End of training
Browse files- all_results.json +14 -0
- eval_results.json +9 -0
- runs/Jun23_03-37-27_e5d9982fab87/events.out.tfevents.1656062871.e5d9982fab87.2021.2 +3 -0
- train_results.json +8 -0
- trainer_state.json +2311 -0
all_results.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 30.0,
|
| 3 |
+
"eval_loss": 1.7582672834396362,
|
| 4 |
+
"eval_runtime": 228.4233,
|
| 5 |
+
"eval_samples": 1994,
|
| 6 |
+
"eval_samples_per_second": 8.729,
|
| 7 |
+
"eval_steps_per_second": 0.876,
|
| 8 |
+
"eval_wer": 0.9386031533857885,
|
| 9 |
+
"train_loss": 3.5984562555948894,
|
| 10 |
+
"train_runtime": 102250.8509,
|
| 11 |
+
"train_samples": 17937,
|
| 12 |
+
"train_samples_per_second": 5.263,
|
| 13 |
+
"train_steps_per_second": 0.033
|
| 14 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 30.0,
|
| 3 |
+
"eval_loss": 1.7582672834396362,
|
| 4 |
+
"eval_runtime": 228.4233,
|
| 5 |
+
"eval_samples": 1994,
|
| 6 |
+
"eval_samples_per_second": 8.729,
|
| 7 |
+
"eval_steps_per_second": 0.876,
|
| 8 |
+
"eval_wer": 0.9386031533857885
|
| 9 |
+
}
|
runs/Jun23_03-37-27_e5d9982fab87/events.out.tfevents.1656062871.e5d9982fab87.2021.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:873523a4b9bf981438317cf56333f771606aae0cd0f0520c0b3c48d7f414d884
|
| 3 |
+
size 358
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 30.0,
|
| 3 |
+
"train_loss": 3.5984562555948894,
|
| 4 |
+
"train_runtime": 102250.8509,
|
| 5 |
+
"train_samples": 17937,
|
| 6 |
+
"train_samples_per_second": 5.263,
|
| 7 |
+
"train_steps_per_second": 0.033
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2311 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 1.7582672834396362,
|
| 3 |
+
"best_model_checkpoint": "ai-light-dance_singing2_ft_wav2vec2-large-xlsr-53/checkpoint-3360",
|
| 4 |
+
"epoch": 29.998885172798218,
|
| 5 |
+
"global_step": 3360,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.09,
|
| 12 |
+
"learning_rate": 2.8e-07,
|
| 13 |
+
"loss": 34.7195,
|
| 14 |
+
"step": 10
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.18,
|
| 18 |
+
"learning_rate": 6.800000000000001e-07,
|
| 19 |
+
"loss": 35.9093,
|
| 20 |
+
"step": 20
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.27,
|
| 24 |
+
"learning_rate": 1.08e-06,
|
| 25 |
+
"loss": 31.3284,
|
| 26 |
+
"step": 30
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.36,
|
| 30 |
+
"learning_rate": 1.48e-06,
|
| 31 |
+
"loss": 34.058,
|
| 32 |
+
"step": 40
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.45,
|
| 36 |
+
"learning_rate": 1.8799999999999998e-06,
|
| 37 |
+
"loss": 33.406,
|
| 38 |
+
"step": 50
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.54,
|
| 42 |
+
"learning_rate": 2.2799999999999998e-06,
|
| 43 |
+
"loss": 34.702,
|
| 44 |
+
"step": 60
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.62,
|
| 48 |
+
"learning_rate": 2.68e-06,
|
| 49 |
+
"loss": 34.8616,
|
| 50 |
+
"step": 70
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.71,
|
| 54 |
+
"learning_rate": 3.0799999999999997e-06,
|
| 55 |
+
"loss": 30.6447,
|
| 56 |
+
"step": 80
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.8,
|
| 60 |
+
"learning_rate": 3.4799999999999997e-06,
|
| 61 |
+
"loss": 33.7505,
|
| 62 |
+
"step": 90
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"epoch": 0.89,
|
| 66 |
+
"learning_rate": 3.88e-06,
|
| 67 |
+
"loss": 32.2757,
|
| 68 |
+
"step": 100
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.98,
|
| 72 |
+
"learning_rate": 3.991411042944785e-06,
|
| 73 |
+
"loss": 27.4755,
|
| 74 |
+
"step": 110
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 1.0,
|
| 78 |
+
"eval_loss": 23.2618350982666,
|
| 79 |
+
"eval_runtime": 226.4347,
|
| 80 |
+
"eval_samples_per_second": 8.806,
|
| 81 |
+
"eval_steps_per_second": 0.883,
|
| 82 |
+
"eval_wer": 1.0,
|
| 83 |
+
"step": 112
|
| 84 |
+
},
|
| 85 |
+
{
|
| 86 |
+
"epoch": 1.07,
|
| 87 |
+
"learning_rate": 3.979141104294478e-06,
|
| 88 |
+
"loss": 20.6317,
|
| 89 |
+
"step": 120
|
| 90 |
+
},
|
| 91 |
+
{
|
| 92 |
+
"epoch": 1.16,
|
| 93 |
+
"learning_rate": 3.9668711656441715e-06,
|
| 94 |
+
"loss": 16.4159,
|
| 95 |
+
"step": 130
|
| 96 |
+
},
|
| 97 |
+
{
|
| 98 |
+
"epoch": 1.25,
|
| 99 |
+
"learning_rate": 3.954601226993865e-06,
|
| 100 |
+
"loss": 12.3928,
|
| 101 |
+
"step": 140
|
| 102 |
+
},
|
| 103 |
+
{
|
| 104 |
+
"epoch": 1.34,
|
| 105 |
+
"learning_rate": 3.942331288343558e-06,
|
| 106 |
+
"loss": 10.4952,
|
| 107 |
+
"step": 150
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 1.43,
|
| 111 |
+
"learning_rate": 3.930061349693252e-06,
|
| 112 |
+
"loss": 8.7608,
|
| 113 |
+
"step": 160
|
| 114 |
+
},
|
| 115 |
+
{
|
| 116 |
+
"epoch": 1.52,
|
| 117 |
+
"learning_rate": 3.917791411042945e-06,
|
| 118 |
+
"loss": 8.1101,
|
| 119 |
+
"step": 170
|
| 120 |
+
},
|
| 121 |
+
{
|
| 122 |
+
"epoch": 1.61,
|
| 123 |
+
"learning_rate": 3.905521472392638e-06,
|
| 124 |
+
"loss": 7.022,
|
| 125 |
+
"step": 180
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"epoch": 1.7,
|
| 129 |
+
"learning_rate": 3.893251533742331e-06,
|
| 130 |
+
"loss": 6.5763,
|
| 131 |
+
"step": 190
|
| 132 |
+
},
|
| 133 |
+
{
|
| 134 |
+
"epoch": 1.78,
|
| 135 |
+
"learning_rate": 3.880981595092024e-06,
|
| 136 |
+
"loss": 6.0121,
|
| 137 |
+
"step": 200
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 1.87,
|
| 141 |
+
"learning_rate": 3.868711656441717e-06,
|
| 142 |
+
"loss": 5.7837,
|
| 143 |
+
"step": 210
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"epoch": 1.96,
|
| 147 |
+
"learning_rate": 3.8564417177914105e-06,
|
| 148 |
+
"loss": 5.5145,
|
| 149 |
+
"step": 220
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 2.0,
|
| 153 |
+
"eval_loss": 5.221268177032471,
|
| 154 |
+
"eval_runtime": 225.8013,
|
| 155 |
+
"eval_samples_per_second": 8.831,
|
| 156 |
+
"eval_steps_per_second": 0.886,
|
| 157 |
+
"eval_wer": 1.0,
|
| 158 |
+
"step": 224
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 2.05,
|
| 162 |
+
"learning_rate": 3.8441717791411045e-06,
|
| 163 |
+
"loss": 5.52,
|
| 164 |
+
"step": 230
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 2.14,
|
| 168 |
+
"learning_rate": 3.831901840490798e-06,
|
| 169 |
+
"loss": 4.9127,
|
| 170 |
+
"step": 240
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 2.23,
|
| 174 |
+
"learning_rate": 3.819631901840491e-06,
|
| 175 |
+
"loss": 4.7525,
|
| 176 |
+
"step": 250
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 2.32,
|
| 180 |
+
"learning_rate": 3.807361963190184e-06,
|
| 181 |
+
"loss": 4.5823,
|
| 182 |
+
"step": 260
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 2.41,
|
| 186 |
+
"learning_rate": 3.795092024539877e-06,
|
| 187 |
+
"loss": 4.5971,
|
| 188 |
+
"step": 270
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 2.5,
|
| 192 |
+
"learning_rate": 3.7828220858895705e-06,
|
| 193 |
+
"loss": 4.5216,
|
| 194 |
+
"step": 280
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 2.59,
|
| 198 |
+
"learning_rate": 3.7705521472392636e-06,
|
| 199 |
+
"loss": 4.5161,
|
| 200 |
+
"step": 290
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 2.68,
|
| 204 |
+
"learning_rate": 3.7582822085889567e-06,
|
| 205 |
+
"loss": 4.3186,
|
| 206 |
+
"step": 300
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 2.77,
|
| 210 |
+
"learning_rate": 3.7460122699386503e-06,
|
| 211 |
+
"loss": 4.3137,
|
| 212 |
+
"step": 310
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 2.86,
|
| 216 |
+
"learning_rate": 3.7337423312883434e-06,
|
| 217 |
+
"loss": 4.19,
|
| 218 |
+
"step": 320
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 2.95,
|
| 222 |
+
"learning_rate": 3.7214723926380365e-06,
|
| 223 |
+
"loss": 4.2211,
|
| 224 |
+
"step": 330
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 3.0,
|
| 228 |
+
"eval_loss": 4.167309284210205,
|
| 229 |
+
"eval_runtime": 226.6831,
|
| 230 |
+
"eval_samples_per_second": 8.796,
|
| 231 |
+
"eval_steps_per_second": 0.882,
|
| 232 |
+
"eval_wer": 1.0,
|
| 233 |
+
"step": 336
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 3.04,
|
| 237 |
+
"learning_rate": 3.7092024539877297e-06,
|
| 238 |
+
"loss": 4.2938,
|
| 239 |
+
"step": 340
|
| 240 |
+
},
|
| 241 |
+
{
|
| 242 |
+
"epoch": 3.12,
|
| 243 |
+
"learning_rate": 3.6969325153374232e-06,
|
| 244 |
+
"loss": 4.0354,
|
| 245 |
+
"step": 350
|
| 246 |
+
},
|
| 247 |
+
{
|
| 248 |
+
"epoch": 3.21,
|
| 249 |
+
"learning_rate": 3.6846625766871163e-06,
|
| 250 |
+
"loss": 4.209,
|
| 251 |
+
"step": 360
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 3.3,
|
| 255 |
+
"learning_rate": 3.67239263803681e-06,
|
| 256 |
+
"loss": 3.9982,
|
| 257 |
+
"step": 370
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"epoch": 3.39,
|
| 261 |
+
"learning_rate": 3.660122699386503e-06,
|
| 262 |
+
"loss": 4.0104,
|
| 263 |
+
"step": 380
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 3.48,
|
| 267 |
+
"learning_rate": 3.647852760736196e-06,
|
| 268 |
+
"loss": 3.8911,
|
| 269 |
+
"step": 390
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"epoch": 3.57,
|
| 273 |
+
"learning_rate": 3.6355828220858893e-06,
|
| 274 |
+
"loss": 4.0256,
|
| 275 |
+
"step": 400
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 3.66,
|
| 279 |
+
"learning_rate": 3.623312883435583e-06,
|
| 280 |
+
"loss": 3.8258,
|
| 281 |
+
"step": 410
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 3.75,
|
| 285 |
+
"learning_rate": 3.611042944785276e-06,
|
| 286 |
+
"loss": 3.7316,
|
| 287 |
+
"step": 420
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"epoch": 3.84,
|
| 291 |
+
"learning_rate": 3.598773006134969e-06,
|
| 292 |
+
"loss": 3.8932,
|
| 293 |
+
"step": 430
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 3.93,
|
| 297 |
+
"learning_rate": 3.5865030674846626e-06,
|
| 298 |
+
"loss": 3.8386,
|
| 299 |
+
"step": 440
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"epoch": 4.0,
|
| 303 |
+
"eval_loss": 3.825347661972046,
|
| 304 |
+
"eval_runtime": 224.561,
|
| 305 |
+
"eval_samples_per_second": 8.88,
|
| 306 |
+
"eval_steps_per_second": 0.891,
|
| 307 |
+
"eval_wer": 1.0,
|
| 308 |
+
"step": 448
|
| 309 |
+
},
|
| 310 |
+
{
|
| 311 |
+
"epoch": 4.02,
|
| 312 |
+
"learning_rate": 3.5742331288343553e-06,
|
| 313 |
+
"loss": 3.8842,
|
| 314 |
+
"step": 450
|
| 315 |
+
},
|
| 316 |
+
{
|
| 317 |
+
"epoch": 4.11,
|
| 318 |
+
"learning_rate": 3.561963190184049e-06,
|
| 319 |
+
"loss": 3.8318,
|
| 320 |
+
"step": 460
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 4.2,
|
| 324 |
+
"learning_rate": 3.549693251533742e-06,
|
| 325 |
+
"loss": 3.7975,
|
| 326 |
+
"step": 470
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 4.29,
|
| 330 |
+
"learning_rate": 3.5374233128834355e-06,
|
| 331 |
+
"loss": 3.7912,
|
| 332 |
+
"step": 480
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 4.37,
|
| 336 |
+
"learning_rate": 3.5251533742331287e-06,
|
| 337 |
+
"loss": 3.8111,
|
| 338 |
+
"step": 490
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 4.46,
|
| 342 |
+
"learning_rate": 3.512883435582822e-06,
|
| 343 |
+
"loss": 3.6635,
|
| 344 |
+
"step": 500
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 4.55,
|
| 348 |
+
"learning_rate": 3.5006134969325153e-06,
|
| 349 |
+
"loss": 3.6398,
|
| 350 |
+
"step": 510
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 4.64,
|
| 354 |
+
"learning_rate": 3.4883435582822085e-06,
|
| 355 |
+
"loss": 3.6428,
|
| 356 |
+
"step": 520
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"epoch": 4.73,
|
| 360 |
+
"learning_rate": 3.4760736196319016e-06,
|
| 361 |
+
"loss": 3.6082,
|
| 362 |
+
"step": 530
|
| 363 |
+
},
|
| 364 |
+
{
|
| 365 |
+
"epoch": 4.82,
|
| 366 |
+
"learning_rate": 3.463803680981595e-06,
|
| 367 |
+
"loss": 3.5746,
|
| 368 |
+
"step": 540
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 4.91,
|
| 372 |
+
"learning_rate": 3.4515337423312883e-06,
|
| 373 |
+
"loss": 3.664,
|
| 374 |
+
"step": 550
|
| 375 |
+
},
|
| 376 |
+
{
|
| 377 |
+
"epoch": 5.0,
|
| 378 |
+
"learning_rate": 3.4392638036809814e-06,
|
| 379 |
+
"loss": 3.5531,
|
| 380 |
+
"step": 560
|
| 381 |
+
},
|
| 382 |
+
{
|
| 383 |
+
"epoch": 5.0,
|
| 384 |
+
"eval_loss": 3.6285741329193115,
|
| 385 |
+
"eval_runtime": 224.6131,
|
| 386 |
+
"eval_samples_per_second": 8.877,
|
| 387 |
+
"eval_steps_per_second": 0.89,
|
| 388 |
+
"eval_wer": 1.0,
|
| 389 |
+
"step": 560
|
| 390 |
+
},
|
| 391 |
+
{
|
| 392 |
+
"epoch": 5.09,
|
| 393 |
+
"learning_rate": 3.426993865030675e-06,
|
| 394 |
+
"loss": 3.7207,
|
| 395 |
+
"step": 570
|
| 396 |
+
},
|
| 397 |
+
{
|
| 398 |
+
"epoch": 5.18,
|
| 399 |
+
"learning_rate": 3.4147239263803676e-06,
|
| 400 |
+
"loss": 3.62,
|
| 401 |
+
"step": 580
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 5.27,
|
| 405 |
+
"learning_rate": 3.402453987730061e-06,
|
| 406 |
+
"loss": 3.5644,
|
| 407 |
+
"step": 590
|
| 408 |
+
},
|
| 409 |
+
{
|
| 410 |
+
"epoch": 5.36,
|
| 411 |
+
"learning_rate": 3.3901840490797543e-06,
|
| 412 |
+
"loss": 3.4874,
|
| 413 |
+
"step": 600
|
| 414 |
+
},
|
| 415 |
+
{
|
| 416 |
+
"epoch": 5.45,
|
| 417 |
+
"learning_rate": 3.377914110429448e-06,
|
| 418 |
+
"loss": 3.5444,
|
| 419 |
+
"step": 610
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 5.54,
|
| 423 |
+
"learning_rate": 3.365644171779141e-06,
|
| 424 |
+
"loss": 3.4358,
|
| 425 |
+
"step": 620
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 5.62,
|
| 429 |
+
"learning_rate": 3.3533742331288345e-06,
|
| 430 |
+
"loss": 3.5017,
|
| 431 |
+
"step": 630
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 5.71,
|
| 435 |
+
"learning_rate": 3.3411042944785272e-06,
|
| 436 |
+
"loss": 3.4731,
|
| 437 |
+
"step": 640
|
| 438 |
+
},
|
| 439 |
+
{
|
| 440 |
+
"epoch": 5.8,
|
| 441 |
+
"learning_rate": 3.3288343558282208e-06,
|
| 442 |
+
"loss": 3.3658,
|
| 443 |
+
"step": 650
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 5.89,
|
| 447 |
+
"learning_rate": 3.316564417177914e-06,
|
| 448 |
+
"loss": 3.5377,
|
| 449 |
+
"step": 660
|
| 450 |
+
},
|
| 451 |
+
{
|
| 452 |
+
"epoch": 5.98,
|
| 453 |
+
"learning_rate": 3.304294478527607e-06,
|
| 454 |
+
"loss": 3.5215,
|
| 455 |
+
"step": 670
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"epoch": 6.0,
|
| 459 |
+
"eval_loss": 3.4761970043182373,
|
| 460 |
+
"eval_runtime": 224.497,
|
| 461 |
+
"eval_samples_per_second": 8.882,
|
| 462 |
+
"eval_steps_per_second": 0.891,
|
| 463 |
+
"eval_wer": 0.9864203906795881,
|
| 464 |
+
"step": 672
|
| 465 |
+
},
|
| 466 |
+
{
|
| 467 |
+
"epoch": 6.07,
|
| 468 |
+
"learning_rate": 3.2920245398773006e-06,
|
| 469 |
+
"loss": 3.4964,
|
| 470 |
+
"step": 680
|
| 471 |
+
},
|
| 472 |
+
{
|
| 473 |
+
"epoch": 6.16,
|
| 474 |
+
"learning_rate": 3.2797546012269937e-06,
|
| 475 |
+
"loss": 3.4217,
|
| 476 |
+
"step": 690
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 6.25,
|
| 480 |
+
"learning_rate": 3.267484662576687e-06,
|
| 481 |
+
"loss": 3.385,
|
| 482 |
+
"step": 700
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 6.34,
|
| 486 |
+
"learning_rate": 3.25521472392638e-06,
|
| 487 |
+
"loss": 3.338,
|
| 488 |
+
"step": 710
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 6.43,
|
| 492 |
+
"learning_rate": 3.2429447852760735e-06,
|
| 493 |
+
"loss": 3.3658,
|
| 494 |
+
"step": 720
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 6.52,
|
| 498 |
+
"learning_rate": 3.2306748466257666e-06,
|
| 499 |
+
"loss": 3.3997,
|
| 500 |
+
"step": 730
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 6.61,
|
| 504 |
+
"learning_rate": 3.21840490797546e-06,
|
| 505 |
+
"loss": 3.4159,
|
| 506 |
+
"step": 740
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 6.7,
|
| 510 |
+
"learning_rate": 3.2061349693251533e-06,
|
| 511 |
+
"loss": 3.3652,
|
| 512 |
+
"step": 750
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"epoch": 6.78,
|
| 516 |
+
"learning_rate": 3.193865030674847e-06,
|
| 517 |
+
"loss": 3.3839,
|
| 518 |
+
"step": 760
|
| 519 |
+
},
|
| 520 |
+
{
|
| 521 |
+
"epoch": 6.87,
|
| 522 |
+
"learning_rate": 3.1815950920245395e-06,
|
| 523 |
+
"loss": 3.3874,
|
| 524 |
+
"step": 770
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 6.96,
|
| 528 |
+
"learning_rate": 3.169325153374233e-06,
|
| 529 |
+
"loss": 3.3493,
|
| 530 |
+
"step": 780
|
| 531 |
+
},
|
| 532 |
+
{
|
| 533 |
+
"epoch": 7.0,
|
| 534 |
+
"eval_loss": 3.3548903465270996,
|
| 535 |
+
"eval_runtime": 223.6828,
|
| 536 |
+
"eval_samples_per_second": 8.914,
|
| 537 |
+
"eval_steps_per_second": 0.894,
|
| 538 |
+
"eval_wer": 0.9847191420846371,
|
| 539 |
+
"step": 784
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 7.05,
|
| 543 |
+
"learning_rate": 3.1570552147239262e-06,
|
| 544 |
+
"loss": 3.2476,
|
| 545 |
+
"step": 790
|
| 546 |
+
},
|
| 547 |
+
{
|
| 548 |
+
"epoch": 7.14,
|
| 549 |
+
"learning_rate": 3.1447852760736193e-06,
|
| 550 |
+
"loss": 3.3659,
|
| 551 |
+
"step": 800
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"epoch": 7.23,
|
| 555 |
+
"learning_rate": 3.132515337423313e-06,
|
| 556 |
+
"loss": 3.2943,
|
| 557 |
+
"step": 810
|
| 558 |
+
},
|
| 559 |
+
{
|
| 560 |
+
"epoch": 7.32,
|
| 561 |
+
"learning_rate": 3.120245398773006e-06,
|
| 562 |
+
"loss": 3.3103,
|
| 563 |
+
"step": 820
|
| 564 |
+
},
|
| 565 |
+
{
|
| 566 |
+
"epoch": 7.41,
|
| 567 |
+
"learning_rate": 3.107975460122699e-06,
|
| 568 |
+
"loss": 3.2368,
|
| 569 |
+
"step": 830
|
| 570 |
+
},
|
| 571 |
+
{
|
| 572 |
+
"epoch": 7.5,
|
| 573 |
+
"learning_rate": 3.0957055214723923e-06,
|
| 574 |
+
"loss": 3.2216,
|
| 575 |
+
"step": 840
|
| 576 |
+
},
|
| 577 |
+
{
|
| 578 |
+
"epoch": 7.59,
|
| 579 |
+
"learning_rate": 3.083435582822086e-06,
|
| 580 |
+
"loss": 3.2502,
|
| 581 |
+
"step": 850
|
| 582 |
+
},
|
| 583 |
+
{
|
| 584 |
+
"epoch": 7.68,
|
| 585 |
+
"learning_rate": 3.071165644171779e-06,
|
| 586 |
+
"loss": 3.2379,
|
| 587 |
+
"step": 860
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"epoch": 7.77,
|
| 591 |
+
"learning_rate": 3.0588957055214725e-06,
|
| 592 |
+
"loss": 3.2768,
|
| 593 |
+
"step": 870
|
| 594 |
+
},
|
| 595 |
+
{
|
| 596 |
+
"epoch": 7.86,
|
| 597 |
+
"learning_rate": 3.0466257668711656e-06,
|
| 598 |
+
"loss": 3.2413,
|
| 599 |
+
"step": 880
|
| 600 |
+
},
|
| 601 |
+
{
|
| 602 |
+
"epoch": 7.95,
|
| 603 |
+
"learning_rate": 3.0343558282208587e-06,
|
| 604 |
+
"loss": 3.1264,
|
| 605 |
+
"step": 890
|
| 606 |
+
},
|
| 607 |
+
{
|
| 608 |
+
"epoch": 8.0,
|
| 609 |
+
"eval_loss": 3.1797306537628174,
|
| 610 |
+
"eval_runtime": 226.8173,
|
| 611 |
+
"eval_samples_per_second": 8.791,
|
| 612 |
+
"eval_steps_per_second": 0.882,
|
| 613 |
+
"eval_wer": 0.9758787252787313,
|
| 614 |
+
"step": 896
|
| 615 |
+
},
|
| 616 |
+
{
|
| 617 |
+
"epoch": 8.04,
|
| 618 |
+
"learning_rate": 3.022085889570552e-06,
|
| 619 |
+
"loss": 3.2241,
|
| 620 |
+
"step": 900
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 8.12,
|
| 624 |
+
"learning_rate": 3.0098159509202454e-06,
|
| 625 |
+
"loss": 3.0777,
|
| 626 |
+
"step": 910
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 8.21,
|
| 630 |
+
"learning_rate": 2.9975460122699385e-06,
|
| 631 |
+
"loss": 3.015,
|
| 632 |
+
"step": 920
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 8.3,
|
| 636 |
+
"learning_rate": 2.9852760736196317e-06,
|
| 637 |
+
"loss": 2.9733,
|
| 638 |
+
"step": 930
|
| 639 |
+
},
|
| 640 |
+
{
|
| 641 |
+
"epoch": 8.39,
|
| 642 |
+
"learning_rate": 2.973006134969325e-06,
|
| 643 |
+
"loss": 2.97,
|
| 644 |
+
"step": 940
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 8.48,
|
| 648 |
+
"learning_rate": 2.960736196319018e-06,
|
| 649 |
+
"loss": 2.8979,
|
| 650 |
+
"step": 950
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 8.57,
|
| 654 |
+
"learning_rate": 2.9484662576687115e-06,
|
| 655 |
+
"loss": 2.9348,
|
| 656 |
+
"step": 960
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 8.66,
|
| 660 |
+
"learning_rate": 2.9361963190184046e-06,
|
| 661 |
+
"loss": 2.847,
|
| 662 |
+
"step": 970
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 8.75,
|
| 666 |
+
"learning_rate": 2.923926380368098e-06,
|
| 667 |
+
"loss": 2.9345,
|
| 668 |
+
"step": 980
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 8.84,
|
| 672 |
+
"learning_rate": 2.9116564417177913e-06,
|
| 673 |
+
"loss": 2.9118,
|
| 674 |
+
"step": 990
|
| 675 |
+
},
|
| 676 |
+
{
|
| 677 |
+
"epoch": 8.93,
|
| 678 |
+
"learning_rate": 2.899386503067485e-06,
|
| 679 |
+
"loss": 2.7557,
|
| 680 |
+
"step": 1000
|
| 681 |
+
},
|
| 682 |
+
{
|
| 683 |
+
"epoch": 9.0,
|
| 684 |
+
"eval_loss": 2.8702940940856934,
|
| 685 |
+
"eval_runtime": 226.6648,
|
| 686 |
+
"eval_samples_per_second": 8.797,
|
| 687 |
+
"eval_steps_per_second": 0.882,
|
| 688 |
+
"eval_wer": 0.9865115289971748,
|
| 689 |
+
"step": 1008
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 9.02,
|
| 693 |
+
"learning_rate": 2.887116564417178e-06,
|
| 694 |
+
"loss": 2.8224,
|
| 695 |
+
"step": 1010
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"epoch": 9.11,
|
| 699 |
+
"learning_rate": 2.874846625766871e-06,
|
| 700 |
+
"loss": 2.756,
|
| 701 |
+
"step": 1020
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"epoch": 9.2,
|
| 705 |
+
"learning_rate": 2.862576687116564e-06,
|
| 706 |
+
"loss": 2.8302,
|
| 707 |
+
"step": 1030
|
| 708 |
+
},
|
| 709 |
+
{
|
| 710 |
+
"epoch": 9.29,
|
| 711 |
+
"learning_rate": 2.8503067484662577e-06,
|
| 712 |
+
"loss": 2.7484,
|
| 713 |
+
"step": 1040
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 9.37,
|
| 717 |
+
"learning_rate": 2.838036809815951e-06,
|
| 718 |
+
"loss": 2.7708,
|
| 719 |
+
"step": 1050
|
| 720 |
+
},
|
| 721 |
+
{
|
| 722 |
+
"epoch": 9.46,
|
| 723 |
+
"learning_rate": 2.825766871165644e-06,
|
| 724 |
+
"loss": 2.7069,
|
| 725 |
+
"step": 1060
|
| 726 |
+
},
|
| 727 |
+
{
|
| 728 |
+
"epoch": 9.55,
|
| 729 |
+
"learning_rate": 2.8134969325153375e-06,
|
| 730 |
+
"loss": 2.7055,
|
| 731 |
+
"step": 1070
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 9.64,
|
| 735 |
+
"learning_rate": 2.8012269938650302e-06,
|
| 736 |
+
"loss": 2.6739,
|
| 737 |
+
"step": 1080
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 9.73,
|
| 741 |
+
"learning_rate": 2.7889570552147238e-06,
|
| 742 |
+
"loss": 2.6802,
|
| 743 |
+
"step": 1090
|
| 744 |
+
},
|
| 745 |
+
{
|
| 746 |
+
"epoch": 9.82,
|
| 747 |
+
"learning_rate": 2.776687116564417e-06,
|
| 748 |
+
"loss": 2.6417,
|
| 749 |
+
"step": 1100
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 9.91,
|
| 753 |
+
"learning_rate": 2.7644171779141104e-06,
|
| 754 |
+
"loss": 2.6462,
|
| 755 |
+
"step": 1110
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 10.0,
|
| 759 |
+
"learning_rate": 2.7521472392638036e-06,
|
| 760 |
+
"loss": 2.6345,
|
| 761 |
+
"step": 1120
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 10.0,
|
| 765 |
+
"eval_loss": 2.6736390590667725,
|
| 766 |
+
"eval_runtime": 229.3255,
|
| 767 |
+
"eval_samples_per_second": 8.695,
|
| 768 |
+
"eval_steps_per_second": 0.872,
|
| 769 |
+
"eval_wer": 0.9969924355196403,
|
| 770 |
+
"step": 1120
|
| 771 |
+
},
|
| 772 |
+
{
|
| 773 |
+
"epoch": 10.09,
|
| 774 |
+
"learning_rate": 2.739877300613497e-06,
|
| 775 |
+
"loss": 2.5804,
|
| 776 |
+
"step": 1130
|
| 777 |
+
},
|
| 778 |
+
{
|
| 779 |
+
"epoch": 10.18,
|
| 780 |
+
"learning_rate": 2.72760736196319e-06,
|
| 781 |
+
"loss": 2.5819,
|
| 782 |
+
"step": 1140
|
| 783 |
+
},
|
| 784 |
+
{
|
| 785 |
+
"epoch": 10.27,
|
| 786 |
+
"learning_rate": 2.7153374233128834e-06,
|
| 787 |
+
"loss": 2.568,
|
| 788 |
+
"step": 1150
|
| 789 |
+
},
|
| 790 |
+
{
|
| 791 |
+
"epoch": 10.36,
|
| 792 |
+
"learning_rate": 2.7030674846625765e-06,
|
| 793 |
+
"loss": 2.5261,
|
| 794 |
+
"step": 1160
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 10.45,
|
| 798 |
+
"learning_rate": 2.6907975460122696e-06,
|
| 799 |
+
"loss": 2.5213,
|
| 800 |
+
"step": 1170
|
| 801 |
+
},
|
| 802 |
+
{
|
| 803 |
+
"epoch": 10.54,
|
| 804 |
+
"learning_rate": 2.678527607361963e-06,
|
| 805 |
+
"loss": 2.54,
|
| 806 |
+
"step": 1180
|
| 807 |
+
},
|
| 808 |
+
{
|
| 809 |
+
"epoch": 10.62,
|
| 810 |
+
"learning_rate": 2.6662576687116563e-06,
|
| 811 |
+
"loss": 2.465,
|
| 812 |
+
"step": 1190
|
| 813 |
+
},
|
| 814 |
+
{
|
| 815 |
+
"epoch": 10.71,
|
| 816 |
+
"learning_rate": 2.65398773006135e-06,
|
| 817 |
+
"loss": 2.4933,
|
| 818 |
+
"step": 1200
|
| 819 |
+
},
|
| 820 |
+
{
|
| 821 |
+
"epoch": 10.8,
|
| 822 |
+
"learning_rate": 2.6417177914110425e-06,
|
| 823 |
+
"loss": 2.4469,
|
| 824 |
+
"step": 1210
|
| 825 |
+
},
|
| 826 |
+
{
|
| 827 |
+
"epoch": 10.89,
|
| 828 |
+
"learning_rate": 2.629447852760736e-06,
|
| 829 |
+
"loss": 2.4227,
|
| 830 |
+
"step": 1220
|
| 831 |
+
},
|
| 832 |
+
{
|
| 833 |
+
"epoch": 10.98,
|
| 834 |
+
"learning_rate": 2.617177914110429e-06,
|
| 835 |
+
"loss": 2.4297,
|
| 836 |
+
"step": 1230
|
| 837 |
+
},
|
| 838 |
+
{
|
| 839 |
+
"epoch": 11.0,
|
| 840 |
+
"eval_loss": 2.563849925994873,
|
| 841 |
+
"eval_runtime": 228.8987,
|
| 842 |
+
"eval_samples_per_second": 8.711,
|
| 843 |
+
"eval_steps_per_second": 0.874,
|
| 844 |
+
"eval_wer": 1.0336907980678676,
|
| 845 |
+
"step": 1232
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 11.07,
|
| 849 |
+
"learning_rate": 2.6049079754601228e-06,
|
| 850 |
+
"loss": 2.4706,
|
| 851 |
+
"step": 1240
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 11.16,
|
| 855 |
+
"learning_rate": 2.592638036809816e-06,
|
| 856 |
+
"loss": 2.4096,
|
| 857 |
+
"step": 1250
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 11.25,
|
| 861 |
+
"learning_rate": 2.5803680981595094e-06,
|
| 862 |
+
"loss": 2.3846,
|
| 863 |
+
"step": 1260
|
| 864 |
+
},
|
| 865 |
+
{
|
| 866 |
+
"epoch": 11.34,
|
| 867 |
+
"learning_rate": 2.568098159509202e-06,
|
| 868 |
+
"loss": 2.3384,
|
| 869 |
+
"step": 1270
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 11.43,
|
| 873 |
+
"learning_rate": 2.5558282208588957e-06,
|
| 874 |
+
"loss": 2.3856,
|
| 875 |
+
"step": 1280
|
| 876 |
+
},
|
| 877 |
+
{
|
| 878 |
+
"epoch": 11.52,
|
| 879 |
+
"learning_rate": 2.543558282208589e-06,
|
| 880 |
+
"loss": 2.3349,
|
| 881 |
+
"step": 1290
|
| 882 |
+
},
|
| 883 |
+
{
|
| 884 |
+
"epoch": 11.61,
|
| 885 |
+
"learning_rate": 2.531288343558282e-06,
|
| 886 |
+
"loss": 2.2744,
|
| 887 |
+
"step": 1300
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 11.7,
|
| 891 |
+
"learning_rate": 2.5190184049079755e-06,
|
| 892 |
+
"loss": 2.3033,
|
| 893 |
+
"step": 1310
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"epoch": 11.78,
|
| 897 |
+
"learning_rate": 2.5067484662576686e-06,
|
| 898 |
+
"loss": 2.3258,
|
| 899 |
+
"step": 1320
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 11.87,
|
| 903 |
+
"learning_rate": 2.4944785276073617e-06,
|
| 904 |
+
"loss": 2.3106,
|
| 905 |
+
"step": 1330
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 11.96,
|
| 909 |
+
"learning_rate": 2.482208588957055e-06,
|
| 910 |
+
"loss": 2.3057,
|
| 911 |
+
"step": 1340
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 12.0,
|
| 915 |
+
"eval_loss": 2.3679769039154053,
|
| 916 |
+
"eval_runtime": 226.8372,
|
| 917 |
+
"eval_samples_per_second": 8.79,
|
| 918 |
+
"eval_steps_per_second": 0.882,
|
| 919 |
+
"eval_wer": 0.9838685177871617,
|
| 920 |
+
"step": 1344
|
| 921 |
+
},
|
| 922 |
+
{
|
| 923 |
+
"epoch": 12.05,
|
| 924 |
+
"learning_rate": 2.4699386503067484e-06,
|
| 925 |
+
"loss": 2.2966,
|
| 926 |
+
"step": 1350
|
| 927 |
+
},
|
| 928 |
+
{
|
| 929 |
+
"epoch": 12.14,
|
| 930 |
+
"learning_rate": 2.4576687116564415e-06,
|
| 931 |
+
"loss": 2.2726,
|
| 932 |
+
"step": 1360
|
| 933 |
+
},
|
| 934 |
+
{
|
| 935 |
+
"epoch": 12.23,
|
| 936 |
+
"learning_rate": 2.445398773006135e-06,
|
| 937 |
+
"loss": 2.2556,
|
| 938 |
+
"step": 1370
|
| 939 |
+
},
|
| 940 |
+
{
|
| 941 |
+
"epoch": 12.32,
|
| 942 |
+
"learning_rate": 2.433128834355828e-06,
|
| 943 |
+
"loss": 2.2264,
|
| 944 |
+
"step": 1380
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 12.41,
|
| 948 |
+
"learning_rate": 2.4208588957055213e-06,
|
| 949 |
+
"loss": 2.229,
|
| 950 |
+
"step": 1390
|
| 951 |
+
},
|
| 952 |
+
{
|
| 953 |
+
"epoch": 12.5,
|
| 954 |
+
"learning_rate": 2.4085889570552144e-06,
|
| 955 |
+
"loss": 2.202,
|
| 956 |
+
"step": 1400
|
| 957 |
+
},
|
| 958 |
+
{
|
| 959 |
+
"epoch": 12.59,
|
| 960 |
+
"learning_rate": 2.396319018404908e-06,
|
| 961 |
+
"loss": 2.2028,
|
| 962 |
+
"step": 1410
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 12.68,
|
| 966 |
+
"learning_rate": 2.384049079754601e-06,
|
| 967 |
+
"loss": 2.1755,
|
| 968 |
+
"step": 1420
|
| 969 |
+
},
|
| 970 |
+
{
|
| 971 |
+
"epoch": 12.77,
|
| 972 |
+
"learning_rate": 2.3717791411042942e-06,
|
| 973 |
+
"loss": 2.1927,
|
| 974 |
+
"step": 1430
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 12.86,
|
| 978 |
+
"learning_rate": 2.359509202453988e-06,
|
| 979 |
+
"loss": 2.1783,
|
| 980 |
+
"step": 1440
|
| 981 |
+
},
|
| 982 |
+
{
|
| 983 |
+
"epoch": 12.95,
|
| 984 |
+
"learning_rate": 2.347239263803681e-06,
|
| 985 |
+
"loss": 2.1436,
|
| 986 |
+
"step": 1450
|
| 987 |
+
},
|
| 988 |
+
{
|
| 989 |
+
"epoch": 13.0,
|
| 990 |
+
"eval_loss": 2.2366645336151123,
|
| 991 |
+
"eval_runtime": 227.8374,
|
| 992 |
+
"eval_samples_per_second": 8.752,
|
| 993 |
+
"eval_steps_per_second": 0.878,
|
| 994 |
+
"eval_wer": 0.9647902299723548,
|
| 995 |
+
"step": 1456
|
| 996 |
+
},
|
| 997 |
+
{
|
| 998 |
+
"epoch": 13.04,
|
| 999 |
+
"learning_rate": 2.334969325153374e-06,
|
| 1000 |
+
"loss": 2.1661,
|
| 1001 |
+
"step": 1460
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 13.12,
|
| 1005 |
+
"learning_rate": 2.322699386503067e-06,
|
| 1006 |
+
"loss": 2.1329,
|
| 1007 |
+
"step": 1470
|
| 1008 |
+
},
|
| 1009 |
+
{
|
| 1010 |
+
"epoch": 13.21,
|
| 1011 |
+
"learning_rate": 2.3104294478527607e-06,
|
| 1012 |
+
"loss": 2.1202,
|
| 1013 |
+
"step": 1480
|
| 1014 |
+
},
|
| 1015 |
+
{
|
| 1016 |
+
"epoch": 13.3,
|
| 1017 |
+
"learning_rate": 2.298159509202454e-06,
|
| 1018 |
+
"loss": 2.1499,
|
| 1019 |
+
"step": 1490
|
| 1020 |
+
},
|
| 1021 |
+
{
|
| 1022 |
+
"epoch": 13.39,
|
| 1023 |
+
"learning_rate": 2.2858895705521474e-06,
|
| 1024 |
+
"loss": 2.1047,
|
| 1025 |
+
"step": 1500
|
| 1026 |
+
},
|
| 1027 |
+
{
|
| 1028 |
+
"epoch": 13.48,
|
| 1029 |
+
"learning_rate": 2.2736196319018405e-06,
|
| 1030 |
+
"loss": 2.1364,
|
| 1031 |
+
"step": 1510
|
| 1032 |
+
},
|
| 1033 |
+
{
|
| 1034 |
+
"epoch": 13.57,
|
| 1035 |
+
"learning_rate": 2.2613496932515336e-06,
|
| 1036 |
+
"loss": 2.1224,
|
| 1037 |
+
"step": 1520
|
| 1038 |
+
},
|
| 1039 |
+
{
|
| 1040 |
+
"epoch": 13.66,
|
| 1041 |
+
"learning_rate": 2.2490797546012268e-06,
|
| 1042 |
+
"loss": 2.0951,
|
| 1043 |
+
"step": 1530
|
| 1044 |
+
},
|
| 1045 |
+
{
|
| 1046 |
+
"epoch": 13.75,
|
| 1047 |
+
"learning_rate": 2.2368098159509203e-06,
|
| 1048 |
+
"loss": 2.11,
|
| 1049 |
+
"step": 1540
|
| 1050 |
+
},
|
| 1051 |
+
{
|
| 1052 |
+
"epoch": 13.84,
|
| 1053 |
+
"learning_rate": 2.2245398773006134e-06,
|
| 1054 |
+
"loss": 2.1022,
|
| 1055 |
+
"step": 1550
|
| 1056 |
+
},
|
| 1057 |
+
{
|
| 1058 |
+
"epoch": 13.93,
|
| 1059 |
+
"learning_rate": 2.2122699386503066e-06,
|
| 1060 |
+
"loss": 2.0856,
|
| 1061 |
+
"step": 1560
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 14.0,
|
| 1065 |
+
"eval_loss": 2.163454294204712,
|
| 1066 |
+
"eval_runtime": 228.8577,
|
| 1067 |
+
"eval_samples_per_second": 8.713,
|
| 1068 |
+
"eval_steps_per_second": 0.874,
|
| 1069 |
+
"eval_wer": 0.9585624449372665,
|
| 1070 |
+
"step": 1568
|
| 1071 |
+
},
|
| 1072 |
+
{
|
| 1073 |
+
"epoch": 14.02,
|
| 1074 |
+
"learning_rate": 2.2e-06,
|
| 1075 |
+
"loss": 2.1466,
|
| 1076 |
+
"step": 1570
|
| 1077 |
+
},
|
| 1078 |
+
{
|
| 1079 |
+
"epoch": 14.11,
|
| 1080 |
+
"learning_rate": 2.187730061349693e-06,
|
| 1081 |
+
"loss": 2.0836,
|
| 1082 |
+
"step": 1580
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"epoch": 14.2,
|
| 1086 |
+
"learning_rate": 2.1754601226993864e-06,
|
| 1087 |
+
"loss": 2.0697,
|
| 1088 |
+
"step": 1590
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"epoch": 14.29,
|
| 1092 |
+
"learning_rate": 2.1631901840490795e-06,
|
| 1093 |
+
"loss": 2.064,
|
| 1094 |
+
"step": 1600
|
| 1095 |
+
},
|
| 1096 |
+
{
|
| 1097 |
+
"epoch": 14.37,
|
| 1098 |
+
"learning_rate": 2.150920245398773e-06,
|
| 1099 |
+
"loss": 2.0571,
|
| 1100 |
+
"step": 1610
|
| 1101 |
+
},
|
| 1102 |
+
{
|
| 1103 |
+
"epoch": 14.46,
|
| 1104 |
+
"learning_rate": 2.138650306748466e-06,
|
| 1105 |
+
"loss": 2.0054,
|
| 1106 |
+
"step": 1620
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 14.55,
|
| 1110 |
+
"learning_rate": 2.1263803680981597e-06,
|
| 1111 |
+
"loss": 2.0765,
|
| 1112 |
+
"step": 1630
|
| 1113 |
+
},
|
| 1114 |
+
{
|
| 1115 |
+
"epoch": 14.64,
|
| 1116 |
+
"learning_rate": 2.1141104294478524e-06,
|
| 1117 |
+
"loss": 2.0276,
|
| 1118 |
+
"step": 1640
|
| 1119 |
+
},
|
| 1120 |
+
{
|
| 1121 |
+
"epoch": 14.73,
|
| 1122 |
+
"learning_rate": 2.101840490797546e-06,
|
| 1123 |
+
"loss": 2.0324,
|
| 1124 |
+
"step": 1650
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 14.82,
|
| 1128 |
+
"learning_rate": 2.089570552147239e-06,
|
| 1129 |
+
"loss": 2.0354,
|
| 1130 |
+
"step": 1660
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"epoch": 14.91,
|
| 1134 |
+
"learning_rate": 2.0773006134969326e-06,
|
| 1135 |
+
"loss": 2.0212,
|
| 1136 |
+
"step": 1670
|
| 1137 |
+
},
|
| 1138 |
+
{
|
| 1139 |
+
"epoch": 15.0,
|
| 1140 |
+
"learning_rate": 2.0650306748466258e-06,
|
| 1141 |
+
"loss": 2.0035,
|
| 1142 |
+
"step": 1680
|
| 1143 |
+
},
|
| 1144 |
+
{
|
| 1145 |
+
"epoch": 15.0,
|
| 1146 |
+
"eval_loss": 2.0944879055023193,
|
| 1147 |
+
"eval_runtime": 229.1871,
|
| 1148 |
+
"eval_samples_per_second": 8.7,
|
| 1149 |
+
"eval_steps_per_second": 0.873,
|
| 1150 |
+
"eval_wer": 0.9645471944587903,
|
| 1151 |
+
"step": 1680
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 15.09,
|
| 1155 |
+
"learning_rate": 2.052760736196319e-06,
|
| 1156 |
+
"loss": 2.0578,
|
| 1157 |
+
"step": 1690
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 15.18,
|
| 1161 |
+
"learning_rate": 2.0404907975460124e-06,
|
| 1162 |
+
"loss": 1.9819,
|
| 1163 |
+
"step": 1700
|
| 1164 |
+
},
|
| 1165 |
+
{
|
| 1166 |
+
"epoch": 15.27,
|
| 1167 |
+
"learning_rate": 2.028220858895705e-06,
|
| 1168 |
+
"loss": 2.0174,
|
| 1169 |
+
"step": 1710
|
| 1170 |
+
},
|
| 1171 |
+
{
|
| 1172 |
+
"epoch": 15.36,
|
| 1173 |
+
"learning_rate": 2.0159509202453987e-06,
|
| 1174 |
+
"loss": 2.0332,
|
| 1175 |
+
"step": 1720
|
| 1176 |
+
},
|
| 1177 |
+
{
|
| 1178 |
+
"epoch": 15.45,
|
| 1179 |
+
"learning_rate": 2.003680981595092e-06,
|
| 1180 |
+
"loss": 1.923,
|
| 1181 |
+
"step": 1730
|
| 1182 |
+
},
|
| 1183 |
+
{
|
| 1184 |
+
"epoch": 15.54,
|
| 1185 |
+
"learning_rate": 1.9914110429447854e-06,
|
| 1186 |
+
"loss": 1.9894,
|
| 1187 |
+
"step": 1740
|
| 1188 |
+
},
|
| 1189 |
+
{
|
| 1190 |
+
"epoch": 15.62,
|
| 1191 |
+
"learning_rate": 1.9791411042944785e-06,
|
| 1192 |
+
"loss": 1.9868,
|
| 1193 |
+
"step": 1750
|
| 1194 |
+
},
|
| 1195 |
+
{
|
| 1196 |
+
"epoch": 15.71,
|
| 1197 |
+
"learning_rate": 1.9668711656441716e-06,
|
| 1198 |
+
"loss": 1.9463,
|
| 1199 |
+
"step": 1760
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 15.8,
|
| 1203 |
+
"learning_rate": 1.954601226993865e-06,
|
| 1204 |
+
"loss": 1.9641,
|
| 1205 |
+
"step": 1770
|
| 1206 |
+
},
|
| 1207 |
+
{
|
| 1208 |
+
"epoch": 15.89,
|
| 1209 |
+
"learning_rate": 1.9423312883435583e-06,
|
| 1210 |
+
"loss": 1.9391,
|
| 1211 |
+
"step": 1780
|
| 1212 |
+
},
|
| 1213 |
+
{
|
| 1214 |
+
"epoch": 15.98,
|
| 1215 |
+
"learning_rate": 1.9300613496932514e-06,
|
| 1216 |
+
"loss": 1.9134,
|
| 1217 |
+
"step": 1790
|
| 1218 |
+
},
|
| 1219 |
+
{
|
| 1220 |
+
"epoch": 16.0,
|
| 1221 |
+
"eval_loss": 2.039499044418335,
|
| 1222 |
+
"eval_runtime": 230.4499,
|
| 1223 |
+
"eval_samples_per_second": 8.653,
|
| 1224 |
+
"eval_steps_per_second": 0.868,
|
| 1225 |
+
"eval_wer": 0.9630282224990127,
|
| 1226 |
+
"step": 1792
|
| 1227 |
+
},
|
| 1228 |
+
{
|
| 1229 |
+
"epoch": 16.07,
|
| 1230 |
+
"learning_rate": 1.9177914110429445e-06,
|
| 1231 |
+
"loss": 2.0007,
|
| 1232 |
+
"step": 1800
|
| 1233 |
+
},
|
| 1234 |
+
{
|
| 1235 |
+
"epoch": 16.16,
|
| 1236 |
+
"learning_rate": 1.905521472392638e-06,
|
| 1237 |
+
"loss": 1.9244,
|
| 1238 |
+
"step": 1810
|
| 1239 |
+
},
|
| 1240 |
+
{
|
| 1241 |
+
"epoch": 16.25,
|
| 1242 |
+
"learning_rate": 1.8932515337423312e-06,
|
| 1243 |
+
"loss": 1.966,
|
| 1244 |
+
"step": 1820
|
| 1245 |
+
},
|
| 1246 |
+
{
|
| 1247 |
+
"epoch": 16.34,
|
| 1248 |
+
"learning_rate": 1.8809815950920245e-06,
|
| 1249 |
+
"loss": 1.9547,
|
| 1250 |
+
"step": 1830
|
| 1251 |
+
},
|
| 1252 |
+
{
|
| 1253 |
+
"epoch": 16.43,
|
| 1254 |
+
"learning_rate": 1.8687116564417179e-06,
|
| 1255 |
+
"loss": 1.918,
|
| 1256 |
+
"step": 1840
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 16.52,
|
| 1260 |
+
"learning_rate": 1.856441717791411e-06,
|
| 1261 |
+
"loss": 1.9317,
|
| 1262 |
+
"step": 1850
|
| 1263 |
+
},
|
| 1264 |
+
{
|
| 1265 |
+
"epoch": 16.61,
|
| 1266 |
+
"learning_rate": 1.8441717791411041e-06,
|
| 1267 |
+
"loss": 1.9608,
|
| 1268 |
+
"step": 1860
|
| 1269 |
+
},
|
| 1270 |
+
{
|
| 1271 |
+
"epoch": 16.7,
|
| 1272 |
+
"learning_rate": 1.8319018404907975e-06,
|
| 1273 |
+
"loss": 1.9207,
|
| 1274 |
+
"step": 1870
|
| 1275 |
+
},
|
| 1276 |
+
{
|
| 1277 |
+
"epoch": 16.78,
|
| 1278 |
+
"learning_rate": 1.8196319018404906e-06,
|
| 1279 |
+
"loss": 1.9173,
|
| 1280 |
+
"step": 1880
|
| 1281 |
+
},
|
| 1282 |
+
{
|
| 1283 |
+
"epoch": 16.87,
|
| 1284 |
+
"learning_rate": 1.807361963190184e-06,
|
| 1285 |
+
"loss": 1.8698,
|
| 1286 |
+
"step": 1890
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"epoch": 16.96,
|
| 1290 |
+
"learning_rate": 1.7950920245398773e-06,
|
| 1291 |
+
"loss": 1.9443,
|
| 1292 |
+
"step": 1900
|
| 1293 |
+
},
|
| 1294 |
+
{
|
| 1295 |
+
"epoch": 17.0,
|
| 1296 |
+
"eval_loss": 2.001734495162964,
|
| 1297 |
+
"eval_runtime": 229.2485,
|
| 1298 |
+
"eval_samples_per_second": 8.698,
|
| 1299 |
+
"eval_steps_per_second": 0.872,
|
| 1300 |
+
"eval_wer": 0.940061366467175,
|
| 1301 |
+
"step": 1904
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 17.05,
|
| 1305 |
+
"learning_rate": 1.7828220858895704e-06,
|
| 1306 |
+
"loss": 1.9318,
|
| 1307 |
+
"step": 1910
|
| 1308 |
+
},
|
| 1309 |
+
{
|
| 1310 |
+
"epoch": 17.14,
|
| 1311 |
+
"learning_rate": 1.7705521472392637e-06,
|
| 1312 |
+
"loss": 1.8787,
|
| 1313 |
+
"step": 1920
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 17.23,
|
| 1317 |
+
"learning_rate": 1.758282208588957e-06,
|
| 1318 |
+
"loss": 1.9085,
|
| 1319 |
+
"step": 1930
|
| 1320 |
+
},
|
| 1321 |
+
{
|
| 1322 |
+
"epoch": 17.32,
|
| 1323 |
+
"learning_rate": 1.7460122699386502e-06,
|
| 1324 |
+
"loss": 1.9192,
|
| 1325 |
+
"step": 1940
|
| 1326 |
+
},
|
| 1327 |
+
{
|
| 1328 |
+
"epoch": 17.41,
|
| 1329 |
+
"learning_rate": 1.7337423312883435e-06,
|
| 1330 |
+
"loss": 1.876,
|
| 1331 |
+
"step": 1950
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"epoch": 17.5,
|
| 1335 |
+
"learning_rate": 1.7214723926380368e-06,
|
| 1336 |
+
"loss": 1.9029,
|
| 1337 |
+
"step": 1960
|
| 1338 |
+
},
|
| 1339 |
+
{
|
| 1340 |
+
"epoch": 17.59,
|
| 1341 |
+
"learning_rate": 1.70920245398773e-06,
|
| 1342 |
+
"loss": 1.8502,
|
| 1343 |
+
"step": 1970
|
| 1344 |
+
},
|
| 1345 |
+
{
|
| 1346 |
+
"epoch": 17.68,
|
| 1347 |
+
"learning_rate": 1.6969325153374233e-06,
|
| 1348 |
+
"loss": 1.8889,
|
| 1349 |
+
"step": 1980
|
| 1350 |
+
},
|
| 1351 |
+
{
|
| 1352 |
+
"epoch": 17.77,
|
| 1353 |
+
"learning_rate": 1.6846625766871164e-06,
|
| 1354 |
+
"loss": 1.8713,
|
| 1355 |
+
"step": 1990
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 17.86,
|
| 1359 |
+
"learning_rate": 1.6723926380368096e-06,
|
| 1360 |
+
"loss": 1.8772,
|
| 1361 |
+
"step": 2000
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"epoch": 17.95,
|
| 1365 |
+
"learning_rate": 1.660122699386503e-06,
|
| 1366 |
+
"loss": 1.8988,
|
| 1367 |
+
"step": 2010
|
| 1368 |
+
},
|
| 1369 |
+
{
|
| 1370 |
+
"epoch": 18.0,
|
| 1371 |
+
"eval_loss": 1.9513905048370361,
|
| 1372 |
+
"eval_runtime": 229.6946,
|
| 1373 |
+
"eval_samples_per_second": 8.681,
|
| 1374 |
+
"eval_steps_per_second": 0.871,
|
| 1375 |
+
"eval_wer": 0.9492663365434274,
|
| 1376 |
+
"step": 2016
|
| 1377 |
+
},
|
| 1378 |
+
{
|
| 1379 |
+
"epoch": 18.04,
|
| 1380 |
+
"learning_rate": 1.6478527607361962e-06,
|
| 1381 |
+
"loss": 1.8788,
|
| 1382 |
+
"step": 2020
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"epoch": 18.12,
|
| 1386 |
+
"learning_rate": 1.6355828220858896e-06,
|
| 1387 |
+
"loss": 1.8656,
|
| 1388 |
+
"step": 2030
|
| 1389 |
+
},
|
| 1390 |
+
{
|
| 1391 |
+
"epoch": 18.21,
|
| 1392 |
+
"learning_rate": 1.6233128834355827e-06,
|
| 1393 |
+
"loss": 1.8622,
|
| 1394 |
+
"step": 2040
|
| 1395 |
+
},
|
| 1396 |
+
{
|
| 1397 |
+
"epoch": 18.3,
|
| 1398 |
+
"learning_rate": 1.611042944785276e-06,
|
| 1399 |
+
"loss": 1.8891,
|
| 1400 |
+
"step": 2050
|
| 1401 |
+
},
|
| 1402 |
+
{
|
| 1403 |
+
"epoch": 18.39,
|
| 1404 |
+
"learning_rate": 1.5987730061349694e-06,
|
| 1405 |
+
"loss": 1.8184,
|
| 1406 |
+
"step": 2060
|
| 1407 |
+
},
|
| 1408 |
+
{
|
| 1409 |
+
"epoch": 18.48,
|
| 1410 |
+
"learning_rate": 1.5865030674846625e-06,
|
| 1411 |
+
"loss": 1.8475,
|
| 1412 |
+
"step": 2070
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 18.57,
|
| 1416 |
+
"learning_rate": 1.5742331288343558e-06,
|
| 1417 |
+
"loss": 1.8573,
|
| 1418 |
+
"step": 2080
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 18.66,
|
| 1422 |
+
"learning_rate": 1.5619631901840492e-06,
|
| 1423 |
+
"loss": 1.819,
|
| 1424 |
+
"step": 2090
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 18.75,
|
| 1428 |
+
"learning_rate": 1.5496932515337423e-06,
|
| 1429 |
+
"loss": 1.8366,
|
| 1430 |
+
"step": 2100
|
| 1431 |
+
},
|
| 1432 |
+
{
|
| 1433 |
+
"epoch": 18.84,
|
| 1434 |
+
"learning_rate": 1.5374233128834354e-06,
|
| 1435 |
+
"loss": 1.8187,
|
| 1436 |
+
"step": 2110
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 18.93,
|
| 1440 |
+
"learning_rate": 1.5251533742331288e-06,
|
| 1441 |
+
"loss": 1.8141,
|
| 1442 |
+
"step": 2120
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 19.0,
|
| 1446 |
+
"eval_loss": 1.9111149311065674,
|
| 1447 |
+
"eval_runtime": 229.8356,
|
| 1448 |
+
"eval_samples_per_second": 8.676,
|
| 1449 |
+
"eval_steps_per_second": 0.87,
|
| 1450 |
+
"eval_wer": 0.9475347085092809,
|
| 1451 |
+
"step": 2128
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"epoch": 19.02,
|
| 1455 |
+
"learning_rate": 1.5128834355828219e-06,
|
| 1456 |
+
"loss": 1.8492,
|
| 1457 |
+
"step": 2130
|
| 1458 |
+
},
|
| 1459 |
+
{
|
| 1460 |
+
"epoch": 19.11,
|
| 1461 |
+
"learning_rate": 1.5006134969325152e-06,
|
| 1462 |
+
"loss": 1.8041,
|
| 1463 |
+
"step": 2140
|
| 1464 |
+
},
|
| 1465 |
+
{
|
| 1466 |
+
"epoch": 19.2,
|
| 1467 |
+
"learning_rate": 1.4883435582822085e-06,
|
| 1468 |
+
"loss": 1.8007,
|
| 1469 |
+
"step": 2150
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 19.29,
|
| 1473 |
+
"learning_rate": 1.4760736196319017e-06,
|
| 1474 |
+
"loss": 1.8104,
|
| 1475 |
+
"step": 2160
|
| 1476 |
+
},
|
| 1477 |
+
{
|
| 1478 |
+
"epoch": 19.37,
|
| 1479 |
+
"learning_rate": 1.463803680981595e-06,
|
| 1480 |
+
"loss": 1.8089,
|
| 1481 |
+
"step": 2170
|
| 1482 |
+
},
|
| 1483 |
+
{
|
| 1484 |
+
"epoch": 19.46,
|
| 1485 |
+
"learning_rate": 1.4515337423312883e-06,
|
| 1486 |
+
"loss": 1.8453,
|
| 1487 |
+
"step": 2180
|
| 1488 |
+
},
|
| 1489 |
+
{
|
| 1490 |
+
"epoch": 19.55,
|
| 1491 |
+
"learning_rate": 1.4392638036809815e-06,
|
| 1492 |
+
"loss": 1.8138,
|
| 1493 |
+
"step": 2190
|
| 1494 |
+
},
|
| 1495 |
+
{
|
| 1496 |
+
"epoch": 19.64,
|
| 1497 |
+
"learning_rate": 1.4269938650306748e-06,
|
| 1498 |
+
"loss": 1.7673,
|
| 1499 |
+
"step": 2200
|
| 1500 |
+
},
|
| 1501 |
+
{
|
| 1502 |
+
"epoch": 19.73,
|
| 1503 |
+
"learning_rate": 1.4147239263803681e-06,
|
| 1504 |
+
"loss": 1.8214,
|
| 1505 |
+
"step": 2210
|
| 1506 |
+
},
|
| 1507 |
+
{
|
| 1508 |
+
"epoch": 19.82,
|
| 1509 |
+
"learning_rate": 1.4024539877300613e-06,
|
| 1510 |
+
"loss": 1.8133,
|
| 1511 |
+
"step": 2220
|
| 1512 |
+
},
|
| 1513 |
+
{
|
| 1514 |
+
"epoch": 19.91,
|
| 1515 |
+
"learning_rate": 1.3901840490797546e-06,
|
| 1516 |
+
"loss": 1.766,
|
| 1517 |
+
"step": 2230
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"epoch": 20.0,
|
| 1521 |
+
"learning_rate": 1.3779141104294477e-06,
|
| 1522 |
+
"loss": 1.8344,
|
| 1523 |
+
"step": 2240
|
| 1524 |
+
},
|
| 1525 |
+
{
|
| 1526 |
+
"epoch": 20.0,
|
| 1527 |
+
"eval_loss": 1.8789727687835693,
|
| 1528 |
+
"eval_runtime": 230.1819,
|
| 1529 |
+
"eval_samples_per_second": 8.663,
|
| 1530 |
+
"eval_steps_per_second": 0.869,
|
| 1531 |
+
"eval_wer": 0.9395449160008507,
|
| 1532 |
+
"step": 2240
|
| 1533 |
+
},
|
| 1534 |
+
{
|
| 1535 |
+
"epoch": 20.09,
|
| 1536 |
+
"learning_rate": 1.365644171779141e-06,
|
| 1537 |
+
"loss": 1.7892,
|
| 1538 |
+
"step": 2250
|
| 1539 |
+
},
|
| 1540 |
+
{
|
| 1541 |
+
"epoch": 20.18,
|
| 1542 |
+
"learning_rate": 1.3533742331288342e-06,
|
| 1543 |
+
"loss": 1.7738,
|
| 1544 |
+
"step": 2260
|
| 1545 |
+
},
|
| 1546 |
+
{
|
| 1547 |
+
"epoch": 20.27,
|
| 1548 |
+
"learning_rate": 1.3411042944785275e-06,
|
| 1549 |
+
"loss": 1.8034,
|
| 1550 |
+
"step": 2270
|
| 1551 |
+
},
|
| 1552 |
+
{
|
| 1553 |
+
"epoch": 20.36,
|
| 1554 |
+
"learning_rate": 1.3288343558282209e-06,
|
| 1555 |
+
"loss": 1.7568,
|
| 1556 |
+
"step": 2280
|
| 1557 |
+
},
|
| 1558 |
+
{
|
| 1559 |
+
"epoch": 20.45,
|
| 1560 |
+
"learning_rate": 1.316564417177914e-06,
|
| 1561 |
+
"loss": 1.7828,
|
| 1562 |
+
"step": 2290
|
| 1563 |
+
},
|
| 1564 |
+
{
|
| 1565 |
+
"epoch": 20.54,
|
| 1566 |
+
"learning_rate": 1.3042944785276073e-06,
|
| 1567 |
+
"loss": 1.7748,
|
| 1568 |
+
"step": 2300
|
| 1569 |
+
},
|
| 1570 |
+
{
|
| 1571 |
+
"epoch": 20.62,
|
| 1572 |
+
"learning_rate": 1.2920245398773007e-06,
|
| 1573 |
+
"loss": 1.7529,
|
| 1574 |
+
"step": 2310
|
| 1575 |
+
},
|
| 1576 |
+
{
|
| 1577 |
+
"epoch": 20.71,
|
| 1578 |
+
"learning_rate": 1.2797546012269938e-06,
|
| 1579 |
+
"loss": 1.7858,
|
| 1580 |
+
"step": 2320
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 20.8,
|
| 1584 |
+
"learning_rate": 1.2674846625766871e-06,
|
| 1585 |
+
"loss": 1.7939,
|
| 1586 |
+
"step": 2330
|
| 1587 |
+
},
|
| 1588 |
+
{
|
| 1589 |
+
"epoch": 20.89,
|
| 1590 |
+
"learning_rate": 1.2552147239263805e-06,
|
| 1591 |
+
"loss": 1.7734,
|
| 1592 |
+
"step": 2340
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 20.98,
|
| 1596 |
+
"learning_rate": 1.2429447852760736e-06,
|
| 1597 |
+
"loss": 1.7775,
|
| 1598 |
+
"step": 2350
|
| 1599 |
+
},
|
| 1600 |
+
{
|
| 1601 |
+
"epoch": 21.0,
|
| 1602 |
+
"eval_loss": 1.8616416454315186,
|
| 1603 |
+
"eval_runtime": 230.4184,
|
| 1604 |
+
"eval_samples_per_second": 8.654,
|
| 1605 |
+
"eval_steps_per_second": 0.868,
|
| 1606 |
+
"eval_wer": 0.9502992374760761,
|
| 1607 |
+
"step": 2352
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"epoch": 21.07,
|
| 1611 |
+
"learning_rate": 1.2306748466257667e-06,
|
| 1612 |
+
"loss": 1.7807,
|
| 1613 |
+
"step": 2360
|
| 1614 |
+
},
|
| 1615 |
+
{
|
| 1616 |
+
"epoch": 21.16,
|
| 1617 |
+
"learning_rate": 1.21840490797546e-06,
|
| 1618 |
+
"loss": 1.7896,
|
| 1619 |
+
"step": 2370
|
| 1620 |
+
},
|
| 1621 |
+
{
|
| 1622 |
+
"epoch": 21.25,
|
| 1623 |
+
"learning_rate": 1.2061349693251532e-06,
|
| 1624 |
+
"loss": 1.7772,
|
| 1625 |
+
"step": 2380
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 21.34,
|
| 1629 |
+
"learning_rate": 1.1938650306748465e-06,
|
| 1630 |
+
"loss": 1.7337,
|
| 1631 |
+
"step": 2390
|
| 1632 |
+
},
|
| 1633 |
+
{
|
| 1634 |
+
"epoch": 21.43,
|
| 1635 |
+
"learning_rate": 1.1815950920245398e-06,
|
| 1636 |
+
"loss": 1.7481,
|
| 1637 |
+
"step": 2400
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 21.52,
|
| 1641 |
+
"learning_rate": 1.169325153374233e-06,
|
| 1642 |
+
"loss": 1.7551,
|
| 1643 |
+
"step": 2410
|
| 1644 |
+
},
|
| 1645 |
+
{
|
| 1646 |
+
"epoch": 21.61,
|
| 1647 |
+
"learning_rate": 1.1570552147239263e-06,
|
| 1648 |
+
"loss": 1.7527,
|
| 1649 |
+
"step": 2420
|
| 1650 |
+
},
|
| 1651 |
+
{
|
| 1652 |
+
"epoch": 21.7,
|
| 1653 |
+
"learning_rate": 1.1447852760736196e-06,
|
| 1654 |
+
"loss": 1.7565,
|
| 1655 |
+
"step": 2430
|
| 1656 |
+
},
|
| 1657 |
+
{
|
| 1658 |
+
"epoch": 21.78,
|
| 1659 |
+
"learning_rate": 1.1325153374233128e-06,
|
| 1660 |
+
"loss": 1.7389,
|
| 1661 |
+
"step": 2440
|
| 1662 |
+
},
|
| 1663 |
+
{
|
| 1664 |
+
"epoch": 21.87,
|
| 1665 |
+
"learning_rate": 1.120245398773006e-06,
|
| 1666 |
+
"loss": 1.73,
|
| 1667 |
+
"step": 2450
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 21.96,
|
| 1671 |
+
"learning_rate": 1.1079754601226994e-06,
|
| 1672 |
+
"loss": 1.7517,
|
| 1673 |
+
"step": 2460
|
| 1674 |
+
},
|
| 1675 |
+
{
|
| 1676 |
+
"epoch": 22.0,
|
| 1677 |
+
"eval_loss": 1.8332698345184326,
|
| 1678 |
+
"eval_runtime": 231.0169,
|
| 1679 |
+
"eval_samples_per_second": 8.631,
|
| 1680 |
+
"eval_steps_per_second": 0.866,
|
| 1681 |
+
"eval_wer": 0.9433423459002946,
|
| 1682 |
+
"step": 2464
|
| 1683 |
+
},
|
| 1684 |
+
{
|
| 1685 |
+
"epoch": 22.05,
|
| 1686 |
+
"learning_rate": 1.0957055214723926e-06,
|
| 1687 |
+
"loss": 1.798,
|
| 1688 |
+
"step": 2470
|
| 1689 |
+
},
|
| 1690 |
+
{
|
| 1691 |
+
"epoch": 22.14,
|
| 1692 |
+
"learning_rate": 1.083435582822086e-06,
|
| 1693 |
+
"loss": 1.7413,
|
| 1694 |
+
"step": 2480
|
| 1695 |
+
},
|
| 1696 |
+
{
|
| 1697 |
+
"epoch": 22.23,
|
| 1698 |
+
"learning_rate": 1.071165644171779e-06,
|
| 1699 |
+
"loss": 1.7061,
|
| 1700 |
+
"step": 2490
|
| 1701 |
+
},
|
| 1702 |
+
{
|
| 1703 |
+
"epoch": 22.32,
|
| 1704 |
+
"learning_rate": 1.0588957055214724e-06,
|
| 1705 |
+
"loss": 1.7485,
|
| 1706 |
+
"step": 2500
|
| 1707 |
+
},
|
| 1708 |
+
{
|
| 1709 |
+
"epoch": 22.41,
|
| 1710 |
+
"learning_rate": 1.0466257668711655e-06,
|
| 1711 |
+
"loss": 1.722,
|
| 1712 |
+
"step": 2510
|
| 1713 |
+
},
|
| 1714 |
+
{
|
| 1715 |
+
"epoch": 22.5,
|
| 1716 |
+
"learning_rate": 1.0343558282208588e-06,
|
| 1717 |
+
"loss": 1.7105,
|
| 1718 |
+
"step": 2520
|
| 1719 |
+
},
|
| 1720 |
+
{
|
| 1721 |
+
"epoch": 22.59,
|
| 1722 |
+
"learning_rate": 1.0220858895705522e-06,
|
| 1723 |
+
"loss": 1.7343,
|
| 1724 |
+
"step": 2530
|
| 1725 |
+
},
|
| 1726 |
+
{
|
| 1727 |
+
"epoch": 22.68,
|
| 1728 |
+
"learning_rate": 1.0098159509202453e-06,
|
| 1729 |
+
"loss": 1.7403,
|
| 1730 |
+
"step": 2540
|
| 1731 |
+
},
|
| 1732 |
+
{
|
| 1733 |
+
"epoch": 22.77,
|
| 1734 |
+
"learning_rate": 9.975460122699386e-07,
|
| 1735 |
+
"loss": 1.7422,
|
| 1736 |
+
"step": 2550
|
| 1737 |
+
},
|
| 1738 |
+
{
|
| 1739 |
+
"epoch": 22.86,
|
| 1740 |
+
"learning_rate": 9.852760736196317e-07,
|
| 1741 |
+
"loss": 1.7464,
|
| 1742 |
+
"step": 2560
|
| 1743 |
+
},
|
| 1744 |
+
{
|
| 1745 |
+
"epoch": 22.95,
|
| 1746 |
+
"learning_rate": 9.73006134969325e-07,
|
| 1747 |
+
"loss": 1.7037,
|
| 1748 |
+
"step": 2570
|
| 1749 |
+
},
|
| 1750 |
+
{
|
| 1751 |
+
"epoch": 23.0,
|
| 1752 |
+
"eval_loss": 1.8155736923217773,
|
| 1753 |
+
"eval_runtime": 230.3013,
|
| 1754 |
+
"eval_samples_per_second": 8.658,
|
| 1755 |
+
"eval_steps_per_second": 0.868,
|
| 1756 |
+
"eval_wer": 0.9372056991827931,
|
| 1757 |
+
"step": 2576
|
| 1758 |
+
},
|
| 1759 |
+
{
|
| 1760 |
+
"epoch": 23.04,
|
| 1761 |
+
"learning_rate": 9.607361963190184e-07,
|
| 1762 |
+
"loss": 1.7301,
|
| 1763 |
+
"step": 2580
|
| 1764 |
+
},
|
| 1765 |
+
{
|
| 1766 |
+
"epoch": 23.12,
|
| 1767 |
+
"learning_rate": 9.484662576687115e-07,
|
| 1768 |
+
"loss": 1.7254,
|
| 1769 |
+
"step": 2590
|
| 1770 |
+
},
|
| 1771 |
+
{
|
| 1772 |
+
"epoch": 23.21,
|
| 1773 |
+
"learning_rate": 9.361963190184049e-07,
|
| 1774 |
+
"loss": 1.7005,
|
| 1775 |
+
"step": 2600
|
| 1776 |
+
},
|
| 1777 |
+
{
|
| 1778 |
+
"epoch": 23.3,
|
| 1779 |
+
"learning_rate": 9.239263803680981e-07,
|
| 1780 |
+
"loss": 1.726,
|
| 1781 |
+
"step": 2610
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 23.39,
|
| 1785 |
+
"learning_rate": 9.116564417177913e-07,
|
| 1786 |
+
"loss": 1.7074,
|
| 1787 |
+
"step": 2620
|
| 1788 |
+
},
|
| 1789 |
+
{
|
| 1790 |
+
"epoch": 23.48,
|
| 1791 |
+
"learning_rate": 8.993865030674847e-07,
|
| 1792 |
+
"loss": 1.7301,
|
| 1793 |
+
"step": 2630
|
| 1794 |
+
},
|
| 1795 |
+
{
|
| 1796 |
+
"epoch": 23.57,
|
| 1797 |
+
"learning_rate": 8.871165644171779e-07,
|
| 1798 |
+
"loss": 1.7189,
|
| 1799 |
+
"step": 2640
|
| 1800 |
+
},
|
| 1801 |
+
{
|
| 1802 |
+
"epoch": 23.66,
|
| 1803 |
+
"learning_rate": 8.74846625766871e-07,
|
| 1804 |
+
"loss": 1.6932,
|
| 1805 |
+
"step": 2650
|
| 1806 |
+
},
|
| 1807 |
+
{
|
| 1808 |
+
"epoch": 23.75,
|
| 1809 |
+
"learning_rate": 8.625766871165644e-07,
|
| 1810 |
+
"loss": 1.7322,
|
| 1811 |
+
"step": 2660
|
| 1812 |
+
},
|
| 1813 |
+
{
|
| 1814 |
+
"epoch": 23.84,
|
| 1815 |
+
"learning_rate": 8.503067484662576e-07,
|
| 1816 |
+
"loss": 1.7056,
|
| 1817 |
+
"step": 2670
|
| 1818 |
+
},
|
| 1819 |
+
{
|
| 1820 |
+
"epoch": 23.93,
|
| 1821 |
+
"learning_rate": 8.380368098159509e-07,
|
| 1822 |
+
"loss": 1.7158,
|
| 1823 |
+
"step": 2680
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 24.0,
|
| 1827 |
+
"eval_loss": 1.796068787574768,
|
| 1828 |
+
"eval_runtime": 229.8221,
|
| 1829 |
+
"eval_samples_per_second": 8.676,
|
| 1830 |
+
"eval_steps_per_second": 0.87,
|
| 1831 |
+
"eval_wer": 0.9482334356107787,
|
| 1832 |
+
"step": 2688
|
| 1833 |
+
},
|
| 1834 |
+
{
|
| 1835 |
+
"epoch": 24.02,
|
| 1836 |
+
"learning_rate": 8.257668711656442e-07,
|
| 1837 |
+
"loss": 1.7244,
|
| 1838 |
+
"step": 2690
|
| 1839 |
+
},
|
| 1840 |
+
{
|
| 1841 |
+
"epoch": 24.11,
|
| 1842 |
+
"learning_rate": 8.134969325153374e-07,
|
| 1843 |
+
"loss": 1.7245,
|
| 1844 |
+
"step": 2700
|
| 1845 |
+
},
|
| 1846 |
+
{
|
| 1847 |
+
"epoch": 24.2,
|
| 1848 |
+
"learning_rate": 8.012269938650306e-07,
|
| 1849 |
+
"loss": 1.6697,
|
| 1850 |
+
"step": 2710
|
| 1851 |
+
},
|
| 1852 |
+
{
|
| 1853 |
+
"epoch": 24.29,
|
| 1854 |
+
"learning_rate": 7.889570552147239e-07,
|
| 1855 |
+
"loss": 1.7091,
|
| 1856 |
+
"step": 2720
|
| 1857 |
+
},
|
| 1858 |
+
{
|
| 1859 |
+
"epoch": 24.37,
|
| 1860 |
+
"learning_rate": 7.766871165644171e-07,
|
| 1861 |
+
"loss": 1.7122,
|
| 1862 |
+
"step": 2730
|
| 1863 |
+
},
|
| 1864 |
+
{
|
| 1865 |
+
"epoch": 24.46,
|
| 1866 |
+
"learning_rate": 7.644171779141104e-07,
|
| 1867 |
+
"loss": 1.7035,
|
| 1868 |
+
"step": 2740
|
| 1869 |
+
},
|
| 1870 |
+
{
|
| 1871 |
+
"epoch": 24.55,
|
| 1872 |
+
"learning_rate": 7.521472392638037e-07,
|
| 1873 |
+
"loss": 1.6943,
|
| 1874 |
+
"step": 2750
|
| 1875 |
+
},
|
| 1876 |
+
{
|
| 1877 |
+
"epoch": 24.64,
|
| 1878 |
+
"learning_rate": 7.398773006134969e-07,
|
| 1879 |
+
"loss": 1.6933,
|
| 1880 |
+
"step": 2760
|
| 1881 |
+
},
|
| 1882 |
+
{
|
| 1883 |
+
"epoch": 24.73,
|
| 1884 |
+
"learning_rate": 7.276073619631901e-07,
|
| 1885 |
+
"loss": 1.6776,
|
| 1886 |
+
"step": 2770
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"epoch": 24.82,
|
| 1890 |
+
"learning_rate": 7.153374233128834e-07,
|
| 1891 |
+
"loss": 1.7123,
|
| 1892 |
+
"step": 2780
|
| 1893 |
+
},
|
| 1894 |
+
{
|
| 1895 |
+
"epoch": 24.91,
|
| 1896 |
+
"learning_rate": 7.030674846625767e-07,
|
| 1897 |
+
"loss": 1.6912,
|
| 1898 |
+
"step": 2790
|
| 1899 |
+
},
|
| 1900 |
+
{
|
| 1901 |
+
"epoch": 25.0,
|
| 1902 |
+
"learning_rate": 6.907975460122699e-07,
|
| 1903 |
+
"loss": 1.7111,
|
| 1904 |
+
"step": 2800
|
| 1905 |
+
},
|
| 1906 |
+
{
|
| 1907 |
+
"epoch": 25.0,
|
| 1908 |
+
"eval_loss": 1.7816540002822876,
|
| 1909 |
+
"eval_runtime": 230.1276,
|
| 1910 |
+
"eval_samples_per_second": 8.665,
|
| 1911 |
+
"eval_steps_per_second": 0.869,
|
| 1912 |
+
"eval_wer": 0.9422183066500592,
|
| 1913 |
+
"step": 2800
|
| 1914 |
+
},
|
| 1915 |
+
{
|
| 1916 |
+
"epoch": 25.09,
|
| 1917 |
+
"learning_rate": 6.785276073619631e-07,
|
| 1918 |
+
"loss": 1.7032,
|
| 1919 |
+
"step": 2810
|
| 1920 |
+
},
|
| 1921 |
+
{
|
| 1922 |
+
"epoch": 25.18,
|
| 1923 |
+
"learning_rate": 6.662576687116565e-07,
|
| 1924 |
+
"loss": 1.684,
|
| 1925 |
+
"step": 2820
|
| 1926 |
+
},
|
| 1927 |
+
{
|
| 1928 |
+
"epoch": 25.27,
|
| 1929 |
+
"learning_rate": 6.539877300613497e-07,
|
| 1930 |
+
"loss": 1.6899,
|
| 1931 |
+
"step": 2830
|
| 1932 |
+
},
|
| 1933 |
+
{
|
| 1934 |
+
"epoch": 25.36,
|
| 1935 |
+
"learning_rate": 6.417177914110428e-07,
|
| 1936 |
+
"loss": 1.6875,
|
| 1937 |
+
"step": 2840
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 25.45,
|
| 1941 |
+
"learning_rate": 6.294478527607362e-07,
|
| 1942 |
+
"loss": 1.6621,
|
| 1943 |
+
"step": 2850
|
| 1944 |
+
},
|
| 1945 |
+
{
|
| 1946 |
+
"epoch": 25.54,
|
| 1947 |
+
"learning_rate": 6.171779141104294e-07,
|
| 1948 |
+
"loss": 1.6926,
|
| 1949 |
+
"step": 2860
|
| 1950 |
+
},
|
| 1951 |
+
{
|
| 1952 |
+
"epoch": 25.62,
|
| 1953 |
+
"learning_rate": 6.049079754601226e-07,
|
| 1954 |
+
"loss": 1.7014,
|
| 1955 |
+
"step": 2870
|
| 1956 |
+
},
|
| 1957 |
+
{
|
| 1958 |
+
"epoch": 25.71,
|
| 1959 |
+
"learning_rate": 5.92638036809816e-07,
|
| 1960 |
+
"loss": 1.7038,
|
| 1961 |
+
"step": 2880
|
| 1962 |
+
},
|
| 1963 |
+
{
|
| 1964 |
+
"epoch": 25.8,
|
| 1965 |
+
"learning_rate": 5.803680981595092e-07,
|
| 1966 |
+
"loss": 1.6665,
|
| 1967 |
+
"step": 2890
|
| 1968 |
+
},
|
| 1969 |
+
{
|
| 1970 |
+
"epoch": 25.89,
|
| 1971 |
+
"learning_rate": 5.680981595092024e-07,
|
| 1972 |
+
"loss": 1.6654,
|
| 1973 |
+
"step": 2900
|
| 1974 |
+
},
|
| 1975 |
+
{
|
| 1976 |
+
"epoch": 25.98,
|
| 1977 |
+
"learning_rate": 5.558282208588957e-07,
|
| 1978 |
+
"loss": 1.69,
|
| 1979 |
+
"step": 2910
|
| 1980 |
+
},
|
| 1981 |
+
{
|
| 1982 |
+
"epoch": 26.0,
|
| 1983 |
+
"eval_loss": 1.7818827629089355,
|
| 1984 |
+
"eval_runtime": 230.0994,
|
| 1985 |
+
"eval_samples_per_second": 8.666,
|
| 1986 |
+
"eval_steps_per_second": 0.869,
|
| 1987 |
+
"eval_wer": 0.9430081720691436,
|
| 1988 |
+
"step": 2912
|
| 1989 |
+
},
|
| 1990 |
+
{
|
| 1991 |
+
"epoch": 26.07,
|
| 1992 |
+
"learning_rate": 5.435582822085889e-07,
|
| 1993 |
+
"loss": 1.7023,
|
| 1994 |
+
"step": 2920
|
| 1995 |
+
},
|
| 1996 |
+
{
|
| 1997 |
+
"epoch": 26.16,
|
| 1998 |
+
"learning_rate": 5.312883435582822e-07,
|
| 1999 |
+
"loss": 1.6845,
|
| 2000 |
+
"step": 2930
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 26.25,
|
| 2004 |
+
"learning_rate": 5.190184049079755e-07,
|
| 2005 |
+
"loss": 1.6649,
|
| 2006 |
+
"step": 2940
|
| 2007 |
+
},
|
| 2008 |
+
{
|
| 2009 |
+
"epoch": 26.34,
|
| 2010 |
+
"learning_rate": 5.067484662576687e-07,
|
| 2011 |
+
"loss": 1.6683,
|
| 2012 |
+
"step": 2950
|
| 2013 |
+
},
|
| 2014 |
+
{
|
| 2015 |
+
"epoch": 26.43,
|
| 2016 |
+
"learning_rate": 4.944785276073619e-07,
|
| 2017 |
+
"loss": 1.6638,
|
| 2018 |
+
"step": 2960
|
| 2019 |
+
},
|
| 2020 |
+
{
|
| 2021 |
+
"epoch": 26.52,
|
| 2022 |
+
"learning_rate": 4.822085889570552e-07,
|
| 2023 |
+
"loss": 1.6879,
|
| 2024 |
+
"step": 2970
|
| 2025 |
+
},
|
| 2026 |
+
{
|
| 2027 |
+
"epoch": 26.61,
|
| 2028 |
+
"learning_rate": 4.6993865030674844e-07,
|
| 2029 |
+
"loss": 1.6527,
|
| 2030 |
+
"step": 2980
|
| 2031 |
+
},
|
| 2032 |
+
{
|
| 2033 |
+
"epoch": 26.7,
|
| 2034 |
+
"learning_rate": 4.576687116564417e-07,
|
| 2035 |
+
"loss": 1.6745,
|
| 2036 |
+
"step": 2990
|
| 2037 |
+
},
|
| 2038 |
+
{
|
| 2039 |
+
"epoch": 26.78,
|
| 2040 |
+
"learning_rate": 4.4539877300613495e-07,
|
| 2041 |
+
"loss": 1.6944,
|
| 2042 |
+
"step": 3000
|
| 2043 |
+
},
|
| 2044 |
+
{
|
| 2045 |
+
"epoch": 26.87,
|
| 2046 |
+
"learning_rate": 4.331288343558282e-07,
|
| 2047 |
+
"loss": 1.6737,
|
| 2048 |
+
"step": 3010
|
| 2049 |
+
},
|
| 2050 |
+
{
|
| 2051 |
+
"epoch": 26.96,
|
| 2052 |
+
"learning_rate": 4.2085889570552147e-07,
|
| 2053 |
+
"loss": 1.6889,
|
| 2054 |
+
"step": 3020
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"epoch": 27.0,
|
| 2058 |
+
"eval_loss": 1.77213454246521,
|
| 2059 |
+
"eval_runtime": 230.2415,
|
| 2060 |
+
"eval_samples_per_second": 8.66,
|
| 2061 |
+
"eval_steps_per_second": 0.869,
|
| 2062 |
+
"eval_wer": 0.9386031533857885,
|
| 2063 |
+
"step": 3024
|
| 2064 |
+
},
|
| 2065 |
+
{
|
| 2066 |
+
"epoch": 27.05,
|
| 2067 |
+
"learning_rate": 4.085889570552147e-07,
|
| 2068 |
+
"loss": 1.6938,
|
| 2069 |
+
"step": 3030
|
| 2070 |
+
},
|
| 2071 |
+
{
|
| 2072 |
+
"epoch": 27.14,
|
| 2073 |
+
"learning_rate": 3.9631901840490793e-07,
|
| 2074 |
+
"loss": 1.6539,
|
| 2075 |
+
"step": 3040
|
| 2076 |
+
},
|
| 2077 |
+
{
|
| 2078 |
+
"epoch": 27.23,
|
| 2079 |
+
"learning_rate": 3.840490797546012e-07,
|
| 2080 |
+
"loss": 1.6533,
|
| 2081 |
+
"step": 3050
|
| 2082 |
+
},
|
| 2083 |
+
{
|
| 2084 |
+
"epoch": 27.32,
|
| 2085 |
+
"learning_rate": 3.717791411042945e-07,
|
| 2086 |
+
"loss": 1.6803,
|
| 2087 |
+
"step": 3060
|
| 2088 |
+
},
|
| 2089 |
+
{
|
| 2090 |
+
"epoch": 27.41,
|
| 2091 |
+
"learning_rate": 3.595092024539877e-07,
|
| 2092 |
+
"loss": 1.6623,
|
| 2093 |
+
"step": 3070
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 27.5,
|
| 2097 |
+
"learning_rate": 3.4723926380368096e-07,
|
| 2098 |
+
"loss": 1.6803,
|
| 2099 |
+
"step": 3080
|
| 2100 |
+
},
|
| 2101 |
+
{
|
| 2102 |
+
"epoch": 27.59,
|
| 2103 |
+
"learning_rate": 3.3496932515337424e-07,
|
| 2104 |
+
"loss": 1.6694,
|
| 2105 |
+
"step": 3090
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 27.68,
|
| 2109 |
+
"learning_rate": 3.2269938650306747e-07,
|
| 2110 |
+
"loss": 1.6578,
|
| 2111 |
+
"step": 3100
|
| 2112 |
+
},
|
| 2113 |
+
{
|
| 2114 |
+
"epoch": 27.77,
|
| 2115 |
+
"learning_rate": 3.104294478527607e-07,
|
| 2116 |
+
"loss": 1.6606,
|
| 2117 |
+
"step": 3110
|
| 2118 |
+
},
|
| 2119 |
+
{
|
| 2120 |
+
"epoch": 27.86,
|
| 2121 |
+
"learning_rate": 2.98159509202454e-07,
|
| 2122 |
+
"loss": 1.6786,
|
| 2123 |
+
"step": 3120
|
| 2124 |
+
},
|
| 2125 |
+
{
|
| 2126 |
+
"epoch": 27.95,
|
| 2127 |
+
"learning_rate": 2.858895705521472e-07,
|
| 2128 |
+
"loss": 1.6546,
|
| 2129 |
+
"step": 3130
|
| 2130 |
+
},
|
| 2131 |
+
{
|
| 2132 |
+
"epoch": 28.0,
|
| 2133 |
+
"eval_loss": 1.764729619026184,
|
| 2134 |
+
"eval_runtime": 230.6298,
|
| 2135 |
+
"eval_samples_per_second": 8.646,
|
| 2136 |
+
"eval_steps_per_second": 0.867,
|
| 2137 |
+
"eval_wer": 0.9452866300088101,
|
| 2138 |
+
"step": 3136
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"epoch": 28.04,
|
| 2142 |
+
"learning_rate": 2.7361963190184045e-07,
|
| 2143 |
+
"loss": 1.7145,
|
| 2144 |
+
"step": 3140
|
| 2145 |
+
},
|
| 2146 |
+
{
|
| 2147 |
+
"epoch": 28.12,
|
| 2148 |
+
"learning_rate": 2.6134969325153373e-07,
|
| 2149 |
+
"loss": 1.6281,
|
| 2150 |
+
"step": 3150
|
| 2151 |
+
},
|
| 2152 |
+
{
|
| 2153 |
+
"epoch": 28.21,
|
| 2154 |
+
"learning_rate": 2.4907975460122696e-07,
|
| 2155 |
+
"loss": 1.6589,
|
| 2156 |
+
"step": 3160
|
| 2157 |
+
},
|
| 2158 |
+
{
|
| 2159 |
+
"epoch": 28.3,
|
| 2160 |
+
"learning_rate": 2.3680981595092022e-07,
|
| 2161 |
+
"loss": 1.6808,
|
| 2162 |
+
"step": 3170
|
| 2163 |
+
},
|
| 2164 |
+
{
|
| 2165 |
+
"epoch": 28.39,
|
| 2166 |
+
"learning_rate": 2.245398773006135e-07,
|
| 2167 |
+
"loss": 1.6921,
|
| 2168 |
+
"step": 3180
|
| 2169 |
+
},
|
| 2170 |
+
{
|
| 2171 |
+
"epoch": 28.48,
|
| 2172 |
+
"learning_rate": 2.1226993865030673e-07,
|
| 2173 |
+
"loss": 1.6476,
|
| 2174 |
+
"step": 3190
|
| 2175 |
+
},
|
| 2176 |
+
{
|
| 2177 |
+
"epoch": 28.57,
|
| 2178 |
+
"learning_rate": 2e-07,
|
| 2179 |
+
"loss": 1.6768,
|
| 2180 |
+
"step": 3200
|
| 2181 |
+
},
|
| 2182 |
+
{
|
| 2183 |
+
"epoch": 28.66,
|
| 2184 |
+
"learning_rate": 1.8773006134969325e-07,
|
| 2185 |
+
"loss": 1.6676,
|
| 2186 |
+
"step": 3210
|
| 2187 |
+
},
|
| 2188 |
+
{
|
| 2189 |
+
"epoch": 28.75,
|
| 2190 |
+
"learning_rate": 1.7546012269938648e-07,
|
| 2191 |
+
"loss": 1.6758,
|
| 2192 |
+
"step": 3220
|
| 2193 |
+
},
|
| 2194 |
+
{
|
| 2195 |
+
"epoch": 28.84,
|
| 2196 |
+
"learning_rate": 1.6319018404907974e-07,
|
| 2197 |
+
"loss": 1.6158,
|
| 2198 |
+
"step": 3230
|
| 2199 |
+
},
|
| 2200 |
+
{
|
| 2201 |
+
"epoch": 28.93,
|
| 2202 |
+
"learning_rate": 1.50920245398773e-07,
|
| 2203 |
+
"loss": 1.6542,
|
| 2204 |
+
"step": 3240
|
| 2205 |
+
},
|
| 2206 |
+
{
|
| 2207 |
+
"epoch": 29.0,
|
| 2208 |
+
"eval_loss": 1.765263557434082,
|
| 2209 |
+
"eval_runtime": 230.3263,
|
| 2210 |
+
"eval_samples_per_second": 8.657,
|
| 2211 |
+
"eval_steps_per_second": 0.868,
|
| 2212 |
+
"eval_wer": 0.9375094935747487,
|
| 2213 |
+
"step": 3248
|
| 2214 |
+
},
|
| 2215 |
+
{
|
| 2216 |
+
"epoch": 29.02,
|
| 2217 |
+
"learning_rate": 1.3865030674846625e-07,
|
| 2218 |
+
"loss": 1.694,
|
| 2219 |
+
"step": 3250
|
| 2220 |
+
},
|
| 2221 |
+
{
|
| 2222 |
+
"epoch": 29.11,
|
| 2223 |
+
"learning_rate": 1.263803680981595e-07,
|
| 2224 |
+
"loss": 1.6854,
|
| 2225 |
+
"step": 3260
|
| 2226 |
+
},
|
| 2227 |
+
{
|
| 2228 |
+
"epoch": 29.2,
|
| 2229 |
+
"learning_rate": 1.1411042944785275e-07,
|
| 2230 |
+
"loss": 1.6717,
|
| 2231 |
+
"step": 3270
|
| 2232 |
+
},
|
| 2233 |
+
{
|
| 2234 |
+
"epoch": 29.29,
|
| 2235 |
+
"learning_rate": 1.0184049079754601e-07,
|
| 2236 |
+
"loss": 1.6737,
|
| 2237 |
+
"step": 3280
|
| 2238 |
+
},
|
| 2239 |
+
{
|
| 2240 |
+
"epoch": 29.37,
|
| 2241 |
+
"learning_rate": 8.957055214723925e-08,
|
| 2242 |
+
"loss": 1.6468,
|
| 2243 |
+
"step": 3290
|
| 2244 |
+
},
|
| 2245 |
+
{
|
| 2246 |
+
"epoch": 29.46,
|
| 2247 |
+
"learning_rate": 7.730061349693251e-08,
|
| 2248 |
+
"loss": 1.6409,
|
| 2249 |
+
"step": 3300
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 29.55,
|
| 2253 |
+
"learning_rate": 6.503067484662577e-08,
|
| 2254 |
+
"loss": 1.66,
|
| 2255 |
+
"step": 3310
|
| 2256 |
+
},
|
| 2257 |
+
{
|
| 2258 |
+
"epoch": 29.64,
|
| 2259 |
+
"learning_rate": 5.276073619631902e-08,
|
| 2260 |
+
"loss": 1.6526,
|
| 2261 |
+
"step": 3320
|
| 2262 |
+
},
|
| 2263 |
+
{
|
| 2264 |
+
"epoch": 29.73,
|
| 2265 |
+
"learning_rate": 4.049079754601227e-08,
|
| 2266 |
+
"loss": 1.6681,
|
| 2267 |
+
"step": 3330
|
| 2268 |
+
},
|
| 2269 |
+
{
|
| 2270 |
+
"epoch": 29.82,
|
| 2271 |
+
"learning_rate": 2.822085889570552e-08,
|
| 2272 |
+
"loss": 1.6626,
|
| 2273 |
+
"step": 3340
|
| 2274 |
+
},
|
| 2275 |
+
{
|
| 2276 |
+
"epoch": 29.91,
|
| 2277 |
+
"learning_rate": 1.5950920245398775e-08,
|
| 2278 |
+
"loss": 1.6352,
|
| 2279 |
+
"step": 3350
|
| 2280 |
+
},
|
| 2281 |
+
{
|
| 2282 |
+
"epoch": 30.0,
|
| 2283 |
+
"learning_rate": 3.6809815950920243e-09,
|
| 2284 |
+
"loss": 1.647,
|
| 2285 |
+
"step": 3360
|
| 2286 |
+
},
|
| 2287 |
+
{
|
| 2288 |
+
"epoch": 30.0,
|
| 2289 |
+
"eval_loss": 1.7582672834396362,
|
| 2290 |
+
"eval_runtime": 230.7695,
|
| 2291 |
+
"eval_samples_per_second": 8.641,
|
| 2292 |
+
"eval_steps_per_second": 0.867,
|
| 2293 |
+
"eval_wer": 0.9386031533857885,
|
| 2294 |
+
"step": 3360
|
| 2295 |
+
},
|
| 2296 |
+
{
|
| 2297 |
+
"epoch": 30.0,
|
| 2298 |
+
"step": 3360,
|
| 2299 |
+
"total_flos": 1.6308106473516027e+20,
|
| 2300 |
+
"train_loss": 3.5984562555948894,
|
| 2301 |
+
"train_runtime": 102250.8509,
|
| 2302 |
+
"train_samples_per_second": 5.263,
|
| 2303 |
+
"train_steps_per_second": 0.033
|
| 2304 |
+
}
|
| 2305 |
+
],
|
| 2306 |
+
"max_steps": 3360,
|
| 2307 |
+
"num_train_epochs": 30,
|
| 2308 |
+
"total_flos": 1.6308106473516027e+20,
|
| 2309 |
+
"trial_name": null,
|
| 2310 |
+
"trial_params": null
|
| 2311 |
+
}
|