End of training
Browse files- all_results.json +14 -0
- eval_results.json +9 -0
- runs/Jun27_14-40-12_cb8663b38859/events.out.tfevents.1656416947.cb8663b38859.2284.2 +3 -0
- train_results.json +8 -0
- trainer_state.json +2371 -0
all_results.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"eval_loss": 1.0162557363510132,
|
| 4 |
+
"eval_runtime": 341.9054,
|
| 5 |
+
"eval_samples": 2674,
|
| 6 |
+
"eval_samples_per_second": 7.821,
|
| 7 |
+
"eval_steps_per_second": 1.957,
|
| 8 |
+
"eval_wer": 0.6621886963032024,
|
| 9 |
+
"train_loss": 0.8711832551245994,
|
| 10 |
+
"train_runtime": 72833.4219,
|
| 11 |
+
"train_samples": 24061,
|
| 12 |
+
"train_samples_per_second": 3.304,
|
| 13 |
+
"train_steps_per_second": 0.052
|
| 14 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"eval_loss": 1.0162557363510132,
|
| 4 |
+
"eval_runtime": 341.9054,
|
| 5 |
+
"eval_samples": 2674,
|
| 6 |
+
"eval_samples_per_second": 7.821,
|
| 7 |
+
"eval_steps_per_second": 1.957,
|
| 8 |
+
"eval_wer": 0.6621886963032024
|
| 9 |
+
}
|
runs/Jun27_14-40-12_cb8663b38859/events.out.tfevents.1656416947.cb8663b38859.2284.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3569a8f500829f4a66cc15c96cb6bf105e79e80ec420ea456a1d7e056d444efa
|
| 3 |
+
size 358
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 10.0,
|
| 3 |
+
"train_loss": 0.8711832551245994,
|
| 4 |
+
"train_runtime": 72833.4219,
|
| 5 |
+
"train_samples": 24061,
|
| 6 |
+
"train_samples_per_second": 3.304,
|
| 7 |
+
"train_steps_per_second": 0.052
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2371 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 1.0162557363510132,
|
| 3 |
+
"best_model_checkpoint": "ai-light-dance_stepmania_ft_wav2vec2-large-xlsr-53-v5/checkpoint-2256",
|
| 4 |
+
"epoch": 10.0,
|
| 5 |
+
"global_step": 3760,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.03,
|
| 12 |
+
"learning_rate": 4.000000000000001e-06,
|
| 13 |
+
"loss": 0.9198,
|
| 14 |
+
"step": 10
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.05,
|
| 18 |
+
"learning_rate": 8.000000000000001e-06,
|
| 19 |
+
"loss": 0.879,
|
| 20 |
+
"step": 20
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.08,
|
| 24 |
+
"learning_rate": 1.2e-05,
|
| 25 |
+
"loss": 0.8647,
|
| 26 |
+
"step": 30
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.11,
|
| 30 |
+
"learning_rate": 1.6000000000000003e-05,
|
| 31 |
+
"loss": 0.8746,
|
| 32 |
+
"step": 40
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.13,
|
| 36 |
+
"learning_rate": 2e-05,
|
| 37 |
+
"loss": 0.8716,
|
| 38 |
+
"step": 50
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.16,
|
| 42 |
+
"learning_rate": 2.4e-05,
|
| 43 |
+
"loss": 0.9164,
|
| 44 |
+
"step": 60
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.19,
|
| 48 |
+
"learning_rate": 2.8e-05,
|
| 49 |
+
"loss": 0.8862,
|
| 50 |
+
"step": 70
|
| 51 |
+
},
|
| 52 |
+
{
|
| 53 |
+
"epoch": 0.21,
|
| 54 |
+
"learning_rate": 3.2000000000000005e-05,
|
| 55 |
+
"loss": 0.8818,
|
| 56 |
+
"step": 80
|
| 57 |
+
},
|
| 58 |
+
{
|
| 59 |
+
"epoch": 0.24,
|
| 60 |
+
"learning_rate": 3.6e-05,
|
| 61 |
+
"loss": 0.8818,
|
| 62 |
+
"step": 90
|
| 63 |
+
},
|
| 64 |
+
{
|
| 65 |
+
"epoch": 0.27,
|
| 66 |
+
"learning_rate": 4e-05,
|
| 67 |
+
"loss": 0.8894,
|
| 68 |
+
"step": 100
|
| 69 |
+
},
|
| 70 |
+
{
|
| 71 |
+
"epoch": 0.29,
|
| 72 |
+
"learning_rate": 3.989071038251366e-05,
|
| 73 |
+
"loss": 0.9504,
|
| 74 |
+
"step": 110
|
| 75 |
+
},
|
| 76 |
+
{
|
| 77 |
+
"epoch": 0.32,
|
| 78 |
+
"learning_rate": 3.9781420765027323e-05,
|
| 79 |
+
"loss": 0.8916,
|
| 80 |
+
"step": 120
|
| 81 |
+
},
|
| 82 |
+
{
|
| 83 |
+
"epoch": 0.35,
|
| 84 |
+
"learning_rate": 3.9672131147540983e-05,
|
| 85 |
+
"loss": 0.8946,
|
| 86 |
+
"step": 130
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.37,
|
| 90 |
+
"learning_rate": 3.956284153005465e-05,
|
| 91 |
+
"loss": 0.8959,
|
| 92 |
+
"step": 140
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 0.4,
|
| 96 |
+
"learning_rate": 3.945355191256831e-05,
|
| 97 |
+
"loss": 0.8869,
|
| 98 |
+
"step": 150
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 0.43,
|
| 102 |
+
"learning_rate": 3.934426229508197e-05,
|
| 103 |
+
"loss": 0.9374,
|
| 104 |
+
"step": 160
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 0.45,
|
| 108 |
+
"learning_rate": 3.923497267759563e-05,
|
| 109 |
+
"loss": 0.9024,
|
| 110 |
+
"step": 170
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 0.48,
|
| 114 |
+
"learning_rate": 3.912568306010929e-05,
|
| 115 |
+
"loss": 0.8983,
|
| 116 |
+
"step": 180
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 0.51,
|
| 120 |
+
"learning_rate": 3.901639344262295e-05,
|
| 121 |
+
"loss": 0.8933,
|
| 122 |
+
"step": 190
|
| 123 |
+
},
|
| 124 |
+
{
|
| 125 |
+
"epoch": 0.53,
|
| 126 |
+
"learning_rate": 3.890710382513661e-05,
|
| 127 |
+
"loss": 0.9064,
|
| 128 |
+
"step": 200
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.56,
|
| 132 |
+
"learning_rate": 3.879781420765028e-05,
|
| 133 |
+
"loss": 0.9612,
|
| 134 |
+
"step": 210
|
| 135 |
+
},
|
| 136 |
+
{
|
| 137 |
+
"epoch": 0.59,
|
| 138 |
+
"learning_rate": 3.868852459016394e-05,
|
| 139 |
+
"loss": 0.8932,
|
| 140 |
+
"step": 220
|
| 141 |
+
},
|
| 142 |
+
{
|
| 143 |
+
"epoch": 0.61,
|
| 144 |
+
"learning_rate": 3.85792349726776e-05,
|
| 145 |
+
"loss": 0.9063,
|
| 146 |
+
"step": 230
|
| 147 |
+
},
|
| 148 |
+
{
|
| 149 |
+
"epoch": 0.64,
|
| 150 |
+
"learning_rate": 3.846994535519126e-05,
|
| 151 |
+
"loss": 0.8773,
|
| 152 |
+
"step": 240
|
| 153 |
+
},
|
| 154 |
+
{
|
| 155 |
+
"epoch": 0.66,
|
| 156 |
+
"learning_rate": 3.836065573770492e-05,
|
| 157 |
+
"loss": 0.8908,
|
| 158 |
+
"step": 250
|
| 159 |
+
},
|
| 160 |
+
{
|
| 161 |
+
"epoch": 0.69,
|
| 162 |
+
"learning_rate": 3.825136612021858e-05,
|
| 163 |
+
"loss": 0.939,
|
| 164 |
+
"step": 260
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 0.72,
|
| 168 |
+
"learning_rate": 3.8142076502732245e-05,
|
| 169 |
+
"loss": 0.8949,
|
| 170 |
+
"step": 270
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 0.74,
|
| 174 |
+
"learning_rate": 3.8032786885245905e-05,
|
| 175 |
+
"loss": 0.8847,
|
| 176 |
+
"step": 280
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 0.77,
|
| 180 |
+
"learning_rate": 3.7923497267759565e-05,
|
| 181 |
+
"loss": 0.8946,
|
| 182 |
+
"step": 290
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 0.8,
|
| 186 |
+
"learning_rate": 3.781420765027323e-05,
|
| 187 |
+
"loss": 0.8865,
|
| 188 |
+
"step": 300
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 0.82,
|
| 192 |
+
"learning_rate": 3.770491803278689e-05,
|
| 193 |
+
"loss": 0.9316,
|
| 194 |
+
"step": 310
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 0.85,
|
| 198 |
+
"learning_rate": 3.759562841530055e-05,
|
| 199 |
+
"loss": 0.8822,
|
| 200 |
+
"step": 320
|
| 201 |
+
},
|
| 202 |
+
{
|
| 203 |
+
"epoch": 0.88,
|
| 204 |
+
"learning_rate": 3.748633879781421e-05,
|
| 205 |
+
"loss": 0.8983,
|
| 206 |
+
"step": 330
|
| 207 |
+
},
|
| 208 |
+
{
|
| 209 |
+
"epoch": 0.9,
|
| 210 |
+
"learning_rate": 3.737704918032787e-05,
|
| 211 |
+
"loss": 0.8889,
|
| 212 |
+
"step": 340
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 0.93,
|
| 216 |
+
"learning_rate": 3.726775956284153e-05,
|
| 217 |
+
"loss": 0.8957,
|
| 218 |
+
"step": 350
|
| 219 |
+
},
|
| 220 |
+
{
|
| 221 |
+
"epoch": 0.96,
|
| 222 |
+
"learning_rate": 3.71584699453552e-05,
|
| 223 |
+
"loss": 0.9067,
|
| 224 |
+
"step": 360
|
| 225 |
+
},
|
| 226 |
+
{
|
| 227 |
+
"epoch": 0.98,
|
| 228 |
+
"learning_rate": 3.704918032786886e-05,
|
| 229 |
+
"loss": 0.8867,
|
| 230 |
+
"step": 370
|
| 231 |
+
},
|
| 232 |
+
{
|
| 233 |
+
"epoch": 1.0,
|
| 234 |
+
"eval_loss": 1.0381758213043213,
|
| 235 |
+
"eval_runtime": 348.9535,
|
| 236 |
+
"eval_samples_per_second": 7.663,
|
| 237 |
+
"eval_steps_per_second": 1.917,
|
| 238 |
+
"eval_wer": 0.6820688128905276,
|
| 239 |
+
"step": 376
|
| 240 |
+
},
|
| 241 |
+
{
|
| 242 |
+
"epoch": 1.01,
|
| 243 |
+
"learning_rate": 3.693989071038252e-05,
|
| 244 |
+
"loss": 0.9229,
|
| 245 |
+
"step": 380
|
| 246 |
+
},
|
| 247 |
+
{
|
| 248 |
+
"epoch": 1.04,
|
| 249 |
+
"learning_rate": 3.683060109289618e-05,
|
| 250 |
+
"loss": 0.9061,
|
| 251 |
+
"step": 390
|
| 252 |
+
},
|
| 253 |
+
{
|
| 254 |
+
"epoch": 1.06,
|
| 255 |
+
"learning_rate": 3.672131147540984e-05,
|
| 256 |
+
"loss": 0.8896,
|
| 257 |
+
"step": 400
|
| 258 |
+
},
|
| 259 |
+
{
|
| 260 |
+
"epoch": 1.09,
|
| 261 |
+
"learning_rate": 3.66120218579235e-05,
|
| 262 |
+
"loss": 0.8908,
|
| 263 |
+
"step": 410
|
| 264 |
+
},
|
| 265 |
+
{
|
| 266 |
+
"epoch": 1.12,
|
| 267 |
+
"learning_rate": 3.650273224043716e-05,
|
| 268 |
+
"loss": 0.8999,
|
| 269 |
+
"step": 420
|
| 270 |
+
},
|
| 271 |
+
{
|
| 272 |
+
"epoch": 1.14,
|
| 273 |
+
"learning_rate": 3.6393442622950826e-05,
|
| 274 |
+
"loss": 0.9126,
|
| 275 |
+
"step": 430
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 1.17,
|
| 279 |
+
"learning_rate": 3.6284153005464486e-05,
|
| 280 |
+
"loss": 0.9274,
|
| 281 |
+
"step": 440
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 1.2,
|
| 285 |
+
"learning_rate": 3.6174863387978146e-05,
|
| 286 |
+
"loss": 0.8986,
|
| 287 |
+
"step": 450
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"epoch": 1.22,
|
| 291 |
+
"learning_rate": 3.6065573770491806e-05,
|
| 292 |
+
"loss": 0.8889,
|
| 293 |
+
"step": 460
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 1.25,
|
| 297 |
+
"learning_rate": 3.5956284153005466e-05,
|
| 298 |
+
"loss": 0.89,
|
| 299 |
+
"step": 470
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"epoch": 1.28,
|
| 303 |
+
"learning_rate": 3.5846994535519126e-05,
|
| 304 |
+
"loss": 0.9142,
|
| 305 |
+
"step": 480
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 1.3,
|
| 309 |
+
"learning_rate": 3.5737704918032786e-05,
|
| 310 |
+
"loss": 0.9122,
|
| 311 |
+
"step": 490
|
| 312 |
+
},
|
| 313 |
+
{
|
| 314 |
+
"epoch": 1.33,
|
| 315 |
+
"learning_rate": 3.562841530054645e-05,
|
| 316 |
+
"loss": 0.8805,
|
| 317 |
+
"step": 500
|
| 318 |
+
},
|
| 319 |
+
{
|
| 320 |
+
"epoch": 1.36,
|
| 321 |
+
"learning_rate": 3.551912568306011e-05,
|
| 322 |
+
"loss": 0.878,
|
| 323 |
+
"step": 510
|
| 324 |
+
},
|
| 325 |
+
{
|
| 326 |
+
"epoch": 1.38,
|
| 327 |
+
"learning_rate": 3.5409836065573773e-05,
|
| 328 |
+
"loss": 0.8903,
|
| 329 |
+
"step": 520
|
| 330 |
+
},
|
| 331 |
+
{
|
| 332 |
+
"epoch": 1.41,
|
| 333 |
+
"learning_rate": 3.5300546448087433e-05,
|
| 334 |
+
"loss": 0.9098,
|
| 335 |
+
"step": 530
|
| 336 |
+
},
|
| 337 |
+
{
|
| 338 |
+
"epoch": 1.44,
|
| 339 |
+
"learning_rate": 3.5191256830601094e-05,
|
| 340 |
+
"loss": 0.918,
|
| 341 |
+
"step": 540
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"epoch": 1.46,
|
| 345 |
+
"learning_rate": 3.5081967213114754e-05,
|
| 346 |
+
"loss": 0.8911,
|
| 347 |
+
"step": 550
|
| 348 |
+
},
|
| 349 |
+
{
|
| 350 |
+
"epoch": 1.49,
|
| 351 |
+
"learning_rate": 3.4972677595628414e-05,
|
| 352 |
+
"loss": 0.8767,
|
| 353 |
+
"step": 560
|
| 354 |
+
},
|
| 355 |
+
{
|
| 356 |
+
"epoch": 1.52,
|
| 357 |
+
"learning_rate": 3.486338797814208e-05,
|
| 358 |
+
"loss": 0.8821,
|
| 359 |
+
"step": 570
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 1.54,
|
| 363 |
+
"learning_rate": 3.475409836065574e-05,
|
| 364 |
+
"loss": 0.9001,
|
| 365 |
+
"step": 580
|
| 366 |
+
},
|
| 367 |
+
{
|
| 368 |
+
"epoch": 1.57,
|
| 369 |
+
"learning_rate": 3.46448087431694e-05,
|
| 370 |
+
"loss": 0.8989,
|
| 371 |
+
"step": 590
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 1.6,
|
| 375 |
+
"learning_rate": 3.453551912568306e-05,
|
| 376 |
+
"loss": 0.8716,
|
| 377 |
+
"step": 600
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 1.62,
|
| 381 |
+
"learning_rate": 3.442622950819672e-05,
|
| 382 |
+
"loss": 0.8965,
|
| 383 |
+
"step": 610
|
| 384 |
+
},
|
| 385 |
+
{
|
| 386 |
+
"epoch": 1.65,
|
| 387 |
+
"learning_rate": 3.431693989071038e-05,
|
| 388 |
+
"loss": 0.9048,
|
| 389 |
+
"step": 620
|
| 390 |
+
},
|
| 391 |
+
{
|
| 392 |
+
"epoch": 1.68,
|
| 393 |
+
"learning_rate": 3.420765027322405e-05,
|
| 394 |
+
"loss": 0.9245,
|
| 395 |
+
"step": 630
|
| 396 |
+
},
|
| 397 |
+
{
|
| 398 |
+
"epoch": 1.7,
|
| 399 |
+
"learning_rate": 3.409836065573771e-05,
|
| 400 |
+
"loss": 0.9237,
|
| 401 |
+
"step": 640
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"epoch": 1.73,
|
| 405 |
+
"learning_rate": 3.398907103825137e-05,
|
| 406 |
+
"loss": 0.9033,
|
| 407 |
+
"step": 650
|
| 408 |
+
},
|
| 409 |
+
{
|
| 410 |
+
"epoch": 1.76,
|
| 411 |
+
"learning_rate": 3.3879781420765035e-05,
|
| 412 |
+
"loss": 0.8863,
|
| 413 |
+
"step": 660
|
| 414 |
+
},
|
| 415 |
+
{
|
| 416 |
+
"epoch": 1.78,
|
| 417 |
+
"learning_rate": 3.3770491803278695e-05,
|
| 418 |
+
"loss": 0.8961,
|
| 419 |
+
"step": 670
|
| 420 |
+
},
|
| 421 |
+
{
|
| 422 |
+
"epoch": 1.81,
|
| 423 |
+
"learning_rate": 3.3661202185792355e-05,
|
| 424 |
+
"loss": 0.9243,
|
| 425 |
+
"step": 680
|
| 426 |
+
},
|
| 427 |
+
{
|
| 428 |
+
"epoch": 1.84,
|
| 429 |
+
"learning_rate": 3.3551912568306015e-05,
|
| 430 |
+
"loss": 0.9223,
|
| 431 |
+
"step": 690
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 1.86,
|
| 435 |
+
"learning_rate": 3.3442622950819675e-05,
|
| 436 |
+
"loss": 0.8865,
|
| 437 |
+
"step": 700
|
| 438 |
+
},
|
| 439 |
+
{
|
| 440 |
+
"epoch": 1.89,
|
| 441 |
+
"learning_rate": 3.3333333333333335e-05,
|
| 442 |
+
"loss": 0.893,
|
| 443 |
+
"step": 710
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 1.91,
|
| 447 |
+
"learning_rate": 3.3224043715847e-05,
|
| 448 |
+
"loss": 0.892,
|
| 449 |
+
"step": 720
|
| 450 |
+
},
|
| 451 |
+
{
|
| 452 |
+
"epoch": 1.94,
|
| 453 |
+
"learning_rate": 3.311475409836066e-05,
|
| 454 |
+
"loss": 0.9133,
|
| 455 |
+
"step": 730
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"epoch": 1.97,
|
| 459 |
+
"learning_rate": 3.300546448087432e-05,
|
| 460 |
+
"loss": 0.8801,
|
| 461 |
+
"step": 740
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 1.99,
|
| 465 |
+
"learning_rate": 3.289617486338798e-05,
|
| 466 |
+
"loss": 0.8861,
|
| 467 |
+
"step": 750
|
| 468 |
+
},
|
| 469 |
+
{
|
| 470 |
+
"epoch": 2.0,
|
| 471 |
+
"eval_loss": 1.0259734392166138,
|
| 472 |
+
"eval_runtime": 341.2083,
|
| 473 |
+
"eval_samples_per_second": 7.837,
|
| 474 |
+
"eval_steps_per_second": 1.961,
|
| 475 |
+
"eval_wer": 0.6686301931036431,
|
| 476 |
+
"step": 752
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 2.02,
|
| 480 |
+
"learning_rate": 3.278688524590164e-05,
|
| 481 |
+
"loss": 0.9331,
|
| 482 |
+
"step": 760
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 2.05,
|
| 486 |
+
"learning_rate": 3.26775956284153e-05,
|
| 487 |
+
"loss": 0.8923,
|
| 488 |
+
"step": 770
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 2.07,
|
| 492 |
+
"learning_rate": 3.256830601092896e-05,
|
| 493 |
+
"loss": 0.871,
|
| 494 |
+
"step": 780
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 2.1,
|
| 498 |
+
"learning_rate": 3.245901639344263e-05,
|
| 499 |
+
"loss": 0.8728,
|
| 500 |
+
"step": 790
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 2.13,
|
| 504 |
+
"learning_rate": 3.234972677595629e-05,
|
| 505 |
+
"loss": 0.897,
|
| 506 |
+
"step": 800
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 2.15,
|
| 510 |
+
"learning_rate": 3.224043715846995e-05,
|
| 511 |
+
"loss": 0.9309,
|
| 512 |
+
"step": 810
|
| 513 |
+
},
|
| 514 |
+
{
|
| 515 |
+
"epoch": 2.18,
|
| 516 |
+
"learning_rate": 3.213114754098361e-05,
|
| 517 |
+
"loss": 0.8783,
|
| 518 |
+
"step": 820
|
| 519 |
+
},
|
| 520 |
+
{
|
| 521 |
+
"epoch": 2.21,
|
| 522 |
+
"learning_rate": 3.202185792349727e-05,
|
| 523 |
+
"loss": 0.8631,
|
| 524 |
+
"step": 830
|
| 525 |
+
},
|
| 526 |
+
{
|
| 527 |
+
"epoch": 2.23,
|
| 528 |
+
"learning_rate": 3.191256830601093e-05,
|
| 529 |
+
"loss": 0.8903,
|
| 530 |
+
"step": 840
|
| 531 |
+
},
|
| 532 |
+
{
|
| 533 |
+
"epoch": 2.26,
|
| 534 |
+
"learning_rate": 3.180327868852459e-05,
|
| 535 |
+
"loss": 0.8706,
|
| 536 |
+
"step": 850
|
| 537 |
+
},
|
| 538 |
+
{
|
| 539 |
+
"epoch": 2.29,
|
| 540 |
+
"learning_rate": 3.1693989071038256e-05,
|
| 541 |
+
"loss": 0.9252,
|
| 542 |
+
"step": 860
|
| 543 |
+
},
|
| 544 |
+
{
|
| 545 |
+
"epoch": 2.31,
|
| 546 |
+
"learning_rate": 3.1584699453551916e-05,
|
| 547 |
+
"loss": 0.8834,
|
| 548 |
+
"step": 870
|
| 549 |
+
},
|
| 550 |
+
{
|
| 551 |
+
"epoch": 2.34,
|
| 552 |
+
"learning_rate": 3.1475409836065576e-05,
|
| 553 |
+
"loss": 0.8802,
|
| 554 |
+
"step": 880
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 2.37,
|
| 558 |
+
"learning_rate": 3.1366120218579236e-05,
|
| 559 |
+
"loss": 0.8935,
|
| 560 |
+
"step": 890
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 2.39,
|
| 564 |
+
"learning_rate": 3.1256830601092896e-05,
|
| 565 |
+
"loss": 0.8759,
|
| 566 |
+
"step": 900
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 2.42,
|
| 570 |
+
"learning_rate": 3.1147540983606557e-05,
|
| 571 |
+
"loss": 0.9278,
|
| 572 |
+
"step": 910
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 2.45,
|
| 576 |
+
"learning_rate": 3.103825136612022e-05,
|
| 577 |
+
"loss": 0.8689,
|
| 578 |
+
"step": 920
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 2.47,
|
| 582 |
+
"learning_rate": 3.0928961748633883e-05,
|
| 583 |
+
"loss": 0.8675,
|
| 584 |
+
"step": 930
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 2.5,
|
| 588 |
+
"learning_rate": 3.0819672131147544e-05,
|
| 589 |
+
"loss": 0.8751,
|
| 590 |
+
"step": 940
|
| 591 |
+
},
|
| 592 |
+
{
|
| 593 |
+
"epoch": 2.53,
|
| 594 |
+
"learning_rate": 3.0710382513661204e-05,
|
| 595 |
+
"loss": 0.877,
|
| 596 |
+
"step": 950
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"epoch": 2.55,
|
| 600 |
+
"learning_rate": 3.0601092896174864e-05,
|
| 601 |
+
"loss": 0.9216,
|
| 602 |
+
"step": 960
|
| 603 |
+
},
|
| 604 |
+
{
|
| 605 |
+
"epoch": 2.58,
|
| 606 |
+
"learning_rate": 3.0491803278688527e-05,
|
| 607 |
+
"loss": 0.8636,
|
| 608 |
+
"step": 970
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"epoch": 2.61,
|
| 612 |
+
"learning_rate": 3.0382513661202187e-05,
|
| 613 |
+
"loss": 0.8748,
|
| 614 |
+
"step": 980
|
| 615 |
+
},
|
| 616 |
+
{
|
| 617 |
+
"epoch": 2.63,
|
| 618 |
+
"learning_rate": 3.0273224043715847e-05,
|
| 619 |
+
"loss": 0.8876,
|
| 620 |
+
"step": 990
|
| 621 |
+
},
|
| 622 |
+
{
|
| 623 |
+
"epoch": 2.66,
|
| 624 |
+
"learning_rate": 3.0163934426229507e-05,
|
| 625 |
+
"loss": 0.8637,
|
| 626 |
+
"step": 1000
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"epoch": 2.69,
|
| 630 |
+
"learning_rate": 3.005464480874317e-05,
|
| 631 |
+
"loss": 0.9266,
|
| 632 |
+
"step": 1010
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 2.71,
|
| 636 |
+
"learning_rate": 2.9945355191256834e-05,
|
| 637 |
+
"loss": 0.8768,
|
| 638 |
+
"step": 1020
|
| 639 |
+
},
|
| 640 |
+
{
|
| 641 |
+
"epoch": 2.74,
|
| 642 |
+
"learning_rate": 2.9836065573770498e-05,
|
| 643 |
+
"loss": 0.8833,
|
| 644 |
+
"step": 1030
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 2.77,
|
| 648 |
+
"learning_rate": 2.9726775956284158e-05,
|
| 649 |
+
"loss": 0.878,
|
| 650 |
+
"step": 1040
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 2.79,
|
| 654 |
+
"learning_rate": 2.9617486338797818e-05,
|
| 655 |
+
"loss": 0.8862,
|
| 656 |
+
"step": 1050
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 2.82,
|
| 660 |
+
"learning_rate": 2.9508196721311478e-05,
|
| 661 |
+
"loss": 0.9319,
|
| 662 |
+
"step": 1060
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 2.85,
|
| 666 |
+
"learning_rate": 2.939890710382514e-05,
|
| 667 |
+
"loss": 0.8807,
|
| 668 |
+
"step": 1070
|
| 669 |
+
},
|
| 670 |
+
{
|
| 671 |
+
"epoch": 2.87,
|
| 672 |
+
"learning_rate": 2.92896174863388e-05,
|
| 673 |
+
"loss": 0.9018,
|
| 674 |
+
"step": 1080
|
| 675 |
+
},
|
| 676 |
+
{
|
| 677 |
+
"epoch": 2.9,
|
| 678 |
+
"learning_rate": 2.918032786885246e-05,
|
| 679 |
+
"loss": 0.8827,
|
| 680 |
+
"step": 1090
|
| 681 |
+
},
|
| 682 |
+
{
|
| 683 |
+
"epoch": 2.93,
|
| 684 |
+
"learning_rate": 2.9071038251366125e-05,
|
| 685 |
+
"loss": 0.8759,
|
| 686 |
+
"step": 1100
|
| 687 |
+
},
|
| 688 |
+
{
|
| 689 |
+
"epoch": 2.95,
|
| 690 |
+
"learning_rate": 2.8961748633879785e-05,
|
| 691 |
+
"loss": 0.9162,
|
| 692 |
+
"step": 1110
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 2.98,
|
| 696 |
+
"learning_rate": 2.8852459016393445e-05,
|
| 697 |
+
"loss": 0.8682,
|
| 698 |
+
"step": 1120
|
| 699 |
+
},
|
| 700 |
+
{
|
| 701 |
+
"epoch": 3.0,
|
| 702 |
+
"eval_loss": 1.0357706546783447,
|
| 703 |
+
"eval_runtime": 341.6621,
|
| 704 |
+
"eval_samples_per_second": 7.826,
|
| 705 |
+
"eval_steps_per_second": 1.958,
|
| 706 |
+
"eval_wer": 0.6603899742434302,
|
| 707 |
+
"step": 1128
|
| 708 |
+
},
|
| 709 |
+
{
|
| 710 |
+
"epoch": 3.01,
|
| 711 |
+
"learning_rate": 2.874316939890711e-05,
|
| 712 |
+
"loss": 0.9038,
|
| 713 |
+
"step": 1130
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 3.03,
|
| 717 |
+
"learning_rate": 2.863387978142077e-05,
|
| 718 |
+
"loss": 0.9097,
|
| 719 |
+
"step": 1140
|
| 720 |
+
},
|
| 721 |
+
{
|
| 722 |
+
"epoch": 3.06,
|
| 723 |
+
"learning_rate": 2.852459016393443e-05,
|
| 724 |
+
"loss": 0.891,
|
| 725 |
+
"step": 1150
|
| 726 |
+
},
|
| 727 |
+
{
|
| 728 |
+
"epoch": 3.09,
|
| 729 |
+
"learning_rate": 2.841530054644809e-05,
|
| 730 |
+
"loss": 0.8733,
|
| 731 |
+
"step": 1160
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"epoch": 3.11,
|
| 735 |
+
"learning_rate": 2.8306010928961752e-05,
|
| 736 |
+
"loss": 0.8789,
|
| 737 |
+
"step": 1170
|
| 738 |
+
},
|
| 739 |
+
{
|
| 740 |
+
"epoch": 3.14,
|
| 741 |
+
"learning_rate": 2.8196721311475412e-05,
|
| 742 |
+
"loss": 0.8846,
|
| 743 |
+
"step": 1180
|
| 744 |
+
},
|
| 745 |
+
{
|
| 746 |
+
"epoch": 3.16,
|
| 747 |
+
"learning_rate": 2.8087431693989072e-05,
|
| 748 |
+
"loss": 0.9069,
|
| 749 |
+
"step": 1190
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 3.19,
|
| 753 |
+
"learning_rate": 2.7978142076502736e-05,
|
| 754 |
+
"loss": 0.8742,
|
| 755 |
+
"step": 1200
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 3.22,
|
| 759 |
+
"learning_rate": 2.7868852459016396e-05,
|
| 760 |
+
"loss": 0.8769,
|
| 761 |
+
"step": 1210
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 3.24,
|
| 765 |
+
"learning_rate": 2.7759562841530056e-05,
|
| 766 |
+
"loss": 0.876,
|
| 767 |
+
"step": 1220
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 3.27,
|
| 771 |
+
"learning_rate": 2.765027322404372e-05,
|
| 772 |
+
"loss": 0.8952,
|
| 773 |
+
"step": 1230
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 3.3,
|
| 777 |
+
"learning_rate": 2.754098360655738e-05,
|
| 778 |
+
"loss": 0.898,
|
| 779 |
+
"step": 1240
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"epoch": 3.32,
|
| 783 |
+
"learning_rate": 2.743169398907104e-05,
|
| 784 |
+
"loss": 0.8642,
|
| 785 |
+
"step": 1250
|
| 786 |
+
},
|
| 787 |
+
{
|
| 788 |
+
"epoch": 3.35,
|
| 789 |
+
"learning_rate": 2.73224043715847e-05,
|
| 790 |
+
"loss": 0.8772,
|
| 791 |
+
"step": 1260
|
| 792 |
+
},
|
| 793 |
+
{
|
| 794 |
+
"epoch": 3.38,
|
| 795 |
+
"learning_rate": 2.7213114754098363e-05,
|
| 796 |
+
"loss": 0.8589,
|
| 797 |
+
"step": 1270
|
| 798 |
+
},
|
| 799 |
+
{
|
| 800 |
+
"epoch": 3.4,
|
| 801 |
+
"learning_rate": 2.7103825136612023e-05,
|
| 802 |
+
"loss": 0.8754,
|
| 803 |
+
"step": 1280
|
| 804 |
+
},
|
| 805 |
+
{
|
| 806 |
+
"epoch": 3.43,
|
| 807 |
+
"learning_rate": 2.6994535519125683e-05,
|
| 808 |
+
"loss": 0.9035,
|
| 809 |
+
"step": 1290
|
| 810 |
+
},
|
| 811 |
+
{
|
| 812 |
+
"epoch": 3.46,
|
| 813 |
+
"learning_rate": 2.6885245901639346e-05,
|
| 814 |
+
"loss": 0.8784,
|
| 815 |
+
"step": 1300
|
| 816 |
+
},
|
| 817 |
+
{
|
| 818 |
+
"epoch": 3.48,
|
| 819 |
+
"learning_rate": 2.6775956284153007e-05,
|
| 820 |
+
"loss": 0.8698,
|
| 821 |
+
"step": 1310
|
| 822 |
+
},
|
| 823 |
+
{
|
| 824 |
+
"epoch": 3.51,
|
| 825 |
+
"learning_rate": 2.6666666666666667e-05,
|
| 826 |
+
"loss": 0.8671,
|
| 827 |
+
"step": 1320
|
| 828 |
+
},
|
| 829 |
+
{
|
| 830 |
+
"epoch": 3.54,
|
| 831 |
+
"learning_rate": 2.655737704918033e-05,
|
| 832 |
+
"loss": 0.8834,
|
| 833 |
+
"step": 1330
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 3.56,
|
| 837 |
+
"learning_rate": 2.644808743169399e-05,
|
| 838 |
+
"loss": 0.9126,
|
| 839 |
+
"step": 1340
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 3.59,
|
| 843 |
+
"learning_rate": 2.633879781420765e-05,
|
| 844 |
+
"loss": 0.8678,
|
| 845 |
+
"step": 1350
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 3.62,
|
| 849 |
+
"learning_rate": 2.622950819672131e-05,
|
| 850 |
+
"loss": 0.8769,
|
| 851 |
+
"step": 1360
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 3.64,
|
| 855 |
+
"learning_rate": 2.6120218579234974e-05,
|
| 856 |
+
"loss": 0.8687,
|
| 857 |
+
"step": 1370
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 3.67,
|
| 861 |
+
"learning_rate": 2.6010928961748634e-05,
|
| 862 |
+
"loss": 0.8817,
|
| 863 |
+
"step": 1380
|
| 864 |
+
},
|
| 865 |
+
{
|
| 866 |
+
"epoch": 3.7,
|
| 867 |
+
"learning_rate": 2.59016393442623e-05,
|
| 868 |
+
"loss": 0.9162,
|
| 869 |
+
"step": 1390
|
| 870 |
+
},
|
| 871 |
+
{
|
| 872 |
+
"epoch": 3.72,
|
| 873 |
+
"learning_rate": 2.579234972677596e-05,
|
| 874 |
+
"loss": 0.8587,
|
| 875 |
+
"step": 1400
|
| 876 |
+
},
|
| 877 |
+
{
|
| 878 |
+
"epoch": 3.75,
|
| 879 |
+
"learning_rate": 2.568306010928962e-05,
|
| 880 |
+
"loss": 0.8572,
|
| 881 |
+
"step": 1410
|
| 882 |
+
},
|
| 883 |
+
{
|
| 884 |
+
"epoch": 3.78,
|
| 885 |
+
"learning_rate": 2.5573770491803284e-05,
|
| 886 |
+
"loss": 0.8629,
|
| 887 |
+
"step": 1420
|
| 888 |
+
},
|
| 889 |
+
{
|
| 890 |
+
"epoch": 3.8,
|
| 891 |
+
"learning_rate": 2.5464480874316944e-05,
|
| 892 |
+
"loss": 0.8755,
|
| 893 |
+
"step": 1430
|
| 894 |
+
},
|
| 895 |
+
{
|
| 896 |
+
"epoch": 3.83,
|
| 897 |
+
"learning_rate": 2.5355191256830604e-05,
|
| 898 |
+
"loss": 0.9048,
|
| 899 |
+
"step": 1440
|
| 900 |
+
},
|
| 901 |
+
{
|
| 902 |
+
"epoch": 3.86,
|
| 903 |
+
"learning_rate": 2.5245901639344264e-05,
|
| 904 |
+
"loss": 0.8635,
|
| 905 |
+
"step": 1450
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 3.88,
|
| 909 |
+
"learning_rate": 2.5136612021857928e-05,
|
| 910 |
+
"loss": 0.8618,
|
| 911 |
+
"step": 1460
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 3.91,
|
| 915 |
+
"learning_rate": 2.5027322404371588e-05,
|
| 916 |
+
"loss": 0.8637,
|
| 917 |
+
"step": 1470
|
| 918 |
+
},
|
| 919 |
+
{
|
| 920 |
+
"epoch": 3.94,
|
| 921 |
+
"learning_rate": 2.4918032786885248e-05,
|
| 922 |
+
"loss": 0.8885,
|
| 923 |
+
"step": 1480
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"epoch": 3.96,
|
| 927 |
+
"learning_rate": 2.480874316939891e-05,
|
| 928 |
+
"loss": 0.8829,
|
| 929 |
+
"step": 1490
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 3.99,
|
| 933 |
+
"learning_rate": 2.469945355191257e-05,
|
| 934 |
+
"loss": 0.8662,
|
| 935 |
+
"step": 1500
|
| 936 |
+
},
|
| 937 |
+
{
|
| 938 |
+
"epoch": 4.0,
|
| 939 |
+
"eval_loss": 1.0233858823776245,
|
| 940 |
+
"eval_runtime": 343.724,
|
| 941 |
+
"eval_samples_per_second": 7.779,
|
| 942 |
+
"eval_steps_per_second": 1.946,
|
| 943 |
+
"eval_wer": 0.6664547753245469,
|
| 944 |
+
"step": 1504
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 4.02,
|
| 948 |
+
"learning_rate": 2.459016393442623e-05,
|
| 949 |
+
"loss": 0.9061,
|
| 950 |
+
"step": 1510
|
| 951 |
+
},
|
| 952 |
+
{
|
| 953 |
+
"epoch": 4.04,
|
| 954 |
+
"learning_rate": 2.448087431693989e-05,
|
| 955 |
+
"loss": 0.8684,
|
| 956 |
+
"step": 1520
|
| 957 |
+
},
|
| 958 |
+
{
|
| 959 |
+
"epoch": 4.07,
|
| 960 |
+
"learning_rate": 2.4371584699453555e-05,
|
| 961 |
+
"loss": 0.8616,
|
| 962 |
+
"step": 1530
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 4.1,
|
| 966 |
+
"learning_rate": 2.4262295081967215e-05,
|
| 967 |
+
"loss": 0.8718,
|
| 968 |
+
"step": 1540
|
| 969 |
+
},
|
| 970 |
+
{
|
| 971 |
+
"epoch": 4.12,
|
| 972 |
+
"learning_rate": 2.4153005464480875e-05,
|
| 973 |
+
"loss": 0.8663,
|
| 974 |
+
"step": 1550
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 4.15,
|
| 978 |
+
"learning_rate": 2.404371584699454e-05,
|
| 979 |
+
"loss": 0.9004,
|
| 980 |
+
"step": 1560
|
| 981 |
+
},
|
| 982 |
+
{
|
| 983 |
+
"epoch": 4.18,
|
| 984 |
+
"learning_rate": 2.39344262295082e-05,
|
| 985 |
+
"loss": 0.8659,
|
| 986 |
+
"step": 1570
|
| 987 |
+
},
|
| 988 |
+
{
|
| 989 |
+
"epoch": 4.2,
|
| 990 |
+
"learning_rate": 2.382513661202186e-05,
|
| 991 |
+
"loss": 0.8569,
|
| 992 |
+
"step": 1580
|
| 993 |
+
},
|
| 994 |
+
{
|
| 995 |
+
"epoch": 4.23,
|
| 996 |
+
"learning_rate": 2.3715846994535522e-05,
|
| 997 |
+
"loss": 0.8574,
|
| 998 |
+
"step": 1590
|
| 999 |
+
},
|
| 1000 |
+
{
|
| 1001 |
+
"epoch": 4.26,
|
| 1002 |
+
"learning_rate": 2.3606557377049182e-05,
|
| 1003 |
+
"loss": 0.8702,
|
| 1004 |
+
"step": 1600
|
| 1005 |
+
},
|
| 1006 |
+
{
|
| 1007 |
+
"epoch": 4.28,
|
| 1008 |
+
"learning_rate": 2.3497267759562842e-05,
|
| 1009 |
+
"loss": 0.9054,
|
| 1010 |
+
"step": 1610
|
| 1011 |
+
},
|
| 1012 |
+
{
|
| 1013 |
+
"epoch": 4.31,
|
| 1014 |
+
"learning_rate": 2.3387978142076502e-05,
|
| 1015 |
+
"loss": 0.8777,
|
| 1016 |
+
"step": 1620
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 4.34,
|
| 1020 |
+
"learning_rate": 2.3278688524590166e-05,
|
| 1021 |
+
"loss": 0.8657,
|
| 1022 |
+
"step": 1630
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 4.36,
|
| 1026 |
+
"learning_rate": 2.3169398907103826e-05,
|
| 1027 |
+
"loss": 0.8712,
|
| 1028 |
+
"step": 1640
|
| 1029 |
+
},
|
| 1030 |
+
{
|
| 1031 |
+
"epoch": 4.39,
|
| 1032 |
+
"learning_rate": 2.3060109289617486e-05,
|
| 1033 |
+
"loss": 0.8578,
|
| 1034 |
+
"step": 1650
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 4.41,
|
| 1038 |
+
"learning_rate": 2.295081967213115e-05,
|
| 1039 |
+
"loss": 0.8931,
|
| 1040 |
+
"step": 1660
|
| 1041 |
+
},
|
| 1042 |
+
{
|
| 1043 |
+
"epoch": 4.44,
|
| 1044 |
+
"learning_rate": 2.284153005464481e-05,
|
| 1045 |
+
"loss": 0.8734,
|
| 1046 |
+
"step": 1670
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 4.47,
|
| 1050 |
+
"learning_rate": 2.273224043715847e-05,
|
| 1051 |
+
"loss": 0.8715,
|
| 1052 |
+
"step": 1680
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 4.49,
|
| 1056 |
+
"learning_rate": 2.2622950819672133e-05,
|
| 1057 |
+
"loss": 0.8578,
|
| 1058 |
+
"step": 1690
|
| 1059 |
+
},
|
| 1060 |
+
{
|
| 1061 |
+
"epoch": 4.52,
|
| 1062 |
+
"learning_rate": 2.2513661202185793e-05,
|
| 1063 |
+
"loss": 0.8612,
|
| 1064 |
+
"step": 1700
|
| 1065 |
+
},
|
| 1066 |
+
{
|
| 1067 |
+
"epoch": 4.55,
|
| 1068 |
+
"learning_rate": 2.2404371584699453e-05,
|
| 1069 |
+
"loss": 0.8929,
|
| 1070 |
+
"step": 1710
|
| 1071 |
+
},
|
| 1072 |
+
{
|
| 1073 |
+
"epoch": 4.57,
|
| 1074 |
+
"learning_rate": 2.2295081967213113e-05,
|
| 1075 |
+
"loss": 0.8684,
|
| 1076 |
+
"step": 1720
|
| 1077 |
+
},
|
| 1078 |
+
{
|
| 1079 |
+
"epoch": 4.6,
|
| 1080 |
+
"learning_rate": 2.2185792349726777e-05,
|
| 1081 |
+
"loss": 0.87,
|
| 1082 |
+
"step": 1730
|
| 1083 |
+
},
|
| 1084 |
+
{
|
| 1085 |
+
"epoch": 4.63,
|
| 1086 |
+
"learning_rate": 2.2076502732240437e-05,
|
| 1087 |
+
"loss": 0.8621,
|
| 1088 |
+
"step": 1740
|
| 1089 |
+
},
|
| 1090 |
+
{
|
| 1091 |
+
"epoch": 4.65,
|
| 1092 |
+
"learning_rate": 2.1967213114754104e-05,
|
| 1093 |
+
"loss": 0.8569,
|
| 1094 |
+
"step": 1750
|
| 1095 |
+
},
|
| 1096 |
+
{
|
| 1097 |
+
"epoch": 4.68,
|
| 1098 |
+
"learning_rate": 2.1857923497267764e-05,
|
| 1099 |
+
"loss": 0.9089,
|
| 1100 |
+
"step": 1760
|
| 1101 |
+
},
|
| 1102 |
+
{
|
| 1103 |
+
"epoch": 4.71,
|
| 1104 |
+
"learning_rate": 2.1748633879781424e-05,
|
| 1105 |
+
"loss": 0.8664,
|
| 1106 |
+
"step": 1770
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 4.73,
|
| 1110 |
+
"learning_rate": 2.1639344262295087e-05,
|
| 1111 |
+
"loss": 0.872,
|
| 1112 |
+
"step": 1780
|
| 1113 |
+
},
|
| 1114 |
+
{
|
| 1115 |
+
"epoch": 4.76,
|
| 1116 |
+
"learning_rate": 2.1530054644808747e-05,
|
| 1117 |
+
"loss": 0.8537,
|
| 1118 |
+
"step": 1790
|
| 1119 |
+
},
|
| 1120 |
+
{
|
| 1121 |
+
"epoch": 4.79,
|
| 1122 |
+
"learning_rate": 2.1420765027322407e-05,
|
| 1123 |
+
"loss": 0.8592,
|
| 1124 |
+
"step": 1800
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 4.81,
|
| 1128 |
+
"learning_rate": 2.1311475409836067e-05,
|
| 1129 |
+
"loss": 0.9015,
|
| 1130 |
+
"step": 1810
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"epoch": 4.84,
|
| 1134 |
+
"learning_rate": 2.120218579234973e-05,
|
| 1135 |
+
"loss": 0.8788,
|
| 1136 |
+
"step": 1820
|
| 1137 |
+
},
|
| 1138 |
+
{
|
| 1139 |
+
"epoch": 4.87,
|
| 1140 |
+
"learning_rate": 2.109289617486339e-05,
|
| 1141 |
+
"loss": 0.8552,
|
| 1142 |
+
"step": 1830
|
| 1143 |
+
},
|
| 1144 |
+
{
|
| 1145 |
+
"epoch": 4.89,
|
| 1146 |
+
"learning_rate": 2.098360655737705e-05,
|
| 1147 |
+
"loss": 0.8587,
|
| 1148 |
+
"step": 1840
|
| 1149 |
+
},
|
| 1150 |
+
{
|
| 1151 |
+
"epoch": 4.92,
|
| 1152 |
+
"learning_rate": 2.0874316939890714e-05,
|
| 1153 |
+
"loss": 0.8552,
|
| 1154 |
+
"step": 1850
|
| 1155 |
+
},
|
| 1156 |
+
{
|
| 1157 |
+
"epoch": 4.95,
|
| 1158 |
+
"learning_rate": 2.0765027322404374e-05,
|
| 1159 |
+
"loss": 0.8833,
|
| 1160 |
+
"step": 1860
|
| 1161 |
+
},
|
| 1162 |
+
{
|
| 1163 |
+
"epoch": 4.97,
|
| 1164 |
+
"learning_rate": 2.0655737704918034e-05,
|
| 1165 |
+
"loss": 0.8501,
|
| 1166 |
+
"step": 1870
|
| 1167 |
+
},
|
| 1168 |
+
{
|
| 1169 |
+
"epoch": 5.0,
|
| 1170 |
+
"learning_rate": 2.0546448087431698e-05,
|
| 1171 |
+
"loss": 0.8463,
|
| 1172 |
+
"step": 1880
|
| 1173 |
+
},
|
| 1174 |
+
{
|
| 1175 |
+
"epoch": 5.0,
|
| 1176 |
+
"eval_loss": 1.0332772731781006,
|
| 1177 |
+
"eval_runtime": 349.3455,
|
| 1178 |
+
"eval_samples_per_second": 7.654,
|
| 1179 |
+
"eval_steps_per_second": 1.915,
|
| 1180 |
+
"eval_wer": 0.6666054536122765,
|
| 1181 |
+
"step": 1880
|
| 1182 |
+
},
|
| 1183 |
+
{
|
| 1184 |
+
"epoch": 5.03,
|
| 1185 |
+
"learning_rate": 2.0437158469945358e-05,
|
| 1186 |
+
"loss": 0.9174,
|
| 1187 |
+
"step": 1890
|
| 1188 |
+
},
|
| 1189 |
+
{
|
| 1190 |
+
"epoch": 5.05,
|
| 1191 |
+
"learning_rate": 2.0327868852459018e-05,
|
| 1192 |
+
"loss": 0.8478,
|
| 1193 |
+
"step": 1900
|
| 1194 |
+
},
|
| 1195 |
+
{
|
| 1196 |
+
"epoch": 5.08,
|
| 1197 |
+
"learning_rate": 2.0218579234972678e-05,
|
| 1198 |
+
"loss": 0.8535,
|
| 1199 |
+
"step": 1910
|
| 1200 |
+
},
|
| 1201 |
+
{
|
| 1202 |
+
"epoch": 5.11,
|
| 1203 |
+
"learning_rate": 2.010928961748634e-05,
|
| 1204 |
+
"loss": 0.8651,
|
| 1205 |
+
"step": 1920
|
| 1206 |
+
},
|
| 1207 |
+
{
|
| 1208 |
+
"epoch": 5.13,
|
| 1209 |
+
"learning_rate": 2e-05,
|
| 1210 |
+
"loss": 0.8535,
|
| 1211 |
+
"step": 1930
|
| 1212 |
+
},
|
| 1213 |
+
{
|
| 1214 |
+
"epoch": 5.16,
|
| 1215 |
+
"learning_rate": 1.9890710382513662e-05,
|
| 1216 |
+
"loss": 0.9069,
|
| 1217 |
+
"step": 1940
|
| 1218 |
+
},
|
| 1219 |
+
{
|
| 1220 |
+
"epoch": 5.19,
|
| 1221 |
+
"learning_rate": 1.9781420765027325e-05,
|
| 1222 |
+
"loss": 0.8565,
|
| 1223 |
+
"step": 1950
|
| 1224 |
+
},
|
| 1225 |
+
{
|
| 1226 |
+
"epoch": 5.21,
|
| 1227 |
+
"learning_rate": 1.9672131147540985e-05,
|
| 1228 |
+
"loss": 0.8713,
|
| 1229 |
+
"step": 1960
|
| 1230 |
+
},
|
| 1231 |
+
{
|
| 1232 |
+
"epoch": 5.24,
|
| 1233 |
+
"learning_rate": 1.9562841530054645e-05,
|
| 1234 |
+
"loss": 0.8545,
|
| 1235 |
+
"step": 1970
|
| 1236 |
+
},
|
| 1237 |
+
{
|
| 1238 |
+
"epoch": 5.27,
|
| 1239 |
+
"learning_rate": 1.9453551912568305e-05,
|
| 1240 |
+
"loss": 0.8637,
|
| 1241 |
+
"step": 1980
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 5.29,
|
| 1245 |
+
"learning_rate": 1.934426229508197e-05,
|
| 1246 |
+
"loss": 0.9046,
|
| 1247 |
+
"step": 1990
|
| 1248 |
+
},
|
| 1249 |
+
{
|
| 1250 |
+
"epoch": 5.32,
|
| 1251 |
+
"learning_rate": 1.923497267759563e-05,
|
| 1252 |
+
"loss": 0.8542,
|
| 1253 |
+
"step": 2000
|
| 1254 |
+
},
|
| 1255 |
+
{
|
| 1256 |
+
"epoch": 5.35,
|
| 1257 |
+
"learning_rate": 1.912568306010929e-05,
|
| 1258 |
+
"loss": 0.8442,
|
| 1259 |
+
"step": 2010
|
| 1260 |
+
},
|
| 1261 |
+
{
|
| 1262 |
+
"epoch": 5.37,
|
| 1263 |
+
"learning_rate": 1.9027322404371588e-05,
|
| 1264 |
+
"loss": 0.8502,
|
| 1265 |
+
"step": 2020
|
| 1266 |
+
},
|
| 1267 |
+
{
|
| 1268 |
+
"epoch": 5.4,
|
| 1269 |
+
"learning_rate": 1.8918032786885248e-05,
|
| 1270 |
+
"loss": 0.8663,
|
| 1271 |
+
"step": 2030
|
| 1272 |
+
},
|
| 1273 |
+
{
|
| 1274 |
+
"epoch": 5.43,
|
| 1275 |
+
"learning_rate": 1.8808743169398908e-05,
|
| 1276 |
+
"loss": 0.8946,
|
| 1277 |
+
"step": 2040
|
| 1278 |
+
},
|
| 1279 |
+
{
|
| 1280 |
+
"epoch": 5.45,
|
| 1281 |
+
"learning_rate": 1.869945355191257e-05,
|
| 1282 |
+
"loss": 0.8449,
|
| 1283 |
+
"step": 2050
|
| 1284 |
+
},
|
| 1285 |
+
{
|
| 1286 |
+
"epoch": 5.48,
|
| 1287 |
+
"learning_rate": 1.859016393442623e-05,
|
| 1288 |
+
"loss": 0.8443,
|
| 1289 |
+
"step": 2060
|
| 1290 |
+
},
|
| 1291 |
+
{
|
| 1292 |
+
"epoch": 5.51,
|
| 1293 |
+
"learning_rate": 1.848087431693989e-05,
|
| 1294 |
+
"loss": 0.8446,
|
| 1295 |
+
"step": 2070
|
| 1296 |
+
},
|
| 1297 |
+
{
|
| 1298 |
+
"epoch": 5.53,
|
| 1299 |
+
"learning_rate": 1.837158469945355e-05,
|
| 1300 |
+
"loss": 0.859,
|
| 1301 |
+
"step": 2080
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 5.56,
|
| 1305 |
+
"learning_rate": 1.8262295081967215e-05,
|
| 1306 |
+
"loss": 0.9006,
|
| 1307 |
+
"step": 2090
|
| 1308 |
+
},
|
| 1309 |
+
{
|
| 1310 |
+
"epoch": 5.59,
|
| 1311 |
+
"learning_rate": 1.8153005464480875e-05,
|
| 1312 |
+
"loss": 0.8588,
|
| 1313 |
+
"step": 2100
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 5.61,
|
| 1317 |
+
"learning_rate": 1.8043715846994535e-05,
|
| 1318 |
+
"loss": 0.8568,
|
| 1319 |
+
"step": 2110
|
| 1320 |
+
},
|
| 1321 |
+
{
|
| 1322 |
+
"epoch": 5.64,
|
| 1323 |
+
"learning_rate": 1.79344262295082e-05,
|
| 1324 |
+
"loss": 0.8614,
|
| 1325 |
+
"step": 2120
|
| 1326 |
+
},
|
| 1327 |
+
{
|
| 1328 |
+
"epoch": 5.66,
|
| 1329 |
+
"learning_rate": 1.782513661202186e-05,
|
| 1330 |
+
"loss": 0.8447,
|
| 1331 |
+
"step": 2130
|
| 1332 |
+
},
|
| 1333 |
+
{
|
| 1334 |
+
"epoch": 5.69,
|
| 1335 |
+
"learning_rate": 1.7715846994535522e-05,
|
| 1336 |
+
"loss": 0.9098,
|
| 1337 |
+
"step": 2140
|
| 1338 |
+
},
|
| 1339 |
+
{
|
| 1340 |
+
"epoch": 5.72,
|
| 1341 |
+
"learning_rate": 1.7606557377049182e-05,
|
| 1342 |
+
"loss": 0.8408,
|
| 1343 |
+
"step": 2150
|
| 1344 |
+
},
|
| 1345 |
+
{
|
| 1346 |
+
"epoch": 5.74,
|
| 1347 |
+
"learning_rate": 1.7497267759562842e-05,
|
| 1348 |
+
"loss": 0.8611,
|
| 1349 |
+
"step": 2160
|
| 1350 |
+
},
|
| 1351 |
+
{
|
| 1352 |
+
"epoch": 5.77,
|
| 1353 |
+
"learning_rate": 1.7387978142076506e-05,
|
| 1354 |
+
"loss": 0.8524,
|
| 1355 |
+
"step": 2170
|
| 1356 |
+
},
|
| 1357 |
+
{
|
| 1358 |
+
"epoch": 5.8,
|
| 1359 |
+
"learning_rate": 1.7278688524590166e-05,
|
| 1360 |
+
"loss": 0.859,
|
| 1361 |
+
"step": 2180
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"epoch": 5.82,
|
| 1365 |
+
"learning_rate": 1.7169398907103826e-05,
|
| 1366 |
+
"loss": 0.9099,
|
| 1367 |
+
"step": 2190
|
| 1368 |
+
},
|
| 1369 |
+
{
|
| 1370 |
+
"epoch": 5.85,
|
| 1371 |
+
"learning_rate": 1.706010928961749e-05,
|
| 1372 |
+
"loss": 0.8534,
|
| 1373 |
+
"step": 2200
|
| 1374 |
+
},
|
| 1375 |
+
{
|
| 1376 |
+
"epoch": 5.88,
|
| 1377 |
+
"learning_rate": 1.695081967213115e-05,
|
| 1378 |
+
"loss": 0.8525,
|
| 1379 |
+
"step": 2210
|
| 1380 |
+
},
|
| 1381 |
+
{
|
| 1382 |
+
"epoch": 5.9,
|
| 1383 |
+
"learning_rate": 1.684153005464481e-05,
|
| 1384 |
+
"loss": 0.8426,
|
| 1385 |
+
"step": 2220
|
| 1386 |
+
},
|
| 1387 |
+
{
|
| 1388 |
+
"epoch": 5.93,
|
| 1389 |
+
"learning_rate": 1.6732240437158473e-05,
|
| 1390 |
+
"loss": 0.8708,
|
| 1391 |
+
"step": 2230
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"epoch": 5.96,
|
| 1395 |
+
"learning_rate": 1.6622950819672133e-05,
|
| 1396 |
+
"loss": 0.8648,
|
| 1397 |
+
"step": 2240
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 5.98,
|
| 1401 |
+
"learning_rate": 1.6513661202185793e-05,
|
| 1402 |
+
"loss": 0.8573,
|
| 1403 |
+
"step": 2250
|
| 1404 |
+
},
|
| 1405 |
+
{
|
| 1406 |
+
"epoch": 6.0,
|
| 1407 |
+
"eval_loss": 1.0162557363510132,
|
| 1408 |
+
"eval_runtime": 338.1768,
|
| 1409 |
+
"eval_samples_per_second": 7.907,
|
| 1410 |
+
"eval_steps_per_second": 1.978,
|
| 1411 |
+
"eval_wer": 0.6621886963032024,
|
| 1412 |
+
"step": 2256
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 6.01,
|
| 1416 |
+
"learning_rate": 1.6404371584699453e-05,
|
| 1417 |
+
"loss": 0.8838,
|
| 1418 |
+
"step": 2260
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 6.04,
|
| 1422 |
+
"learning_rate": 1.6295081967213116e-05,
|
| 1423 |
+
"loss": 0.8716,
|
| 1424 |
+
"step": 2270
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 6.06,
|
| 1428 |
+
"learning_rate": 1.6185792349726776e-05,
|
| 1429 |
+
"loss": 0.8522,
|
| 1430 |
+
"step": 2280
|
| 1431 |
+
},
|
| 1432 |
+
{
|
| 1433 |
+
"epoch": 6.09,
|
| 1434 |
+
"learning_rate": 1.6076502732240437e-05,
|
| 1435 |
+
"loss": 0.8475,
|
| 1436 |
+
"step": 2290
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 6.12,
|
| 1440 |
+
"learning_rate": 1.59672131147541e-05,
|
| 1441 |
+
"loss": 0.8533,
|
| 1442 |
+
"step": 2300
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 6.14,
|
| 1446 |
+
"learning_rate": 1.585792349726776e-05,
|
| 1447 |
+
"loss": 0.8867,
|
| 1448 |
+
"step": 2310
|
| 1449 |
+
},
|
| 1450 |
+
{
|
| 1451 |
+
"epoch": 6.17,
|
| 1452 |
+
"learning_rate": 1.5748633879781424e-05,
|
| 1453 |
+
"loss": 0.8741,
|
| 1454 |
+
"step": 2320
|
| 1455 |
+
},
|
| 1456 |
+
{
|
| 1457 |
+
"epoch": 6.2,
|
| 1458 |
+
"learning_rate": 1.5639344262295084e-05,
|
| 1459 |
+
"loss": 0.8564,
|
| 1460 |
+
"step": 2330
|
| 1461 |
+
},
|
| 1462 |
+
{
|
| 1463 |
+
"epoch": 6.22,
|
| 1464 |
+
"learning_rate": 1.5530054644808744e-05,
|
| 1465 |
+
"loss": 0.8445,
|
| 1466 |
+
"step": 2340
|
| 1467 |
+
},
|
| 1468 |
+
{
|
| 1469 |
+
"epoch": 6.25,
|
| 1470 |
+
"learning_rate": 1.5420765027322407e-05,
|
| 1471 |
+
"loss": 0.8345,
|
| 1472 |
+
"step": 2350
|
| 1473 |
+
},
|
| 1474 |
+
{
|
| 1475 |
+
"epoch": 6.28,
|
| 1476 |
+
"learning_rate": 1.5311475409836067e-05,
|
| 1477 |
+
"loss": 0.8805,
|
| 1478 |
+
"step": 2360
|
| 1479 |
+
},
|
| 1480 |
+
{
|
| 1481 |
+
"epoch": 6.3,
|
| 1482 |
+
"learning_rate": 1.5202185792349729e-05,
|
| 1483 |
+
"loss": 0.8609,
|
| 1484 |
+
"step": 2370
|
| 1485 |
+
},
|
| 1486 |
+
{
|
| 1487 |
+
"epoch": 6.33,
|
| 1488 |
+
"learning_rate": 1.5092896174863389e-05,
|
| 1489 |
+
"loss": 0.8438,
|
| 1490 |
+
"step": 2380
|
| 1491 |
+
},
|
| 1492 |
+
{
|
| 1493 |
+
"epoch": 6.36,
|
| 1494 |
+
"learning_rate": 1.498360655737705e-05,
|
| 1495 |
+
"loss": 0.8471,
|
| 1496 |
+
"step": 2390
|
| 1497 |
+
},
|
| 1498 |
+
{
|
| 1499 |
+
"epoch": 6.38,
|
| 1500 |
+
"learning_rate": 1.4874316939890713e-05,
|
| 1501 |
+
"loss": 0.8487,
|
| 1502 |
+
"step": 2400
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 6.41,
|
| 1506 |
+
"learning_rate": 1.4765027322404373e-05,
|
| 1507 |
+
"loss": 0.8833,
|
| 1508 |
+
"step": 2410
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"epoch": 6.44,
|
| 1512 |
+
"learning_rate": 1.4655737704918034e-05,
|
| 1513 |
+
"loss": 0.8878,
|
| 1514 |
+
"step": 2420
|
| 1515 |
+
},
|
| 1516 |
+
{
|
| 1517 |
+
"epoch": 6.46,
|
| 1518 |
+
"learning_rate": 1.4546448087431694e-05,
|
| 1519 |
+
"loss": 0.8349,
|
| 1520 |
+
"step": 2430
|
| 1521 |
+
},
|
| 1522 |
+
{
|
| 1523 |
+
"epoch": 6.49,
|
| 1524 |
+
"learning_rate": 1.4437158469945356e-05,
|
| 1525 |
+
"loss": 0.8425,
|
| 1526 |
+
"step": 2440
|
| 1527 |
+
},
|
| 1528 |
+
{
|
| 1529 |
+
"epoch": 6.52,
|
| 1530 |
+
"learning_rate": 1.4327868852459016e-05,
|
| 1531 |
+
"loss": 0.8523,
|
| 1532 |
+
"step": 2450
|
| 1533 |
+
},
|
| 1534 |
+
{
|
| 1535 |
+
"epoch": 6.54,
|
| 1536 |
+
"learning_rate": 1.4218579234972678e-05,
|
| 1537 |
+
"loss": 0.879,
|
| 1538 |
+
"step": 2460
|
| 1539 |
+
},
|
| 1540 |
+
{
|
| 1541 |
+
"epoch": 6.57,
|
| 1542 |
+
"learning_rate": 1.410928961748634e-05,
|
| 1543 |
+
"loss": 0.865,
|
| 1544 |
+
"step": 2470
|
| 1545 |
+
},
|
| 1546 |
+
{
|
| 1547 |
+
"epoch": 6.6,
|
| 1548 |
+
"learning_rate": 1.4e-05,
|
| 1549 |
+
"loss": 0.8514,
|
| 1550 |
+
"step": 2480
|
| 1551 |
+
},
|
| 1552 |
+
{
|
| 1553 |
+
"epoch": 6.62,
|
| 1554 |
+
"learning_rate": 1.3890710382513662e-05,
|
| 1555 |
+
"loss": 0.8625,
|
| 1556 |
+
"step": 2490
|
| 1557 |
+
},
|
| 1558 |
+
{
|
| 1559 |
+
"epoch": 6.65,
|
| 1560 |
+
"learning_rate": 1.3781420765027325e-05,
|
| 1561 |
+
"loss": 0.8543,
|
| 1562 |
+
"step": 2500
|
| 1563 |
+
},
|
| 1564 |
+
{
|
| 1565 |
+
"epoch": 6.68,
|
| 1566 |
+
"learning_rate": 1.3672131147540985e-05,
|
| 1567 |
+
"loss": 0.8793,
|
| 1568 |
+
"step": 2510
|
| 1569 |
+
},
|
| 1570 |
+
{
|
| 1571 |
+
"epoch": 6.7,
|
| 1572 |
+
"learning_rate": 1.3562841530054647e-05,
|
| 1573 |
+
"loss": 0.8603,
|
| 1574 |
+
"step": 2520
|
| 1575 |
+
},
|
| 1576 |
+
{
|
| 1577 |
+
"epoch": 6.73,
|
| 1578 |
+
"learning_rate": 1.3453551912568309e-05,
|
| 1579 |
+
"loss": 0.8584,
|
| 1580 |
+
"step": 2530
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 6.76,
|
| 1584 |
+
"learning_rate": 1.3344262295081969e-05,
|
| 1585 |
+
"loss": 0.8384,
|
| 1586 |
+
"step": 2540
|
| 1587 |
+
},
|
| 1588 |
+
{
|
| 1589 |
+
"epoch": 6.78,
|
| 1590 |
+
"learning_rate": 1.323497267759563e-05,
|
| 1591 |
+
"loss": 0.8335,
|
| 1592 |
+
"step": 2550
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 6.81,
|
| 1596 |
+
"learning_rate": 1.312568306010929e-05,
|
| 1597 |
+
"loss": 0.8594,
|
| 1598 |
+
"step": 2560
|
| 1599 |
+
},
|
| 1600 |
+
{
|
| 1601 |
+
"epoch": 6.84,
|
| 1602 |
+
"learning_rate": 1.3016393442622952e-05,
|
| 1603 |
+
"loss": 0.8789,
|
| 1604 |
+
"step": 2570
|
| 1605 |
+
},
|
| 1606 |
+
{
|
| 1607 |
+
"epoch": 6.86,
|
| 1608 |
+
"learning_rate": 1.2907103825136614e-05,
|
| 1609 |
+
"loss": 0.8446,
|
| 1610 |
+
"step": 2580
|
| 1611 |
+
},
|
| 1612 |
+
{
|
| 1613 |
+
"epoch": 6.89,
|
| 1614 |
+
"learning_rate": 1.2797814207650274e-05,
|
| 1615 |
+
"loss": 0.8577,
|
| 1616 |
+
"step": 2590
|
| 1617 |
+
},
|
| 1618 |
+
{
|
| 1619 |
+
"epoch": 6.91,
|
| 1620 |
+
"learning_rate": 1.2688524590163936e-05,
|
| 1621 |
+
"loss": 0.8502,
|
| 1622 |
+
"step": 2600
|
| 1623 |
+
},
|
| 1624 |
+
{
|
| 1625 |
+
"epoch": 6.94,
|
| 1626 |
+
"learning_rate": 1.2579234972677596e-05,
|
| 1627 |
+
"loss": 0.8872,
|
| 1628 |
+
"step": 2610
|
| 1629 |
+
},
|
| 1630 |
+
{
|
| 1631 |
+
"epoch": 6.97,
|
| 1632 |
+
"learning_rate": 1.2469945355191258e-05,
|
| 1633 |
+
"loss": 0.8502,
|
| 1634 |
+
"step": 2620
|
| 1635 |
+
},
|
| 1636 |
+
{
|
| 1637 |
+
"epoch": 6.99,
|
| 1638 |
+
"learning_rate": 1.236065573770492e-05,
|
| 1639 |
+
"loss": 0.8628,
|
| 1640 |
+
"step": 2630
|
| 1641 |
+
},
|
| 1642 |
+
{
|
| 1643 |
+
"epoch": 7.0,
|
| 1644 |
+
"eval_loss": 1.0209393501281738,
|
| 1645 |
+
"eval_runtime": 339.6154,
|
| 1646 |
+
"eval_samples_per_second": 7.874,
|
| 1647 |
+
"eval_steps_per_second": 1.97,
|
| 1648 |
+
"eval_wer": 0.6550550211185038,
|
| 1649 |
+
"step": 2632
|
| 1650 |
+
},
|
| 1651 |
+
{
|
| 1652 |
+
"epoch": 7.02,
|
| 1653 |
+
"learning_rate": 1.225136612021858e-05,
|
| 1654 |
+
"loss": 0.9001,
|
| 1655 |
+
"step": 2640
|
| 1656 |
+
},
|
| 1657 |
+
{
|
| 1658 |
+
"epoch": 7.05,
|
| 1659 |
+
"learning_rate": 1.2142076502732241e-05,
|
| 1660 |
+
"loss": 0.8424,
|
| 1661 |
+
"step": 2650
|
| 1662 |
+
},
|
| 1663 |
+
{
|
| 1664 |
+
"epoch": 7.07,
|
| 1665 |
+
"learning_rate": 1.2032786885245901e-05,
|
| 1666 |
+
"loss": 0.8444,
|
| 1667 |
+
"step": 2660
|
| 1668 |
+
},
|
| 1669 |
+
{
|
| 1670 |
+
"epoch": 7.1,
|
| 1671 |
+
"learning_rate": 1.1923497267759563e-05,
|
| 1672 |
+
"loss": 0.8413,
|
| 1673 |
+
"step": 2670
|
| 1674 |
+
},
|
| 1675 |
+
{
|
| 1676 |
+
"epoch": 7.13,
|
| 1677 |
+
"learning_rate": 1.1814207650273223e-05,
|
| 1678 |
+
"loss": 0.8429,
|
| 1679 |
+
"step": 2680
|
| 1680 |
+
},
|
| 1681 |
+
{
|
| 1682 |
+
"epoch": 7.15,
|
| 1683 |
+
"learning_rate": 1.1704918032786887e-05,
|
| 1684 |
+
"loss": 0.904,
|
| 1685 |
+
"step": 2690
|
| 1686 |
+
},
|
| 1687 |
+
{
|
| 1688 |
+
"epoch": 7.18,
|
| 1689 |
+
"learning_rate": 1.1595628415300548e-05,
|
| 1690 |
+
"loss": 0.8378,
|
| 1691 |
+
"step": 2700
|
| 1692 |
+
},
|
| 1693 |
+
{
|
| 1694 |
+
"epoch": 7.21,
|
| 1695 |
+
"learning_rate": 1.148633879781421e-05,
|
| 1696 |
+
"loss": 0.8401,
|
| 1697 |
+
"step": 2710
|
| 1698 |
+
},
|
| 1699 |
+
{
|
| 1700 |
+
"epoch": 7.23,
|
| 1701 |
+
"learning_rate": 1.137704918032787e-05,
|
| 1702 |
+
"loss": 0.8533,
|
| 1703 |
+
"step": 2720
|
| 1704 |
+
},
|
| 1705 |
+
{
|
| 1706 |
+
"epoch": 7.26,
|
| 1707 |
+
"learning_rate": 1.1267759562841532e-05,
|
| 1708 |
+
"loss": 0.8341,
|
| 1709 |
+
"step": 2730
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 7.29,
|
| 1713 |
+
"learning_rate": 1.1158469945355192e-05,
|
| 1714 |
+
"loss": 0.8987,
|
| 1715 |
+
"step": 2740
|
| 1716 |
+
},
|
| 1717 |
+
{
|
| 1718 |
+
"epoch": 7.31,
|
| 1719 |
+
"learning_rate": 1.1049180327868854e-05,
|
| 1720 |
+
"loss": 0.8522,
|
| 1721 |
+
"step": 2750
|
| 1722 |
+
},
|
| 1723 |
+
{
|
| 1724 |
+
"epoch": 7.34,
|
| 1725 |
+
"learning_rate": 1.0939890710382515e-05,
|
| 1726 |
+
"loss": 0.8437,
|
| 1727 |
+
"step": 2760
|
| 1728 |
+
},
|
| 1729 |
+
{
|
| 1730 |
+
"epoch": 7.37,
|
| 1731 |
+
"learning_rate": 1.0830601092896176e-05,
|
| 1732 |
+
"loss": 0.8466,
|
| 1733 |
+
"step": 2770
|
| 1734 |
+
},
|
| 1735 |
+
{
|
| 1736 |
+
"epoch": 7.39,
|
| 1737 |
+
"learning_rate": 1.0721311475409837e-05,
|
| 1738 |
+
"loss": 0.8477,
|
| 1739 |
+
"step": 2780
|
| 1740 |
+
},
|
| 1741 |
+
{
|
| 1742 |
+
"epoch": 7.42,
|
| 1743 |
+
"learning_rate": 1.0612021857923497e-05,
|
| 1744 |
+
"loss": 0.9094,
|
| 1745 |
+
"step": 2790
|
| 1746 |
+
},
|
| 1747 |
+
{
|
| 1748 |
+
"epoch": 7.45,
|
| 1749 |
+
"learning_rate": 1.0502732240437159e-05,
|
| 1750 |
+
"loss": 0.8399,
|
| 1751 |
+
"step": 2800
|
| 1752 |
+
},
|
| 1753 |
+
{
|
| 1754 |
+
"epoch": 7.47,
|
| 1755 |
+
"learning_rate": 1.0393442622950821e-05,
|
| 1756 |
+
"loss": 0.855,
|
| 1757 |
+
"step": 2810
|
| 1758 |
+
},
|
| 1759 |
+
{
|
| 1760 |
+
"epoch": 7.5,
|
| 1761 |
+
"learning_rate": 1.0284153005464481e-05,
|
| 1762 |
+
"loss": 0.8448,
|
| 1763 |
+
"step": 2820
|
| 1764 |
+
},
|
| 1765 |
+
{
|
| 1766 |
+
"epoch": 7.53,
|
| 1767 |
+
"learning_rate": 1.0174863387978143e-05,
|
| 1768 |
+
"loss": 0.8438,
|
| 1769 |
+
"step": 2830
|
| 1770 |
+
},
|
| 1771 |
+
{
|
| 1772 |
+
"epoch": 7.55,
|
| 1773 |
+
"learning_rate": 1.0065573770491803e-05,
|
| 1774 |
+
"loss": 0.8951,
|
| 1775 |
+
"step": 2840
|
| 1776 |
+
},
|
| 1777 |
+
{
|
| 1778 |
+
"epoch": 7.58,
|
| 1779 |
+
"learning_rate": 9.956284153005466e-06,
|
| 1780 |
+
"loss": 0.858,
|
| 1781 |
+
"step": 2850
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 7.61,
|
| 1785 |
+
"learning_rate": 9.846994535519126e-06,
|
| 1786 |
+
"loss": 0.8505,
|
| 1787 |
+
"step": 2860
|
| 1788 |
+
},
|
| 1789 |
+
{
|
| 1790 |
+
"epoch": 7.63,
|
| 1791 |
+
"learning_rate": 9.737704918032788e-06,
|
| 1792 |
+
"loss": 0.8452,
|
| 1793 |
+
"step": 2870
|
| 1794 |
+
},
|
| 1795 |
+
{
|
| 1796 |
+
"epoch": 7.66,
|
| 1797 |
+
"learning_rate": 9.628415300546448e-06,
|
| 1798 |
+
"loss": 0.8374,
|
| 1799 |
+
"step": 2880
|
| 1800 |
+
},
|
| 1801 |
+
{
|
| 1802 |
+
"epoch": 7.69,
|
| 1803 |
+
"learning_rate": 9.51912568306011e-06,
|
| 1804 |
+
"loss": 0.8905,
|
| 1805 |
+
"step": 2890
|
| 1806 |
+
},
|
| 1807 |
+
{
|
| 1808 |
+
"epoch": 7.71,
|
| 1809 |
+
"learning_rate": 9.409836065573772e-06,
|
| 1810 |
+
"loss": 0.8538,
|
| 1811 |
+
"step": 2900
|
| 1812 |
+
},
|
| 1813 |
+
{
|
| 1814 |
+
"epoch": 7.74,
|
| 1815 |
+
"learning_rate": 9.300546448087432e-06,
|
| 1816 |
+
"loss": 0.848,
|
| 1817 |
+
"step": 2910
|
| 1818 |
+
},
|
| 1819 |
+
{
|
| 1820 |
+
"epoch": 7.77,
|
| 1821 |
+
"learning_rate": 9.191256830601093e-06,
|
| 1822 |
+
"loss": 0.844,
|
| 1823 |
+
"step": 2920
|
| 1824 |
+
},
|
| 1825 |
+
{
|
| 1826 |
+
"epoch": 7.79,
|
| 1827 |
+
"learning_rate": 9.081967213114755e-06,
|
| 1828 |
+
"loss": 0.8375,
|
| 1829 |
+
"step": 2930
|
| 1830 |
+
},
|
| 1831 |
+
{
|
| 1832 |
+
"epoch": 7.82,
|
| 1833 |
+
"learning_rate": 8.972677595628417e-06,
|
| 1834 |
+
"loss": 0.8901,
|
| 1835 |
+
"step": 2940
|
| 1836 |
+
},
|
| 1837 |
+
{
|
| 1838 |
+
"epoch": 7.85,
|
| 1839 |
+
"learning_rate": 8.863387978142077e-06,
|
| 1840 |
+
"loss": 0.8564,
|
| 1841 |
+
"step": 2950
|
| 1842 |
+
},
|
| 1843 |
+
{
|
| 1844 |
+
"epoch": 7.87,
|
| 1845 |
+
"learning_rate": 8.754098360655739e-06,
|
| 1846 |
+
"loss": 0.8372,
|
| 1847 |
+
"step": 2960
|
| 1848 |
+
},
|
| 1849 |
+
{
|
| 1850 |
+
"epoch": 7.9,
|
| 1851 |
+
"learning_rate": 8.644808743169399e-06,
|
| 1852 |
+
"loss": 0.8338,
|
| 1853 |
+
"step": 2970
|
| 1854 |
+
},
|
| 1855 |
+
{
|
| 1856 |
+
"epoch": 7.93,
|
| 1857 |
+
"learning_rate": 8.53551912568306e-06,
|
| 1858 |
+
"loss": 0.8386,
|
| 1859 |
+
"step": 2980
|
| 1860 |
+
},
|
| 1861 |
+
{
|
| 1862 |
+
"epoch": 7.95,
|
| 1863 |
+
"learning_rate": 8.426229508196722e-06,
|
| 1864 |
+
"loss": 0.8699,
|
| 1865 |
+
"step": 2990
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 7.98,
|
| 1869 |
+
"learning_rate": 8.316939890710382e-06,
|
| 1870 |
+
"loss": 0.8493,
|
| 1871 |
+
"step": 3000
|
| 1872 |
+
},
|
| 1873 |
+
{
|
| 1874 |
+
"epoch": 8.0,
|
| 1875 |
+
"eval_loss": 1.0525248050689697,
|
| 1876 |
+
"eval_runtime": 343.4159,
|
| 1877 |
+
"eval_samples_per_second": 7.786,
|
| 1878 |
+
"eval_steps_per_second": 1.948,
|
| 1879 |
+
"eval_wer": 0.6581815955888931,
|
| 1880 |
+
"step": 3008
|
| 1881 |
+
},
|
| 1882 |
+
{
|
| 1883 |
+
"epoch": 8.01,
|
| 1884 |
+
"learning_rate": 8.207650273224044e-06,
|
| 1885 |
+
"loss": 0.8601,
|
| 1886 |
+
"step": 3010
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"epoch": 8.03,
|
| 1890 |
+
"learning_rate": 8.098360655737706e-06,
|
| 1891 |
+
"loss": 0.8718,
|
| 1892 |
+
"step": 3020
|
| 1893 |
+
},
|
| 1894 |
+
{
|
| 1895 |
+
"epoch": 8.06,
|
| 1896 |
+
"learning_rate": 7.989071038251368e-06,
|
| 1897 |
+
"loss": 0.8381,
|
| 1898 |
+
"step": 3030
|
| 1899 |
+
},
|
| 1900 |
+
{
|
| 1901 |
+
"epoch": 8.09,
|
| 1902 |
+
"learning_rate": 7.879781420765028e-06,
|
| 1903 |
+
"loss": 0.8315,
|
| 1904 |
+
"step": 3040
|
| 1905 |
+
},
|
| 1906 |
+
{
|
| 1907 |
+
"epoch": 8.11,
|
| 1908 |
+
"learning_rate": 7.77049180327869e-06,
|
| 1909 |
+
"loss": 0.8324,
|
| 1910 |
+
"step": 3050
|
| 1911 |
+
},
|
| 1912 |
+
{
|
| 1913 |
+
"epoch": 8.14,
|
| 1914 |
+
"learning_rate": 7.66120218579235e-06,
|
| 1915 |
+
"loss": 0.8601,
|
| 1916 |
+
"step": 3060
|
| 1917 |
+
},
|
| 1918 |
+
{
|
| 1919 |
+
"epoch": 8.16,
|
| 1920 |
+
"learning_rate": 7.551912568306011e-06,
|
| 1921 |
+
"loss": 0.8776,
|
| 1922 |
+
"step": 3070
|
| 1923 |
+
},
|
| 1924 |
+
{
|
| 1925 |
+
"epoch": 8.19,
|
| 1926 |
+
"learning_rate": 7.442622950819672e-06,
|
| 1927 |
+
"loss": 0.8488,
|
| 1928 |
+
"step": 3080
|
| 1929 |
+
},
|
| 1930 |
+
{
|
| 1931 |
+
"epoch": 8.22,
|
| 1932 |
+
"learning_rate": 7.333333333333333e-06,
|
| 1933 |
+
"loss": 0.8432,
|
| 1934 |
+
"step": 3090
|
| 1935 |
+
},
|
| 1936 |
+
{
|
| 1937 |
+
"epoch": 8.24,
|
| 1938 |
+
"learning_rate": 7.224043715846996e-06,
|
| 1939 |
+
"loss": 0.8341,
|
| 1940 |
+
"step": 3100
|
| 1941 |
+
},
|
| 1942 |
+
{
|
| 1943 |
+
"epoch": 8.27,
|
| 1944 |
+
"learning_rate": 7.114754098360657e-06,
|
| 1945 |
+
"loss": 0.8486,
|
| 1946 |
+
"step": 3110
|
| 1947 |
+
},
|
| 1948 |
+
{
|
| 1949 |
+
"epoch": 8.3,
|
| 1950 |
+
"learning_rate": 7.0054644808743176e-06,
|
| 1951 |
+
"loss": 0.885,
|
| 1952 |
+
"step": 3120
|
| 1953 |
+
},
|
| 1954 |
+
{
|
| 1955 |
+
"epoch": 8.32,
|
| 1956 |
+
"learning_rate": 6.8961748633879785e-06,
|
| 1957 |
+
"loss": 0.8465,
|
| 1958 |
+
"step": 3130
|
| 1959 |
+
},
|
| 1960 |
+
{
|
| 1961 |
+
"epoch": 8.35,
|
| 1962 |
+
"learning_rate": 6.78688524590164e-06,
|
| 1963 |
+
"loss": 0.8395,
|
| 1964 |
+
"step": 3140
|
| 1965 |
+
},
|
| 1966 |
+
{
|
| 1967 |
+
"epoch": 8.38,
|
| 1968 |
+
"learning_rate": 6.677595628415301e-06,
|
| 1969 |
+
"loss": 0.8381,
|
| 1970 |
+
"step": 3150
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"epoch": 8.4,
|
| 1974 |
+
"learning_rate": 6.568306010928962e-06,
|
| 1975 |
+
"loss": 0.8614,
|
| 1976 |
+
"step": 3160
|
| 1977 |
+
},
|
| 1978 |
+
{
|
| 1979 |
+
"epoch": 8.43,
|
| 1980 |
+
"learning_rate": 6.459016393442623e-06,
|
| 1981 |
+
"loss": 0.8787,
|
| 1982 |
+
"step": 3170
|
| 1983 |
+
},
|
| 1984 |
+
{
|
| 1985 |
+
"epoch": 8.46,
|
| 1986 |
+
"learning_rate": 6.349726775956284e-06,
|
| 1987 |
+
"loss": 0.83,
|
| 1988 |
+
"step": 3180
|
| 1989 |
+
},
|
| 1990 |
+
{
|
| 1991 |
+
"epoch": 8.48,
|
| 1992 |
+
"learning_rate": 6.2404371584699465e-06,
|
| 1993 |
+
"loss": 0.848,
|
| 1994 |
+
"step": 3190
|
| 1995 |
+
},
|
| 1996 |
+
{
|
| 1997 |
+
"epoch": 8.51,
|
| 1998 |
+
"learning_rate": 6.131147540983607e-06,
|
| 1999 |
+
"loss": 0.8501,
|
| 2000 |
+
"step": 3200
|
| 2001 |
+
},
|
| 2002 |
+
{
|
| 2003 |
+
"epoch": 8.54,
|
| 2004 |
+
"learning_rate": 6.021857923497268e-06,
|
| 2005 |
+
"loss": 0.8618,
|
| 2006 |
+
"step": 3210
|
| 2007 |
+
},
|
| 2008 |
+
{
|
| 2009 |
+
"epoch": 8.56,
|
| 2010 |
+
"learning_rate": 5.912568306010929e-06,
|
| 2011 |
+
"loss": 0.8855,
|
| 2012 |
+
"step": 3220
|
| 2013 |
+
},
|
| 2014 |
+
{
|
| 2015 |
+
"epoch": 8.59,
|
| 2016 |
+
"learning_rate": 5.803278688524591e-06,
|
| 2017 |
+
"loss": 0.8357,
|
| 2018 |
+
"step": 3230
|
| 2019 |
+
},
|
| 2020 |
+
{
|
| 2021 |
+
"epoch": 8.62,
|
| 2022 |
+
"learning_rate": 5.693989071038252e-06,
|
| 2023 |
+
"loss": 0.8365,
|
| 2024 |
+
"step": 3240
|
| 2025 |
+
},
|
| 2026 |
+
{
|
| 2027 |
+
"epoch": 8.64,
|
| 2028 |
+
"learning_rate": 5.584699453551913e-06,
|
| 2029 |
+
"loss": 0.8537,
|
| 2030 |
+
"step": 3250
|
| 2031 |
+
},
|
| 2032 |
+
{
|
| 2033 |
+
"epoch": 8.67,
|
| 2034 |
+
"learning_rate": 5.475409836065574e-06,
|
| 2035 |
+
"loss": 0.8565,
|
| 2036 |
+
"step": 3260
|
| 2037 |
+
},
|
| 2038 |
+
{
|
| 2039 |
+
"epoch": 8.7,
|
| 2040 |
+
"learning_rate": 5.366120218579235e-06,
|
| 2041 |
+
"loss": 0.8691,
|
| 2042 |
+
"step": 3270
|
| 2043 |
+
},
|
| 2044 |
+
{
|
| 2045 |
+
"epoch": 8.72,
|
| 2046 |
+
"learning_rate": 5.256830601092897e-06,
|
| 2047 |
+
"loss": 0.8333,
|
| 2048 |
+
"step": 3280
|
| 2049 |
+
},
|
| 2050 |
+
{
|
| 2051 |
+
"epoch": 8.75,
|
| 2052 |
+
"learning_rate": 5.147540983606558e-06,
|
| 2053 |
+
"loss": 0.8268,
|
| 2054 |
+
"step": 3290
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"epoch": 8.78,
|
| 2058 |
+
"learning_rate": 5.038251366120219e-06,
|
| 2059 |
+
"loss": 0.8405,
|
| 2060 |
+
"step": 3300
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 8.8,
|
| 2064 |
+
"learning_rate": 4.92896174863388e-06,
|
| 2065 |
+
"loss": 0.8438,
|
| 2066 |
+
"step": 3310
|
| 2067 |
+
},
|
| 2068 |
+
{
|
| 2069 |
+
"epoch": 8.83,
|
| 2070 |
+
"learning_rate": 4.819672131147542e-06,
|
| 2071 |
+
"loss": 0.8763,
|
| 2072 |
+
"step": 3320
|
| 2073 |
+
},
|
| 2074 |
+
{
|
| 2075 |
+
"epoch": 8.86,
|
| 2076 |
+
"learning_rate": 4.710382513661203e-06,
|
| 2077 |
+
"loss": 0.841,
|
| 2078 |
+
"step": 3330
|
| 2079 |
+
},
|
| 2080 |
+
{
|
| 2081 |
+
"epoch": 8.88,
|
| 2082 |
+
"learning_rate": 4.6010928961748635e-06,
|
| 2083 |
+
"loss": 0.8371,
|
| 2084 |
+
"step": 3340
|
| 2085 |
+
},
|
| 2086 |
+
{
|
| 2087 |
+
"epoch": 8.91,
|
| 2088 |
+
"learning_rate": 4.491803278688525e-06,
|
| 2089 |
+
"loss": 0.8287,
|
| 2090 |
+
"step": 3350
|
| 2091 |
+
},
|
| 2092 |
+
{
|
| 2093 |
+
"epoch": 8.94,
|
| 2094 |
+
"learning_rate": 4.382513661202186e-06,
|
| 2095 |
+
"loss": 0.8526,
|
| 2096 |
+
"step": 3360
|
| 2097 |
+
},
|
| 2098 |
+
{
|
| 2099 |
+
"epoch": 8.96,
|
| 2100 |
+
"learning_rate": 4.273224043715847e-06,
|
| 2101 |
+
"loss": 0.8507,
|
| 2102 |
+
"step": 3370
|
| 2103 |
+
},
|
| 2104 |
+
{
|
| 2105 |
+
"epoch": 8.99,
|
| 2106 |
+
"learning_rate": 4.163934426229508e-06,
|
| 2107 |
+
"loss": 0.8371,
|
| 2108 |
+
"step": 3380
|
| 2109 |
+
},
|
| 2110 |
+
{
|
| 2111 |
+
"epoch": 9.0,
|
| 2112 |
+
"eval_loss": 1.0408543348312378,
|
| 2113 |
+
"eval_runtime": 348.0788,
|
| 2114 |
+
"eval_samples_per_second": 7.682,
|
| 2115 |
+
"eval_steps_per_second": 1.922,
|
| 2116 |
+
"eval_wer": 0.651471703088434,
|
| 2117 |
+
"step": 3384
|
| 2118 |
+
},
|
| 2119 |
+
{
|
| 2120 |
+
"epoch": 9.02,
|
| 2121 |
+
"learning_rate": 4.05464480874317e-06,
|
| 2122 |
+
"loss": 0.8875,
|
| 2123 |
+
"step": 3390
|
| 2124 |
+
},
|
| 2125 |
+
{
|
| 2126 |
+
"epoch": 9.04,
|
| 2127 |
+
"learning_rate": 3.945355191256831e-06,
|
| 2128 |
+
"loss": 0.8507,
|
| 2129 |
+
"step": 3400
|
| 2130 |
+
},
|
| 2131 |
+
{
|
| 2132 |
+
"epoch": 9.07,
|
| 2133 |
+
"learning_rate": 3.8360655737704925e-06,
|
| 2134 |
+
"loss": 0.8481,
|
| 2135 |
+
"step": 3410
|
| 2136 |
+
},
|
| 2137 |
+
{
|
| 2138 |
+
"epoch": 9.1,
|
| 2139 |
+
"learning_rate": 3.7267759562841534e-06,
|
| 2140 |
+
"loss": 0.843,
|
| 2141 |
+
"step": 3420
|
| 2142 |
+
},
|
| 2143 |
+
{
|
| 2144 |
+
"epoch": 9.12,
|
| 2145 |
+
"learning_rate": 3.6174863387978143e-06,
|
| 2146 |
+
"loss": 0.8397,
|
| 2147 |
+
"step": 3430
|
| 2148 |
+
},
|
| 2149 |
+
{
|
| 2150 |
+
"epoch": 9.15,
|
| 2151 |
+
"learning_rate": 3.508196721311476e-06,
|
| 2152 |
+
"loss": 0.8743,
|
| 2153 |
+
"step": 3440
|
| 2154 |
+
},
|
| 2155 |
+
{
|
| 2156 |
+
"epoch": 9.18,
|
| 2157 |
+
"learning_rate": 3.398907103825137e-06,
|
| 2158 |
+
"loss": 0.8453,
|
| 2159 |
+
"step": 3450
|
| 2160 |
+
},
|
| 2161 |
+
{
|
| 2162 |
+
"epoch": 9.2,
|
| 2163 |
+
"learning_rate": 3.2896174863387983e-06,
|
| 2164 |
+
"loss": 0.8378,
|
| 2165 |
+
"step": 3460
|
| 2166 |
+
},
|
| 2167 |
+
{
|
| 2168 |
+
"epoch": 9.23,
|
| 2169 |
+
"learning_rate": 3.180327868852459e-06,
|
| 2170 |
+
"loss": 0.8245,
|
| 2171 |
+
"step": 3470
|
| 2172 |
+
},
|
| 2173 |
+
{
|
| 2174 |
+
"epoch": 9.26,
|
| 2175 |
+
"learning_rate": 3.07103825136612e-06,
|
| 2176 |
+
"loss": 0.8337,
|
| 2177 |
+
"step": 3480
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 9.28,
|
| 2181 |
+
"learning_rate": 2.961748633879782e-06,
|
| 2182 |
+
"loss": 0.8748,
|
| 2183 |
+
"step": 3490
|
| 2184 |
+
},
|
| 2185 |
+
{
|
| 2186 |
+
"epoch": 9.31,
|
| 2187 |
+
"learning_rate": 2.8524590163934428e-06,
|
| 2188 |
+
"loss": 0.8398,
|
| 2189 |
+
"step": 3500
|
| 2190 |
+
},
|
| 2191 |
+
{
|
| 2192 |
+
"epoch": 9.34,
|
| 2193 |
+
"learning_rate": 2.743169398907104e-06,
|
| 2194 |
+
"loss": 0.8387,
|
| 2195 |
+
"step": 3510
|
| 2196 |
+
},
|
| 2197 |
+
{
|
| 2198 |
+
"epoch": 9.36,
|
| 2199 |
+
"learning_rate": 2.633879781420765e-06,
|
| 2200 |
+
"loss": 0.8413,
|
| 2201 |
+
"step": 3520
|
| 2202 |
+
},
|
| 2203 |
+
{
|
| 2204 |
+
"epoch": 9.39,
|
| 2205 |
+
"learning_rate": 2.5245901639344268e-06,
|
| 2206 |
+
"loss": 0.8338,
|
| 2207 |
+
"step": 3530
|
| 2208 |
+
},
|
| 2209 |
+
{
|
| 2210 |
+
"epoch": 9.41,
|
| 2211 |
+
"learning_rate": 2.4153005464480877e-06,
|
| 2212 |
+
"loss": 0.8767,
|
| 2213 |
+
"step": 3540
|
| 2214 |
+
},
|
| 2215 |
+
{
|
| 2216 |
+
"epoch": 9.44,
|
| 2217 |
+
"learning_rate": 2.306010928961749e-06,
|
| 2218 |
+
"loss": 0.8293,
|
| 2219 |
+
"step": 3550
|
| 2220 |
+
},
|
| 2221 |
+
{
|
| 2222 |
+
"epoch": 9.47,
|
| 2223 |
+
"learning_rate": 2.19672131147541e-06,
|
| 2224 |
+
"loss": 0.839,
|
| 2225 |
+
"step": 3560
|
| 2226 |
+
},
|
| 2227 |
+
{
|
| 2228 |
+
"epoch": 9.49,
|
| 2229 |
+
"learning_rate": 2.0874316939890713e-06,
|
| 2230 |
+
"loss": 0.8318,
|
| 2231 |
+
"step": 3570
|
| 2232 |
+
},
|
| 2233 |
+
{
|
| 2234 |
+
"epoch": 9.52,
|
| 2235 |
+
"learning_rate": 1.978142076502732e-06,
|
| 2236 |
+
"loss": 0.8353,
|
| 2237 |
+
"step": 3580
|
| 2238 |
+
},
|
| 2239 |
+
{
|
| 2240 |
+
"epoch": 9.55,
|
| 2241 |
+
"learning_rate": 1.8688524590163937e-06,
|
| 2242 |
+
"loss": 0.8795,
|
| 2243 |
+
"step": 3590
|
| 2244 |
+
},
|
| 2245 |
+
{
|
| 2246 |
+
"epoch": 9.57,
|
| 2247 |
+
"learning_rate": 1.7595628415300548e-06,
|
| 2248 |
+
"loss": 0.8441,
|
| 2249 |
+
"step": 3600
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 9.6,
|
| 2253 |
+
"learning_rate": 1.6502732240437162e-06,
|
| 2254 |
+
"loss": 0.8318,
|
| 2255 |
+
"step": 3610
|
| 2256 |
+
},
|
| 2257 |
+
{
|
| 2258 |
+
"epoch": 9.63,
|
| 2259 |
+
"learning_rate": 1.540983606557377e-06,
|
| 2260 |
+
"loss": 0.8313,
|
| 2261 |
+
"step": 3620
|
| 2262 |
+
},
|
| 2263 |
+
{
|
| 2264 |
+
"epoch": 9.65,
|
| 2265 |
+
"learning_rate": 1.4316939890710382e-06,
|
| 2266 |
+
"loss": 0.8372,
|
| 2267 |
+
"step": 3630
|
| 2268 |
+
},
|
| 2269 |
+
{
|
| 2270 |
+
"epoch": 9.68,
|
| 2271 |
+
"learning_rate": 1.3224043715846995e-06,
|
| 2272 |
+
"loss": 0.8708,
|
| 2273 |
+
"step": 3640
|
| 2274 |
+
},
|
| 2275 |
+
{
|
| 2276 |
+
"epoch": 9.71,
|
| 2277 |
+
"learning_rate": 1.2131147540983609e-06,
|
| 2278 |
+
"loss": 0.8448,
|
| 2279 |
+
"step": 3650
|
| 2280 |
+
},
|
| 2281 |
+
{
|
| 2282 |
+
"epoch": 9.73,
|
| 2283 |
+
"learning_rate": 1.103825136612022e-06,
|
| 2284 |
+
"loss": 0.837,
|
| 2285 |
+
"step": 3660
|
| 2286 |
+
},
|
| 2287 |
+
{
|
| 2288 |
+
"epoch": 9.76,
|
| 2289 |
+
"learning_rate": 9.945355191256831e-07,
|
| 2290 |
+
"loss": 0.8315,
|
| 2291 |
+
"step": 3670
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 9.79,
|
| 2295 |
+
"learning_rate": 8.852459016393443e-07,
|
| 2296 |
+
"loss": 0.8322,
|
| 2297 |
+
"step": 3680
|
| 2298 |
+
},
|
| 2299 |
+
{
|
| 2300 |
+
"epoch": 9.81,
|
| 2301 |
+
"learning_rate": 7.759562841530055e-07,
|
| 2302 |
+
"loss": 0.8701,
|
| 2303 |
+
"step": 3690
|
| 2304 |
+
},
|
| 2305 |
+
{
|
| 2306 |
+
"epoch": 9.84,
|
| 2307 |
+
"learning_rate": 6.666666666666667e-07,
|
| 2308 |
+
"loss": 0.8334,
|
| 2309 |
+
"step": 3700
|
| 2310 |
+
},
|
| 2311 |
+
{
|
| 2312 |
+
"epoch": 9.87,
|
| 2313 |
+
"learning_rate": 5.573770491803279e-07,
|
| 2314 |
+
"loss": 0.8416,
|
| 2315 |
+
"step": 3710
|
| 2316 |
+
},
|
| 2317 |
+
{
|
| 2318 |
+
"epoch": 9.89,
|
| 2319 |
+
"learning_rate": 4.480874316939891e-07,
|
| 2320 |
+
"loss": 0.8386,
|
| 2321 |
+
"step": 3720
|
| 2322 |
+
},
|
| 2323 |
+
{
|
| 2324 |
+
"epoch": 9.92,
|
| 2325 |
+
"learning_rate": 3.387978142076503e-07,
|
| 2326 |
+
"loss": 0.8322,
|
| 2327 |
+
"step": 3730
|
| 2328 |
+
},
|
| 2329 |
+
{
|
| 2330 |
+
"epoch": 9.95,
|
| 2331 |
+
"learning_rate": 2.295081967213115e-07,
|
| 2332 |
+
"loss": 0.8574,
|
| 2333 |
+
"step": 3740
|
| 2334 |
+
},
|
| 2335 |
+
{
|
| 2336 |
+
"epoch": 9.97,
|
| 2337 |
+
"learning_rate": 1.202185792349727e-07,
|
| 2338 |
+
"loss": 0.8294,
|
| 2339 |
+
"step": 3750
|
| 2340 |
+
},
|
| 2341 |
+
{
|
| 2342 |
+
"epoch": 10.0,
|
| 2343 |
+
"learning_rate": 1.0928961748633881e-08,
|
| 2344 |
+
"loss": 0.8229,
|
| 2345 |
+
"step": 3760
|
| 2346 |
+
},
|
| 2347 |
+
{
|
| 2348 |
+
"epoch": 10.0,
|
| 2349 |
+
"eval_loss": 1.059747576713562,
|
| 2350 |
+
"eval_runtime": 346.8481,
|
| 2351 |
+
"eval_samples_per_second": 7.709,
|
| 2352 |
+
"eval_steps_per_second": 1.929,
|
| 2353 |
+
"eval_wer": 0.6522768901884891,
|
| 2354 |
+
"step": 3760
|
| 2355 |
+
},
|
| 2356 |
+
{
|
| 2357 |
+
"epoch": 10.0,
|
| 2358 |
+
"step": 3760,
|
| 2359 |
+
"total_flos": 7.29061419637639e+19,
|
| 2360 |
+
"train_loss": 0.8711832551245994,
|
| 2361 |
+
"train_runtime": 72833.4219,
|
| 2362 |
+
"train_samples_per_second": 3.304,
|
| 2363 |
+
"train_steps_per_second": 0.052
|
| 2364 |
+
}
|
| 2365 |
+
],
|
| 2366 |
+
"max_steps": 3760,
|
| 2367 |
+
"num_train_epochs": 10,
|
| 2368 |
+
"total_flos": 7.29061419637639e+19,
|
| 2369 |
+
"trial_name": null,
|
| 2370 |
+
"trial_params": null
|
| 2371 |
+
}
|