_steps_per_second": 0.903,
|
| 1801 |
+
"eval_wer": 0.189823161788747,
|
| 1802 |
+
"step": 11500
|
| 1803 |
+
},
|
| 1804 |
+
{
|
| 1805 |
+
"epoch": 13.05,
|
| 1806 |
+
"learning_rate": 2.9407643312101907e-05,
|
| 1807 |
+
"loss": 0.9543,
|
| 1808 |
+
"step": 11550
|
| 1809 |
+
},
|
| 1810 |
+
{
|
| 1811 |
+
"epoch": 13.11,
|
| 1812 |
+
"learning_rate": 2.9168789808917194e-05,
|
| 1813 |
+
"loss": 0.9386,
|
| 1814 |
+
"step": 11600
|
| 1815 |
+
},
|
| 1816 |
+
{
|
| 1817 |
+
"epoch": 13.16,
|
| 1818 |
+
"learning_rate": 2.892993630573248e-05,
|
| 1819 |
+
"loss": 0.9662,
|
| 1820 |
+
"step": 11650
|
| 1821 |
+
},
|
| 1822 |
+
{
|
| 1823 |
+
"epoch": 13.22,
|
| 1824 |
+
"learning_rate": 2.8691082802547767e-05,
|
| 1825 |
+
"loss": 0.9426,
|
| 1826 |
+
"step": 11700
|
| 1827 |
+
},
|
| 1828 |
+
{
|
| 1829 |
+
"epoch": 13.28,
|
| 1830 |
+
"learning_rate": 2.8452229299363054e-05,
|
| 1831 |
+
"loss": 0.9347,
|
| 1832 |
+
"step": 11750
|
| 1833 |
+
},
|
| 1834 |
+
{
|
| 1835 |
+
"epoch": 13.28,
|
| 1836 |
+
"eval_loss": 0.2053879350423813,
|
| 1837 |
+
"eval_runtime": 427.5266,
|
| 1838 |
+
"eval_samples_per_second": 14.521,
|
| 1839 |
+
"eval_steps_per_second": 0.908,
|
| 1840 |
+
"eval_wer": 0.1842835390062309,
|
| 1841 |
+
"step": 11750
|
| 1842 |
+
},
|
| 1843 |
+
{
|
| 1844 |
+
"epoch": 13.33,
|
| 1845 |
+
"learning_rate": 2.821337579617834e-05,
|
| 1846 |
+
"loss": 0.9579,
|
| 1847 |
+
"step": 11800
|
| 1848 |
+
},
|
| 1849 |
+
{
|
| 1850 |
+
"epoch": 13.39,
|
| 1851 |
+
"learning_rate": 2.7974522292993628e-05,
|
| 1852 |
+
"loss": 0.9313,
|
| 1853 |
+
"step": 11850
|
| 1854 |
+
},
|
| 1855 |
+
{
|
| 1856 |
+
"epoch": 13.45,
|
| 1857 |
+
"learning_rate": 2.7735668789808914e-05,
|
| 1858 |
+
"loss": 0.9295,
|
| 1859 |
+
"step": 11900
|
| 1860 |
+
},
|
| 1861 |
+
{
|
| 1862 |
+
"epoch": 13.5,
|
| 1863 |
+
"learning_rate": 2.74968152866242e-05,
|
| 1864 |
+
"loss": 0.9437,
|
| 1865 |
+
"step": 11950
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 13.56,
|
| 1869 |
+
"learning_rate": 2.7257961783439488e-05,
|
| 1870 |
+
"loss": 0.9377,
|
| 1871 |
+
"step": 12000
|
| 1872 |
+
},
|
| 1873 |
+
{
|
| 1874 |
+
"epoch": 13.56,
|
| 1875 |
+
"eval_loss": 0.20305366814136505,
|
| 1876 |
+
"eval_runtime": 429.8935,
|
| 1877 |
+
"eval_samples_per_second": 14.441,
|
| 1878 |
+
"eval_steps_per_second": 0.903,
|
| 1879 |
+
"eval_wer": 0.18423683055106635,
|
| 1880 |
+
"step": 12000
|
| 1881 |
+
},
|
| 1882 |
+
{
|
| 1883 |
+
"epoch": 13.62,
|
| 1884 |
+
"learning_rate": 2.7019108280254775e-05,
|
| 1885 |
+
"loss": 0.9273,
|
| 1886 |
+
"step": 12050
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"epoch": 13.67,
|
| 1890 |
+
"learning_rate": 2.678025477707006e-05,
|
| 1891 |
+
"loss": 0.9804,
|
| 1892 |
+
"step": 12100
|
| 1893 |
+
},
|
| 1894 |
+
{
|
| 1895 |
+
"epoch": 13.73,
|
| 1896 |
+
"learning_rate": 2.6541401273885348e-05,
|
| 1897 |
+
"loss": 0.9392,
|
| 1898 |
+
"step": 12150
|
| 1899 |
+
},
|
| 1900 |
+
{
|
| 1901 |
+
"epoch": 13.78,
|
| 1902 |
+
"learning_rate": 2.6302547770700635e-05,
|
| 1903 |
+
"loss": 0.9379,
|
| 1904 |
+
"step": 12200
|
| 1905 |
+
},
|
| 1906 |
+
{
|
| 1907 |
+
"epoch": 13.84,
|
| 1908 |
+
"learning_rate": 2.606369426751592e-05,
|
| 1909 |
+
"loss": 0.934,
|
| 1910 |
+
"step": 12250
|
| 1911 |
+
},
|
| 1912 |
+
{
|
| 1913 |
+
"epoch": 13.84,
|
| 1914 |
+
"eval_loss": 0.20586800575256348,
|
| 1915 |
+
"eval_runtime": 428.3313,
|
| 1916 |
+
"eval_samples_per_second": 14.493,
|
| 1917 |
+
"eval_steps_per_second": 0.906,
|
| 1918 |
+
"eval_wer": 0.18060291273926407,
|
| 1919 |
+
"step": 12250
|
| 1920 |
+
},
|
| 1921 |
+
{
|
| 1922 |
+
"epoch": 13.9,
|
| 1923 |
+
"learning_rate": 2.5824840764331208e-05,
|
| 1924 |
+
"loss": 0.9177,
|
| 1925 |
+
"step": 12300
|
| 1926 |
+
},
|
| 1927 |
+
{
|
| 1928 |
+
"epoch": 13.95,
|
| 1929 |
+
"learning_rate": 2.5585987261146495e-05,
|
| 1930 |
+
"loss": 0.9369,
|
| 1931 |
+
"step": 12350
|
| 1932 |
+
},
|
| 1933 |
+
{
|
| 1934 |
+
"epoch": 14.01,
|
| 1935 |
+
"learning_rate": 2.534713375796178e-05,
|
| 1936 |
+
"loss": 0.9438,
|
| 1937 |
+
"step": 12400
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 14.07,
|
| 1941 |
+
"learning_rate": 2.510828025477707e-05,
|
| 1942 |
+
"loss": 0.9341,
|
| 1943 |
+
"step": 12450
|
| 1944 |
+
},
|
| 1945 |
+
{
|
| 1946 |
+
"epoch": 14.12,
|
| 1947 |
+
"learning_rate": 2.4869426751592355e-05,
|
| 1948 |
+
"loss": 0.9295,
|
| 1949 |
+
"step": 12500
|
| 1950 |
+
},
|
| 1951 |
+
{
|
| 1952 |
+
"epoch": 14.12,
|
| 1953 |
+
"eval_loss": 0.21221554279327393,
|
| 1954 |
+
"eval_runtime": 432.7246,
|
| 1955 |
+
"eval_samples_per_second": 14.346,
|
| 1956 |
+
"eval_steps_per_second": 0.897,
|
| 1957 |
+
"eval_wer": 0.18605846030248396,
|
| 1958 |
+
"step": 12500
|
| 1959 |
+
},
|
| 1960 |
+
{
|
| 1961 |
+
"epoch": 14.18,
|
| 1962 |
+
"learning_rate": 2.4630573248407642e-05,
|
| 1963 |
+
"loss": 0.9239,
|
| 1964 |
+
"step": 12550
|
| 1965 |
+
},
|
| 1966 |
+
{
|
| 1967 |
+
"epoch": 14.24,
|
| 1968 |
+
"learning_rate": 2.439171974522293e-05,
|
| 1969 |
+
"loss": 0.9235,
|
| 1970 |
+
"step": 12600
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"epoch": 14.29,
|
| 1974 |
+
"learning_rate": 2.4152866242038215e-05,
|
| 1975 |
+
"loss": 0.9631,
|
| 1976 |
+
"step": 12650
|
| 1977 |
+
},
|
| 1978 |
+
{
|
| 1979 |
+
"epoch": 14.35,
|
| 1980 |
+
"learning_rate": 2.3914012738853502e-05,
|
| 1981 |
+
"loss": 0.9467,
|
| 1982 |
+
"step": 12700
|
| 1983 |
+
},
|
| 1984 |
+
{
|
| 1985 |
+
"epoch": 14.41,
|
| 1986 |
+
"learning_rate": 2.367515923566879e-05,
|
| 1987 |
+
"loss": 0.935,
|
| 1988 |
+
"step": 12750
|
| 1989 |
+
},
|
| 1990 |
+
{
|
| 1991 |
+
"epoch": 14.41,
|
| 1992 |
+
"eval_loss": 0.20723822712898254,
|
| 1993 |
+
"eval_runtime": 429.493,
|
| 1994 |
+
"eval_samples_per_second": 14.454,
|
| 1995 |
+
"eval_steps_per_second": 0.903,
|
| 1996 |
+
"eval_wer": 0.17866918269545154,
|
| 1997 |
+
"step": 12750
|
| 1998 |
+
},
|
| 1999 |
+
{
|
| 2000 |
+
"epoch": 14.46,
|
| 2001 |
+
"learning_rate": 2.3436305732484076e-05,
|
| 2002 |
+
"loss": 0.9319,
|
| 2003 |
+
"step": 12800
|
| 2004 |
+
},
|
| 2005 |
+
{
|
| 2006 |
+
"epoch": 14.52,
|
| 2007 |
+
"learning_rate": 2.3197452229299362e-05,
|
| 2008 |
+
"loss": 0.9337,
|
| 2009 |
+
"step": 12850
|
| 2010 |
+
},
|
| 2011 |
+
{
|
| 2012 |
+
"epoch": 14.58,
|
| 2013 |
+
"learning_rate": 2.295859872611465e-05,
|
| 2014 |
+
"loss": 0.9259,
|
| 2015 |
+
"step": 12900
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 14.63,
|
| 2019 |
+
"learning_rate": 2.2719745222929936e-05,
|
| 2020 |
+
"loss": 0.9228,
|
| 2021 |
+
"step": 12950
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 14.69,
|
| 2025 |
+
"learning_rate": 2.2480891719745222e-05,
|
| 2026 |
+
"loss": 0.9021,
|
| 2027 |
+
"step": 13000
|
| 2028 |
+
},
|
| 2029 |
+
{
|
| 2030 |
+
"epoch": 14.69,
|
| 2031 |
+
"eval_loss": 0.21045178174972534,
|
| 2032 |
+
"eval_runtime": 428.8167,
|
| 2033 |
+
"eval_samples_per_second": 14.477,
|
| 2034 |
+
"eval_steps_per_second": 0.905,
|
| 2035 |
+
"eval_wer": 0.1781273646155427,
|
| 2036 |
+
"step": 13000
|
| 2037 |
+
},
|
| 2038 |
+
{
|
| 2039 |
+
"epoch": 14.75,
|
| 2040 |
+
"learning_rate": 2.224203821656051e-05,
|
| 2041 |
+
"loss": 0.9238,
|
| 2042 |
+
"step": 13050
|
| 2043 |
+
},
|
| 2044 |
+
{
|
| 2045 |
+
"epoch": 14.8,
|
| 2046 |
+
"learning_rate": 2.2003184713375796e-05,
|
| 2047 |
+
"loss": 0.9373,
|
| 2048 |
+
"step": 13100
|
| 2049 |
+
},
|
| 2050 |
+
{
|
| 2051 |
+
"epoch": 14.86,
|
| 2052 |
+
"learning_rate": 2.1764331210191083e-05,
|
| 2053 |
+
"loss": 0.9365,
|
| 2054 |
+
"step": 13150
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"epoch": 14.91,
|
| 2058 |
+
"learning_rate": 2.152547770700637e-05,
|
| 2059 |
+
"loss": 0.9656,
|
| 2060 |
+
"step": 13200
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 14.97,
|
| 2064 |
+
"learning_rate": 2.1286624203821656e-05,
|
| 2065 |
+
"loss": 0.9193,
|
| 2066 |
+
"step": 13250
|
| 2067 |
+
},
|
| 2068 |
+
{
|
| 2069 |
+
"epoch": 14.97,
|
| 2070 |
+
"eval_loss": 0.20348267257213593,
|
| 2071 |
+
"eval_runtime": 430.2042,
|
| 2072 |
+
"eval_samples_per_second": 14.43,
|
| 2073 |
+
"eval_steps_per_second": 0.902,
|
| 2074 |
+
"eval_wer": 0.17860379085822115,
|
| 2075 |
+
"step": 13250
|
| 2076 |
+
},
|
| 2077 |
+
{
|
| 2078 |
+
"epoch": 15.03,
|
| 2079 |
+
"learning_rate": 2.1047770700636943e-05,
|
| 2080 |
+
"loss": 0.9366,
|
| 2081 |
+
"step": 13300
|
| 2082 |
+
},
|
| 2083 |
+
{
|
| 2084 |
+
"epoch": 15.08,
|
| 2085 |
+
"learning_rate": 2.080891719745223e-05,
|
| 2086 |
+
"loss": 0.9129,
|
| 2087 |
+
"step": 13350
|
| 2088 |
+
},
|
| 2089 |
+
{
|
| 2090 |
+
"epoch": 15.14,
|
| 2091 |
+
"learning_rate": 2.0570063694267513e-05,
|
| 2092 |
+
"loss": 0.9032,
|
| 2093 |
+
"step": 13400
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 15.2,
|
| 2097 |
+
"learning_rate": 2.03312101910828e-05,
|
| 2098 |
+
"loss": 0.9152,
|
| 2099 |
+
"step": 13450
|
| 2100 |
+
},
|
| 2101 |
+
{
|
| 2102 |
+
"epoch": 15.25,
|
| 2103 |
+
"learning_rate": 2.0092356687898086e-05,
|
| 2104 |
+
"loss": 0.9214,
|
| 2105 |
+
"step": 13500
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 15.25,
|
| 2109 |
+
"eval_loss": 0.2034832239151001,
|
| 2110 |
+
"eval_runtime": 432.039,
|
| 2111 |
+
"eval_samples_per_second": 14.369,
|
| 2112 |
+
"eval_steps_per_second": 0.898,
|
| 2113 |
+
"eval_wer": 0.17661401066821117,
|
| 2114 |
+
"step": 13500
|
| 2115 |
+
},
|
| 2116 |
+
{
|
| 2117 |
+
"epoch": 15.31,
|
| 2118 |
+
"learning_rate": 1.9853503184713373e-05,
|
| 2119 |
+
"loss": 0.9438,
|
| 2120 |
+
"step": 13550
|
| 2121 |
+
},
|
| 2122 |
+
{
|
| 2123 |
+
"epoch": 15.37,
|
| 2124 |
+
"learning_rate": 1.961464968152866e-05,
|
| 2125 |
+
"loss": 0.9262,
|
| 2126 |
+
"step": 13600
|
| 2127 |
+
},
|
| 2128 |
+
{
|
| 2129 |
+
"epoch": 15.42,
|
| 2130 |
+
"learning_rate": 1.9375796178343947e-05,
|
| 2131 |
+
"loss": 0.9157,
|
| 2132 |
+
"step": 13650
|
| 2133 |
+
},
|
| 2134 |
+
{
|
| 2135 |
+
"epoch": 15.48,
|
| 2136 |
+
"learning_rate": 1.9136942675159233e-05,
|
| 2137 |
+
"loss": 0.9299,
|
| 2138 |
+
"step": 13700
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"epoch": 15.54,
|
| 2142 |
+
"learning_rate": 1.889808917197452e-05,
|
| 2143 |
+
"loss": 0.9048,
|
| 2144 |
+
"step": 13750
|
| 2145 |
+
},
|
| 2146 |
+
{
|
| 2147 |
+
"epoch": 15.54,
|
| 2148 |
+
"eval_loss": 0.19639889895915985,
|
| 2149 |
+
"eval_runtime": 438.8483,
|
| 2150 |
+
"eval_samples_per_second": 14.146,
|
| 2151 |
+
"eval_steps_per_second": 0.884,
|
| 2152 |
+
"eval_wer": 0.17581062523938082,
|
| 2153 |
+
"step": 13750
|
| 2154 |
+
},
|
| 2155 |
+
{
|
| 2156 |
+
"epoch": 15.59,
|
| 2157 |
+
"learning_rate": 1.8659235668789807e-05,
|
| 2158 |
+
"loss": 0.9399,
|
| 2159 |
+
"step": 13800
|
| 2160 |
+
},
|
| 2161 |
+
{
|
| 2162 |
+
"epoch": 15.65,
|
| 2163 |
+
"learning_rate": 1.8420382165605094e-05,
|
| 2164 |
+
"loss": 0.9309,
|
| 2165 |
+
"step": 13850
|
| 2166 |
+
},
|
| 2167 |
+
{
|
| 2168 |
+
"epoch": 15.71,
|
| 2169 |
+
"learning_rate": 1.818152866242038e-05,
|
| 2170 |
+
"loss": 0.9646,
|
| 2171 |
+
"step": 13900
|
| 2172 |
+
},
|
| 2173 |
+
{
|
| 2174 |
+
"epoch": 15.76,
|
| 2175 |
+
"learning_rate": 1.7942675159235667e-05,
|
| 2176 |
+
"loss": 0.9095,
|
| 2177 |
+
"step": 13950
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 15.82,
|
| 2181 |
+
"learning_rate": 1.7703821656050954e-05,
|
| 2182 |
+
"loss": 0.9006,
|
| 2183 |
+
"step": 14000
|
| 2184 |
+
},
|
| 2185 |
+
{
|
| 2186 |
+
"epoch": 15.82,
|
| 2187 |
+
"eval_loss": 0.19844159483909607,
|
| 2188 |
+
"eval_runtime": 435.4721,
|
| 2189 |
+
"eval_samples_per_second": 14.256,
|
| 2190 |
+
"eval_steps_per_second": 0.891,
|
| 2191 |
+
"eval_wer": 0.17574523340215045,
|
| 2192 |
+
"step": 14000
|
| 2193 |
+
},
|
| 2194 |
+
{
|
| 2195 |
+
"epoch": 15.87,
|
| 2196 |
+
"learning_rate": 1.746496815286624e-05,
|
| 2197 |
+
"loss": 0.8845,
|
| 2198 |
+
"step": 14050
|
| 2199 |
+
},
|
| 2200 |
+
{
|
| 2201 |
+
"epoch": 15.93,
|
| 2202 |
+
"learning_rate": 1.7226114649681527e-05,
|
| 2203 |
+
"loss": 0.8991,
|
| 2204 |
+
"step": 14100
|
| 2205 |
+
},
|
| 2206 |
+
{
|
| 2207 |
+
"epoch": 15.99,
|
| 2208 |
+
"learning_rate": 1.6987261146496814e-05,
|
| 2209 |
+
"loss": 0.9266,
|
| 2210 |
+
"step": 14150
|
| 2211 |
+
},
|
| 2212 |
+
{
|
| 2213 |
+
"epoch": 16.05,
|
| 2214 |
+
"learning_rate": 1.67484076433121e-05,
|
| 2215 |
+
"loss": 0.9535,
|
| 2216 |
+
"step": 14200
|
| 2217 |
+
},
|
| 2218 |
+
{
|
| 2219 |
+
"epoch": 16.1,
|
| 2220 |
+
"learning_rate": 1.6509554140127387e-05,
|
| 2221 |
+
"loss": 0.9027,
|
| 2222 |
+
"step": 14250
|
| 2223 |
+
},
|
| 2224 |
+
{
|
| 2225 |
+
"epoch": 16.1,
|
| 2226 |
+
"eval_loss": 0.20223206281661987,
|
| 2227 |
+
"eval_runtime": 434.737,
|
| 2228 |
+
"eval_samples_per_second": 14.28,
|
| 2229 |
+
"eval_steps_per_second": 0.892,
|
| 2230 |
+
"eval_wer": 0.17431595467411512,
|
| 2231 |
+
"step": 14250
|
| 2232 |
+
},
|
| 2233 |
+
{
|
| 2234 |
+
"epoch": 16.16,
|
| 2235 |
+
"learning_rate": 1.6270700636942674e-05,
|
| 2236 |
+
"loss": 0.9095,
|
| 2237 |
+
"step": 14300
|
| 2238 |
+
},
|
| 2239 |
+
{
|
| 2240 |
+
"epoch": 16.21,
|
| 2241 |
+
"learning_rate": 1.603184713375796e-05,
|
| 2242 |
+
"loss": 0.9024,
|
| 2243 |
+
"step": 14350
|
| 2244 |
+
},
|
| 2245 |
+
{
|
| 2246 |
+
"epoch": 16.27,
|
| 2247 |
+
"learning_rate": 1.5792993630573248e-05,
|
| 2248 |
+
"loss": 0.9135,
|
| 2249 |
+
"step": 14400
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 16.33,
|
| 2253 |
+
"learning_rate": 1.5554140127388534e-05,
|
| 2254 |
+
"loss": 0.9013,
|
| 2255 |
+
"step": 14450
|
| 2256 |
+
},
|
| 2257 |
+
{
|
| 2258 |
+
"epoch": 16.38,
|
| 2259 |
+
"learning_rate": 1.531528662420382e-05,
|
| 2260 |
+
"loss": 0.9083,
|
| 2261 |
+
"step": 14500
|
| 2262 |
+
},
|
| 2263 |
+
{
|
| 2264 |
+
"epoch": 16.38,
|
| 2265 |
+
"eval_loss": 0.19693595170974731,
|
| 2266 |
+
"eval_runtime": 437.2683,
|
| 2267 |
+
"eval_samples_per_second": 14.197,
|
| 2268 |
+
"eval_steps_per_second": 0.887,
|
| 2269 |
+
"eval_wer": 0.1744000298934113,
|
| 2270 |
+
"step": 14500
|
| 2271 |
+
},
|
| 2272 |
+
{
|
| 2273 |
+
"epoch": 16.44,
|
| 2274 |
+
"learning_rate": 1.5076433121019106e-05,
|
| 2275 |
+
"loss": 0.9173,
|
| 2276 |
+
"step": 14550
|
| 2277 |
+
},
|
| 2278 |
+
{
|
| 2279 |
+
"epoch": 16.5,
|
| 2280 |
+
"learning_rate": 1.4837579617834393e-05,
|
| 2281 |
+
"loss": 0.9133,
|
| 2282 |
+
"step": 14600
|
| 2283 |
+
},
|
| 2284 |
+
{
|
| 2285 |
+
"epoch": 16.55,
|
| 2286 |
+
"learning_rate": 1.459872611464968e-05,
|
| 2287 |
+
"loss": 0.9161,
|
| 2288 |
+
"step": 14650
|
| 2289 |
+
},
|
| 2290 |
+
{
|
| 2291 |
+
"epoch": 16.61,
|
| 2292 |
+
"learning_rate": 1.4359872611464966e-05,
|
| 2293 |
+
"loss": 0.8844,
|
| 2294 |
+
"step": 14700
|
| 2295 |
+
},
|
| 2296 |
+
{
|
| 2297 |
+
"epoch": 16.67,
|
| 2298 |
+
"learning_rate": 1.4121019108280253e-05,
|
| 2299 |
+
"loss": 0.9761,
|
| 2300 |
+
"step": 14750
|
| 2301 |
+
},
|
| 2302 |
+
{
|
| 2303 |
+
"epoch": 16.67,
|
| 2304 |
+
"eval_loss": 0.19631367921829224,
|
| 2305 |
+
"eval_runtime": 434.2237,
|
| 2306 |
+
"eval_samples_per_second": 14.297,
|
| 2307 |
+
"eval_steps_per_second": 0.894,
|
| 2308 |
+
"eval_wer": 0.17276523396265192,
|
| 2309 |
+
"step": 14750
|
| 2310 |
+
},
|
| 2311 |
+
{
|
| 2312 |
+
"epoch": 16.72,
|
| 2313 |
+
"learning_rate": 1.388216560509554e-05,
|
| 2314 |
+
"loss": 0.9057,
|
| 2315 |
+
"step": 14800
|
| 2316 |
+
},
|
| 2317 |
+
{
|
| 2318 |
+
"epoch": 16.78,
|
| 2319 |
+
"learning_rate": 1.3643312101910826e-05,
|
| 2320 |
+
"loss": 0.9128,
|
| 2321 |
+
"step": 14850
|
| 2322 |
+
},
|
| 2323 |
+
{
|
| 2324 |
+
"epoch": 16.84,
|
| 2325 |
+
"learning_rate": 1.3404458598726113e-05,
|
| 2326 |
+
"loss": 0.9056,
|
| 2327 |
+
"step": 14900
|
| 2328 |
+
},
|
| 2329 |
+
{
|
| 2330 |
+
"epoch": 16.89,
|
| 2331 |
+
"learning_rate": 1.31656050955414e-05,
|
| 2332 |
+
"loss": 0.9024,
|
| 2333 |
+
"step": 14950
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 16.95,
|
| 2337 |
+
"learning_rate": 1.2926751592356687e-05,
|
| 2338 |
+
"loss": 0.9311,
|
| 2339 |
+
"step": 15000
|
| 2340 |
+
},
|
| 2341 |
+
{
|
| 2342 |
+
"epoch": 16.95,
|
| 2343 |
+
"eval_loss": 0.19600756466388702,
|
| 2344 |
+
"eval_runtime": 438.9128,
|
| 2345 |
+
"eval_samples_per_second": 14.144,
|
| 2346 |
+
"eval_steps_per_second": 0.884,
|
| 2347 |
+
"eval_wer": 0.1736807196838772,
|
| 2348 |
+
"step": 15000
|
| 2349 |
+
},
|
| 2350 |
+
{
|
| 2351 |
+
"epoch": 17.01,
|
| 2352 |
+
"learning_rate": 1.2687898089171973e-05,
|
| 2353 |
+
"loss": 0.9372,
|
| 2354 |
+
"step": 15050
|
| 2355 |
+
},
|
| 2356 |
+
{
|
| 2357 |
+
"epoch": 17.06,
|
| 2358 |
+
"learning_rate": 1.244904458598726e-05,
|
| 2359 |
+
"loss": 0.8955,
|
| 2360 |
+
"step": 15100
|
| 2361 |
+
},
|
| 2362 |
+
{
|
| 2363 |
+
"epoch": 17.12,
|
| 2364 |
+
"learning_rate": 1.2210191082802547e-05,
|
| 2365 |
+
"loss": 0.909,
|
| 2366 |
+
"step": 15150
|
| 2367 |
+
},
|
| 2368 |
+
{
|
| 2369 |
+
"epoch": 17.17,
|
| 2370 |
+
"learning_rate": 1.1971337579617834e-05,
|
| 2371 |
+
"loss": 0.9092,
|
| 2372 |
+
"step": 15200
|
| 2373 |
+
},
|
| 2374 |
+
{
|
| 2375 |
+
"epoch": 17.23,
|
| 2376 |
+
"learning_rate": 1.173248407643312e-05,
|
| 2377 |
+
"loss": 0.886,
|
| 2378 |
+
"step": 15250
|
| 2379 |
+
},
|
| 2380 |
+
{
|
| 2381 |
+
"epoch": 17.23,
|
| 2382 |
+
"eval_loss": 0.1928754597902298,
|
| 2383 |
+
"eval_runtime": 438.297,
|
| 2384 |
+
"eval_samples_per_second": 14.164,
|
| 2385 |
+
"eval_steps_per_second": 0.885,
|
| 2386 |
+
"eval_wer": 0.17263445028819116,
|
| 2387 |
+
"step": 15250
|
| 2388 |
+
},
|
| 2389 |
+
{
|
| 2390 |
+
"epoch": 17.29,
|
| 2391 |
+
"learning_rate": 1.1493630573248407e-05,
|
| 2392 |
+
"loss": 0.9053,
|
| 2393 |
+
"step": 15300
|
| 2394 |
+
},
|
| 2395 |
+
{
|
| 2396 |
+
"epoch": 17.34,
|
| 2397 |
+
"learning_rate": 1.1254777070063694e-05,
|
| 2398 |
+
"loss": 0.9056,
|
| 2399 |
+
"step": 15350
|
| 2400 |
+
},
|
| 2401 |
+
{
|
| 2402 |
+
"epoch": 17.4,
|
| 2403 |
+
"learning_rate": 1.101592356687898e-05,
|
| 2404 |
+
"loss": 0.9219,
|
| 2405 |
+
"step": 15400
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 17.46,
|
| 2409 |
+
"learning_rate": 1.0777070063694267e-05,
|
| 2410 |
+
"loss": 0.8967,
|
| 2411 |
+
"step": 15450
|
| 2412 |
+
},
|
| 2413 |
+
{
|
| 2414 |
+
"epoch": 17.51,
|
| 2415 |
+
"learning_rate": 1.0538216560509554e-05,
|
| 2416 |
+
"loss": 0.8969,
|
| 2417 |
+
"step": 15500
|
| 2418 |
+
},
|
| 2419 |
+
{
|
| 2420 |
+
"epoch": 17.51,
|
| 2421 |
+
"eval_loss": 0.1928360015153885,
|
| 2422 |
+
"eval_runtime": 442.1109,
|
| 2423 |
+
"eval_samples_per_second": 14.042,
|
| 2424 |
+
"eval_steps_per_second": 0.878,
|
| 2425 |
+
"eval_wer": 0.17337244387979112,
|
| 2426 |
+
"step": 15500
|
| 2427 |
+
},
|
| 2428 |
+
{
|
| 2429 |
+
"epoch": 17.57,
|
| 2430 |
+
"learning_rate": 1.029936305732484e-05,
|
| 2431 |
+
"loss": 0.8899,
|
| 2432 |
+
"step": 15550
|
| 2433 |
+
},
|
| 2434 |
+
{
|
| 2435 |
+
"epoch": 17.63,
|
| 2436 |
+
"learning_rate": 1.0060509554140127e-05,
|
| 2437 |
+
"loss": 0.9056,
|
| 2438 |
+
"step": 15600
|
| 2439 |
+
},
|
| 2440 |
+
{
|
| 2441 |
+
"epoch": 17.68,
|
| 2442 |
+
"learning_rate": 9.821656050955414e-06,
|
| 2443 |
+
"loss": 0.9048,
|
| 2444 |
+
"step": 15650
|
| 2445 |
+
},
|
| 2446 |
+
{
|
| 2447 |
+
"epoch": 17.74,
|
| 2448 |
+
"learning_rate": 9.582802547770701e-06,
|
| 2449 |
+
"loss": 0.9572,
|
| 2450 |
+
"step": 15700
|
| 2451 |
+
},
|
| 2452 |
+
{
|
| 2453 |
+
"epoch": 17.8,
|
| 2454 |
+
"learning_rate": 9.34872611464968e-06,
|
| 2455 |
+
"loss": 0.9084,
|
| 2456 |
+
"step": 15750
|
| 2457 |
+
},
|
| 2458 |
+
{
|
| 2459 |
+
"epoch": 17.8,
|
| 2460 |
+
"eval_loss": 0.19373278319835663,
|
| 2461 |
+
"eval_runtime": 446.3693,
|
| 2462 |
+
"eval_samples_per_second": 13.908,
|
| 2463 |
+
"eval_steps_per_second": 0.869,
|
| 2464 |
+
"eval_wer": 0.17133595523461656,
|
| 2465 |
+
"step": 15750
|
| 2466 |
+
},
|
| 2467 |
+
{
|
| 2468 |
+
"epoch": 17.85,
|
| 2469 |
+
"learning_rate": 9.109872611464967e-06,
|
| 2470 |
+
"loss": 0.8861,
|
| 2471 |
+
"step": 15800
|
| 2472 |
+
},
|
| 2473 |
+
{
|
| 2474 |
+
"epoch": 17.91,
|
| 2475 |
+
"learning_rate": 8.871019108280254e-06,
|
| 2476 |
+
"loss": 0.8842,
|
| 2477 |
+
"step": 15850
|
| 2478 |
+
},
|
| 2479 |
+
{
|
| 2480 |
+
"epoch": 17.97,
|
| 2481 |
+
"learning_rate": 8.63216560509554e-06,
|
| 2482 |
+
"loss": 0.8949,
|
| 2483 |
+
"step": 15900
|
| 2484 |
+
},
|
| 2485 |
+
{
|
| 2486 |
+
"epoch": 18.02,
|
| 2487 |
+
"learning_rate": 8.398089171974522e-06,
|
| 2488 |
+
"loss": 0.8977,
|
| 2489 |
+
"step": 15950
|
| 2490 |
+
},
|
| 2491 |
+
{
|
| 2492 |
+
"epoch": 18.08,
|
| 2493 |
+
"learning_rate": 8.159235668789809e-06,
|
| 2494 |
+
"loss": 0.8795,
|
| 2495 |
+
"step": 16000
|
| 2496 |
+
},
|
| 2497 |
+
{
|
| 2498 |
+
"epoch": 18.08,
|
| 2499 |
+
"eval_loss": 0.1977699100971222,
|
| 2500 |
+
"eval_runtime": 437.7611,
|
| 2501 |
+
"eval_samples_per_second": 14.181,
|
| 2502 |
+
"eval_steps_per_second": 0.886,
|
| 2503 |
+
"eval_wer": 0.17086887068297102,
|
| 2504 |
+
"step": 16000
|
| 2505 |
+
},
|
| 2506 |
+
{
|
| 2507 |
+
"epoch": 18.14,
|
| 2508 |
+
"learning_rate": 7.920382165605094e-06,
|
| 2509 |
+
"loss": 0.8984,
|
| 2510 |
+
"step": 16050
|
| 2511 |
+
},
|
| 2512 |
+
{
|
| 2513 |
+
"epoch": 18.19,
|
| 2514 |
+
"learning_rate": 7.68152866242038e-06,
|
| 2515 |
+
"loss": 0.9005,
|
| 2516 |
+
"step": 16100
|
| 2517 |
+
},
|
| 2518 |
+
{
|
| 2519 |
+
"epoch": 18.25,
|
| 2520 |
+
"learning_rate": 7.4426751592356675e-06,
|
| 2521 |
+
"loss": 0.8981,
|
| 2522 |
+
"step": 16150
|
| 2523 |
+
},
|
| 2524 |
+
{
|
| 2525 |
+
"epoch": 18.3,
|
| 2526 |
+
"learning_rate": 7.203821656050954e-06,
|
| 2527 |
+
"loss": 0.9029,
|
| 2528 |
+
"step": 16200
|
| 2529 |
+
},
|
| 2530 |
+
{
|
| 2531 |
+
"epoch": 18.36,
|
| 2532 |
+
"learning_rate": 6.964968152866241e-06,
|
| 2533 |
+
"loss": 0.8883,
|
| 2534 |
+
"step": 16250
|
| 2535 |
+
},
|
| 2536 |
+
{
|
| 2537 |
+
"epoch": 18.36,
|
| 2538 |
+
"eval_loss": 0.19563348591327667,
|
| 2539 |
+
"eval_runtime": 434.7761,
|
| 2540 |
+
"eval_samples_per_second": 14.279,
|
| 2541 |
+
"eval_steps_per_second": 0.892,
|
| 2542 |
+
"eval_wer": 0.17032705260306222,
|
| 2543 |
+
"step": 16250
|
| 2544 |
+
},
|
| 2545 |
+
{
|
| 2546 |
+
"epoch": 18.42,
|
| 2547 |
+
"learning_rate": 6.726114649681528e-06,
|
| 2548 |
+
"loss": 0.8919,
|
| 2549 |
+
"step": 16300
|
| 2550 |
+
},
|
| 2551 |
+
{
|
| 2552 |
+
"epoch": 18.47,
|
| 2553 |
+
"learning_rate": 6.4872611464968145e-06,
|
| 2554 |
+
"loss": 0.8978,
|
| 2555 |
+
"step": 16350
|
| 2556 |
+
},
|
| 2557 |
+
{
|
| 2558 |
+
"epoch": 18.53,
|
| 2559 |
+
"learning_rate": 6.248407643312101e-06,
|
| 2560 |
+
"loss": 0.8897,
|
| 2561 |
+
"step": 16400
|
| 2562 |
+
},
|
| 2563 |
+
{
|
| 2564 |
+
"epoch": 18.59,
|
| 2565 |
+
"learning_rate": 6.009554140127388e-06,
|
| 2566 |
+
"loss": 0.9477,
|
| 2567 |
+
"step": 16450
|
| 2568 |
+
},
|
| 2569 |
+
{
|
| 2570 |
+
"epoch": 18.64,
|
| 2571 |
+
"learning_rate": 5.770700636942675e-06,
|
| 2572 |
+
"loss": 0.8901,
|
| 2573 |
+
"step": 16500
|
| 2574 |
+
},
|
| 2575 |
+
{
|
| 2576 |
+
"epoch": 18.64,
|
| 2577 |
+
"eval_loss": 0.19332656264305115,
|
| 2578 |
+
"eval_runtime": 439.133,
|
| 2579 |
+
"eval_samples_per_second": 14.137,
|
| 2580 |
+
"eval_steps_per_second": 0.884,
|
| 2581 |
+
"eval_wer": 0.17053256980578624,
|
| 2582 |
+
"step": 16500
|
| 2583 |
+
},
|
| 2584 |
+
{
|
| 2585 |
+
"epoch": 18.7,
|
| 2586 |
+
"learning_rate": 5.531847133757961e-06,
|
| 2587 |
+
"loss": 0.8992,
|
| 2588 |
+
"step": 16550
|
| 2589 |
+
},
|
| 2590 |
+
{
|
| 2591 |
+
"epoch": 18.76,
|
| 2592 |
+
"learning_rate": 5.292993630573248e-06,
|
| 2593 |
+
"loss": 0.8988,
|
| 2594 |
+
"step": 16600
|
| 2595 |
+
},
|
| 2596 |
+
{
|
| 2597 |
+
"epoch": 18.81,
|
| 2598 |
+
"learning_rate": 5.054140127388535e-06,
|
| 2599 |
+
"loss": 0.8885,
|
| 2600 |
+
"step": 16650
|
| 2601 |
+
},
|
| 2602 |
+
{
|
| 2603 |
+
"epoch": 18.87,
|
| 2604 |
+
"learning_rate": 4.815286624203822e-06,
|
| 2605 |
+
"loss": 0.8837,
|
| 2606 |
+
"step": 16700
|
| 2607 |
+
},
|
| 2608 |
+
{
|
| 2609 |
+
"epoch": 18.93,
|
| 2610 |
+
"learning_rate": 4.576433121019108e-06,
|
| 2611 |
+
"loss": 0.8922,
|
| 2612 |
+
"step": 16750
|
| 2613 |
+
},
|
| 2614 |
+
{
|
| 2615 |
+
"epoch": 18.93,
|
| 2616 |
+
"eval_loss": 0.1962379515171051,
|
| 2617 |
+
"eval_runtime": 444.5287,
|
| 2618 |
+
"eval_samples_per_second": 13.965,
|
| 2619 |
+
"eval_steps_per_second": 0.873,
|
| 2620 |
+
"eval_wer": 0.17109307126776088,
|
| 2621 |
+
"step": 16750
|
| 2622 |
+
},
|
| 2623 |
+
{
|
| 2624 |
+
"epoch": 18.98,
|
| 2625 |
+
"learning_rate": 4.337579617834394e-06,
|
| 2626 |
+
"loss": 0.8943,
|
| 2627 |
+
"step": 16800
|
| 2628 |
+
},
|
| 2629 |
+
{
|
| 2630 |
+
"epoch": 19.04,
|
| 2631 |
+
"learning_rate": 4.098726114649681e-06,
|
| 2632 |
+
"loss": 0.9171,
|
| 2633 |
+
"step": 16850
|
| 2634 |
+
},
|
| 2635 |
+
{
|
| 2636 |
+
"epoch": 19.1,
|
| 2637 |
+
"learning_rate": 3.859872611464968e-06,
|
| 2638 |
+
"loss": 0.9144,
|
| 2639 |
+
"step": 16900
|
| 2640 |
+
},
|
| 2641 |
+
{
|
| 2642 |
+
"epoch": 19.15,
|
| 2643 |
+
"learning_rate": 3.6210191082802544e-06,
|
| 2644 |
+
"loss": 0.9517,
|
| 2645 |
+
"step": 16950
|
| 2646 |
+
},
|
| 2647 |
+
{
|
| 2648 |
+
"epoch": 19.21,
|
| 2649 |
+
"learning_rate": 3.382165605095541e-06,
|
| 2650 |
+
"loss": 0.8765,
|
| 2651 |
+
"step": 17000
|
| 2652 |
+
},
|
| 2653 |
+
{
|
| 2654 |
+
"epoch": 19.21,
|
| 2655 |
+
"eval_loss": 0.19622743129730225,
|
| 2656 |
+
"eval_runtime": 445.8046,
|
| 2657 |
+
"eval_samples_per_second": 13.925,
|
| 2658 |
+
"eval_steps_per_second": 0.87,
|
| 2659 |
+
"eval_wer": 0.17106504619466215,
|
| 2660 |
+
"step": 17000
|
| 2661 |
+
},
|
| 2662 |
+
{
|
| 2663 |
+
"epoch": 19.27,
|
| 2664 |
+
"learning_rate": 3.143312101910828e-06,
|
| 2665 |
+
"loss": 0.9072,
|
| 2666 |
+
"step": 17050
|
| 2667 |
+
},
|
| 2668 |
+
{
|
| 2669 |
+
"epoch": 19.32,
|
| 2670 |
+
"learning_rate": 2.9044585987261146e-06,
|
| 2671 |
+
"loss": 0.8897,
|
| 2672 |
+
"step": 17100
|
| 2673 |
+
},
|
| 2674 |
+
{
|
| 2675 |
+
"epoch": 19.38,
|
| 2676 |
+
"learning_rate": 2.6656050955414013e-06,
|
| 2677 |
+
"loss": 0.8879,
|
| 2678 |
+
"step": 17150
|
| 2679 |
+
},
|
| 2680 |
+
{
|
| 2681 |
+
"epoch": 19.43,
|
| 2682 |
+
"learning_rate": 2.426751592356688e-06,
|
| 2683 |
+
"loss": 0.883,
|
| 2684 |
+
"step": 17200
|
| 2685 |
+
},
|
| 2686 |
+
{
|
| 2687 |
+
"epoch": 19.49,
|
| 2688 |
+
"learning_rate": 2.1878980891719744e-06,
|
| 2689 |
+
"loss": 0.8992,
|
| 2690 |
+
"step": 17250
|
| 2691 |
+
},
|
| 2692 |
+
{
|
| 2693 |
+
"epoch": 19.49,
|
| 2694 |
+
"eval_loss": 0.19645148515701294,
|
| 2695 |
+
"eval_runtime": 447.1526,
|
| 2696 |
+
"eval_samples_per_second": 13.883,
|
| 2697 |
+
"eval_steps_per_second": 0.868,
|
| 2698 |
+
"eval_wer": 0.17034573598512803,
|
| 2699 |
+
"step": 17250
|
| 2700 |
+
},
|
| 2701 |
+
{
|
| 2702 |
+
"epoch": 19.55,
|
| 2703 |
+
"learning_rate": 1.949044585987261e-06,
|
| 2704 |
+
"loss": 0.8969,
|
| 2705 |
+
"step": 17300
|
| 2706 |
+
},
|
| 2707 |
+
{
|
| 2708 |
+
"epoch": 19.6,
|
| 2709 |
+
"learning_rate": 1.7101910828025476e-06,
|
| 2710 |
+
"loss": 0.872,
|
| 2711 |
+
"step": 17350
|
| 2712 |
+
},
|
| 2713 |
+
{
|
| 2714 |
+
"epoch": 19.66,
|
| 2715 |
+
"learning_rate": 1.4713375796178341e-06,
|
| 2716 |
+
"loss": 0.8984,
|
| 2717 |
+
"step": 17400
|
| 2718 |
+
},
|
| 2719 |
+
{
|
| 2720 |
+
"epoch": 19.72,
|
| 2721 |
+
"learning_rate": 1.2324840764331209e-06,
|
| 2722 |
+
"loss": 0.8913,
|
| 2723 |
+
"step": 17450
|
| 2724 |
+
},
|
| 2725 |
+
{
|
| 2726 |
+
"epoch": 19.77,
|
| 2727 |
+
"learning_rate": 9.936305732484076e-07,
|
| 2728 |
+
"loss": 0.8778,
|
| 2729 |
+
"step": 17500
|
| 2730 |
+
},
|
| 2731 |
+
{
|
| 2732 |
+
"epoch": 19.77,
|
| 2733 |
+
"eval_loss": 0.19571013748645782,
|
| 2734 |
+
"eval_runtime": 442.6975,
|
| 2735 |
+
"eval_samples_per_second": 14.023,
|
| 2736 |
+
"eval_steps_per_second": 0.876,
|
| 2737 |
+
"eval_wer": 0.16990667650658123,
|
| 2738 |
+
"step": 17500
|
| 2739 |
+
},
|
| 2740 |
+
{
|
| 2741 |
+
"epoch": 19.83,
|
| 2742 |
+
"learning_rate": 7.547770700636942e-07,
|
| 2743 |
+
"loss": 0.8687,
|
| 2744 |
+
"step": 17550
|
| 2745 |
+
},
|
| 2746 |
+
{
|
| 2747 |
+
"epoch": 19.89,
|
| 2748 |
+
"learning_rate": 5.159235668789809e-07,
|
| 2749 |
+
"loss": 0.8858,
|
| 2750 |
+
"step": 17600
|
| 2751 |
+
},
|
| 2752 |
+
{
|
| 2753 |
+
"epoch": 19.94,
|
| 2754 |
+
"learning_rate": 2.770700636942675e-07,
|
| 2755 |
+
"loss": 0.8854,
|
| 2756 |
+
"step": 17650
|
| 2757 |
+
},
|
| 2758 |
+
{
|
| 2759 |
+
"epoch": 20.0,
|
| 2760 |
+
"learning_rate": 3.821656050955413e-08,
|
| 2761 |
+
"loss": 0.8898,
|
| 2762 |
+
"step": 17700
|
| 2763 |
+
},
|
| 2764 |
+
{
|
| 2765 |
+
"epoch": 20.0,
|
| 2766 |
+
"step": 17700,
|
| 2767 |
+
"total_flos": 2.3221664293970497e+20,
|
| 2768 |
+
"train_loss": 1.2649466082470566,
|
| 2769 |
+
"train_runtime": 115228.5602,
|
| 2770 |
+
"train_samples_per_second": 9.837,
|
| 2771 |
+
"train_steps_per_second": 0.154
|
| 2772 |
+
}
|
| 2773 |
+
],
|
| 2774 |
+
"max_steps": 17700,
|
| 2775 |
+
"num_train_epochs": 20,
|
| 2776 |
+
"total_flos": 2.3221664293970497e+20,
|
| 2777 |
+
"trial_name": null,
|
| 2778 |
+
"trial_params": null
|
| 2779 |
+
}
|