Commit
•
fe56317
1
Parent(s):
0115efe
End of training
Browse files- all_results.json +14 -0
- eval_results.json +9 -0
- runs/May21_09-02-48_bioman2/events.out.tfevents.1653250459.bioman2.19214.2 +3 -0
- train_results.json +8 -0
- trainer_state.json +2221 -0
all_results.json
ADDED
|
@@ -0,0 +1,14 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 50.0,
|
| 3 |
+
"eval_loss": 0.10580451041460037,
|
| 4 |
+
"eval_runtime": 31.0996,
|
| 5 |
+
"eval_samples": 178,
|
| 6 |
+
"eval_samples_per_second": 5.724,
|
| 7 |
+
"eval_steps_per_second": 1.929,
|
| 8 |
+
"eval_wer": 0.10749076251259658,
|
| 9 |
+
"train_loss": 0.6889870901987062,
|
| 10 |
+
"train_runtime": 132713.9902,
|
| 11 |
+
"train_samples": 6770,
|
| 12 |
+
"train_samples_per_second": 2.551,
|
| 13 |
+
"train_steps_per_second": 0.212
|
| 14 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 50.0,
|
| 3 |
+
"eval_loss": 0.10580451041460037,
|
| 4 |
+
"eval_runtime": 31.0996,
|
| 5 |
+
"eval_samples": 178,
|
| 6 |
+
"eval_samples_per_second": 5.724,
|
| 7 |
+
"eval_steps_per_second": 1.929,
|
| 8 |
+
"eval_wer": 0.10749076251259658
|
| 9 |
+
}
|
runs/May21_09-02-48_bioman2/events.out.tfevents.1653250459.bioman2.19214.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1e0d53e20940f64d80f7ff85ddefa514142b54a7fec3fbf3af653b1603f5407c
|
| 3 |
+
size 364
|
train_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 50.0,
|
| 3 |
+
"train_loss": 0.6889870901987062,
|
| 4 |
+
"train_runtime": 132713.9902,
|
| 5 |
+
"train_samples": 6770,
|
| 6 |
+
"train_samples_per_second": 2.551,
|
| 7 |
+
"train_steps_per_second": 0.212
|
| 8 |
+
}
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2221 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": null,
|
| 3 |
+
"best_model_checkpoint": null,
|
| 4 |
+
"epoch": 49.999556933983165,
|
| 5 |
+
"global_step": 28200,
|
| 6 |
+
"is_hyper_param_search": false,
|
| 7 |
+
"is_local_process_zero": true,
|
| 8 |
+
"is_world_process_zero": true,
|
| 9 |
+
"log_history": [
|
| 10 |
+
{
|
| 11 |
+
"epoch": 0.18,
|
| 12 |
+
"learning_rate": 3.7125e-06,
|
| 13 |
+
"loss": 9.6392,
|
| 14 |
+
"step": 100
|
| 15 |
+
},
|
| 16 |
+
{
|
| 17 |
+
"epoch": 0.35,
|
| 18 |
+
"learning_rate": 7.4625e-06,
|
| 19 |
+
"loss": 5.3472,
|
| 20 |
+
"step": 200
|
| 21 |
+
},
|
| 22 |
+
{
|
| 23 |
+
"epoch": 0.53,
|
| 24 |
+
"learning_rate": 1.1212499999999998e-05,
|
| 25 |
+
"loss": 3.7052,
|
| 26 |
+
"step": 300
|
| 27 |
+
},
|
| 28 |
+
{
|
| 29 |
+
"epoch": 0.71,
|
| 30 |
+
"learning_rate": 1.49625e-05,
|
| 31 |
+
"loss": 3.35,
|
| 32 |
+
"step": 400
|
| 33 |
+
},
|
| 34 |
+
{
|
| 35 |
+
"epoch": 0.89,
|
| 36 |
+
"learning_rate": 1.8712499999999997e-05,
|
| 37 |
+
"loss": 3.1484,
|
| 38 |
+
"step": 500
|
| 39 |
+
},
|
| 40 |
+
{
|
| 41 |
+
"epoch": 0.89,
|
| 42 |
+
"eval_loss": 3.084352731704712,
|
| 43 |
+
"eval_runtime": 31.0524,
|
| 44 |
+
"eval_samples_per_second": 5.732,
|
| 45 |
+
"eval_steps_per_second": 1.932,
|
| 46 |
+
"eval_wer": 1.0,
|
| 47 |
+
"step": 500
|
| 48 |
+
},
|
| 49 |
+
{
|
| 50 |
+
"epoch": 1.06,
|
| 51 |
+
"learning_rate": 2.2462499999999997e-05,
|
| 52 |
+
"loss": 3.0531,
|
| 53 |
+
"step": 600
|
| 54 |
+
},
|
| 55 |
+
{
|
| 56 |
+
"epoch": 1.24,
|
| 57 |
+
"learning_rate": 2.6212499999999997e-05,
|
| 58 |
+
"loss": 2.9911,
|
| 59 |
+
"step": 700
|
| 60 |
+
},
|
| 61 |
+
{
|
| 62 |
+
"epoch": 1.42,
|
| 63 |
+
"learning_rate": 2.99625e-05,
|
| 64 |
+
"loss": 2.9576,
|
| 65 |
+
"step": 800
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 1.6,
|
| 69 |
+
"learning_rate": 3.37125e-05,
|
| 70 |
+
"loss": 2.9406,
|
| 71 |
+
"step": 900
|
| 72 |
+
},
|
| 73 |
+
{
|
| 74 |
+
"epoch": 1.77,
|
| 75 |
+
"learning_rate": 3.7462499999999996e-05,
|
| 76 |
+
"loss": 2.6539,
|
| 77 |
+
"step": 1000
|
| 78 |
+
},
|
| 79 |
+
{
|
| 80 |
+
"epoch": 1.77,
|
| 81 |
+
"eval_loss": 1.7271525859832764,
|
| 82 |
+
"eval_runtime": 30.9027,
|
| 83 |
+
"eval_samples_per_second": 5.76,
|
| 84 |
+
"eval_steps_per_second": 1.942,
|
| 85 |
+
"eval_wer": 0.9358414511252939,
|
| 86 |
+
"step": 1000
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 1.95,
|
| 90 |
+
"learning_rate": 4.12125e-05,
|
| 91 |
+
"loss": 1.5429,
|
| 92 |
+
"step": 1100
|
| 93 |
+
},
|
| 94 |
+
{
|
| 95 |
+
"epoch": 2.13,
|
| 96 |
+
"learning_rate": 4.4962499999999995e-05,
|
| 97 |
+
"loss": 1.0875,
|
| 98 |
+
"step": 1200
|
| 99 |
+
},
|
| 100 |
+
{
|
| 101 |
+
"epoch": 2.3,
|
| 102 |
+
"learning_rate": 4.871249999999999e-05,
|
| 103 |
+
"loss": 0.9728,
|
| 104 |
+
"step": 1300
|
| 105 |
+
},
|
| 106 |
+
{
|
| 107 |
+
"epoch": 2.48,
|
| 108 |
+
"learning_rate": 5.2424999999999994e-05,
|
| 109 |
+
"loss": 0.9117,
|
| 110 |
+
"step": 1400
|
| 111 |
+
},
|
| 112 |
+
{
|
| 113 |
+
"epoch": 2.66,
|
| 114 |
+
"learning_rate": 5.6175e-05,
|
| 115 |
+
"loss": 0.8732,
|
| 116 |
+
"step": 1500
|
| 117 |
+
},
|
| 118 |
+
{
|
| 119 |
+
"epoch": 2.66,
|
| 120 |
+
"eval_loss": 0.19752363860607147,
|
| 121 |
+
"eval_runtime": 31.0609,
|
| 122 |
+
"eval_samples_per_second": 5.731,
|
| 123 |
+
"eval_steps_per_second": 1.932,
|
| 124 |
+
"eval_wer": 0.16090023513604298,
|
| 125 |
+
"step": 1500
|
| 126 |
+
},
|
| 127 |
+
{
|
| 128 |
+
"epoch": 2.84,
|
| 129 |
+
"learning_rate": 5.9925e-05,
|
| 130 |
+
"loss": 0.8518,
|
| 131 |
+
"step": 1600
|
| 132 |
+
},
|
| 133 |
+
{
|
| 134 |
+
"epoch": 3.01,
|
| 135 |
+
"learning_rate": 6.367499999999999e-05,
|
| 136 |
+
"loss": 0.8373,
|
| 137 |
+
"step": 1700
|
| 138 |
+
},
|
| 139 |
+
{
|
| 140 |
+
"epoch": 3.19,
|
| 141 |
+
"learning_rate": 6.7425e-05,
|
| 142 |
+
"loss": 0.7958,
|
| 143 |
+
"step": 1800
|
| 144 |
+
},
|
| 145 |
+
{
|
| 146 |
+
"epoch": 3.37,
|
| 147 |
+
"learning_rate": 7.1175e-05,
|
| 148 |
+
"loss": 0.8058,
|
| 149 |
+
"step": 1900
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 3.55,
|
| 153 |
+
"learning_rate": 7.492499999999999e-05,
|
| 154 |
+
"loss": 0.8075,
|
| 155 |
+
"step": 2000
|
| 156 |
+
},
|
| 157 |
+
{
|
| 158 |
+
"epoch": 3.55,
|
| 159 |
+
"eval_loss": 0.1483038365840912,
|
| 160 |
+
"eval_runtime": 31.0609,
|
| 161 |
+
"eval_samples_per_second": 5.731,
|
| 162 |
+
"eval_steps_per_second": 1.932,
|
| 163 |
+
"eval_wer": 0.1467920725562647,
|
| 164 |
+
"step": 2000
|
| 165 |
+
},
|
| 166 |
+
{
|
| 167 |
+
"epoch": 3.72,
|
| 168 |
+
"learning_rate": 7.471946564885497e-05,
|
| 169 |
+
"loss": 0.785,
|
| 170 |
+
"step": 2100
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 3.9,
|
| 174 |
+
"learning_rate": 7.443320610687022e-05,
|
| 175 |
+
"loss": 0.7656,
|
| 176 |
+
"step": 2200
|
| 177 |
+
},
|
| 178 |
+
{
|
| 179 |
+
"epoch": 4.08,
|
| 180 |
+
"learning_rate": 7.414694656488549e-05,
|
| 181 |
+
"loss": 0.7681,
|
| 182 |
+
"step": 2300
|
| 183 |
+
},
|
| 184 |
+
{
|
| 185 |
+
"epoch": 4.26,
|
| 186 |
+
"learning_rate": 7.386068702290076e-05,
|
| 187 |
+
"loss": 0.7535,
|
| 188 |
+
"step": 2400
|
| 189 |
+
},
|
| 190 |
+
{
|
| 191 |
+
"epoch": 4.43,
|
| 192 |
+
"learning_rate": 7.357442748091603e-05,
|
| 193 |
+
"loss": 0.7358,
|
| 194 |
+
"step": 2500
|
| 195 |
+
},
|
| 196 |
+
{
|
| 197 |
+
"epoch": 4.43,
|
| 198 |
+
"eval_loss": 0.1331482231616974,
|
| 199 |
+
"eval_runtime": 30.9431,
|
| 200 |
+
"eval_samples_per_second": 5.752,
|
| 201 |
+
"eval_steps_per_second": 1.939,
|
| 202 |
+
"eval_wer": 0.14007389989922742,
|
| 203 |
+
"step": 2500
|
| 204 |
+
},
|
| 205 |
+
{
|
| 206 |
+
"epoch": 4.61,
|
| 207 |
+
"learning_rate": 7.328816793893128e-05,
|
| 208 |
+
"loss": 0.7478,
|
| 209 |
+
"step": 2600
|
| 210 |
+
},
|
| 211 |
+
{
|
| 212 |
+
"epoch": 4.79,
|
| 213 |
+
"learning_rate": 7.300190839694656e-05,
|
| 214 |
+
"loss": 0.746,
|
| 215 |
+
"step": 2700
|
| 216 |
+
},
|
| 217 |
+
{
|
| 218 |
+
"epoch": 4.96,
|
| 219 |
+
"learning_rate": 7.271564885496182e-05,
|
| 220 |
+
"loss": 0.7545,
|
| 221 |
+
"step": 2800
|
| 222 |
+
},
|
| 223 |
+
{
|
| 224 |
+
"epoch": 5.14,
|
| 225 |
+
"learning_rate": 7.242938931297709e-05,
|
| 226 |
+
"loss": 0.7268,
|
| 227 |
+
"step": 2900
|
| 228 |
+
},
|
| 229 |
+
{
|
| 230 |
+
"epoch": 5.32,
|
| 231 |
+
"learning_rate": 7.214312977099236e-05,
|
| 232 |
+
"loss": 0.7079,
|
| 233 |
+
"step": 3000
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 5.32,
|
| 237 |
+
"eval_loss": 0.12731623649597168,
|
| 238 |
+
"eval_runtime": 30.9985,
|
| 239 |
+
"eval_samples_per_second": 5.742,
|
| 240 |
+
"eval_steps_per_second": 1.936,
|
| 241 |
+
"eval_wer": 0.1363789049378569,
|
| 242 |
+
"step": 3000
|
| 243 |
+
},
|
| 244 |
+
{
|
| 245 |
+
"epoch": 5.5,
|
| 246 |
+
"learning_rate": 7.185687022900763e-05,
|
| 247 |
+
"loss": 0.7344,
|
| 248 |
+
"step": 3100
|
| 249 |
+
},
|
| 250 |
+
{
|
| 251 |
+
"epoch": 5.67,
|
| 252 |
+
"learning_rate": 7.15706106870229e-05,
|
| 253 |
+
"loss": 0.7334,
|
| 254 |
+
"step": 3200
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 5.85,
|
| 258 |
+
"learning_rate": 7.128435114503815e-05,
|
| 259 |
+
"loss": 0.7274,
|
| 260 |
+
"step": 3300
|
| 261 |
+
},
|
| 262 |
+
{
|
| 263 |
+
"epoch": 6.03,
|
| 264 |
+
"learning_rate": 7.099809160305343e-05,
|
| 265 |
+
"loss": 0.7271,
|
| 266 |
+
"step": 3400
|
| 267 |
+
},
|
| 268 |
+
{
|
| 269 |
+
"epoch": 6.21,
|
| 270 |
+
"learning_rate": 7.071469465648854e-05,
|
| 271 |
+
"loss": 0.7032,
|
| 272 |
+
"step": 3500
|
| 273 |
+
},
|
| 274 |
+
{
|
| 275 |
+
"epoch": 6.21,
|
| 276 |
+
"eval_loss": 0.11333052814006805,
|
| 277 |
+
"eval_runtime": 31.1935,
|
| 278 |
+
"eval_samples_per_second": 5.706,
|
| 279 |
+
"eval_steps_per_second": 1.923,
|
| 280 |
+
"eval_wer": 0.12395028552233793,
|
| 281 |
+
"step": 3500
|
| 282 |
+
},
|
| 283 |
+
{
|
| 284 |
+
"epoch": 6.38,
|
| 285 |
+
"learning_rate": 7.04284351145038e-05,
|
| 286 |
+
"loss": 0.6852,
|
| 287 |
+
"step": 3600
|
| 288 |
+
},
|
| 289 |
+
{
|
| 290 |
+
"epoch": 6.56,
|
| 291 |
+
"learning_rate": 7.014217557251907e-05,
|
| 292 |
+
"loss": 0.699,
|
| 293 |
+
"step": 3700
|
| 294 |
+
},
|
| 295 |
+
{
|
| 296 |
+
"epoch": 6.74,
|
| 297 |
+
"learning_rate": 6.985591603053435e-05,
|
| 298 |
+
"loss": 0.6894,
|
| 299 |
+
"step": 3800
|
| 300 |
+
},
|
| 301 |
+
{
|
| 302 |
+
"epoch": 6.91,
|
| 303 |
+
"learning_rate": 6.956965648854962e-05,
|
| 304 |
+
"loss": 0.7014,
|
| 305 |
+
"step": 3900
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 7.09,
|
| 309 |
+
"learning_rate": 6.928339694656487e-05,
|
| 310 |
+
"loss": 0.7129,
|
| 311 |
+
"step": 4000
|
| 312 |
+
},
|
| 313 |
+
{
|
| 314 |
+
"epoch": 7.09,
|
| 315 |
+
"eval_loss": 0.11242285370826721,
|
| 316 |
+
"eval_runtime": 31.0672,
|
| 317 |
+
"eval_samples_per_second": 5.73,
|
| 318 |
+
"eval_steps_per_second": 1.931,
|
| 319 |
+
"eval_wer": 0.1289889150151159,
|
| 320 |
+
"step": 4000
|
| 321 |
+
},
|
| 322 |
+
{
|
| 323 |
+
"epoch": 7.27,
|
| 324 |
+
"learning_rate": 6.899713740458015e-05,
|
| 325 |
+
"loss": 0.6902,
|
| 326 |
+
"step": 4100
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 7.45,
|
| 330 |
+
"learning_rate": 6.871087786259541e-05,
|
| 331 |
+
"loss": 0.6898,
|
| 332 |
+
"step": 4200
|
| 333 |
+
},
|
| 334 |
+
{
|
| 335 |
+
"epoch": 7.62,
|
| 336 |
+
"learning_rate": 6.842461832061069e-05,
|
| 337 |
+
"loss": 0.6785,
|
| 338 |
+
"step": 4300
|
| 339 |
+
},
|
| 340 |
+
{
|
| 341 |
+
"epoch": 7.8,
|
| 342 |
+
"learning_rate": 6.813835877862594e-05,
|
| 343 |
+
"loss": 0.7024,
|
| 344 |
+
"step": 4400
|
| 345 |
+
},
|
| 346 |
+
{
|
| 347 |
+
"epoch": 7.98,
|
| 348 |
+
"learning_rate": 6.785209923664122e-05,
|
| 349 |
+
"loss": 0.6771,
|
| 350 |
+
"step": 4500
|
| 351 |
+
},
|
| 352 |
+
{
|
| 353 |
+
"epoch": 7.98,
|
| 354 |
+
"eval_loss": 0.11212227493524551,
|
| 355 |
+
"eval_runtime": 31.0691,
|
| 356 |
+
"eval_samples_per_second": 5.729,
|
| 357 |
+
"eval_steps_per_second": 1.931,
|
| 358 |
+
"eval_wer": 0.12999664091367147,
|
| 359 |
+
"step": 4500
|
| 360 |
+
},
|
| 361 |
+
{
|
| 362 |
+
"epoch": 8.16,
|
| 363 |
+
"learning_rate": 6.756583969465648e-05,
|
| 364 |
+
"loss": 0.6555,
|
| 365 |
+
"step": 4600
|
| 366 |
+
},
|
| 367 |
+
{
|
| 368 |
+
"epoch": 8.33,
|
| 369 |
+
"learning_rate": 6.727958015267174e-05,
|
| 370 |
+
"loss": 0.6897,
|
| 371 |
+
"step": 4700
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"epoch": 8.51,
|
| 375 |
+
"learning_rate": 6.699618320610687e-05,
|
| 376 |
+
"loss": 0.6809,
|
| 377 |
+
"step": 4800
|
| 378 |
+
},
|
| 379 |
+
{
|
| 380 |
+
"epoch": 8.69,
|
| 381 |
+
"learning_rate": 6.670992366412213e-05,
|
| 382 |
+
"loss": 0.6821,
|
| 383 |
+
"step": 4900
|
| 384 |
+
},
|
| 385 |
+
{
|
| 386 |
+
"epoch": 8.86,
|
| 387 |
+
"learning_rate": 6.642366412213741e-05,
|
| 388 |
+
"loss": 0.6859,
|
| 389 |
+
"step": 5000
|
| 390 |
+
},
|
| 391 |
+
{
|
| 392 |
+
"epoch": 8.86,
|
| 393 |
+
"eval_loss": 0.109534852206707,
|
| 394 |
+
"eval_runtime": 31.3176,
|
| 395 |
+
"eval_samples_per_second": 5.684,
|
| 396 |
+
"eval_steps_per_second": 1.916,
|
| 397 |
+
"eval_wer": 0.13134027544507895,
|
| 398 |
+
"step": 5000
|
| 399 |
+
},
|
| 400 |
+
{
|
| 401 |
+
"epoch": 9.04,
|
| 402 |
+
"learning_rate": 6.613740458015266e-05,
|
| 403 |
+
"loss": 0.6849,
|
| 404 |
+
"step": 5100
|
| 405 |
+
},
|
| 406 |
+
{
|
| 407 |
+
"epoch": 9.22,
|
| 408 |
+
"learning_rate": 6.585114503816793e-05,
|
| 409 |
+
"loss": 0.6615,
|
| 410 |
+
"step": 5200
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 9.4,
|
| 414 |
+
"learning_rate": 6.55648854961832e-05,
|
| 415 |
+
"loss": 0.6587,
|
| 416 |
+
"step": 5300
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"epoch": 9.57,
|
| 420 |
+
"learning_rate": 6.527862595419846e-05,
|
| 421 |
+
"loss": 0.6741,
|
| 422 |
+
"step": 5400
|
| 423 |
+
},
|
| 424 |
+
{
|
| 425 |
+
"epoch": 9.75,
|
| 426 |
+
"learning_rate": 6.499236641221373e-05,
|
| 427 |
+
"loss": 0.6496,
|
| 428 |
+
"step": 5500
|
| 429 |
+
},
|
| 430 |
+
{
|
| 431 |
+
"epoch": 9.75,
|
| 432 |
+
"eval_loss": 0.1090548112988472,
|
| 433 |
+
"eval_runtime": 31.0434,
|
| 434 |
+
"eval_samples_per_second": 5.734,
|
| 435 |
+
"eval_steps_per_second": 1.933,
|
| 436 |
+
"eval_wer": 0.12495801142089352,
|
| 437 |
+
"step": 5500
|
| 438 |
+
},
|
| 439 |
+
{
|
| 440 |
+
"epoch": 9.93,
|
| 441 |
+
"learning_rate": 6.4706106870229e-05,
|
| 442 |
+
"loss": 0.6672,
|
| 443 |
+
"step": 5600
|
| 444 |
+
},
|
| 445 |
+
{
|
| 446 |
+
"epoch": 10.11,
|
| 447 |
+
"learning_rate": 6.441984732824428e-05,
|
| 448 |
+
"loss": 0.6511,
|
| 449 |
+
"step": 5700
|
| 450 |
+
},
|
| 451 |
+
{
|
| 452 |
+
"epoch": 10.28,
|
| 453 |
+
"learning_rate": 6.413645038167938e-05,
|
| 454 |
+
"loss": 0.6393,
|
| 455 |
+
"step": 5800
|
| 456 |
+
},
|
| 457 |
+
{
|
| 458 |
+
"epoch": 10.46,
|
| 459 |
+
"learning_rate": 6.385019083969465e-05,
|
| 460 |
+
"loss": 0.6531,
|
| 461 |
+
"step": 5900
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"epoch": 10.64,
|
| 465 |
+
"learning_rate": 6.356393129770992e-05,
|
| 466 |
+
"loss": 0.6431,
|
| 467 |
+
"step": 6000
|
| 468 |
+
},
|
| 469 |
+
{
|
| 470 |
+
"epoch": 10.64,
|
| 471 |
+
"eval_loss": 0.11021342128515244,
|
| 472 |
+
"eval_runtime": 30.9854,
|
| 473 |
+
"eval_samples_per_second": 5.745,
|
| 474 |
+
"eval_steps_per_second": 1.936,
|
| 475 |
+
"eval_wer": 0.12932482364796774,
|
| 476 |
+
"step": 6000
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"epoch": 10.82,
|
| 480 |
+
"learning_rate": 6.32776717557252e-05,
|
| 481 |
+
"loss": 0.6476,
|
| 482 |
+
"step": 6100
|
| 483 |
+
},
|
| 484 |
+
{
|
| 485 |
+
"epoch": 10.99,
|
| 486 |
+
"learning_rate": 6.299141221374044e-05,
|
| 487 |
+
"loss": 0.6504,
|
| 488 |
+
"step": 6200
|
| 489 |
+
},
|
| 490 |
+
{
|
| 491 |
+
"epoch": 11.17,
|
| 492 |
+
"learning_rate": 6.270515267175572e-05,
|
| 493 |
+
"loss": 0.6366,
|
| 494 |
+
"step": 6300
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 11.35,
|
| 498 |
+
"learning_rate": 6.241889312977098e-05,
|
| 499 |
+
"loss": 0.6435,
|
| 500 |
+
"step": 6400
|
| 501 |
+
},
|
| 502 |
+
{
|
| 503 |
+
"epoch": 11.52,
|
| 504 |
+
"learning_rate": 6.213263358778625e-05,
|
| 505 |
+
"loss": 0.6422,
|
| 506 |
+
"step": 6500
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"epoch": 11.52,
|
| 510 |
+
"eval_loss": 0.11072035878896713,
|
| 511 |
+
"eval_runtime": 31.126,
|
| 512 |
+
"eval_samples_per_second": 5.719,
|
| 513 |
+
"eval_steps_per_second": 1.928,
|
| 514 |
+
"eval_wer": 0.11790393013100436,
|
| 515 |
+
"step": 6500
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 11.7,
|
| 519 |
+
"learning_rate": 6.184637404580152e-05,
|
| 520 |
+
"loss": 0.636,
|
| 521 |
+
"step": 6600
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"epoch": 11.88,
|
| 525 |
+
"learning_rate": 6.156011450381679e-05,
|
| 526 |
+
"loss": 0.647,
|
| 527 |
+
"step": 6700
|
| 528 |
+
},
|
| 529 |
+
{
|
| 530 |
+
"epoch": 12.06,
|
| 531 |
+
"learning_rate": 6.127385496183207e-05,
|
| 532 |
+
"loss": 0.644,
|
| 533 |
+
"step": 6800
|
| 534 |
+
},
|
| 535 |
+
{
|
| 536 |
+
"epoch": 12.23,
|
| 537 |
+
"learning_rate": 6.098759541984732e-05,
|
| 538 |
+
"loss": 0.611,
|
| 539 |
+
"step": 6900
|
| 540 |
+
},
|
| 541 |
+
{
|
| 542 |
+
"epoch": 12.41,
|
| 543 |
+
"learning_rate": 6.070133587786259e-05,
|
| 544 |
+
"loss": 0.6334,
|
| 545 |
+
"step": 7000
|
| 546 |
+
},
|
| 547 |
+
{
|
| 548 |
+
"epoch": 12.41,
|
| 549 |
+
"eval_loss": 0.10494451969861984,
|
| 550 |
+
"eval_runtime": 31.1005,
|
| 551 |
+
"eval_samples_per_second": 5.723,
|
| 552 |
+
"eval_steps_per_second": 1.929,
|
| 553 |
+
"eval_wer": 0.12361437688948607,
|
| 554 |
+
"step": 7000
|
| 555 |
+
},
|
| 556 |
+
{
|
| 557 |
+
"epoch": 12.59,
|
| 558 |
+
"learning_rate": 6.0415076335877855e-05,
|
| 559 |
+
"loss": 0.6305,
|
| 560 |
+
"step": 7100
|
| 561 |
+
},
|
| 562 |
+
{
|
| 563 |
+
"epoch": 12.77,
|
| 564 |
+
"learning_rate": 6.0128816793893126e-05,
|
| 565 |
+
"loss": 0.6292,
|
| 566 |
+
"step": 7200
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"epoch": 12.94,
|
| 570 |
+
"learning_rate": 5.984255725190839e-05,
|
| 571 |
+
"loss": 0.6341,
|
| 572 |
+
"step": 7300
|
| 573 |
+
},
|
| 574 |
+
{
|
| 575 |
+
"epoch": 13.12,
|
| 576 |
+
"learning_rate": 5.955629770992366e-05,
|
| 577 |
+
"loss": 0.625,
|
| 578 |
+
"step": 7400
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 13.3,
|
| 582 |
+
"learning_rate": 5.927003816793893e-05,
|
| 583 |
+
"loss": 0.599,
|
| 584 |
+
"step": 7500
|
| 585 |
+
},
|
| 586 |
+
{
|
| 587 |
+
"epoch": 13.3,
|
| 588 |
+
"eval_loss": 0.10921534150838852,
|
| 589 |
+
"eval_runtime": 31.0896,
|
| 590 |
+
"eval_samples_per_second": 5.725,
|
| 591 |
+
"eval_steps_per_second": 1.93,
|
| 592 |
+
"eval_wer": 0.11521666106818945,
|
| 593 |
+
"step": 7500
|
| 594 |
+
},
|
| 595 |
+
{
|
| 596 |
+
"epoch": 13.47,
|
| 597 |
+
"learning_rate": 5.898377862595419e-05,
|
| 598 |
+
"loss": 0.6144,
|
| 599 |
+
"step": 7600
|
| 600 |
+
},
|
| 601 |
+
{
|
| 602 |
+
"epoch": 13.65,
|
| 603 |
+
"learning_rate": 5.869751908396946e-05,
|
| 604 |
+
"loss": 0.6314,
|
| 605 |
+
"step": 7700
|
| 606 |
+
},
|
| 607 |
+
{
|
| 608 |
+
"epoch": 13.83,
|
| 609 |
+
"learning_rate": 5.8411259541984726e-05,
|
| 610 |
+
"loss": 0.6185,
|
| 611 |
+
"step": 7800
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"epoch": 14.01,
|
| 615 |
+
"learning_rate": 5.8124999999999997e-05,
|
| 616 |
+
"loss": 0.6122,
|
| 617 |
+
"step": 7900
|
| 618 |
+
},
|
| 619 |
+
{
|
| 620 |
+
"epoch": 14.18,
|
| 621 |
+
"learning_rate": 5.783874045801526e-05,
|
| 622 |
+
"loss": 0.6205,
|
| 623 |
+
"step": 8000
|
| 624 |
+
},
|
| 625 |
+
{
|
| 626 |
+
"epoch": 14.18,
|
| 627 |
+
"eval_loss": 0.1046827957034111,
|
| 628 |
+
"eval_runtime": 31.2229,
|
| 629 |
+
"eval_samples_per_second": 5.701,
|
| 630 |
+
"eval_steps_per_second": 1.922,
|
| 631 |
+
"eval_wer": 0.12193483372522673,
|
| 632 |
+
"step": 8000
|
| 633 |
+
},
|
| 634 |
+
{
|
| 635 |
+
"epoch": 14.36,
|
| 636 |
+
"learning_rate": 5.755248091603053e-05,
|
| 637 |
+
"loss": 0.612,
|
| 638 |
+
"step": 8100
|
| 639 |
+
},
|
| 640 |
+
{
|
| 641 |
+
"epoch": 14.54,
|
| 642 |
+
"learning_rate": 5.7266221374045794e-05,
|
| 643 |
+
"loss": 0.6196,
|
| 644 |
+
"step": 8200
|
| 645 |
+
},
|
| 646 |
+
{
|
| 647 |
+
"epoch": 14.72,
|
| 648 |
+
"learning_rate": 5.6979961832061064e-05,
|
| 649 |
+
"loss": 0.6114,
|
| 650 |
+
"step": 8300
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 14.89,
|
| 654 |
+
"learning_rate": 5.6693702290076334e-05,
|
| 655 |
+
"loss": 0.6005,
|
| 656 |
+
"step": 8400
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"epoch": 15.07,
|
| 660 |
+
"learning_rate": 5.64074427480916e-05,
|
| 661 |
+
"loss": 0.5944,
|
| 662 |
+
"step": 8500
|
| 663 |
+
},
|
| 664 |
+
{
|
| 665 |
+
"epoch": 15.07,
|
| 666 |
+
"eval_loss": 0.10684490948915482,
|
| 667 |
+
"eval_runtime": 31.2028,
|
| 668 |
+
"eval_samples_per_second": 5.705,
|
| 669 |
+
"eval_steps_per_second": 1.923,
|
| 670 |
+
"eval_wer": 0.12025529056096741,
|
| 671 |
+
"step": 8500
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 15.25,
|
| 675 |
+
"learning_rate": 5.612118320610687e-05,
|
| 676 |
+
"loss": 0.6108,
|
| 677 |
+
"step": 8600
|
| 678 |
+
},
|
| 679 |
+
{
|
| 680 |
+
"epoch": 15.43,
|
| 681 |
+
"learning_rate": 5.583492366412213e-05,
|
| 682 |
+
"loss": 0.5995,
|
| 683 |
+
"step": 8700
|
| 684 |
+
},
|
| 685 |
+
{
|
| 686 |
+
"epoch": 15.6,
|
| 687 |
+
"learning_rate": 5.55486641221374e-05,
|
| 688 |
+
"loss": 0.6117,
|
| 689 |
+
"step": 8800
|
| 690 |
+
},
|
| 691 |
+
{
|
| 692 |
+
"epoch": 15.78,
|
| 693 |
+
"learning_rate": 5.5262404580152665e-05,
|
| 694 |
+
"loss": 0.6006,
|
| 695 |
+
"step": 8900
|
| 696 |
+
},
|
| 697 |
+
{
|
| 698 |
+
"epoch": 15.96,
|
| 699 |
+
"learning_rate": 5.4976145038167935e-05,
|
| 700 |
+
"loss": 0.6102,
|
| 701 |
+
"step": 9000
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"epoch": 15.96,
|
| 705 |
+
"eval_loss": 0.10560546070337296,
|
| 706 |
+
"eval_runtime": 31.2837,
|
| 707 |
+
"eval_samples_per_second": 5.69,
|
| 708 |
+
"eval_steps_per_second": 1.918,
|
| 709 |
+
"eval_wer": 0.11588847833389318,
|
| 710 |
+
"step": 9000
|
| 711 |
+
},
|
| 712 |
+
{
|
| 713 |
+
"epoch": 16.13,
|
| 714 |
+
"learning_rate": 5.46898854961832e-05,
|
| 715 |
+
"loss": 0.5968,
|
| 716 |
+
"step": 9100
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"epoch": 16.31,
|
| 720 |
+
"learning_rate": 5.440362595419847e-05,
|
| 721 |
+
"loss": 0.5912,
|
| 722 |
+
"step": 9200
|
| 723 |
+
},
|
| 724 |
+
{
|
| 725 |
+
"epoch": 16.49,
|
| 726 |
+
"learning_rate": 5.411736641221374e-05,
|
| 727 |
+
"loss": 0.5889,
|
| 728 |
+
"step": 9300
|
| 729 |
+
},
|
| 730 |
+
{
|
| 731 |
+
"epoch": 16.67,
|
| 732 |
+
"learning_rate": 5.3831106870229e-05,
|
| 733 |
+
"loss": 0.6014,
|
| 734 |
+
"step": 9400
|
| 735 |
+
},
|
| 736 |
+
{
|
| 737 |
+
"epoch": 16.84,
|
| 738 |
+
"learning_rate": 5.354484732824427e-05,
|
| 739 |
+
"loss": 0.5983,
|
| 740 |
+
"step": 9500
|
| 741 |
+
},
|
| 742 |
+
{
|
| 743 |
+
"epoch": 16.84,
|
| 744 |
+
"eval_loss": 0.1061149537563324,
|
| 745 |
+
"eval_runtime": 31.4987,
|
| 746 |
+
"eval_samples_per_second": 5.651,
|
| 747 |
+
"eval_steps_per_second": 1.905,
|
| 748 |
+
"eval_wer": 0.11521666106818945,
|
| 749 |
+
"step": 9500
|
| 750 |
+
},
|
| 751 |
+
{
|
| 752 |
+
"epoch": 17.02,
|
| 753 |
+
"learning_rate": 5.3258587786259536e-05,
|
| 754 |
+
"loss": 0.5992,
|
| 755 |
+
"step": 9600
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 17.2,
|
| 759 |
+
"learning_rate": 5.2972328244274806e-05,
|
| 760 |
+
"loss": 0.5887,
|
| 761 |
+
"step": 9700
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"epoch": 17.38,
|
| 765 |
+
"learning_rate": 5.268606870229007e-05,
|
| 766 |
+
"loss": 0.6012,
|
| 767 |
+
"step": 9800
|
| 768 |
+
},
|
| 769 |
+
{
|
| 770 |
+
"epoch": 17.55,
|
| 771 |
+
"learning_rate": 5.239980916030534e-05,
|
| 772 |
+
"loss": 0.5985,
|
| 773 |
+
"step": 9900
|
| 774 |
+
},
|
| 775 |
+
{
|
| 776 |
+
"epoch": 17.73,
|
| 777 |
+
"learning_rate": 5.211354961832061e-05,
|
| 778 |
+
"loss": 0.5882,
|
| 779 |
+
"step": 10000
|
| 780 |
+
},
|
| 781 |
+
{
|
| 782 |
+
"epoch": 17.73,
|
| 783 |
+
"eval_loss": 0.10430345684289932,
|
| 784 |
+
"eval_runtime": 30.988,
|
| 785 |
+
"eval_samples_per_second": 5.744,
|
| 786 |
+
"eval_steps_per_second": 1.936,
|
| 787 |
+
"eval_wer": 0.11353711790393013,
|
| 788 |
+
"step": 10000
|
| 789 |
+
},
|
| 790 |
+
{
|
| 791 |
+
"epoch": 17.91,
|
| 792 |
+
"learning_rate": 5.182729007633587e-05,
|
| 793 |
+
"loss": 0.5906,
|
| 794 |
+
"step": 10100
|
| 795 |
+
},
|
| 796 |
+
{
|
| 797 |
+
"epoch": 18.09,
|
| 798 |
+
"learning_rate": 5.1541030534351143e-05,
|
| 799 |
+
"loss": 0.5843,
|
| 800 |
+
"step": 10200
|
| 801 |
+
},
|
| 802 |
+
{
|
| 803 |
+
"epoch": 18.26,
|
| 804 |
+
"learning_rate": 5.1257633587786254e-05,
|
| 805 |
+
"loss": 0.5976,
|
| 806 |
+
"step": 10300
|
| 807 |
+
},
|
| 808 |
+
{
|
| 809 |
+
"epoch": 18.44,
|
| 810 |
+
"learning_rate": 5.0971374045801525e-05,
|
| 811 |
+
"loss": 0.59,
|
| 812 |
+
"step": 10400
|
| 813 |
+
},
|
| 814 |
+
{
|
| 815 |
+
"epoch": 18.62,
|
| 816 |
+
"learning_rate": 5.068511450381679e-05,
|
| 817 |
+
"loss": 0.5876,
|
| 818 |
+
"step": 10500
|
| 819 |
+
},
|
| 820 |
+
{
|
| 821 |
+
"epoch": 18.62,
|
| 822 |
+
"eval_loss": 0.10231117904186249,
|
| 823 |
+
"eval_runtime": 31.0041,
|
| 824 |
+
"eval_samples_per_second": 5.741,
|
| 825 |
+
"eval_steps_per_second": 1.935,
|
| 826 |
+
"eval_wer": 0.11588847833389318,
|
| 827 |
+
"step": 10500
|
| 828 |
+
},
|
| 829 |
+
{
|
| 830 |
+
"epoch": 18.79,
|
| 831 |
+
"learning_rate": 5.039885496183206e-05,
|
| 832 |
+
"loss": 0.5845,
|
| 833 |
+
"step": 10600
|
| 834 |
+
},
|
| 835 |
+
{
|
| 836 |
+
"epoch": 18.97,
|
| 837 |
+
"learning_rate": 5.011259541984732e-05,
|
| 838 |
+
"loss": 0.5812,
|
| 839 |
+
"step": 10700
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 19.15,
|
| 843 |
+
"learning_rate": 4.982633587786259e-05,
|
| 844 |
+
"loss": 0.5695,
|
| 845 |
+
"step": 10800
|
| 846 |
+
},
|
| 847 |
+
{
|
| 848 |
+
"epoch": 19.33,
|
| 849 |
+
"learning_rate": 4.954007633587786e-05,
|
| 850 |
+
"loss": 0.5665,
|
| 851 |
+
"step": 10900
|
| 852 |
+
},
|
| 853 |
+
{
|
| 854 |
+
"epoch": 19.5,
|
| 855 |
+
"learning_rate": 4.9253816793893125e-05,
|
| 856 |
+
"loss": 0.5717,
|
| 857 |
+
"step": 11000
|
| 858 |
+
},
|
| 859 |
+
{
|
| 860 |
+
"epoch": 19.5,
|
| 861 |
+
"eval_loss": 0.10367337614297867,
|
| 862 |
+
"eval_runtime": 31.2366,
|
| 863 |
+
"eval_samples_per_second": 5.698,
|
| 864 |
+
"eval_steps_per_second": 1.921,
|
| 865 |
+
"eval_wer": 0.1232784682566342,
|
| 866 |
+
"step": 11000
|
| 867 |
+
},
|
| 868 |
+
{
|
| 869 |
+
"epoch": 19.68,
|
| 870 |
+
"learning_rate": 4.8967557251908396e-05,
|
| 871 |
+
"loss": 0.5657,
|
| 872 |
+
"step": 11100
|
| 873 |
+
},
|
| 874 |
+
{
|
| 875 |
+
"epoch": 19.86,
|
| 876 |
+
"learning_rate": 4.868129770992366e-05,
|
| 877 |
+
"loss": 0.5733,
|
| 878 |
+
"step": 11200
|
| 879 |
+
},
|
| 880 |
+
{
|
| 881 |
+
"epoch": 20.04,
|
| 882 |
+
"learning_rate": 4.839503816793893e-05,
|
| 883 |
+
"loss": 0.5816,
|
| 884 |
+
"step": 11300
|
| 885 |
+
},
|
| 886 |
+
{
|
| 887 |
+
"epoch": 20.21,
|
| 888 |
+
"learning_rate": 4.810877862595419e-05,
|
| 889 |
+
"loss": 0.5694,
|
| 890 |
+
"step": 11400
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 20.39,
|
| 894 |
+
"learning_rate": 4.782251908396946e-05,
|
| 895 |
+
"loss": 0.5537,
|
| 896 |
+
"step": 11500
|
| 897 |
+
},
|
| 898 |
+
{
|
| 899 |
+
"epoch": 20.39,
|
| 900 |
+
"eval_loss": 0.10704291611909866,
|
| 901 |
+
"eval_runtime": 31.0837,
|
| 902 |
+
"eval_samples_per_second": 5.726,
|
| 903 |
+
"eval_steps_per_second": 1.93,
|
| 904 |
+
"eval_wer": 0.11924756466241182,
|
| 905 |
+
"step": 11500
|
| 906 |
+
},
|
| 907 |
+
{
|
| 908 |
+
"epoch": 20.57,
|
| 909 |
+
"learning_rate": 4.7536259541984726e-05,
|
| 910 |
+
"loss": 0.5726,
|
| 911 |
+
"step": 11600
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 20.74,
|
| 915 |
+
"learning_rate": 4.7249999999999997e-05,
|
| 916 |
+
"loss": 0.5742,
|
| 917 |
+
"step": 11700
|
| 918 |
+
},
|
| 919 |
+
{
|
| 920 |
+
"epoch": 20.92,
|
| 921 |
+
"learning_rate": 4.696374045801527e-05,
|
| 922 |
+
"loss": 0.563,
|
| 923 |
+
"step": 11800
|
| 924 |
+
},
|
| 925 |
+
{
|
| 926 |
+
"epoch": 21.1,
|
| 927 |
+
"learning_rate": 4.667748091603053e-05,
|
| 928 |
+
"loss": 0.5655,
|
| 929 |
+
"step": 11900
|
| 930 |
+
},
|
| 931 |
+
{
|
| 932 |
+
"epoch": 21.28,
|
| 933 |
+
"learning_rate": 4.63912213740458e-05,
|
| 934 |
+
"loss": 0.5636,
|
| 935 |
+
"step": 12000
|
| 936 |
+
},
|
| 937 |
+
{
|
| 938 |
+
"epoch": 21.28,
|
| 939 |
+
"eval_loss": 0.10363561660051346,
|
| 940 |
+
"eval_runtime": 31.1987,
|
| 941 |
+
"eval_samples_per_second": 5.705,
|
| 942 |
+
"eval_steps_per_second": 1.923,
|
| 943 |
+
"eval_wer": 0.11689620423244877,
|
| 944 |
+
"step": 12000
|
| 945 |
+
},
|
| 946 |
+
{
|
| 947 |
+
"epoch": 21.45,
|
| 948 |
+
"learning_rate": 4.6104961832061064e-05,
|
| 949 |
+
"loss": 0.5612,
|
| 950 |
+
"step": 12100
|
| 951 |
+
},
|
| 952 |
+
{
|
| 953 |
+
"epoch": 21.63,
|
| 954 |
+
"learning_rate": 4.5818702290076334e-05,
|
| 955 |
+
"loss": 0.5732,
|
| 956 |
+
"step": 12200
|
| 957 |
+
},
|
| 958 |
+
{
|
| 959 |
+
"epoch": 21.81,
|
| 960 |
+
"learning_rate": 4.55324427480916e-05,
|
| 961 |
+
"loss": 0.5726,
|
| 962 |
+
"step": 12300
|
| 963 |
+
},
|
| 964 |
+
{
|
| 965 |
+
"epoch": 21.99,
|
| 966 |
+
"learning_rate": 4.524618320610687e-05,
|
| 967 |
+
"loss": 0.5547,
|
| 968 |
+
"step": 12400
|
| 969 |
+
},
|
| 970 |
+
{
|
| 971 |
+
"epoch": 22.16,
|
| 972 |
+
"learning_rate": 4.495992366412213e-05,
|
| 973 |
+
"loss": 0.5536,
|
| 974 |
+
"step": 12500
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 22.16,
|
| 978 |
+
"eval_loss": 0.10082551091909409,
|
| 979 |
+
"eval_runtime": 31.121,
|
| 980 |
+
"eval_samples_per_second": 5.72,
|
| 981 |
+
"eval_steps_per_second": 1.928,
|
| 982 |
+
"eval_wer": 0.11823983876385623,
|
| 983 |
+
"step": 12500
|
| 984 |
+
},
|
| 985 |
+
{
|
| 986 |
+
"epoch": 22.34,
|
| 987 |
+
"learning_rate": 4.46736641221374e-05,
|
| 988 |
+
"loss": 0.5509,
|
| 989 |
+
"step": 12600
|
| 990 |
+
},
|
| 991 |
+
{
|
| 992 |
+
"epoch": 22.52,
|
| 993 |
+
"learning_rate": 4.439026717557252e-05,
|
| 994 |
+
"loss": 0.5645,
|
| 995 |
+
"step": 12700
|
| 996 |
+
},
|
| 997 |
+
{
|
| 998 |
+
"epoch": 22.69,
|
| 999 |
+
"learning_rate": 4.410400763358778e-05,
|
| 1000 |
+
"loss": 0.5527,
|
| 1001 |
+
"step": 12800
|
| 1002 |
+
},
|
| 1003 |
+
{
|
| 1004 |
+
"epoch": 22.87,
|
| 1005 |
+
"learning_rate": 4.381774809160305e-05,
|
| 1006 |
+
"loss": 0.547,
|
| 1007 |
+
"step": 12900
|
| 1008 |
+
},
|
| 1009 |
+
{
|
| 1010 |
+
"epoch": 23.05,
|
| 1011 |
+
"learning_rate": 4.3534351145038163e-05,
|
| 1012 |
+
"loss": 0.5656,
|
| 1013 |
+
"step": 13000
|
| 1014 |
+
},
|
| 1015 |
+
{
|
| 1016 |
+
"epoch": 23.05,
|
| 1017 |
+
"eval_loss": 0.10101909935474396,
|
| 1018 |
+
"eval_runtime": 30.9319,
|
| 1019 |
+
"eval_samples_per_second": 5.755,
|
| 1020 |
+
"eval_steps_per_second": 1.94,
|
| 1021 |
+
"eval_wer": 0.11723211286530064,
|
| 1022 |
+
"step": 13000
|
| 1023 |
+
},
|
| 1024 |
+
{
|
| 1025 |
+
"epoch": 23.23,
|
| 1026 |
+
"learning_rate": 4.3248091603053434e-05,
|
| 1027 |
+
"loss": 0.5483,
|
| 1028 |
+
"step": 13100
|
| 1029 |
+
},
|
| 1030 |
+
{
|
| 1031 |
+
"epoch": 23.4,
|
| 1032 |
+
"learning_rate": 4.29618320610687e-05,
|
| 1033 |
+
"loss": 0.5501,
|
| 1034 |
+
"step": 13200
|
| 1035 |
+
},
|
| 1036 |
+
{
|
| 1037 |
+
"epoch": 23.58,
|
| 1038 |
+
"learning_rate": 4.267557251908397e-05,
|
| 1039 |
+
"loss": 0.5429,
|
| 1040 |
+
"step": 13300
|
| 1041 |
+
},
|
| 1042 |
+
{
|
| 1043 |
+
"epoch": 23.76,
|
| 1044 |
+
"learning_rate": 4.238931297709923e-05,
|
| 1045 |
+
"loss": 0.5455,
|
| 1046 |
+
"step": 13400
|
| 1047 |
+
},
|
| 1048 |
+
{
|
| 1049 |
+
"epoch": 23.94,
|
| 1050 |
+
"learning_rate": 4.21030534351145e-05,
|
| 1051 |
+
"loss": 0.5504,
|
| 1052 |
+
"step": 13500
|
| 1053 |
+
},
|
| 1054 |
+
{
|
| 1055 |
+
"epoch": 23.94,
|
| 1056 |
+
"eval_loss": 0.10192937403917313,
|
| 1057 |
+
"eval_runtime": 31.0113,
|
| 1058 |
+
"eval_samples_per_second": 5.74,
|
| 1059 |
+
"eval_steps_per_second": 1.935,
|
| 1060 |
+
"eval_wer": 0.11051394020826336,
|
| 1061 |
+
"step": 13500
|
| 1062 |
+
},
|
| 1063 |
+
{
|
| 1064 |
+
"epoch": 24.11,
|
| 1065 |
+
"learning_rate": 4.181679389312977e-05,
|
| 1066 |
+
"loss": 0.5601,
|
| 1067 |
+
"step": 13600
|
| 1068 |
+
},
|
| 1069 |
+
{
|
| 1070 |
+
"epoch": 24.29,
|
| 1071 |
+
"learning_rate": 4.1530534351145035e-05,
|
| 1072 |
+
"loss": 0.5419,
|
| 1073 |
+
"step": 13700
|
| 1074 |
+
},
|
| 1075 |
+
{
|
| 1076 |
+
"epoch": 24.47,
|
| 1077 |
+
"learning_rate": 4.1244274809160305e-05,
|
| 1078 |
+
"loss": 0.5389,
|
| 1079 |
+
"step": 13800
|
| 1080 |
+
},
|
| 1081 |
+
{
|
| 1082 |
+
"epoch": 24.65,
|
| 1083 |
+
"learning_rate": 4.095801526717557e-05,
|
| 1084 |
+
"loss": 0.5572,
|
| 1085 |
+
"step": 13900
|
| 1086 |
+
},
|
| 1087 |
+
{
|
| 1088 |
+
"epoch": 24.82,
|
| 1089 |
+
"learning_rate": 4.067175572519084e-05,
|
| 1090 |
+
"loss": 0.5476,
|
| 1091 |
+
"step": 14000
|
| 1092 |
+
},
|
| 1093 |
+
{
|
| 1094 |
+
"epoch": 24.82,
|
| 1095 |
+
"eval_loss": 0.10260963439941406,
|
| 1096 |
+
"eval_runtime": 31.1565,
|
| 1097 |
+
"eval_samples_per_second": 5.713,
|
| 1098 |
+
"eval_steps_per_second": 1.926,
|
| 1099 |
+
"eval_wer": 0.11656029559959691,
|
| 1100 |
+
"step": 14000
|
| 1101 |
+
},
|
| 1102 |
+
{
|
| 1103 |
+
"epoch": 25.0,
|
| 1104 |
+
"learning_rate": 4.03854961832061e-05,
|
| 1105 |
+
"loss": 0.5554,
|
| 1106 |
+
"step": 14100
|
| 1107 |
+
},
|
| 1108 |
+
{
|
| 1109 |
+
"epoch": 25.18,
|
| 1110 |
+
"learning_rate": 4.009923664122137e-05,
|
| 1111 |
+
"loss": 0.537,
|
| 1112 |
+
"step": 14200
|
| 1113 |
+
},
|
| 1114 |
+
{
|
| 1115 |
+
"epoch": 25.35,
|
| 1116 |
+
"learning_rate": 3.9812977099236635e-05,
|
| 1117 |
+
"loss": 0.5346,
|
| 1118 |
+
"step": 14300
|
| 1119 |
+
},
|
| 1120 |
+
{
|
| 1121 |
+
"epoch": 25.53,
|
| 1122 |
+
"learning_rate": 3.9526717557251906e-05,
|
| 1123 |
+
"loss": 0.5245,
|
| 1124 |
+
"step": 14400
|
| 1125 |
+
},
|
| 1126 |
+
{
|
| 1127 |
+
"epoch": 25.71,
|
| 1128 |
+
"learning_rate": 3.9240458015267176e-05,
|
| 1129 |
+
"loss": 0.5375,
|
| 1130 |
+
"step": 14500
|
| 1131 |
+
},
|
| 1132 |
+
{
|
| 1133 |
+
"epoch": 25.71,
|
| 1134 |
+
"eval_loss": 0.1107296496629715,
|
| 1135 |
+
"eval_runtime": 30.9057,
|
| 1136 |
+
"eval_samples_per_second": 5.759,
|
| 1137 |
+
"eval_steps_per_second": 1.941,
|
| 1138 |
+
"eval_wer": 0.11891165602955996,
|
| 1139 |
+
"step": 14500
|
| 1140 |
+
},
|
| 1141 |
+
{
|
| 1142 |
+
"epoch": 25.89,
|
| 1143 |
+
"learning_rate": 3.895419847328244e-05,
|
| 1144 |
+
"loss": 0.5437,
|
| 1145 |
+
"step": 14600
|
| 1146 |
+
},
|
| 1147 |
+
{
|
| 1148 |
+
"epoch": 26.06,
|
| 1149 |
+
"learning_rate": 3.866793893129771e-05,
|
| 1150 |
+
"loss": 0.5284,
|
| 1151 |
+
"step": 14700
|
| 1152 |
+
},
|
| 1153 |
+
{
|
| 1154 |
+
"epoch": 26.24,
|
| 1155 |
+
"learning_rate": 3.838167938931297e-05,
|
| 1156 |
+
"loss": 0.5276,
|
| 1157 |
+
"step": 14800
|
| 1158 |
+
},
|
| 1159 |
+
{
|
| 1160 |
+
"epoch": 26.42,
|
| 1161 |
+
"learning_rate": 3.809541984732824e-05,
|
| 1162 |
+
"loss": 0.5441,
|
| 1163 |
+
"step": 14900
|
| 1164 |
+
},
|
| 1165 |
+
{
|
| 1166 |
+
"epoch": 26.6,
|
| 1167 |
+
"learning_rate": 3.7809160305343507e-05,
|
| 1168 |
+
"loss": 0.5318,
|
| 1169 |
+
"step": 15000
|
| 1170 |
+
},
|
| 1171 |
+
{
|
| 1172 |
+
"epoch": 26.6,
|
| 1173 |
+
"eval_loss": 0.10514429956674576,
|
| 1174 |
+
"eval_runtime": 31.3716,
|
| 1175 |
+
"eval_samples_per_second": 5.674,
|
| 1176 |
+
"eval_steps_per_second": 1.913,
|
| 1177 |
+
"eval_wer": 0.11420893516963386,
|
| 1178 |
+
"step": 15000
|
| 1179 |
+
},
|
| 1180 |
+
{
|
| 1181 |
+
"epoch": 26.77,
|
| 1182 |
+
"learning_rate": 3.752290076335878e-05,
|
| 1183 |
+
"loss": 0.5343,
|
| 1184 |
+
"step": 15100
|
| 1185 |
+
},
|
| 1186 |
+
{
|
| 1187 |
+
"epoch": 26.95,
|
| 1188 |
+
"learning_rate": 3.723664122137404e-05,
|
| 1189 |
+
"loss": 0.5394,
|
| 1190 |
+
"step": 15200
|
| 1191 |
+
},
|
| 1192 |
+
{
|
| 1193 |
+
"epoch": 27.13,
|
| 1194 |
+
"learning_rate": 3.695038167938931e-05,
|
| 1195 |
+
"loss": 0.5352,
|
| 1196 |
+
"step": 15300
|
| 1197 |
+
},
|
| 1198 |
+
{
|
| 1199 |
+
"epoch": 27.3,
|
| 1200 |
+
"learning_rate": 3.6664122137404574e-05,
|
| 1201 |
+
"loss": 0.5398,
|
| 1202 |
+
"step": 15400
|
| 1203 |
+
},
|
| 1204 |
+
{
|
| 1205 |
+
"epoch": 27.48,
|
| 1206 |
+
"learning_rate": 3.6377862595419844e-05,
|
| 1207 |
+
"loss": 0.5278,
|
| 1208 |
+
"step": 15500
|
| 1209 |
+
},
|
| 1210 |
+
{
|
| 1211 |
+
"epoch": 27.48,
|
| 1212 |
+
"eval_loss": 0.10493182390928268,
|
| 1213 |
+
"eval_runtime": 31.3849,
|
| 1214 |
+
"eval_samples_per_second": 5.672,
|
| 1215 |
+
"eval_steps_per_second": 1.912,
|
| 1216 |
+
"eval_wer": 0.11656029559959691,
|
| 1217 |
+
"step": 15500
|
| 1218 |
+
},
|
| 1219 |
+
{
|
| 1220 |
+
"epoch": 27.66,
|
| 1221 |
+
"learning_rate": 3.609160305343511e-05,
|
| 1222 |
+
"loss": 0.538,
|
| 1223 |
+
"step": 15600
|
| 1224 |
+
},
|
| 1225 |
+
{
|
| 1226 |
+
"epoch": 27.84,
|
| 1227 |
+
"learning_rate": 3.580534351145038e-05,
|
| 1228 |
+
"loss": 0.5143,
|
| 1229 |
+
"step": 15700
|
| 1230 |
+
},
|
| 1231 |
+
{
|
| 1232 |
+
"epoch": 28.01,
|
| 1233 |
+
"learning_rate": 3.551908396946565e-05,
|
| 1234 |
+
"loss": 0.5229,
|
| 1235 |
+
"step": 15800
|
| 1236 |
+
},
|
| 1237 |
+
{
|
| 1238 |
+
"epoch": 28.19,
|
| 1239 |
+
"learning_rate": 3.523282442748091e-05,
|
| 1240 |
+
"loss": 0.522,
|
| 1241 |
+
"step": 15900
|
| 1242 |
+
},
|
| 1243 |
+
{
|
| 1244 |
+
"epoch": 28.37,
|
| 1245 |
+
"learning_rate": 3.494656488549618e-05,
|
| 1246 |
+
"loss": 0.5204,
|
| 1247 |
+
"step": 16000
|
| 1248 |
+
},
|
| 1249 |
+
{
|
| 1250 |
+
"epoch": 28.37,
|
| 1251 |
+
"eval_loss": 0.10806475579738617,
|
| 1252 |
+
"eval_runtime": 31.4114,
|
| 1253 |
+
"eval_samples_per_second": 5.667,
|
| 1254 |
+
"eval_steps_per_second": 1.91,
|
| 1255 |
+
"eval_wer": 0.11823983876385623,
|
| 1256 |
+
"step": 16000
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 28.55,
|
| 1260 |
+
"learning_rate": 3.4660305343511445e-05,
|
| 1261 |
+
"loss": 0.5322,
|
| 1262 |
+
"step": 16100
|
| 1263 |
+
},
|
| 1264 |
+
{
|
| 1265 |
+
"epoch": 28.72,
|
| 1266 |
+
"learning_rate": 3.4374045801526715e-05,
|
| 1267 |
+
"loss": 0.5151,
|
| 1268 |
+
"step": 16200
|
| 1269 |
+
},
|
| 1270 |
+
{
|
| 1271 |
+
"epoch": 28.9,
|
| 1272 |
+
"learning_rate": 3.408778625954198e-05,
|
| 1273 |
+
"loss": 0.5393,
|
| 1274 |
+
"step": 16300
|
| 1275 |
+
},
|
| 1276 |
+
{
|
| 1277 |
+
"epoch": 29.08,
|
| 1278 |
+
"learning_rate": 3.380152671755725e-05,
|
| 1279 |
+
"loss": 0.513,
|
| 1280 |
+
"step": 16400
|
| 1281 |
+
},
|
| 1282 |
+
{
|
| 1283 |
+
"epoch": 29.26,
|
| 1284 |
+
"learning_rate": 3.351526717557251e-05,
|
| 1285 |
+
"loss": 0.512,
|
| 1286 |
+
"step": 16500
|
| 1287 |
+
},
|
| 1288 |
+
{
|
| 1289 |
+
"epoch": 29.26,
|
| 1290 |
+
"eval_loss": 0.10623880475759506,
|
| 1291 |
+
"eval_runtime": 31.0174,
|
| 1292 |
+
"eval_samples_per_second": 5.739,
|
| 1293 |
+
"eval_steps_per_second": 1.934,
|
| 1294 |
+
"eval_wer": 0.11555256970104132,
|
| 1295 |
+
"step": 16500
|
| 1296 |
+
},
|
| 1297 |
+
{
|
| 1298 |
+
"epoch": 29.43,
|
| 1299 |
+
"learning_rate": 3.322900763358778e-05,
|
| 1300 |
+
"loss": 0.5258,
|
| 1301 |
+
"step": 16600
|
| 1302 |
+
},
|
| 1303 |
+
{
|
| 1304 |
+
"epoch": 29.61,
|
| 1305 |
+
"learning_rate": 3.294274809160305e-05,
|
| 1306 |
+
"loss": 0.5366,
|
| 1307 |
+
"step": 16700
|
| 1308 |
+
},
|
| 1309 |
+
{
|
| 1310 |
+
"epoch": 29.79,
|
| 1311 |
+
"learning_rate": 3.2656488549618316e-05,
|
| 1312 |
+
"loss": 0.515,
|
| 1313 |
+
"step": 16800
|
| 1314 |
+
},
|
| 1315 |
+
{
|
| 1316 |
+
"epoch": 29.96,
|
| 1317 |
+
"learning_rate": 3.2370229007633586e-05,
|
| 1318 |
+
"loss": 0.5246,
|
| 1319 |
+
"step": 16900
|
| 1320 |
+
},
|
| 1321 |
+
{
|
| 1322 |
+
"epoch": 30.14,
|
| 1323 |
+
"learning_rate": 3.208396946564885e-05,
|
| 1324 |
+
"loss": 0.5082,
|
| 1325 |
+
"step": 17000
|
| 1326 |
+
},
|
| 1327 |
+
{
|
| 1328 |
+
"epoch": 30.14,
|
| 1329 |
+
"eval_loss": 0.10452543944120407,
|
| 1330 |
+
"eval_runtime": 30.9393,
|
| 1331 |
+
"eval_samples_per_second": 5.753,
|
| 1332 |
+
"eval_steps_per_second": 1.939,
|
| 1333 |
+
"eval_wer": 0.11353711790393013,
|
| 1334 |
+
"step": 17000
|
| 1335 |
+
},
|
| 1336 |
+
{
|
| 1337 |
+
"epoch": 30.32,
|
| 1338 |
+
"learning_rate": 3.179770992366412e-05,
|
| 1339 |
+
"loss": 0.5029,
|
| 1340 |
+
"step": 17100
|
| 1341 |
+
},
|
| 1342 |
+
{
|
| 1343 |
+
"epoch": 30.5,
|
| 1344 |
+
"learning_rate": 3.151431297709923e-05,
|
| 1345 |
+
"loss": 0.5219,
|
| 1346 |
+
"step": 17200
|
| 1347 |
+
},
|
| 1348 |
+
{
|
| 1349 |
+
"epoch": 30.67,
|
| 1350 |
+
"learning_rate": 3.12280534351145e-05,
|
| 1351 |
+
"loss": 0.5235,
|
| 1352 |
+
"step": 17300
|
| 1353 |
+
},
|
| 1354 |
+
{
|
| 1355 |
+
"epoch": 30.85,
|
| 1356 |
+
"learning_rate": 3.0941793893129764e-05,
|
| 1357 |
+
"loss": 0.5141,
|
| 1358 |
+
"step": 17400
|
| 1359 |
+
},
|
| 1360 |
+
{
|
| 1361 |
+
"epoch": 31.03,
|
| 1362 |
+
"learning_rate": 3.0655534351145035e-05,
|
| 1363 |
+
"loss": 0.5193,
|
| 1364 |
+
"step": 17500
|
| 1365 |
+
},
|
| 1366 |
+
{
|
| 1367 |
+
"epoch": 31.03,
|
| 1368 |
+
"eval_loss": 0.10911107808351517,
|
| 1369 |
+
"eval_runtime": 30.9608,
|
| 1370 |
+
"eval_samples_per_second": 5.749,
|
| 1371 |
+
"eval_steps_per_second": 1.938,
|
| 1372 |
+
"eval_wer": 0.11454484380248572,
|
| 1373 |
+
"step": 17500
|
| 1374 |
+
},
|
| 1375 |
+
{
|
| 1376 |
+
"epoch": 31.21,
|
| 1377 |
+
"learning_rate": 3.0369274809160305e-05,
|
| 1378 |
+
"loss": 0.5038,
|
| 1379 |
+
"step": 17600
|
| 1380 |
+
},
|
| 1381 |
+
{
|
| 1382 |
+
"epoch": 31.38,
|
| 1383 |
+
"learning_rate": 3.008301526717557e-05,
|
| 1384 |
+
"loss": 0.5086,
|
| 1385 |
+
"step": 17700
|
| 1386 |
+
},
|
| 1387 |
+
{
|
| 1388 |
+
"epoch": 31.56,
|
| 1389 |
+
"learning_rate": 2.979675572519084e-05,
|
| 1390 |
+
"loss": 0.5001,
|
| 1391 |
+
"step": 17800
|
| 1392 |
+
},
|
| 1393 |
+
{
|
| 1394 |
+
"epoch": 31.74,
|
| 1395 |
+
"learning_rate": 2.9510496183206105e-05,
|
| 1396 |
+
"loss": 0.5206,
|
| 1397 |
+
"step": 17900
|
| 1398 |
+
},
|
| 1399 |
+
{
|
| 1400 |
+
"epoch": 31.91,
|
| 1401 |
+
"learning_rate": 2.9224236641221372e-05,
|
| 1402 |
+
"loss": 0.5129,
|
| 1403 |
+
"step": 18000
|
| 1404 |
+
},
|
| 1405 |
+
{
|
| 1406 |
+
"epoch": 31.91,
|
| 1407 |
+
"eval_loss": 0.10398419201374054,
|
| 1408 |
+
"eval_runtime": 30.9924,
|
| 1409 |
+
"eval_samples_per_second": 5.743,
|
| 1410 |
+
"eval_steps_per_second": 1.936,
|
| 1411 |
+
"eval_wer": 0.10883439704400404,
|
| 1412 |
+
"step": 18000
|
| 1413 |
+
},
|
| 1414 |
+
{
|
| 1415 |
+
"epoch": 32.09,
|
| 1416 |
+
"learning_rate": 2.893797709923664e-05,
|
| 1417 |
+
"loss": 0.5105,
|
| 1418 |
+
"step": 18100
|
| 1419 |
+
},
|
| 1420 |
+
{
|
| 1421 |
+
"epoch": 32.27,
|
| 1422 |
+
"learning_rate": 2.8651717557251906e-05,
|
| 1423 |
+
"loss": 0.5062,
|
| 1424 |
+
"step": 18200
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 32.45,
|
| 1428 |
+
"learning_rate": 2.8365458015267172e-05,
|
| 1429 |
+
"loss": 0.5021,
|
| 1430 |
+
"step": 18300
|
| 1431 |
+
},
|
| 1432 |
+
{
|
| 1433 |
+
"epoch": 32.62,
|
| 1434 |
+
"learning_rate": 2.807919847328244e-05,
|
| 1435 |
+
"loss": 0.5122,
|
| 1436 |
+
"step": 18400
|
| 1437 |
+
},
|
| 1438 |
+
{
|
| 1439 |
+
"epoch": 32.8,
|
| 1440 |
+
"learning_rate": 2.779293893129771e-05,
|
| 1441 |
+
"loss": 0.5126,
|
| 1442 |
+
"step": 18500
|
| 1443 |
+
},
|
| 1444 |
+
{
|
| 1445 |
+
"epoch": 32.8,
|
| 1446 |
+
"eval_loss": 0.10847991704940796,
|
| 1447 |
+
"eval_runtime": 30.8721,
|
| 1448 |
+
"eval_samples_per_second": 5.766,
|
| 1449 |
+
"eval_steps_per_second": 1.944,
|
| 1450 |
+
"eval_wer": 0.11689620423244877,
|
| 1451 |
+
"step": 18500
|
| 1452 |
+
},
|
| 1453 |
+
{
|
| 1454 |
+
"epoch": 32.98,
|
| 1455 |
+
"learning_rate": 2.7506679389312976e-05,
|
| 1456 |
+
"loss": 0.5179,
|
| 1457 |
+
"step": 18600
|
| 1458 |
+
},
|
| 1459 |
+
{
|
| 1460 |
+
"epoch": 33.16,
|
| 1461 |
+
"learning_rate": 2.7220419847328243e-05,
|
| 1462 |
+
"loss": 0.5049,
|
| 1463 |
+
"step": 18700
|
| 1464 |
+
},
|
| 1465 |
+
{
|
| 1466 |
+
"epoch": 33.33,
|
| 1467 |
+
"learning_rate": 2.693416030534351e-05,
|
| 1468 |
+
"loss": 0.5173,
|
| 1469 |
+
"step": 18800
|
| 1470 |
+
},
|
| 1471 |
+
{
|
| 1472 |
+
"epoch": 33.51,
|
| 1473 |
+
"learning_rate": 2.6647900763358777e-05,
|
| 1474 |
+
"loss": 0.5115,
|
| 1475 |
+
"step": 18900
|
| 1476 |
+
},
|
| 1477 |
+
{
|
| 1478 |
+
"epoch": 33.69,
|
| 1479 |
+
"learning_rate": 2.6361641221374043e-05,
|
| 1480 |
+
"loss": 0.496,
|
| 1481 |
+
"step": 19000
|
| 1482 |
+
},
|
| 1483 |
+
{
|
| 1484 |
+
"epoch": 33.69,
|
| 1485 |
+
"eval_loss": 0.10698471963405609,
|
| 1486 |
+
"eval_runtime": 31.005,
|
| 1487 |
+
"eval_samples_per_second": 5.741,
|
| 1488 |
+
"eval_steps_per_second": 1.935,
|
| 1489 |
+
"eval_wer": 0.11656029559959691,
|
| 1490 |
+
"step": 19000
|
| 1491 |
+
},
|
| 1492 |
+
{
|
| 1493 |
+
"epoch": 33.86,
|
| 1494 |
+
"learning_rate": 2.607538167938931e-05,
|
| 1495 |
+
"loss": 0.4938,
|
| 1496 |
+
"step": 19100
|
| 1497 |
+
},
|
| 1498 |
+
{
|
| 1499 |
+
"epoch": 34.04,
|
| 1500 |
+
"learning_rate": 2.5789122137404577e-05,
|
| 1501 |
+
"loss": 0.5128,
|
| 1502 |
+
"step": 19200
|
| 1503 |
+
},
|
| 1504 |
+
{
|
| 1505 |
+
"epoch": 34.22,
|
| 1506 |
+
"learning_rate": 2.550572519083969e-05,
|
| 1507 |
+
"loss": 0.497,
|
| 1508 |
+
"step": 19300
|
| 1509 |
+
},
|
| 1510 |
+
{
|
| 1511 |
+
"epoch": 34.4,
|
| 1512 |
+
"learning_rate": 2.521946564885496e-05,
|
| 1513 |
+
"loss": 0.4879,
|
| 1514 |
+
"step": 19400
|
| 1515 |
+
},
|
| 1516 |
+
{
|
| 1517 |
+
"epoch": 34.57,
|
| 1518 |
+
"learning_rate": 2.493320610687023e-05,
|
| 1519 |
+
"loss": 0.5017,
|
| 1520 |
+
"step": 19500
|
| 1521 |
+
},
|
| 1522 |
+
{
|
| 1523 |
+
"epoch": 34.57,
|
| 1524 |
+
"eval_loss": 0.11190272867679596,
|
| 1525 |
+
"eval_runtime": 30.8238,
|
| 1526 |
+
"eval_samples_per_second": 5.775,
|
| 1527 |
+
"eval_steps_per_second": 1.947,
|
| 1528 |
+
"eval_wer": 0.11622438696674504,
|
| 1529 |
+
"step": 19500
|
| 1530 |
+
},
|
| 1531 |
+
{
|
| 1532 |
+
"epoch": 34.75,
|
| 1533 |
+
"learning_rate": 2.4646946564885495e-05,
|
| 1534 |
+
"loss": 0.4994,
|
| 1535 |
+
"step": 19600
|
| 1536 |
+
},
|
| 1537 |
+
{
|
| 1538 |
+
"epoch": 34.93,
|
| 1539 |
+
"learning_rate": 2.436354961832061e-05,
|
| 1540 |
+
"loss": 0.4973,
|
| 1541 |
+
"step": 19700
|
| 1542 |
+
},
|
| 1543 |
+
{
|
| 1544 |
+
"epoch": 35.11,
|
| 1545 |
+
"learning_rate": 2.4077290076335876e-05,
|
| 1546 |
+
"loss": 0.5012,
|
| 1547 |
+
"step": 19800
|
| 1548 |
+
},
|
| 1549 |
+
{
|
| 1550 |
+
"epoch": 35.28,
|
| 1551 |
+
"learning_rate": 2.3791030534351143e-05,
|
| 1552 |
+
"loss": 0.4889,
|
| 1553 |
+
"step": 19900
|
| 1554 |
+
},
|
| 1555 |
+
{
|
| 1556 |
+
"epoch": 35.46,
|
| 1557 |
+
"learning_rate": 2.350477099236641e-05,
|
| 1558 |
+
"loss": 0.4808,
|
| 1559 |
+
"step": 20000
|
| 1560 |
+
},
|
| 1561 |
+
{
|
| 1562 |
+
"epoch": 35.46,
|
| 1563 |
+
"eval_loss": 0.11005562543869019,
|
| 1564 |
+
"eval_runtime": 31.0411,
|
| 1565 |
+
"eval_samples_per_second": 5.734,
|
| 1566 |
+
"eval_steps_per_second": 1.933,
|
| 1567 |
+
"eval_wer": 0.113873026536782,
|
| 1568 |
+
"step": 20000
|
| 1569 |
+
},
|
| 1570 |
+
{
|
| 1571 |
+
"epoch": 35.64,
|
| 1572 |
+
"learning_rate": 2.3218511450381677e-05,
|
| 1573 |
+
"loss": 0.4984,
|
| 1574 |
+
"step": 20100
|
| 1575 |
+
},
|
| 1576 |
+
{
|
| 1577 |
+
"epoch": 35.82,
|
| 1578 |
+
"learning_rate": 2.2932251908396944e-05,
|
| 1579 |
+
"loss": 0.4929,
|
| 1580 |
+
"step": 20200
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 35.99,
|
| 1584 |
+
"learning_rate": 2.2645992366412214e-05,
|
| 1585 |
+
"loss": 0.4861,
|
| 1586 |
+
"step": 20300
|
| 1587 |
+
},
|
| 1588 |
+
{
|
| 1589 |
+
"epoch": 36.17,
|
| 1590 |
+
"learning_rate": 2.235973282442748e-05,
|
| 1591 |
+
"loss": 0.506,
|
| 1592 |
+
"step": 20400
|
| 1593 |
+
},
|
| 1594 |
+
{
|
| 1595 |
+
"epoch": 36.35,
|
| 1596 |
+
"learning_rate": 2.2073473282442747e-05,
|
| 1597 |
+
"loss": 0.4939,
|
| 1598 |
+
"step": 20500
|
| 1599 |
+
},
|
| 1600 |
+
{
|
| 1601 |
+
"epoch": 36.35,
|
| 1602 |
+
"eval_loss": 0.10813739150762558,
|
| 1603 |
+
"eval_runtime": 31.0724,
|
| 1604 |
+
"eval_samples_per_second": 5.729,
|
| 1605 |
+
"eval_steps_per_second": 1.931,
|
| 1606 |
+
"eval_wer": 0.11252939200537454,
|
| 1607 |
+
"step": 20500
|
| 1608 |
+
},
|
| 1609 |
+
{
|
| 1610 |
+
"epoch": 36.52,
|
| 1611 |
+
"learning_rate": 2.1787213740458014e-05,
|
| 1612 |
+
"loss": 0.4789,
|
| 1613 |
+
"step": 20600
|
| 1614 |
+
},
|
| 1615 |
+
{
|
| 1616 |
+
"epoch": 36.7,
|
| 1617 |
+
"learning_rate": 2.150095419847328e-05,
|
| 1618 |
+
"loss": 0.4904,
|
| 1619 |
+
"step": 20700
|
| 1620 |
+
},
|
| 1621 |
+
{
|
| 1622 |
+
"epoch": 36.88,
|
| 1623 |
+
"learning_rate": 2.1214694656488548e-05,
|
| 1624 |
+
"loss": 0.4916,
|
| 1625 |
+
"step": 20800
|
| 1626 |
+
},
|
| 1627 |
+
{
|
| 1628 |
+
"epoch": 37.06,
|
| 1629 |
+
"learning_rate": 2.0928435114503815e-05,
|
| 1630 |
+
"loss": 0.487,
|
| 1631 |
+
"step": 20900
|
| 1632 |
+
},
|
| 1633 |
+
{
|
| 1634 |
+
"epoch": 37.23,
|
| 1635 |
+
"learning_rate": 2.064217557251908e-05,
|
| 1636 |
+
"loss": 0.4738,
|
| 1637 |
+
"step": 21000
|
| 1638 |
+
},
|
| 1639 |
+
{
|
| 1640 |
+
"epoch": 37.23,
|
| 1641 |
+
"eval_loss": 0.10911141335964203,
|
| 1642 |
+
"eval_runtime": 31.0688,
|
| 1643 |
+
"eval_samples_per_second": 5.729,
|
| 1644 |
+
"eval_steps_per_second": 1.931,
|
| 1645 |
+
"eval_wer": 0.10984212294255963,
|
| 1646 |
+
"step": 21000
|
| 1647 |
+
},
|
| 1648 |
+
{
|
| 1649 |
+
"epoch": 37.41,
|
| 1650 |
+
"learning_rate": 2.0355916030534352e-05,
|
| 1651 |
+
"loss": 0.4757,
|
| 1652 |
+
"step": 21100
|
| 1653 |
+
},
|
| 1654 |
+
{
|
| 1655 |
+
"epoch": 37.59,
|
| 1656 |
+
"learning_rate": 2.006965648854962e-05,
|
| 1657 |
+
"loss": 0.4752,
|
| 1658 |
+
"step": 21200
|
| 1659 |
+
},
|
| 1660 |
+
{
|
| 1661 |
+
"epoch": 37.77,
|
| 1662 |
+
"learning_rate": 1.9783396946564885e-05,
|
| 1663 |
+
"loss": 0.4939,
|
| 1664 |
+
"step": 21300
|
| 1665 |
+
},
|
| 1666 |
+
{
|
| 1667 |
+
"epoch": 37.94,
|
| 1668 |
+
"learning_rate": 1.9497137404580152e-05,
|
| 1669 |
+
"loss": 0.4673,
|
| 1670 |
+
"step": 21400
|
| 1671 |
+
},
|
| 1672 |
+
{
|
| 1673 |
+
"epoch": 38.12,
|
| 1674 |
+
"learning_rate": 1.921087786259542e-05,
|
| 1675 |
+
"loss": 0.4978,
|
| 1676 |
+
"step": 21500
|
| 1677 |
+
},
|
| 1678 |
+
{
|
| 1679 |
+
"epoch": 38.12,
|
| 1680 |
+
"eval_loss": 0.10570317506790161,
|
| 1681 |
+
"eval_runtime": 31.0697,
|
| 1682 |
+
"eval_samples_per_second": 5.729,
|
| 1683 |
+
"eval_steps_per_second": 1.931,
|
| 1684 |
+
"eval_wer": 0.1091703056768559,
|
| 1685 |
+
"step": 21500
|
| 1686 |
+
},
|
| 1687 |
+
{
|
| 1688 |
+
"epoch": 38.3,
|
| 1689 |
+
"learning_rate": 1.8924618320610686e-05,
|
| 1690 |
+
"loss": 0.4829,
|
| 1691 |
+
"step": 21600
|
| 1692 |
+
},
|
| 1693 |
+
{
|
| 1694 |
+
"epoch": 38.47,
|
| 1695 |
+
"learning_rate": 1.8638358778625953e-05,
|
| 1696 |
+
"loss": 0.4857,
|
| 1697 |
+
"step": 21700
|
| 1698 |
+
},
|
| 1699 |
+
{
|
| 1700 |
+
"epoch": 38.65,
|
| 1701 |
+
"learning_rate": 1.835209923664122e-05,
|
| 1702 |
+
"loss": 0.4618,
|
| 1703 |
+
"step": 21800
|
| 1704 |
+
},
|
| 1705 |
+
{
|
| 1706 |
+
"epoch": 38.83,
|
| 1707 |
+
"learning_rate": 1.8065839694656486e-05,
|
| 1708 |
+
"loss": 0.4831,
|
| 1709 |
+
"step": 21900
|
| 1710 |
+
},
|
| 1711 |
+
{
|
| 1712 |
+
"epoch": 39.01,
|
| 1713 |
+
"learning_rate": 1.77824427480916e-05,
|
| 1714 |
+
"loss": 0.4972,
|
| 1715 |
+
"step": 22000
|
| 1716 |
+
},
|
| 1717 |
+
{
|
| 1718 |
+
"epoch": 39.01,
|
| 1719 |
+
"eval_loss": 0.10742757469415665,
|
| 1720 |
+
"eval_runtime": 31.1659,
|
| 1721 |
+
"eval_samples_per_second": 5.711,
|
| 1722 |
+
"eval_steps_per_second": 1.925,
|
| 1723 |
+
"eval_wer": 0.11051394020826336,
|
| 1724 |
+
"step": 22000
|
| 1725 |
+
},
|
| 1726 |
+
{
|
| 1727 |
+
"epoch": 39.18,
|
| 1728 |
+
"learning_rate": 1.7496183206106867e-05,
|
| 1729 |
+
"loss": 0.4974,
|
| 1730 |
+
"step": 22100
|
| 1731 |
+
},
|
| 1732 |
+
{
|
| 1733 |
+
"epoch": 39.36,
|
| 1734 |
+
"learning_rate": 1.7209923664122138e-05,
|
| 1735 |
+
"loss": 0.4778,
|
| 1736 |
+
"step": 22200
|
| 1737 |
+
},
|
| 1738 |
+
{
|
| 1739 |
+
"epoch": 39.54,
|
| 1740 |
+
"learning_rate": 1.6923664122137404e-05,
|
| 1741 |
+
"loss": 0.4863,
|
| 1742 |
+
"step": 22300
|
| 1743 |
+
},
|
| 1744 |
+
{
|
| 1745 |
+
"epoch": 39.72,
|
| 1746 |
+
"learning_rate": 1.663740458015267e-05,
|
| 1747 |
+
"loss": 0.4815,
|
| 1748 |
+
"step": 22400
|
| 1749 |
+
},
|
| 1750 |
+
{
|
| 1751 |
+
"epoch": 39.89,
|
| 1752 |
+
"learning_rate": 1.6351145038167938e-05,
|
| 1753 |
+
"loss": 0.4773,
|
| 1754 |
+
"step": 22500
|
| 1755 |
+
},
|
| 1756 |
+
{
|
| 1757 |
+
"epoch": 39.89,
|
| 1758 |
+
"eval_loss": 0.10622948408126831,
|
| 1759 |
+
"eval_runtime": 31.5999,
|
| 1760 |
+
"eval_samples_per_second": 5.633,
|
| 1761 |
+
"eval_steps_per_second": 1.899,
|
| 1762 |
+
"eval_wer": 0.11084984884111522,
|
| 1763 |
+
"step": 22500
|
| 1764 |
+
},
|
| 1765 |
+
{
|
| 1766 |
+
"epoch": 40.07,
|
| 1767 |
+
"learning_rate": 1.6064885496183205e-05,
|
| 1768 |
+
"loss": 0.4808,
|
| 1769 |
+
"step": 22600
|
| 1770 |
+
},
|
| 1771 |
+
{
|
| 1772 |
+
"epoch": 40.25,
|
| 1773 |
+
"learning_rate": 1.577862595419847e-05,
|
| 1774 |
+
"loss": 0.4804,
|
| 1775 |
+
"step": 22700
|
| 1776 |
+
},
|
| 1777 |
+
{
|
| 1778 |
+
"epoch": 40.43,
|
| 1779 |
+
"learning_rate": 1.549236641221374e-05,
|
| 1780 |
+
"loss": 0.4796,
|
| 1781 |
+
"step": 22800
|
| 1782 |
+
},
|
| 1783 |
+
{
|
| 1784 |
+
"epoch": 40.6,
|
| 1785 |
+
"learning_rate": 1.5206106870229005e-05,
|
| 1786 |
+
"loss": 0.4866,
|
| 1787 |
+
"step": 22900
|
| 1788 |
+
},
|
| 1789 |
+
{
|
| 1790 |
+
"epoch": 40.78,
|
| 1791 |
+
"learning_rate": 1.4919847328244272e-05,
|
| 1792 |
+
"loss": 0.4741,
|
| 1793 |
+
"step": 23000
|
| 1794 |
+
},
|
| 1795 |
+
{
|
| 1796 |
+
"epoch": 40.78,
|
| 1797 |
+
"eval_loss": 0.10567349940538406,
|
| 1798 |
+
"eval_runtime": 31.2551,
|
| 1799 |
+
"eval_samples_per_second": 5.695,
|
| 1800 |
+
"eval_steps_per_second": 1.92,
|
| 1801 |
+
"eval_wer": 0.10849848841115217,
|
| 1802 |
+
"step": 23000
|
| 1803 |
+
},
|
| 1804 |
+
{
|
| 1805 |
+
"epoch": 40.96,
|
| 1806 |
+
"learning_rate": 1.4633587786259542e-05,
|
| 1807 |
+
"loss": 0.4976,
|
| 1808 |
+
"step": 23100
|
| 1809 |
+
},
|
| 1810 |
+
{
|
| 1811 |
+
"epoch": 41.13,
|
| 1812 |
+
"learning_rate": 1.4347328244274809e-05,
|
| 1813 |
+
"loss": 0.4811,
|
| 1814 |
+
"step": 23200
|
| 1815 |
+
},
|
| 1816 |
+
{
|
| 1817 |
+
"epoch": 41.31,
|
| 1818 |
+
"learning_rate": 1.4061068702290076e-05,
|
| 1819 |
+
"loss": 0.4904,
|
| 1820 |
+
"step": 23300
|
| 1821 |
+
},
|
| 1822 |
+
{
|
| 1823 |
+
"epoch": 41.49,
|
| 1824 |
+
"learning_rate": 1.3774809160305341e-05,
|
| 1825 |
+
"loss": 0.4658,
|
| 1826 |
+
"step": 23400
|
| 1827 |
+
},
|
| 1828 |
+
{
|
| 1829 |
+
"epoch": 41.67,
|
| 1830 |
+
"learning_rate": 1.3488549618320608e-05,
|
| 1831 |
+
"loss": 0.4776,
|
| 1832 |
+
"step": 23500
|
| 1833 |
+
},
|
| 1834 |
+
{
|
| 1835 |
+
"epoch": 41.67,
|
| 1836 |
+
"eval_loss": 0.10771991312503815,
|
| 1837 |
+
"eval_runtime": 31.4029,
|
| 1838 |
+
"eval_samples_per_second": 5.668,
|
| 1839 |
+
"eval_steps_per_second": 1.911,
|
| 1840 |
+
"eval_wer": 0.10849848841115217,
|
| 1841 |
+
"step": 23500
|
| 1842 |
+
},
|
| 1843 |
+
{
|
| 1844 |
+
"epoch": 41.84,
|
| 1845 |
+
"learning_rate": 1.3202290076335878e-05,
|
| 1846 |
+
"loss": 0.4745,
|
| 1847 |
+
"step": 23600
|
| 1848 |
+
},
|
| 1849 |
+
{
|
| 1850 |
+
"epoch": 42.02,
|
| 1851 |
+
"learning_rate": 1.2916030534351145e-05,
|
| 1852 |
+
"loss": 0.4746,
|
| 1853 |
+
"step": 23700
|
| 1854 |
+
},
|
| 1855 |
+
{
|
| 1856 |
+
"epoch": 42.2,
|
| 1857 |
+
"learning_rate": 1.262977099236641e-05,
|
| 1858 |
+
"loss": 0.4858,
|
| 1859 |
+
"step": 23800
|
| 1860 |
+
},
|
| 1861 |
+
{
|
| 1862 |
+
"epoch": 42.38,
|
| 1863 |
+
"learning_rate": 1.2343511450381677e-05,
|
| 1864 |
+
"loss": 0.4687,
|
| 1865 |
+
"step": 23900
|
| 1866 |
+
},
|
| 1867 |
+
{
|
| 1868 |
+
"epoch": 42.55,
|
| 1869 |
+
"learning_rate": 1.2057251908396947e-05,
|
| 1870 |
+
"loss": 0.4637,
|
| 1871 |
+
"step": 24000
|
| 1872 |
+
},
|
| 1873 |
+
{
|
| 1874 |
+
"epoch": 42.55,
|
| 1875 |
+
"eval_loss": 0.10607700049877167,
|
| 1876 |
+
"eval_runtime": 31.045,
|
| 1877 |
+
"eval_samples_per_second": 5.734,
|
| 1878 |
+
"eval_steps_per_second": 1.933,
|
| 1879 |
+
"eval_wer": 0.10950621430970776,
|
| 1880 |
+
"step": 24000
|
| 1881 |
+
},
|
| 1882 |
+
{
|
| 1883 |
+
"epoch": 42.73,
|
| 1884 |
+
"learning_rate": 1.1770992366412214e-05,
|
| 1885 |
+
"loss": 0.4462,
|
| 1886 |
+
"step": 24100
|
| 1887 |
+
},
|
| 1888 |
+
{
|
| 1889 |
+
"epoch": 42.91,
|
| 1890 |
+
"learning_rate": 1.1484732824427479e-05,
|
| 1891 |
+
"loss": 0.4815,
|
| 1892 |
+
"step": 24200
|
| 1893 |
+
},
|
| 1894 |
+
{
|
| 1895 |
+
"epoch": 43.09,
|
| 1896 |
+
"learning_rate": 1.1198473282442746e-05,
|
| 1897 |
+
"loss": 0.461,
|
| 1898 |
+
"step": 24300
|
| 1899 |
+
},
|
| 1900 |
+
{
|
| 1901 |
+
"epoch": 43.26,
|
| 1902 |
+
"learning_rate": 1.0912213740458016e-05,
|
| 1903 |
+
"loss": 0.4594,
|
| 1904 |
+
"step": 24400
|
| 1905 |
+
},
|
| 1906 |
+
{
|
| 1907 |
+
"epoch": 43.44,
|
| 1908 |
+
"learning_rate": 1.0625954198473283e-05,
|
| 1909 |
+
"loss": 0.4853,
|
| 1910 |
+
"step": 24500
|
| 1911 |
+
},
|
| 1912 |
+
{
|
| 1913 |
+
"epoch": 43.44,
|
| 1914 |
+
"eval_loss": 0.10806521028280258,
|
| 1915 |
+
"eval_runtime": 31.4451,
|
| 1916 |
+
"eval_samples_per_second": 5.661,
|
| 1917 |
+
"eval_steps_per_second": 1.908,
|
| 1918 |
+
"eval_wer": 0.10749076251259658,
|
| 1919 |
+
"step": 24500
|
| 1920 |
+
},
|
| 1921 |
+
{
|
| 1922 |
+
"epoch": 43.62,
|
| 1923 |
+
"learning_rate": 1.033969465648855e-05,
|
| 1924 |
+
"loss": 0.4747,
|
| 1925 |
+
"step": 24600
|
| 1926 |
+
},
|
| 1927 |
+
{
|
| 1928 |
+
"epoch": 43.79,
|
| 1929 |
+
"learning_rate": 1.0053435114503815e-05,
|
| 1930 |
+
"loss": 0.4649,
|
| 1931 |
+
"step": 24700
|
| 1932 |
+
},
|
| 1933 |
+
{
|
| 1934 |
+
"epoch": 43.97,
|
| 1935 |
+
"learning_rate": 9.767175572519081e-06,
|
| 1936 |
+
"loss": 0.471,
|
| 1937 |
+
"step": 24800
|
| 1938 |
+
},
|
| 1939 |
+
{
|
| 1940 |
+
"epoch": 44.15,
|
| 1941 |
+
"learning_rate": 9.480916030534352e-06,
|
| 1942 |
+
"loss": 0.453,
|
| 1943 |
+
"step": 24900
|
| 1944 |
+
},
|
| 1945 |
+
{
|
| 1946 |
+
"epoch": 44.33,
|
| 1947 |
+
"learning_rate": 9.194656488549618e-06,
|
| 1948 |
+
"loss": 0.4602,
|
| 1949 |
+
"step": 25000
|
| 1950 |
+
},
|
| 1951 |
+
{
|
| 1952 |
+
"epoch": 44.33,
|
| 1953 |
+
"eval_loss": 0.1075613871216774,
|
| 1954 |
+
"eval_runtime": 31.217,
|
| 1955 |
+
"eval_samples_per_second": 5.702,
|
| 1956 |
+
"eval_steps_per_second": 1.922,
|
| 1957 |
+
"eval_wer": 0.10849848841115217,
|
| 1958 |
+
"step": 25000
|
| 1959 |
+
},
|
| 1960 |
+
{
|
| 1961 |
+
"epoch": 44.5,
|
| 1962 |
+
"learning_rate": 8.908396946564884e-06,
|
| 1963 |
+
"loss": 0.4579,
|
| 1964 |
+
"step": 25100
|
| 1965 |
+
},
|
| 1966 |
+
{
|
| 1967 |
+
"epoch": 44.68,
|
| 1968 |
+
"learning_rate": 8.622137404580152e-06,
|
| 1969 |
+
"loss": 0.4752,
|
| 1970 |
+
"step": 25200
|
| 1971 |
+
},
|
| 1972 |
+
{
|
| 1973 |
+
"epoch": 44.86,
|
| 1974 |
+
"learning_rate": 8.335877862595419e-06,
|
| 1975 |
+
"loss": 0.4708,
|
| 1976 |
+
"step": 25300
|
| 1977 |
+
},
|
| 1978 |
+
{
|
| 1979 |
+
"epoch": 45.04,
|
| 1980 |
+
"learning_rate": 8.049618320610687e-06,
|
| 1981 |
+
"loss": 0.4613,
|
| 1982 |
+
"step": 25400
|
| 1983 |
+
},
|
| 1984 |
+
{
|
| 1985 |
+
"epoch": 45.21,
|
| 1986 |
+
"learning_rate": 7.763358778625954e-06,
|
| 1987 |
+
"loss": 0.4667,
|
| 1988 |
+
"step": 25500
|
| 1989 |
+
},
|
| 1990 |
+
{
|
| 1991 |
+
"epoch": 45.21,
|
| 1992 |
+
"eval_loss": 0.1077902689576149,
|
| 1993 |
+
"eval_runtime": 31.0572,
|
| 1994 |
+
"eval_samples_per_second": 5.731,
|
| 1995 |
+
"eval_steps_per_second": 1.932,
|
| 1996 |
+
"eval_wer": 0.10782667114544844,
|
| 1997 |
+
"step": 25500
|
| 1998 |
+
},
|
| 1999 |
+
{
|
| 2000 |
+
"epoch": 45.39,
|
| 2001 |
+
"learning_rate": 7.477099236641221e-06,
|
| 2002 |
+
"loss": 0.4709,
|
| 2003 |
+
"step": 25600
|
| 2004 |
+
},
|
| 2005 |
+
{
|
| 2006 |
+
"epoch": 45.57,
|
| 2007 |
+
"learning_rate": 7.190839694656488e-06,
|
| 2008 |
+
"loss": 0.4523,
|
| 2009 |
+
"step": 25700
|
| 2010 |
+
},
|
| 2011 |
+
{
|
| 2012 |
+
"epoch": 45.74,
|
| 2013 |
+
"learning_rate": 6.9045801526717555e-06,
|
| 2014 |
+
"loss": 0.4537,
|
| 2015 |
+
"step": 25800
|
| 2016 |
+
},
|
| 2017 |
+
{
|
| 2018 |
+
"epoch": 45.92,
|
| 2019 |
+
"learning_rate": 6.62118320610687e-06,
|
| 2020 |
+
"loss": 0.4642,
|
| 2021 |
+
"step": 25900
|
| 2022 |
+
},
|
| 2023 |
+
{
|
| 2024 |
+
"epoch": 46.1,
|
| 2025 |
+
"learning_rate": 6.334923664122137e-06,
|
| 2026 |
+
"loss": 0.4484,
|
| 2027 |
+
"step": 26000
|
| 2028 |
+
},
|
| 2029 |
+
{
|
| 2030 |
+
"epoch": 46.1,
|
| 2031 |
+
"eval_loss": 0.10564640909433365,
|
| 2032 |
+
"eval_runtime": 31.5659,
|
| 2033 |
+
"eval_samples_per_second": 5.639,
|
| 2034 |
+
"eval_steps_per_second": 1.901,
|
| 2035 |
+
"eval_wer": 0.10816257977830031,
|
| 2036 |
+
"step": 26000
|
| 2037 |
+
},
|
| 2038 |
+
{
|
| 2039 |
+
"epoch": 46.28,
|
| 2040 |
+
"learning_rate": 6.048664122137404e-06,
|
| 2041 |
+
"loss": 0.4505,
|
| 2042 |
+
"step": 26100
|
| 2043 |
+
},
|
| 2044 |
+
{
|
| 2045 |
+
"epoch": 46.45,
|
| 2046 |
+
"learning_rate": 5.762404580152671e-06,
|
| 2047 |
+
"loss": 0.4758,
|
| 2048 |
+
"step": 26200
|
| 2049 |
+
},
|
| 2050 |
+
{
|
| 2051 |
+
"epoch": 46.63,
|
| 2052 |
+
"learning_rate": 5.476145038167939e-06,
|
| 2053 |
+
"loss": 0.4528,
|
| 2054 |
+
"step": 26300
|
| 2055 |
+
},
|
| 2056 |
+
{
|
| 2057 |
+
"epoch": 46.81,
|
| 2058 |
+
"learning_rate": 5.1898854961832056e-06,
|
| 2059 |
+
"loss": 0.4579,
|
| 2060 |
+
"step": 26400
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 46.99,
|
| 2064 |
+
"learning_rate": 4.903625954198473e-06,
|
| 2065 |
+
"loss": 0.4601,
|
| 2066 |
+
"step": 26500
|
| 2067 |
+
},
|
| 2068 |
+
{
|
| 2069 |
+
"epoch": 46.99,
|
| 2070 |
+
"eval_loss": 0.10663167387247086,
|
| 2071 |
+
"eval_runtime": 31.3245,
|
| 2072 |
+
"eval_samples_per_second": 5.682,
|
| 2073 |
+
"eval_steps_per_second": 1.915,
|
| 2074 |
+
"eval_wer": 0.10782667114544844,
|
| 2075 |
+
"step": 26500
|
| 2076 |
+
},
|
| 2077 |
+
{
|
| 2078 |
+
"epoch": 47.16,
|
| 2079 |
+
"learning_rate": 4.61736641221374e-06,
|
| 2080 |
+
"loss": 0.4634,
|
| 2081 |
+
"step": 26600
|
| 2082 |
+
},
|
| 2083 |
+
{
|
| 2084 |
+
"epoch": 47.34,
|
| 2085 |
+
"learning_rate": 4.331106870229007e-06,
|
| 2086 |
+
"loss": 0.4603,
|
| 2087 |
+
"step": 26700
|
| 2088 |
+
},
|
| 2089 |
+
{
|
| 2090 |
+
"epoch": 47.52,
|
| 2091 |
+
"learning_rate": 4.0448473282442745e-06,
|
| 2092 |
+
"loss": 0.4653,
|
| 2093 |
+
"step": 26800
|
| 2094 |
+
},
|
| 2095 |
+
{
|
| 2096 |
+
"epoch": 47.69,
|
| 2097 |
+
"learning_rate": 3.7585877862595418e-06,
|
| 2098 |
+
"loss": 0.4619,
|
| 2099 |
+
"step": 26900
|
| 2100 |
+
},
|
| 2101 |
+
{
|
| 2102 |
+
"epoch": 47.87,
|
| 2103 |
+
"learning_rate": 3.472328244274809e-06,
|
| 2104 |
+
"loss": 0.4691,
|
| 2105 |
+
"step": 27000
|
| 2106 |
+
},
|
| 2107 |
+
{
|
| 2108 |
+
"epoch": 47.87,
|
| 2109 |
+
"eval_loss": 0.10679604858160019,
|
| 2110 |
+
"eval_runtime": 31.1673,
|
| 2111 |
+
"eval_samples_per_second": 5.711,
|
| 2112 |
+
"eval_steps_per_second": 1.925,
|
| 2113 |
+
"eval_wer": 0.10849848841115217,
|
| 2114 |
+
"step": 27000
|
| 2115 |
+
},
|
| 2116 |
+
{
|
| 2117 |
+
"epoch": 48.05,
|
| 2118 |
+
"learning_rate": 3.1860687022900762e-06,
|
| 2119 |
+
"loss": 0.4582,
|
| 2120 |
+
"step": 27100
|
| 2121 |
+
},
|
| 2122 |
+
{
|
| 2123 |
+
"epoch": 48.23,
|
| 2124 |
+
"learning_rate": 2.8998091603053435e-06,
|
| 2125 |
+
"loss": 0.4439,
|
| 2126 |
+
"step": 27200
|
| 2127 |
+
},
|
| 2128 |
+
{
|
| 2129 |
+
"epoch": 48.4,
|
| 2130 |
+
"learning_rate": 2.6135496183206107e-06,
|
| 2131 |
+
"loss": 0.4617,
|
| 2132 |
+
"step": 27300
|
| 2133 |
+
},
|
| 2134 |
+
{
|
| 2135 |
+
"epoch": 48.58,
|
| 2136 |
+
"learning_rate": 2.3272900763358775e-06,
|
| 2137 |
+
"loss": 0.4538,
|
| 2138 |
+
"step": 27400
|
| 2139 |
+
},
|
| 2140 |
+
{
|
| 2141 |
+
"epoch": 48.76,
|
| 2142 |
+
"learning_rate": 2.0410305343511447e-06,
|
| 2143 |
+
"loss": 0.4457,
|
| 2144 |
+
"step": 27500
|
| 2145 |
+
},
|
| 2146 |
+
{
|
| 2147 |
+
"epoch": 48.76,
|
| 2148 |
+
"eval_loss": 0.10663049668073654,
|
| 2149 |
+
"eval_runtime": 31.2143,
|
| 2150 |
+
"eval_samples_per_second": 5.703,
|
| 2151 |
+
"eval_steps_per_second": 1.922,
|
| 2152 |
+
"eval_wer": 0.10782667114544844,
|
| 2153 |
+
"step": 27500
|
| 2154 |
+
},
|
| 2155 |
+
{
|
| 2156 |
+
"epoch": 48.94,
|
| 2157 |
+
"learning_rate": 1.7547709923664122e-06,
|
| 2158 |
+
"loss": 0.4766,
|
| 2159 |
+
"step": 27600
|
| 2160 |
+
},
|
| 2161 |
+
{
|
| 2162 |
+
"epoch": 49.11,
|
| 2163 |
+
"learning_rate": 1.4685114503816792e-06,
|
| 2164 |
+
"loss": 0.4576,
|
| 2165 |
+
"step": 27700
|
| 2166 |
+
},
|
| 2167 |
+
{
|
| 2168 |
+
"epoch": 49.29,
|
| 2169 |
+
"learning_rate": 1.185114503816794e-06,
|
| 2170 |
+
"loss": 0.4616,
|
| 2171 |
+
"step": 27800
|
| 2172 |
+
},
|
| 2173 |
+
{
|
| 2174 |
+
"epoch": 49.47,
|
| 2175 |
+
"learning_rate": 8.98854961832061e-07,
|
| 2176 |
+
"loss": 0.4588,
|
| 2177 |
+
"step": 27900
|
| 2178 |
+
},
|
| 2179 |
+
{
|
| 2180 |
+
"epoch": 49.65,
|
| 2181 |
+
"learning_rate": 6.125954198473282e-07,
|
| 2182 |
+
"loss": 0.475,
|
| 2183 |
+
"step": 28000
|
| 2184 |
+
},
|
| 2185 |
+
{
|
| 2186 |
+
"epoch": 49.65,
|
| 2187 |
+
"eval_loss": 0.1059938296675682,
|
| 2188 |
+
"eval_runtime": 32.0006,
|
| 2189 |
+
"eval_samples_per_second": 5.562,
|
| 2190 |
+
"eval_steps_per_second": 1.875,
|
| 2191 |
+
"eval_wer": 0.10816257977830031,
|
| 2192 |
+
"step": 28000
|
| 2193 |
+
},
|
| 2194 |
+
{
|
| 2195 |
+
"epoch": 49.82,
|
| 2196 |
+
"learning_rate": 3.263358778625954e-07,
|
| 2197 |
+
"loss": 0.4622,
|
| 2198 |
+
"step": 28100
|
| 2199 |
+
},
|
| 2200 |
+
{
|
| 2201 |
+
"epoch": 50.0,
|
| 2202 |
+
"learning_rate": 4.0076335877862596e-08,
|
| 2203 |
+
"loss": 0.4567,
|
| 2204 |
+
"step": 28200
|
| 2205 |
+
},
|
| 2206 |
+
{
|
| 2207 |
+
"epoch": 50.0,
|
| 2208 |
+
"step": 28200,
|
| 2209 |
+
"total_flos": 6.992186583697227e+19,
|
| 2210 |
+
"train_loss": 0.6889870901987062,
|
| 2211 |
+
"train_runtime": 132713.9902,
|
| 2212 |
+
"train_samples_per_second": 2.551,
|
| 2213 |
+
"train_steps_per_second": 0.212
|
| 2214 |
+
}
|
| 2215 |
+
],
|
| 2216 |
+
"max_steps": 28200,
|
| 2217 |
+
"num_train_epochs": 50,
|
| 2218 |
+
"total_flos": 6.992186583697227e+19,
|
| 2219 |
+
"trial_name": null,
|
| 2220 |
+
"trial_params": null
|
| 2221 |
+
}
|