End of training
Browse files- README.md +3 -3
- all_results.json +8 -0
- eval_results.json +8 -0
- runs/Jan14_19-23-29_78ef99605931/events.out.tfevents.1736896660.78ef99605931.300.3 +3 -0
- trainer_state.json +2941 -0
README.md
CHANGED
|
@@ -25,7 +25,7 @@ model-index:
|
|
| 25 |
metrics:
|
| 26 |
- name: Wer
|
| 27 |
type: wer
|
| 28 |
-
value:
|
| 29 |
---
|
| 30 |
|
| 31 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
@@ -35,8 +35,8 @@ should probably proofread and complete it, then remove this comment. -->
|
|
| 35 |
|
| 36 |
This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the Common Voice 17.0 dataset.
|
| 37 |
It achieves the following results on the evaluation set:
|
| 38 |
-
- Loss: 0.
|
| 39 |
-
- Wer:
|
| 40 |
|
| 41 |
## Model description
|
| 42 |
|
|
|
|
| 25 |
metrics:
|
| 26 |
- name: Wer
|
| 27 |
type: wer
|
| 28 |
+
value: 37.98212187043698
|
| 29 |
---
|
| 30 |
|
| 31 |
<!-- This model card has been generated automatically according to the information the Trainer had access to. You
|
|
|
|
| 35 |
|
| 36 |
This model is a fine-tuned version of [openai/whisper-tiny](https://huggingface.co/openai/whisper-tiny) on the Common Voice 17.0 dataset.
|
| 37 |
It achieves the following results on the evaluation set:
|
| 38 |
+
- Loss: 0.4848
|
| 39 |
+
- Wer: 37.9821
|
| 40 |
|
| 41 |
## Model description
|
| 42 |
|
all_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 17.73049645390071,
|
| 3 |
+
"eval_loss": 0.4848020076751709,
|
| 4 |
+
"eval_runtime": 614.3795,
|
| 5 |
+
"eval_samples_per_second": 6.341,
|
| 6 |
+
"eval_steps_per_second": 0.793,
|
| 7 |
+
"eval_wer": 37.98212187043698
|
| 8 |
+
}
|
eval_results.json
ADDED
|
@@ -0,0 +1,8 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 17.73049645390071,
|
| 3 |
+
"eval_loss": 0.4848020076751709,
|
| 4 |
+
"eval_runtime": 614.3795,
|
| 5 |
+
"eval_samples_per_second": 6.341,
|
| 6 |
+
"eval_steps_per_second": 0.793,
|
| 7 |
+
"eval_wer": 37.98212187043698
|
| 8 |
+
}
|
runs/Jan14_19-23-29_78ef99605931/events.out.tfevents.1736896660.78ef99605931.300.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b52856c01fe372526ac27c1b7ea71157ed8c3aa74bdd81b94612c8399b3cc5a1
|
| 3 |
+
size 406
|
trainer_state.json
ADDED
|
@@ -0,0 +1,2941 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"best_metric": 37.96787634887283,
|
| 3 |
+
"best_model_checkpoint": "./whisper-tiny-ro/checkpoint-5000",
|
| 4 |
+
"epoch": 17.73049645390071,
|
| 5 |
+
"eval_steps": 1000,
|
| 6 |
+
"global_step": 10000,
|
| 7 |
+
"is_hyper_param_search": false,
|
| 8 |
+
"is_local_process_zero": true,
|
| 9 |
+
"is_world_process_zero": true,
|
| 10 |
+
"log_history": [
|
| 11 |
+
{
|
| 12 |
+
"epoch": 0.044326241134751775,
|
| 13 |
+
"grad_norm": 28.326566696166992,
|
| 14 |
+
"learning_rate": 2.2e-07,
|
| 15 |
+
"loss": 1.8024,
|
| 16 |
+
"step": 25
|
| 17 |
+
},
|
| 18 |
+
{
|
| 19 |
+
"epoch": 0.08865248226950355,
|
| 20 |
+
"grad_norm": 22.18955421447754,
|
| 21 |
+
"learning_rate": 4.7000000000000005e-07,
|
| 22 |
+
"loss": 1.7794,
|
| 23 |
+
"step": 50
|
| 24 |
+
},
|
| 25 |
+
{
|
| 26 |
+
"epoch": 0.13297872340425532,
|
| 27 |
+
"grad_norm": 21.889328002929688,
|
| 28 |
+
"learning_rate": 7.2e-07,
|
| 29 |
+
"loss": 1.63,
|
| 30 |
+
"step": 75
|
| 31 |
+
},
|
| 32 |
+
{
|
| 33 |
+
"epoch": 0.1773049645390071,
|
| 34 |
+
"grad_norm": 19.02008819580078,
|
| 35 |
+
"learning_rate": 9.7e-07,
|
| 36 |
+
"loss": 1.5239,
|
| 37 |
+
"step": 100
|
| 38 |
+
},
|
| 39 |
+
{
|
| 40 |
+
"epoch": 0.22163120567375885,
|
| 41 |
+
"grad_norm": 16.531150817871094,
|
| 42 |
+
"learning_rate": 1.2200000000000002e-06,
|
| 43 |
+
"loss": 1.3903,
|
| 44 |
+
"step": 125
|
| 45 |
+
},
|
| 46 |
+
{
|
| 47 |
+
"epoch": 0.26595744680851063,
|
| 48 |
+
"grad_norm": 16.245573043823242,
|
| 49 |
+
"learning_rate": 1.4700000000000001e-06,
|
| 50 |
+
"loss": 1.2517,
|
| 51 |
+
"step": 150
|
| 52 |
+
},
|
| 53 |
+
{
|
| 54 |
+
"epoch": 0.3102836879432624,
|
| 55 |
+
"grad_norm": 17.484891891479492,
|
| 56 |
+
"learning_rate": 1.72e-06,
|
| 57 |
+
"loss": 1.1449,
|
| 58 |
+
"step": 175
|
| 59 |
+
},
|
| 60 |
+
{
|
| 61 |
+
"epoch": 0.3546099290780142,
|
| 62 |
+
"grad_norm": 13.317365646362305,
|
| 63 |
+
"learning_rate": 1.97e-06,
|
| 64 |
+
"loss": 1.0231,
|
| 65 |
+
"step": 200
|
| 66 |
+
},
|
| 67 |
+
{
|
| 68 |
+
"epoch": 0.39893617021276595,
|
| 69 |
+
"grad_norm": 16.296846389770508,
|
| 70 |
+
"learning_rate": 2.2200000000000003e-06,
|
| 71 |
+
"loss": 1.0033,
|
| 72 |
+
"step": 225
|
| 73 |
+
},
|
| 74 |
+
{
|
| 75 |
+
"epoch": 0.4432624113475177,
|
| 76 |
+
"grad_norm": 14.858762741088867,
|
| 77 |
+
"learning_rate": 2.47e-06,
|
| 78 |
+
"loss": 0.9183,
|
| 79 |
+
"step": 250
|
| 80 |
+
},
|
| 81 |
+
{
|
| 82 |
+
"epoch": 0.4875886524822695,
|
| 83 |
+
"grad_norm": 15.132709503173828,
|
| 84 |
+
"learning_rate": 2.7200000000000002e-06,
|
| 85 |
+
"loss": 0.9142,
|
| 86 |
+
"step": 275
|
| 87 |
+
},
|
| 88 |
+
{
|
| 89 |
+
"epoch": 0.5319148936170213,
|
| 90 |
+
"grad_norm": 15.102398872375488,
|
| 91 |
+
"learning_rate": 2.97e-06,
|
| 92 |
+
"loss": 0.8795,
|
| 93 |
+
"step": 300
|
| 94 |
+
},
|
| 95 |
+
{
|
| 96 |
+
"epoch": 0.5762411347517731,
|
| 97 |
+
"grad_norm": 15.617897033691406,
|
| 98 |
+
"learning_rate": 3.2200000000000005e-06,
|
| 99 |
+
"loss": 0.863,
|
| 100 |
+
"step": 325
|
| 101 |
+
},
|
| 102 |
+
{
|
| 103 |
+
"epoch": 0.6205673758865248,
|
| 104 |
+
"grad_norm": 15.325774192810059,
|
| 105 |
+
"learning_rate": 3.4700000000000002e-06,
|
| 106 |
+
"loss": 0.8094,
|
| 107 |
+
"step": 350
|
| 108 |
+
},
|
| 109 |
+
{
|
| 110 |
+
"epoch": 0.6648936170212766,
|
| 111 |
+
"grad_norm": 13.950435638427734,
|
| 112 |
+
"learning_rate": 3.7200000000000004e-06,
|
| 113 |
+
"loss": 0.8471,
|
| 114 |
+
"step": 375
|
| 115 |
+
},
|
| 116 |
+
{
|
| 117 |
+
"epoch": 0.7092198581560284,
|
| 118 |
+
"grad_norm": 17.703575134277344,
|
| 119 |
+
"learning_rate": 3.97e-06,
|
| 120 |
+
"loss": 0.83,
|
| 121 |
+
"step": 400
|
| 122 |
+
},
|
| 123 |
+
{
|
| 124 |
+
"epoch": 0.7535460992907801,
|
| 125 |
+
"grad_norm": 13.622574806213379,
|
| 126 |
+
"learning_rate": 4.22e-06,
|
| 127 |
+
"loss": 0.7605,
|
| 128 |
+
"step": 425
|
| 129 |
+
},
|
| 130 |
+
{
|
| 131 |
+
"epoch": 0.7978723404255319,
|
| 132 |
+
"grad_norm": 13.574337005615234,
|
| 133 |
+
"learning_rate": 4.47e-06,
|
| 134 |
+
"loss": 0.7464,
|
| 135 |
+
"step": 450
|
| 136 |
+
},
|
| 137 |
+
{
|
| 138 |
+
"epoch": 0.8421985815602837,
|
| 139 |
+
"grad_norm": 12.981876373291016,
|
| 140 |
+
"learning_rate": 4.7200000000000005e-06,
|
| 141 |
+
"loss": 0.767,
|
| 142 |
+
"step": 475
|
| 143 |
+
},
|
| 144 |
+
{
|
| 145 |
+
"epoch": 0.8865248226950354,
|
| 146 |
+
"grad_norm": 15.00900936126709,
|
| 147 |
+
"learning_rate": 4.970000000000001e-06,
|
| 148 |
+
"loss": 0.7617,
|
| 149 |
+
"step": 500
|
| 150 |
+
},
|
| 151 |
+
{
|
| 152 |
+
"epoch": 0.9308510638297872,
|
| 153 |
+
"grad_norm": 16.31970977783203,
|
| 154 |
+
"learning_rate": 4.988421052631579e-06,
|
| 155 |
+
"loss": 0.6962,
|
| 156 |
+
"step": 525
|
| 157 |
+
},
|
| 158 |
+
{
|
| 159 |
+
"epoch": 0.975177304964539,
|
| 160 |
+
"grad_norm": 10.05798625946045,
|
| 161 |
+
"learning_rate": 4.9752631578947375e-06,
|
| 162 |
+
"loss": 0.6676,
|
| 163 |
+
"step": 550
|
| 164 |
+
},
|
| 165 |
+
{
|
| 166 |
+
"epoch": 1.0195035460992907,
|
| 167 |
+
"grad_norm": 12.080132484436035,
|
| 168 |
+
"learning_rate": 4.962105263157895e-06,
|
| 169 |
+
"loss": 0.6351,
|
| 170 |
+
"step": 575
|
| 171 |
+
},
|
| 172 |
+
{
|
| 173 |
+
"epoch": 1.0638297872340425,
|
| 174 |
+
"grad_norm": 12.87156867980957,
|
| 175 |
+
"learning_rate": 4.948947368421053e-06,
|
| 176 |
+
"loss": 0.6078,
|
| 177 |
+
"step": 600
|
| 178 |
+
},
|
| 179 |
+
{
|
| 180 |
+
"epoch": 1.1081560283687943,
|
| 181 |
+
"grad_norm": 11.931558609008789,
|
| 182 |
+
"learning_rate": 4.935789473684211e-06,
|
| 183 |
+
"loss": 0.5755,
|
| 184 |
+
"step": 625
|
| 185 |
+
},
|
| 186 |
+
{
|
| 187 |
+
"epoch": 1.1524822695035462,
|
| 188 |
+
"grad_norm": 12.827286720275879,
|
| 189 |
+
"learning_rate": 4.922631578947369e-06,
|
| 190 |
+
"loss": 0.5679,
|
| 191 |
+
"step": 650
|
| 192 |
+
},
|
| 193 |
+
{
|
| 194 |
+
"epoch": 1.196808510638298,
|
| 195 |
+
"grad_norm": 12.644274711608887,
|
| 196 |
+
"learning_rate": 4.909473684210527e-06,
|
| 197 |
+
"loss": 0.6122,
|
| 198 |
+
"step": 675
|
| 199 |
+
},
|
| 200 |
+
{
|
| 201 |
+
"epoch": 1.2411347517730495,
|
| 202 |
+
"grad_norm": 12.461518287658691,
|
| 203 |
+
"learning_rate": 4.896315789473685e-06,
|
| 204 |
+
"loss": 0.5522,
|
| 205 |
+
"step": 700
|
| 206 |
+
},
|
| 207 |
+
{
|
| 208 |
+
"epoch": 1.2854609929078014,
|
| 209 |
+
"grad_norm": 14.115540504455566,
|
| 210 |
+
"learning_rate": 4.8831578947368425e-06,
|
| 211 |
+
"loss": 0.5764,
|
| 212 |
+
"step": 725
|
| 213 |
+
},
|
| 214 |
+
{
|
| 215 |
+
"epoch": 1.3297872340425532,
|
| 216 |
+
"grad_norm": 13.1589994430542,
|
| 217 |
+
"learning_rate": 4.87e-06,
|
| 218 |
+
"loss": 0.5421,
|
| 219 |
+
"step": 750
|
| 220 |
+
},
|
| 221 |
+
{
|
| 222 |
+
"epoch": 1.374113475177305,
|
| 223 |
+
"grad_norm": 12.696803092956543,
|
| 224 |
+
"learning_rate": 4.856842105263158e-06,
|
| 225 |
+
"loss": 0.5616,
|
| 226 |
+
"step": 775
|
| 227 |
+
},
|
| 228 |
+
{
|
| 229 |
+
"epoch": 1.4184397163120568,
|
| 230 |
+
"grad_norm": 14.510184288024902,
|
| 231 |
+
"learning_rate": 4.843684210526316e-06,
|
| 232 |
+
"loss": 0.5725,
|
| 233 |
+
"step": 800
|
| 234 |
+
},
|
| 235 |
+
{
|
| 236 |
+
"epoch": 1.4627659574468086,
|
| 237 |
+
"grad_norm": 11.529364585876465,
|
| 238 |
+
"learning_rate": 4.830526315789474e-06,
|
| 239 |
+
"loss": 0.5627,
|
| 240 |
+
"step": 825
|
| 241 |
+
},
|
| 242 |
+
{
|
| 243 |
+
"epoch": 1.5070921985815602,
|
| 244 |
+
"grad_norm": 12.159563064575195,
|
| 245 |
+
"learning_rate": 4.8173684210526324e-06,
|
| 246 |
+
"loss": 0.5452,
|
| 247 |
+
"step": 850
|
| 248 |
+
},
|
| 249 |
+
{
|
| 250 |
+
"epoch": 1.5514184397163122,
|
| 251 |
+
"grad_norm": 10.232617378234863,
|
| 252 |
+
"learning_rate": 4.80421052631579e-06,
|
| 253 |
+
"loss": 0.5192,
|
| 254 |
+
"step": 875
|
| 255 |
+
},
|
| 256 |
+
{
|
| 257 |
+
"epoch": 1.5957446808510638,
|
| 258 |
+
"grad_norm": 10.81043529510498,
|
| 259 |
+
"learning_rate": 4.791052631578948e-06,
|
| 260 |
+
"loss": 0.5151,
|
| 261 |
+
"step": 900
|
| 262 |
+
},
|
| 263 |
+
{
|
| 264 |
+
"epoch": 1.6400709219858156,
|
| 265 |
+
"grad_norm": 14.97497272491455,
|
| 266 |
+
"learning_rate": 4.777894736842106e-06,
|
| 267 |
+
"loss": 0.5263,
|
| 268 |
+
"step": 925
|
| 269 |
+
},
|
| 270 |
+
{
|
| 271 |
+
"epoch": 1.6843971631205674,
|
| 272 |
+
"grad_norm": 14.701244354248047,
|
| 273 |
+
"learning_rate": 4.764736842105264e-06,
|
| 274 |
+
"loss": 0.524,
|
| 275 |
+
"step": 950
|
| 276 |
+
},
|
| 277 |
+
{
|
| 278 |
+
"epoch": 1.728723404255319,
|
| 279 |
+
"grad_norm": 13.469274520874023,
|
| 280 |
+
"learning_rate": 4.7515789473684216e-06,
|
| 281 |
+
"loss": 0.5084,
|
| 282 |
+
"step": 975
|
| 283 |
+
},
|
| 284 |
+
{
|
| 285 |
+
"epoch": 1.773049645390071,
|
| 286 |
+
"grad_norm": 11.251127243041992,
|
| 287 |
+
"learning_rate": 4.738421052631579e-06,
|
| 288 |
+
"loss": 0.5444,
|
| 289 |
+
"step": 1000
|
| 290 |
+
},
|
| 291 |
+
{
|
| 292 |
+
"epoch": 1.773049645390071,
|
| 293 |
+
"eval_loss": 0.604159414768219,
|
| 294 |
+
"eval_runtime": 587.0739,
|
| 295 |
+
"eval_samples_per_second": 6.636,
|
| 296 |
+
"eval_steps_per_second": 0.83,
|
| 297 |
+
"eval_wer": 48.87994586701805,
|
| 298 |
+
"step": 1000
|
| 299 |
+
},
|
| 300 |
+
{
|
| 301 |
+
"epoch": 1.8173758865248226,
|
| 302 |
+
"grad_norm": 13.032508850097656,
|
| 303 |
+
"learning_rate": 4.725263157894737e-06,
|
| 304 |
+
"loss": 0.5314,
|
| 305 |
+
"step": 1025
|
| 306 |
+
},
|
| 307 |
+
{
|
| 308 |
+
"epoch": 1.8617021276595744,
|
| 309 |
+
"grad_norm": 12.22535228729248,
|
| 310 |
+
"learning_rate": 4.712105263157895e-06,
|
| 311 |
+
"loss": 0.4761,
|
| 312 |
+
"step": 1050
|
| 313 |
+
},
|
| 314 |
+
{
|
| 315 |
+
"epoch": 1.9060283687943262,
|
| 316 |
+
"grad_norm": 15.19352912902832,
|
| 317 |
+
"learning_rate": 4.698947368421053e-06,
|
| 318 |
+
"loss": 0.4742,
|
| 319 |
+
"step": 1075
|
| 320 |
+
},
|
| 321 |
+
{
|
| 322 |
+
"epoch": 1.950354609929078,
|
| 323 |
+
"grad_norm": 10.927416801452637,
|
| 324 |
+
"learning_rate": 4.685789473684211e-06,
|
| 325 |
+
"loss": 0.5012,
|
| 326 |
+
"step": 1100
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"epoch": 1.9946808510638299,
|
| 330 |
+
"grad_norm": 10.421246528625488,
|
| 331 |
+
"learning_rate": 4.672631578947369e-06,
|
| 332 |
+
"loss": 0.5225,
|
| 333 |
+
"step": 1125
|
| 334 |
+
},
|
| 335 |
+
{
|
| 336 |
+
"epoch": 2.0390070921985815,
|
| 337 |
+
"grad_norm": 9.38261604309082,
|
| 338 |
+
"learning_rate": 4.6594736842105265e-06,
|
| 339 |
+
"loss": 0.4173,
|
| 340 |
+
"step": 1150
|
| 341 |
+
},
|
| 342 |
+
{
|
| 343 |
+
"epoch": 2.0833333333333335,
|
| 344 |
+
"grad_norm": 10.129746437072754,
|
| 345 |
+
"learning_rate": 4.646315789473684e-06,
|
| 346 |
+
"loss": 0.4153,
|
| 347 |
+
"step": 1175
|
| 348 |
+
},
|
| 349 |
+
{
|
| 350 |
+
"epoch": 2.127659574468085,
|
| 351 |
+
"grad_norm": 11.529908180236816,
|
| 352 |
+
"learning_rate": 4.633157894736842e-06,
|
| 353 |
+
"loss": 0.3981,
|
| 354 |
+
"step": 1200
|
| 355 |
+
},
|
| 356 |
+
{
|
| 357 |
+
"epoch": 2.171985815602837,
|
| 358 |
+
"grad_norm": 11.076881408691406,
|
| 359 |
+
"learning_rate": 4.620000000000001e-06,
|
| 360 |
+
"loss": 0.4152,
|
| 361 |
+
"step": 1225
|
| 362 |
+
},
|
| 363 |
+
{
|
| 364 |
+
"epoch": 2.2163120567375887,
|
| 365 |
+
"grad_norm": 9.994644165039062,
|
| 366 |
+
"learning_rate": 4.606842105263158e-06,
|
| 367 |
+
"loss": 0.3926,
|
| 368 |
+
"step": 1250
|
| 369 |
+
},
|
| 370 |
+
{
|
| 371 |
+
"epoch": 2.2606382978723403,
|
| 372 |
+
"grad_norm": 10.582009315490723,
|
| 373 |
+
"learning_rate": 4.5936842105263165e-06,
|
| 374 |
+
"loss": 0.4378,
|
| 375 |
+
"step": 1275
|
| 376 |
+
},
|
| 377 |
+
{
|
| 378 |
+
"epoch": 2.3049645390070923,
|
| 379 |
+
"grad_norm": 9.771284103393555,
|
| 380 |
+
"learning_rate": 4.580526315789474e-06,
|
| 381 |
+
"loss": 0.3882,
|
| 382 |
+
"step": 1300
|
| 383 |
+
},
|
| 384 |
+
{
|
| 385 |
+
"epoch": 2.349290780141844,
|
| 386 |
+
"grad_norm": 11.286993026733398,
|
| 387 |
+
"learning_rate": 4.567368421052632e-06,
|
| 388 |
+
"loss": 0.3686,
|
| 389 |
+
"step": 1325
|
| 390 |
+
},
|
| 391 |
+
{
|
| 392 |
+
"epoch": 2.393617021276596,
|
| 393 |
+
"grad_norm": 13.008705139160156,
|
| 394 |
+
"learning_rate": 4.55421052631579e-06,
|
| 395 |
+
"loss": 0.3899,
|
| 396 |
+
"step": 1350
|
| 397 |
+
},
|
| 398 |
+
{
|
| 399 |
+
"epoch": 2.4379432624113475,
|
| 400 |
+
"grad_norm": 10.24173355102539,
|
| 401 |
+
"learning_rate": 4.541052631578948e-06,
|
| 402 |
+
"loss": 0.4028,
|
| 403 |
+
"step": 1375
|
| 404 |
+
},
|
| 405 |
+
{
|
| 406 |
+
"epoch": 2.482269503546099,
|
| 407 |
+
"grad_norm": 10.5569486618042,
|
| 408 |
+
"learning_rate": 4.527894736842106e-06,
|
| 409 |
+
"loss": 0.3634,
|
| 410 |
+
"step": 1400
|
| 411 |
+
},
|
| 412 |
+
{
|
| 413 |
+
"epoch": 2.526595744680851,
|
| 414 |
+
"grad_norm": 10.904850006103516,
|
| 415 |
+
"learning_rate": 4.514736842105263e-06,
|
| 416 |
+
"loss": 0.4181,
|
| 417 |
+
"step": 1425
|
| 418 |
+
},
|
| 419 |
+
{
|
| 420 |
+
"epoch": 2.5709219858156027,
|
| 421 |
+
"grad_norm": 10.719099998474121,
|
| 422 |
+
"learning_rate": 4.501578947368421e-06,
|
| 423 |
+
"loss": 0.3803,
|
| 424 |
+
"step": 1450
|
| 425 |
+
},
|
| 426 |
+
{
|
| 427 |
+
"epoch": 2.6152482269503547,
|
| 428 |
+
"grad_norm": 10.873899459838867,
|
| 429 |
+
"learning_rate": 4.488421052631579e-06,
|
| 430 |
+
"loss": 0.3983,
|
| 431 |
+
"step": 1475
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"epoch": 2.6595744680851063,
|
| 435 |
+
"grad_norm": 8.574480056762695,
|
| 436 |
+
"learning_rate": 4.475263157894737e-06,
|
| 437 |
+
"loss": 0.3894,
|
| 438 |
+
"step": 1500
|
| 439 |
+
},
|
| 440 |
+
{
|
| 441 |
+
"epoch": 2.703900709219858,
|
| 442 |
+
"grad_norm": 10.148545265197754,
|
| 443 |
+
"learning_rate": 4.462105263157895e-06,
|
| 444 |
+
"loss": 0.3672,
|
| 445 |
+
"step": 1525
|
| 446 |
+
},
|
| 447 |
+
{
|
| 448 |
+
"epoch": 2.74822695035461,
|
| 449 |
+
"grad_norm": 11.479018211364746,
|
| 450 |
+
"learning_rate": 4.448947368421053e-06,
|
| 451 |
+
"loss": 0.3824,
|
| 452 |
+
"step": 1550
|
| 453 |
+
},
|
| 454 |
+
{
|
| 455 |
+
"epoch": 2.7925531914893615,
|
| 456 |
+
"grad_norm": 10.652966499328613,
|
| 457 |
+
"learning_rate": 4.435789473684211e-06,
|
| 458 |
+
"loss": 0.3849,
|
| 459 |
+
"step": 1575
|
| 460 |
+
},
|
| 461 |
+
{
|
| 462 |
+
"epoch": 2.8368794326241136,
|
| 463 |
+
"grad_norm": 10.057666778564453,
|
| 464 |
+
"learning_rate": 4.422631578947369e-06,
|
| 465 |
+
"loss": 0.3702,
|
| 466 |
+
"step": 1600
|
| 467 |
+
},
|
| 468 |
+
{
|
| 469 |
+
"epoch": 2.881205673758865,
|
| 470 |
+
"grad_norm": 9.985100746154785,
|
| 471 |
+
"learning_rate": 4.409473684210527e-06,
|
| 472 |
+
"loss": 0.4052,
|
| 473 |
+
"step": 1625
|
| 474 |
+
},
|
| 475 |
+
{
|
| 476 |
+
"epoch": 2.925531914893617,
|
| 477 |
+
"grad_norm": 9.165911674499512,
|
| 478 |
+
"learning_rate": 4.396315789473685e-06,
|
| 479 |
+
"loss": 0.3697,
|
| 480 |
+
"step": 1650
|
| 481 |
+
},
|
| 482 |
+
{
|
| 483 |
+
"epoch": 2.969858156028369,
|
| 484 |
+
"grad_norm": 10.057464599609375,
|
| 485 |
+
"learning_rate": 4.383157894736842e-06,
|
| 486 |
+
"loss": 0.3663,
|
| 487 |
+
"step": 1675
|
| 488 |
+
},
|
| 489 |
+
{
|
| 490 |
+
"epoch": 3.0141843971631204,
|
| 491 |
+
"grad_norm": 10.039346694946289,
|
| 492 |
+
"learning_rate": 4.3700000000000005e-06,
|
| 493 |
+
"loss": 0.3443,
|
| 494 |
+
"step": 1700
|
| 495 |
+
},
|
| 496 |
+
{
|
| 497 |
+
"epoch": 3.0585106382978724,
|
| 498 |
+
"grad_norm": 9.51621150970459,
|
| 499 |
+
"learning_rate": 4.356842105263158e-06,
|
| 500 |
+
"loss": 0.2917,
|
| 501 |
+
"step": 1725
|
| 502 |
+
},
|
| 503 |
+
{
|
| 504 |
+
"epoch": 3.102836879432624,
|
| 505 |
+
"grad_norm": 10.015137672424316,
|
| 506 |
+
"learning_rate": 4.343684210526316e-06,
|
| 507 |
+
"loss": 0.3045,
|
| 508 |
+
"step": 1750
|
| 509 |
+
},
|
| 510 |
+
{
|
| 511 |
+
"epoch": 3.147163120567376,
|
| 512 |
+
"grad_norm": 10.268891334533691,
|
| 513 |
+
"learning_rate": 4.330526315789474e-06,
|
| 514 |
+
"loss": 0.2821,
|
| 515 |
+
"step": 1775
|
| 516 |
+
},
|
| 517 |
+
{
|
| 518 |
+
"epoch": 3.1914893617021276,
|
| 519 |
+
"grad_norm": 9.120494842529297,
|
| 520 |
+
"learning_rate": 4.317368421052632e-06,
|
| 521 |
+
"loss": 0.2956,
|
| 522 |
+
"step": 1800
|
| 523 |
+
},
|
| 524 |
+
{
|
| 525 |
+
"epoch": 3.2358156028368796,
|
| 526 |
+
"grad_norm": 10.420275688171387,
|
| 527 |
+
"learning_rate": 4.30421052631579e-06,
|
| 528 |
+
"loss": 0.3086,
|
| 529 |
+
"step": 1825
|
| 530 |
+
},
|
| 531 |
+
{
|
| 532 |
+
"epoch": 3.280141843971631,
|
| 533 |
+
"grad_norm": 9.254629135131836,
|
| 534 |
+
"learning_rate": 4.291052631578947e-06,
|
| 535 |
+
"loss": 0.3083,
|
| 536 |
+
"step": 1850
|
| 537 |
+
},
|
| 538 |
+
{
|
| 539 |
+
"epoch": 3.324468085106383,
|
| 540 |
+
"grad_norm": 9.865363121032715,
|
| 541 |
+
"learning_rate": 4.277894736842106e-06,
|
| 542 |
+
"loss": 0.3071,
|
| 543 |
+
"step": 1875
|
| 544 |
+
},
|
| 545 |
+
{
|
| 546 |
+
"epoch": 3.368794326241135,
|
| 547 |
+
"grad_norm": 11.790287017822266,
|
| 548 |
+
"learning_rate": 4.264736842105264e-06,
|
| 549 |
+
"loss": 0.3014,
|
| 550 |
+
"step": 1900
|
| 551 |
+
},
|
| 552 |
+
{
|
| 553 |
+
"epoch": 3.4131205673758864,
|
| 554 |
+
"grad_norm": 10.183505058288574,
|
| 555 |
+
"learning_rate": 4.251578947368421e-06,
|
| 556 |
+
"loss": 0.293,
|
| 557 |
+
"step": 1925
|
| 558 |
+
},
|
| 559 |
+
{
|
| 560 |
+
"epoch": 3.4574468085106385,
|
| 561 |
+
"grad_norm": 9.69072151184082,
|
| 562 |
+
"learning_rate": 4.23842105263158e-06,
|
| 563 |
+
"loss": 0.2978,
|
| 564 |
+
"step": 1950
|
| 565 |
+
},
|
| 566 |
+
{
|
| 567 |
+
"epoch": 3.50177304964539,
|
| 568 |
+
"grad_norm": 10.96455192565918,
|
| 569 |
+
"learning_rate": 4.225263157894737e-06,
|
| 570 |
+
"loss": 0.311,
|
| 571 |
+
"step": 1975
|
| 572 |
+
},
|
| 573 |
+
{
|
| 574 |
+
"epoch": 3.546099290780142,
|
| 575 |
+
"grad_norm": 11.342255592346191,
|
| 576 |
+
"learning_rate": 4.212105263157895e-06,
|
| 577 |
+
"loss": 0.3042,
|
| 578 |
+
"step": 2000
|
| 579 |
+
},
|
| 580 |
+
{
|
| 581 |
+
"epoch": 3.546099290780142,
|
| 582 |
+
"eval_loss": 0.5099755525588989,
|
| 583 |
+
"eval_runtime": 581.1553,
|
| 584 |
+
"eval_samples_per_second": 6.704,
|
| 585 |
+
"eval_steps_per_second": 0.838,
|
| 586 |
+
"eval_wer": 41.17311870080843,
|
| 587 |
+
"step": 2000
|
| 588 |
+
},
|
| 589 |
+
{
|
| 590 |
+
"epoch": 3.5904255319148937,
|
| 591 |
+
"grad_norm": 10.30219554901123,
|
| 592 |
+
"learning_rate": 4.198947368421053e-06,
|
| 593 |
+
"loss": 0.3039,
|
| 594 |
+
"step": 2025
|
| 595 |
+
},
|
| 596 |
+
{
|
| 597 |
+
"epoch": 3.6347517730496453,
|
| 598 |
+
"grad_norm": 9.825774192810059,
|
| 599 |
+
"learning_rate": 4.185789473684211e-06,
|
| 600 |
+
"loss": 0.292,
|
| 601 |
+
"step": 2050
|
| 602 |
+
},
|
| 603 |
+
{
|
| 604 |
+
"epoch": 3.6790780141843973,
|
| 605 |
+
"grad_norm": 9.612593650817871,
|
| 606 |
+
"learning_rate": 4.172631578947369e-06,
|
| 607 |
+
"loss": 0.3133,
|
| 608 |
+
"step": 2075
|
| 609 |
+
},
|
| 610 |
+
{
|
| 611 |
+
"epoch": 3.723404255319149,
|
| 612 |
+
"grad_norm": 9.864873886108398,
|
| 613 |
+
"learning_rate": 4.159473684210526e-06,
|
| 614 |
+
"loss": 0.2786,
|
| 615 |
+
"step": 2100
|
| 616 |
+
},
|
| 617 |
+
{
|
| 618 |
+
"epoch": 3.7677304964539005,
|
| 619 |
+
"grad_norm": 9.14278507232666,
|
| 620 |
+
"learning_rate": 4.1463157894736845e-06,
|
| 621 |
+
"loss": 0.2992,
|
| 622 |
+
"step": 2125
|
| 623 |
+
},
|
| 624 |
+
{
|
| 625 |
+
"epoch": 3.8120567375886525,
|
| 626 |
+
"grad_norm": 10.981643676757812,
|
| 627 |
+
"learning_rate": 4.133157894736842e-06,
|
| 628 |
+
"loss": 0.2987,
|
| 629 |
+
"step": 2150
|
| 630 |
+
},
|
| 631 |
+
{
|
| 632 |
+
"epoch": 3.8563829787234045,
|
| 633 |
+
"grad_norm": 10.71380615234375,
|
| 634 |
+
"learning_rate": 4.12e-06,
|
| 635 |
+
"loss": 0.3088,
|
| 636 |
+
"step": 2175
|
| 637 |
+
},
|
| 638 |
+
{
|
| 639 |
+
"epoch": 3.900709219858156,
|
| 640 |
+
"grad_norm": 11.365142822265625,
|
| 641 |
+
"learning_rate": 4.106842105263158e-06,
|
| 642 |
+
"loss": 0.302,
|
| 643 |
+
"step": 2200
|
| 644 |
+
},
|
| 645 |
+
{
|
| 646 |
+
"epoch": 3.9450354609929077,
|
| 647 |
+
"grad_norm": 11.918941497802734,
|
| 648 |
+
"learning_rate": 4.093684210526316e-06,
|
| 649 |
+
"loss": 0.31,
|
| 650 |
+
"step": 2225
|
| 651 |
+
},
|
| 652 |
+
{
|
| 653 |
+
"epoch": 3.9893617021276597,
|
| 654 |
+
"grad_norm": 10.240377426147461,
|
| 655 |
+
"learning_rate": 4.0805263157894745e-06,
|
| 656 |
+
"loss": 0.3048,
|
| 657 |
+
"step": 2250
|
| 658 |
+
},
|
| 659 |
+
{
|
| 660 |
+
"epoch": 4.033687943262412,
|
| 661 |
+
"grad_norm": 7.562131404876709,
|
| 662 |
+
"learning_rate": 4.067368421052632e-06,
|
| 663 |
+
"loss": 0.2673,
|
| 664 |
+
"step": 2275
|
| 665 |
+
},
|
| 666 |
+
{
|
| 667 |
+
"epoch": 4.078014184397163,
|
| 668 |
+
"grad_norm": 7.741388320922852,
|
| 669 |
+
"learning_rate": 4.05421052631579e-06,
|
| 670 |
+
"loss": 0.2364,
|
| 671 |
+
"step": 2300
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"epoch": 4.122340425531915,
|
| 675 |
+
"grad_norm": 8.797900199890137,
|
| 676 |
+
"learning_rate": 4.041052631578948e-06,
|
| 677 |
+
"loss": 0.2524,
|
| 678 |
+
"step": 2325
|
| 679 |
+
},
|
| 680 |
+
{
|
| 681 |
+
"epoch": 4.166666666666667,
|
| 682 |
+
"grad_norm": 9.751541137695312,
|
| 683 |
+
"learning_rate": 4.027894736842105e-06,
|
| 684 |
+
"loss": 0.2565,
|
| 685 |
+
"step": 2350
|
| 686 |
+
},
|
| 687 |
+
{
|
| 688 |
+
"epoch": 4.210992907801418,
|
| 689 |
+
"grad_norm": 7.299990653991699,
|
| 690 |
+
"learning_rate": 4.014736842105264e-06,
|
| 691 |
+
"loss": 0.2405,
|
| 692 |
+
"step": 2375
|
| 693 |
+
},
|
| 694 |
+
{
|
| 695 |
+
"epoch": 4.25531914893617,
|
| 696 |
+
"grad_norm": 8.407694816589355,
|
| 697 |
+
"learning_rate": 4.001578947368421e-06,
|
| 698 |
+
"loss": 0.226,
|
| 699 |
+
"step": 2400
|
| 700 |
+
},
|
| 701 |
+
{
|
| 702 |
+
"epoch": 4.299645390070922,
|
| 703 |
+
"grad_norm": 8.874945640563965,
|
| 704 |
+
"learning_rate": 3.9884210526315795e-06,
|
| 705 |
+
"loss": 0.2366,
|
| 706 |
+
"step": 2425
|
| 707 |
+
},
|
| 708 |
+
{
|
| 709 |
+
"epoch": 4.343971631205674,
|
| 710 |
+
"grad_norm": 8.172481536865234,
|
| 711 |
+
"learning_rate": 3.975263157894737e-06,
|
| 712 |
+
"loss": 0.2325,
|
| 713 |
+
"step": 2450
|
| 714 |
+
},
|
| 715 |
+
{
|
| 716 |
+
"epoch": 4.388297872340425,
|
| 717 |
+
"grad_norm": 8.563679695129395,
|
| 718 |
+
"learning_rate": 3.962105263157895e-06,
|
| 719 |
+
"loss": 0.2266,
|
| 720 |
+
"step": 2475
|
| 721 |
+
},
|
| 722 |
+
{
|
| 723 |
+
"epoch": 4.432624113475177,
|
| 724 |
+
"grad_norm": 7.646442413330078,
|
| 725 |
+
"learning_rate": 3.948947368421053e-06,
|
| 726 |
+
"loss": 0.2477,
|
| 727 |
+
"step": 2500
|
| 728 |
+
},
|
| 729 |
+
{
|
| 730 |
+
"epoch": 4.476950354609929,
|
| 731 |
+
"grad_norm": 8.14061164855957,
|
| 732 |
+
"learning_rate": 3.93578947368421e-06,
|
| 733 |
+
"loss": 0.2372,
|
| 734 |
+
"step": 2525
|
| 735 |
+
},
|
| 736 |
+
{
|
| 737 |
+
"epoch": 4.5212765957446805,
|
| 738 |
+
"grad_norm": 6.697457790374756,
|
| 739 |
+
"learning_rate": 3.9226315789473694e-06,
|
| 740 |
+
"loss": 0.2312,
|
| 741 |
+
"step": 2550
|
| 742 |
+
},
|
| 743 |
+
{
|
| 744 |
+
"epoch": 4.565602836879433,
|
| 745 |
+
"grad_norm": 9.178577423095703,
|
| 746 |
+
"learning_rate": 3.909473684210527e-06,
|
| 747 |
+
"loss": 0.239,
|
| 748 |
+
"step": 2575
|
| 749 |
+
},
|
| 750 |
+
{
|
| 751 |
+
"epoch": 4.609929078014185,
|
| 752 |
+
"grad_norm": 7.986817836761475,
|
| 753 |
+
"learning_rate": 3.896315789473684e-06,
|
| 754 |
+
"loss": 0.2266,
|
| 755 |
+
"step": 2600
|
| 756 |
+
},
|
| 757 |
+
{
|
| 758 |
+
"epoch": 4.654255319148936,
|
| 759 |
+
"grad_norm": 9.992223739624023,
|
| 760 |
+
"learning_rate": 3.883157894736843e-06,
|
| 761 |
+
"loss": 0.2422,
|
| 762 |
+
"step": 2625
|
| 763 |
+
},
|
| 764 |
+
{
|
| 765 |
+
"epoch": 4.698581560283688,
|
| 766 |
+
"grad_norm": 8.259024620056152,
|
| 767 |
+
"learning_rate": 3.87e-06,
|
| 768 |
+
"loss": 0.2382,
|
| 769 |
+
"step": 2650
|
| 770 |
+
},
|
| 771 |
+
{
|
| 772 |
+
"epoch": 4.74290780141844,
|
| 773 |
+
"grad_norm": 8.913894653320312,
|
| 774 |
+
"learning_rate": 3.8568421052631585e-06,
|
| 775 |
+
"loss": 0.2404,
|
| 776 |
+
"step": 2675
|
| 777 |
+
},
|
| 778 |
+
{
|
| 779 |
+
"epoch": 4.787234042553192,
|
| 780 |
+
"grad_norm": 8.490303993225098,
|
| 781 |
+
"learning_rate": 3.843684210526316e-06,
|
| 782 |
+
"loss": 0.251,
|
| 783 |
+
"step": 2700
|
| 784 |
+
},
|
| 785 |
+
{
|
| 786 |
+
"epoch": 4.831560283687943,
|
| 787 |
+
"grad_norm": 8.170136451721191,
|
| 788 |
+
"learning_rate": 3.830526315789474e-06,
|
| 789 |
+
"loss": 0.2195,
|
| 790 |
+
"step": 2725
|
| 791 |
+
},
|
| 792 |
+
{
|
| 793 |
+
"epoch": 4.875886524822695,
|
| 794 |
+
"grad_norm": 7.071116924285889,
|
| 795 |
+
"learning_rate": 3.817368421052632e-06,
|
| 796 |
+
"loss": 0.2237,
|
| 797 |
+
"step": 2750
|
| 798 |
+
},
|
| 799 |
+
{
|
| 800 |
+
"epoch": 4.920212765957447,
|
| 801 |
+
"grad_norm": 7.995920181274414,
|
| 802 |
+
"learning_rate": 3.8042105263157898e-06,
|
| 803 |
+
"loss": 0.2377,
|
| 804 |
+
"step": 2775
|
| 805 |
+
},
|
| 806 |
+
{
|
| 807 |
+
"epoch": 4.964539007092198,
|
| 808 |
+
"grad_norm": 8.777873039245605,
|
| 809 |
+
"learning_rate": 3.7910526315789477e-06,
|
| 810 |
+
"loss": 0.2211,
|
| 811 |
+
"step": 2800
|
| 812 |
+
},
|
| 813 |
+
{
|
| 814 |
+
"epoch": 5.00886524822695,
|
| 815 |
+
"grad_norm": 7.402454376220703,
|
| 816 |
+
"learning_rate": 3.7778947368421056e-06,
|
| 817 |
+
"loss": 0.2341,
|
| 818 |
+
"step": 2825
|
| 819 |
+
},
|
| 820 |
+
{
|
| 821 |
+
"epoch": 5.053191489361702,
|
| 822 |
+
"grad_norm": 7.576868057250977,
|
| 823 |
+
"learning_rate": 3.764736842105263e-06,
|
| 824 |
+
"loss": 0.2006,
|
| 825 |
+
"step": 2850
|
| 826 |
+
},
|
| 827 |
+
{
|
| 828 |
+
"epoch": 5.097517730496454,
|
| 829 |
+
"grad_norm": 6.49124002456665,
|
| 830 |
+
"learning_rate": 3.751578947368421e-06,
|
| 831 |
+
"loss": 0.1835,
|
| 832 |
+
"step": 2875
|
| 833 |
+
},
|
| 834 |
+
{
|
| 835 |
+
"epoch": 5.141843971631205,
|
| 836 |
+
"grad_norm": 5.912723064422607,
|
| 837 |
+
"learning_rate": 3.7384210526315793e-06,
|
| 838 |
+
"loss": 0.1962,
|
| 839 |
+
"step": 2900
|
| 840 |
+
},
|
| 841 |
+
{
|
| 842 |
+
"epoch": 5.1861702127659575,
|
| 843 |
+
"grad_norm": 7.608515739440918,
|
| 844 |
+
"learning_rate": 3.7252631578947372e-06,
|
| 845 |
+
"loss": 0.18,
|
| 846 |
+
"step": 2925
|
| 847 |
+
},
|
| 848 |
+
{
|
| 849 |
+
"epoch": 5.2304964539007095,
|
| 850 |
+
"grad_norm": 8.52009105682373,
|
| 851 |
+
"learning_rate": 3.712105263157895e-06,
|
| 852 |
+
"loss": 0.184,
|
| 853 |
+
"step": 2950
|
| 854 |
+
},
|
| 855 |
+
{
|
| 856 |
+
"epoch": 5.274822695035461,
|
| 857 |
+
"grad_norm": 9.251614570617676,
|
| 858 |
+
"learning_rate": 3.698947368421053e-06,
|
| 859 |
+
"loss": 0.2098,
|
| 860 |
+
"step": 2975
|
| 861 |
+
},
|
| 862 |
+
{
|
| 863 |
+
"epoch": 5.319148936170213,
|
| 864 |
+
"grad_norm": 8.74129581451416,
|
| 865 |
+
"learning_rate": 3.685789473684211e-06,
|
| 866 |
+
"loss": 0.1817,
|
| 867 |
+
"step": 3000
|
| 868 |
+
},
|
| 869 |
+
{
|
| 870 |
+
"epoch": 5.319148936170213,
|
| 871 |
+
"eval_loss": 0.4850045144557953,
|
| 872 |
+
"eval_runtime": 588.1549,
|
| 873 |
+
"eval_samples_per_second": 6.624,
|
| 874 |
+
"eval_steps_per_second": 0.828,
|
| 875 |
+
"eval_wer": 40.756437195056805,
|
| 876 |
+
"step": 3000
|
| 877 |
+
},
|
| 878 |
+
{
|
| 879 |
+
"epoch": 5.363475177304965,
|
| 880 |
+
"grad_norm": 9.787571907043457,
|
| 881 |
+
"learning_rate": 3.672631578947369e-06,
|
| 882 |
+
"loss": 0.2017,
|
| 883 |
+
"step": 3025
|
| 884 |
+
},
|
| 885 |
+
{
|
| 886 |
+
"epoch": 5.407801418439716,
|
| 887 |
+
"grad_norm": 9.773175239562988,
|
| 888 |
+
"learning_rate": 3.6594736842105268e-06,
|
| 889 |
+
"loss": 0.1818,
|
| 890 |
+
"step": 3050
|
| 891 |
+
},
|
| 892 |
+
{
|
| 893 |
+
"epoch": 5.452127659574468,
|
| 894 |
+
"grad_norm": 9.025221824645996,
|
| 895 |
+
"learning_rate": 3.6463157894736847e-06,
|
| 896 |
+
"loss": 0.1926,
|
| 897 |
+
"step": 3075
|
| 898 |
+
},
|
| 899 |
+
{
|
| 900 |
+
"epoch": 5.49645390070922,
|
| 901 |
+
"grad_norm": 7.631556987762451,
|
| 902 |
+
"learning_rate": 3.633157894736842e-06,
|
| 903 |
+
"loss": 0.2017,
|
| 904 |
+
"step": 3100
|
| 905 |
+
},
|
| 906 |
+
{
|
| 907 |
+
"epoch": 5.540780141843972,
|
| 908 |
+
"grad_norm": 8.103202819824219,
|
| 909 |
+
"learning_rate": 3.62e-06,
|
| 910 |
+
"loss": 0.1808,
|
| 911 |
+
"step": 3125
|
| 912 |
+
},
|
| 913 |
+
{
|
| 914 |
+
"epoch": 5.585106382978723,
|
| 915 |
+
"grad_norm": 6.022019386291504,
|
| 916 |
+
"learning_rate": 3.606842105263158e-06,
|
| 917 |
+
"loss": 0.2017,
|
| 918 |
+
"step": 3150
|
| 919 |
+
},
|
| 920 |
+
{
|
| 921 |
+
"epoch": 5.629432624113475,
|
| 922 |
+
"grad_norm": 6.922440528869629,
|
| 923 |
+
"learning_rate": 3.593684210526316e-06,
|
| 924 |
+
"loss": 0.2009,
|
| 925 |
+
"step": 3175
|
| 926 |
+
},
|
| 927 |
+
{
|
| 928 |
+
"epoch": 5.673758865248227,
|
| 929 |
+
"grad_norm": 8.611794471740723,
|
| 930 |
+
"learning_rate": 3.580526315789474e-06,
|
| 931 |
+
"loss": 0.1731,
|
| 932 |
+
"step": 3200
|
| 933 |
+
},
|
| 934 |
+
{
|
| 935 |
+
"epoch": 5.718085106382979,
|
| 936 |
+
"grad_norm": 7.4870147705078125,
|
| 937 |
+
"learning_rate": 3.567368421052632e-06,
|
| 938 |
+
"loss": 0.1797,
|
| 939 |
+
"step": 3225
|
| 940 |
+
},
|
| 941 |
+
{
|
| 942 |
+
"epoch": 5.76241134751773,
|
| 943 |
+
"grad_norm": 8.757158279418945,
|
| 944 |
+
"learning_rate": 3.55421052631579e-06,
|
| 945 |
+
"loss": 0.1847,
|
| 946 |
+
"step": 3250
|
| 947 |
+
},
|
| 948 |
+
{
|
| 949 |
+
"epoch": 5.806737588652482,
|
| 950 |
+
"grad_norm": 8.361138343811035,
|
| 951 |
+
"learning_rate": 3.541052631578948e-06,
|
| 952 |
+
"loss": 0.1738,
|
| 953 |
+
"step": 3275
|
| 954 |
+
},
|
| 955 |
+
{
|
| 956 |
+
"epoch": 5.851063829787234,
|
| 957 |
+
"grad_norm": 8.07181453704834,
|
| 958 |
+
"learning_rate": 3.527894736842106e-06,
|
| 959 |
+
"loss": 0.2007,
|
| 960 |
+
"step": 3300
|
| 961 |
+
},
|
| 962 |
+
{
|
| 963 |
+
"epoch": 5.8953900709219855,
|
| 964 |
+
"grad_norm": 7.998460292816162,
|
| 965 |
+
"learning_rate": 3.5147368421052638e-06,
|
| 966 |
+
"loss": 0.1848,
|
| 967 |
+
"step": 3325
|
| 968 |
+
},
|
| 969 |
+
{
|
| 970 |
+
"epoch": 5.939716312056738,
|
| 971 |
+
"grad_norm": 7.463223934173584,
|
| 972 |
+
"learning_rate": 3.5015789473684213e-06,
|
| 973 |
+
"loss": 0.1894,
|
| 974 |
+
"step": 3350
|
| 975 |
+
},
|
| 976 |
+
{
|
| 977 |
+
"epoch": 5.98404255319149,
|
| 978 |
+
"grad_norm": 8.993099212646484,
|
| 979 |
+
"learning_rate": 3.488421052631579e-06,
|
| 980 |
+
"loss": 0.1898,
|
| 981 |
+
"step": 3375
|
| 982 |
+
},
|
| 983 |
+
{
|
| 984 |
+
"epoch": 6.028368794326241,
|
| 985 |
+
"grad_norm": 6.37155294418335,
|
| 986 |
+
"learning_rate": 3.475263157894737e-06,
|
| 987 |
+
"loss": 0.1772,
|
| 988 |
+
"step": 3400
|
| 989 |
+
},
|
| 990 |
+
{
|
| 991 |
+
"epoch": 6.072695035460993,
|
| 992 |
+
"grad_norm": 6.82436990737915,
|
| 993 |
+
"learning_rate": 3.462105263157895e-06,
|
| 994 |
+
"loss": 0.1555,
|
| 995 |
+
"step": 3425
|
| 996 |
+
},
|
| 997 |
+
{
|
| 998 |
+
"epoch": 6.117021276595745,
|
| 999 |
+
"grad_norm": 7.0470428466796875,
|
| 1000 |
+
"learning_rate": 3.448947368421053e-06,
|
| 1001 |
+
"loss": 0.1517,
|
| 1002 |
+
"step": 3450
|
| 1003 |
+
},
|
| 1004 |
+
{
|
| 1005 |
+
"epoch": 6.161347517730497,
|
| 1006 |
+
"grad_norm": 6.5624494552612305,
|
| 1007 |
+
"learning_rate": 3.435789473684211e-06,
|
| 1008 |
+
"loss": 0.149,
|
| 1009 |
+
"step": 3475
|
| 1010 |
+
},
|
| 1011 |
+
{
|
| 1012 |
+
"epoch": 6.205673758865248,
|
| 1013 |
+
"grad_norm": 7.491029262542725,
|
| 1014 |
+
"learning_rate": 3.4226315789473687e-06,
|
| 1015 |
+
"loss": 0.1599,
|
| 1016 |
+
"step": 3500
|
| 1017 |
+
},
|
| 1018 |
+
{
|
| 1019 |
+
"epoch": 6.25,
|
| 1020 |
+
"grad_norm": 6.855647563934326,
|
| 1021 |
+
"learning_rate": 3.409473684210526e-06,
|
| 1022 |
+
"loss": 0.1587,
|
| 1023 |
+
"step": 3525
|
| 1024 |
+
},
|
| 1025 |
+
{
|
| 1026 |
+
"epoch": 6.294326241134752,
|
| 1027 |
+
"grad_norm": 8.074361801147461,
|
| 1028 |
+
"learning_rate": 3.396315789473684e-06,
|
| 1029 |
+
"loss": 0.1674,
|
| 1030 |
+
"step": 3550
|
| 1031 |
+
},
|
| 1032 |
+
{
|
| 1033 |
+
"epoch": 6.338652482269503,
|
| 1034 |
+
"grad_norm": 5.963619709014893,
|
| 1035 |
+
"learning_rate": 3.3831578947368424e-06,
|
| 1036 |
+
"loss": 0.1499,
|
| 1037 |
+
"step": 3575
|
| 1038 |
+
},
|
| 1039 |
+
{
|
| 1040 |
+
"epoch": 6.382978723404255,
|
| 1041 |
+
"grad_norm": 8.021512985229492,
|
| 1042 |
+
"learning_rate": 3.3700000000000003e-06,
|
| 1043 |
+
"loss": 0.1488,
|
| 1044 |
+
"step": 3600
|
| 1045 |
+
},
|
| 1046 |
+
{
|
| 1047 |
+
"epoch": 6.427304964539007,
|
| 1048 |
+
"grad_norm": 5.919581413269043,
|
| 1049 |
+
"learning_rate": 3.3568421052631583e-06,
|
| 1050 |
+
"loss": 0.1498,
|
| 1051 |
+
"step": 3625
|
| 1052 |
+
},
|
| 1053 |
+
{
|
| 1054 |
+
"epoch": 6.471631205673759,
|
| 1055 |
+
"grad_norm": 6.950247287750244,
|
| 1056 |
+
"learning_rate": 3.343684210526316e-06,
|
| 1057 |
+
"loss": 0.1648,
|
| 1058 |
+
"step": 3650
|
| 1059 |
+
},
|
| 1060 |
+
{
|
| 1061 |
+
"epoch": 6.51595744680851,
|
| 1062 |
+
"grad_norm": 6.72702693939209,
|
| 1063 |
+
"learning_rate": 3.330526315789474e-06,
|
| 1064 |
+
"loss": 0.146,
|
| 1065 |
+
"step": 3675
|
| 1066 |
+
},
|
| 1067 |
+
{
|
| 1068 |
+
"epoch": 6.560283687943262,
|
| 1069 |
+
"grad_norm": 7.681860446929932,
|
| 1070 |
+
"learning_rate": 3.317368421052632e-06,
|
| 1071 |
+
"loss": 0.1542,
|
| 1072 |
+
"step": 3700
|
| 1073 |
+
},
|
| 1074 |
+
{
|
| 1075 |
+
"epoch": 6.6046099290780145,
|
| 1076 |
+
"grad_norm": 7.239710330963135,
|
| 1077 |
+
"learning_rate": 3.30421052631579e-06,
|
| 1078 |
+
"loss": 0.1499,
|
| 1079 |
+
"step": 3725
|
| 1080 |
+
},
|
| 1081 |
+
{
|
| 1082 |
+
"epoch": 6.648936170212766,
|
| 1083 |
+
"grad_norm": 7.310706615447998,
|
| 1084 |
+
"learning_rate": 3.291052631578948e-06,
|
| 1085 |
+
"loss": 0.1527,
|
| 1086 |
+
"step": 3750
|
| 1087 |
+
},
|
| 1088 |
+
{
|
| 1089 |
+
"epoch": 6.693262411347518,
|
| 1090 |
+
"grad_norm": 7.060523509979248,
|
| 1091 |
+
"learning_rate": 3.2778947368421053e-06,
|
| 1092 |
+
"loss": 0.1537,
|
| 1093 |
+
"step": 3775
|
| 1094 |
+
},
|
| 1095 |
+
{
|
| 1096 |
+
"epoch": 6.73758865248227,
|
| 1097 |
+
"grad_norm": 8.571366310119629,
|
| 1098 |
+
"learning_rate": 3.264736842105263e-06,
|
| 1099 |
+
"loss": 0.1598,
|
| 1100 |
+
"step": 3800
|
| 1101 |
+
},
|
| 1102 |
+
{
|
| 1103 |
+
"epoch": 6.781914893617021,
|
| 1104 |
+
"grad_norm": 6.046979904174805,
|
| 1105 |
+
"learning_rate": 3.251578947368421e-06,
|
| 1106 |
+
"loss": 0.1416,
|
| 1107 |
+
"step": 3825
|
| 1108 |
+
},
|
| 1109 |
+
{
|
| 1110 |
+
"epoch": 6.826241134751773,
|
| 1111 |
+
"grad_norm": 8.620864868164062,
|
| 1112 |
+
"learning_rate": 3.238421052631579e-06,
|
| 1113 |
+
"loss": 0.1454,
|
| 1114 |
+
"step": 3850
|
| 1115 |
+
},
|
| 1116 |
+
{
|
| 1117 |
+
"epoch": 6.870567375886525,
|
| 1118 |
+
"grad_norm": 8.330490112304688,
|
| 1119 |
+
"learning_rate": 3.225263157894737e-06,
|
| 1120 |
+
"loss": 0.1542,
|
| 1121 |
+
"step": 3875
|
| 1122 |
+
},
|
| 1123 |
+
{
|
| 1124 |
+
"epoch": 6.914893617021277,
|
| 1125 |
+
"grad_norm": 10.611557006835938,
|
| 1126 |
+
"learning_rate": 3.212105263157895e-06,
|
| 1127 |
+
"loss": 0.1676,
|
| 1128 |
+
"step": 3900
|
| 1129 |
+
},
|
| 1130 |
+
{
|
| 1131 |
+
"epoch": 6.959219858156028,
|
| 1132 |
+
"grad_norm": 6.815483570098877,
|
| 1133 |
+
"learning_rate": 3.198947368421053e-06,
|
| 1134 |
+
"loss": 0.1405,
|
| 1135 |
+
"step": 3925
|
| 1136 |
+
},
|
| 1137 |
+
{
|
| 1138 |
+
"epoch": 7.00354609929078,
|
| 1139 |
+
"grad_norm": 5.8308634757995605,
|
| 1140 |
+
"learning_rate": 3.185789473684211e-06,
|
| 1141 |
+
"loss": 0.1479,
|
| 1142 |
+
"step": 3950
|
| 1143 |
+
},
|
| 1144 |
+
{
|
| 1145 |
+
"epoch": 7.047872340425532,
|
| 1146 |
+
"grad_norm": 6.529901027679443,
|
| 1147 |
+
"learning_rate": 3.172631578947369e-06,
|
| 1148 |
+
"loss": 0.12,
|
| 1149 |
+
"step": 3975
|
| 1150 |
+
},
|
| 1151 |
+
{
|
| 1152 |
+
"epoch": 7.092198581560283,
|
| 1153 |
+
"grad_norm": 6.525743007659912,
|
| 1154 |
+
"learning_rate": 3.159473684210527e-06,
|
| 1155 |
+
"loss": 0.1214,
|
| 1156 |
+
"step": 4000
|
| 1157 |
+
},
|
| 1158 |
+
{
|
| 1159 |
+
"epoch": 7.092198581560283,
|
| 1160 |
+
"eval_loss": 0.48074454069137573,
|
| 1161 |
+
"eval_runtime": 589.8236,
|
| 1162 |
+
"eval_samples_per_second": 6.605,
|
| 1163 |
+
"eval_steps_per_second": 0.826,
|
| 1164 |
+
"eval_wer": 41.810605790804516,
|
| 1165 |
+
"step": 4000
|
| 1166 |
+
},
|
| 1167 |
+
{
|
| 1168 |
+
"epoch": 7.136524822695035,
|
| 1169 |
+
"grad_norm": 5.498377799987793,
|
| 1170 |
+
"learning_rate": 3.1463157894736844e-06,
|
| 1171 |
+
"loss": 0.1295,
|
| 1172 |
+
"step": 4025
|
| 1173 |
+
},
|
| 1174 |
+
{
|
| 1175 |
+
"epoch": 7.180851063829787,
|
| 1176 |
+
"grad_norm": 7.085293292999268,
|
| 1177 |
+
"learning_rate": 3.1331578947368423e-06,
|
| 1178 |
+
"loss": 0.1224,
|
| 1179 |
+
"step": 4050
|
| 1180 |
+
},
|
| 1181 |
+
{
|
| 1182 |
+
"epoch": 7.225177304964539,
|
| 1183 |
+
"grad_norm": 4.779361724853516,
|
| 1184 |
+
"learning_rate": 3.12e-06,
|
| 1185 |
+
"loss": 0.1245,
|
| 1186 |
+
"step": 4075
|
| 1187 |
+
},
|
| 1188 |
+
{
|
| 1189 |
+
"epoch": 7.2695035460992905,
|
| 1190 |
+
"grad_norm": 6.457351207733154,
|
| 1191 |
+
"learning_rate": 3.106842105263158e-06,
|
| 1192 |
+
"loss": 0.1192,
|
| 1193 |
+
"step": 4100
|
| 1194 |
+
},
|
| 1195 |
+
{
|
| 1196 |
+
"epoch": 7.3138297872340425,
|
| 1197 |
+
"grad_norm": 4.801368236541748,
|
| 1198 |
+
"learning_rate": 3.093684210526316e-06,
|
| 1199 |
+
"loss": 0.1322,
|
| 1200 |
+
"step": 4125
|
| 1201 |
+
},
|
| 1202 |
+
{
|
| 1203 |
+
"epoch": 7.358156028368795,
|
| 1204 |
+
"grad_norm": 6.449742794036865,
|
| 1205 |
+
"learning_rate": 3.080526315789474e-06,
|
| 1206 |
+
"loss": 0.1218,
|
| 1207 |
+
"step": 4150
|
| 1208 |
+
},
|
| 1209 |
+
{
|
| 1210 |
+
"epoch": 7.402482269503546,
|
| 1211 |
+
"grad_norm": 5.5234456062316895,
|
| 1212 |
+
"learning_rate": 3.067368421052632e-06,
|
| 1213 |
+
"loss": 0.1293,
|
| 1214 |
+
"step": 4175
|
| 1215 |
+
},
|
| 1216 |
+
{
|
| 1217 |
+
"epoch": 7.446808510638298,
|
| 1218 |
+
"grad_norm": 8.489788055419922,
|
| 1219 |
+
"learning_rate": 3.0542105263157893e-06,
|
| 1220 |
+
"loss": 0.1181,
|
| 1221 |
+
"step": 4200
|
| 1222 |
+
},
|
| 1223 |
+
{
|
| 1224 |
+
"epoch": 7.49113475177305,
|
| 1225 |
+
"grad_norm": 6.528730869293213,
|
| 1226 |
+
"learning_rate": 3.0410526315789472e-06,
|
| 1227 |
+
"loss": 0.1263,
|
| 1228 |
+
"step": 4225
|
| 1229 |
+
},
|
| 1230 |
+
{
|
| 1231 |
+
"epoch": 7.535460992907802,
|
| 1232 |
+
"grad_norm": 6.973687648773193,
|
| 1233 |
+
"learning_rate": 3.027894736842106e-06,
|
| 1234 |
+
"loss": 0.1179,
|
| 1235 |
+
"step": 4250
|
| 1236 |
+
},
|
| 1237 |
+
{
|
| 1238 |
+
"epoch": 7.579787234042553,
|
| 1239 |
+
"grad_norm": 6.633789539337158,
|
| 1240 |
+
"learning_rate": 3.0147368421052635e-06,
|
| 1241 |
+
"loss": 0.1314,
|
| 1242 |
+
"step": 4275
|
| 1243 |
+
},
|
| 1244 |
+
{
|
| 1245 |
+
"epoch": 7.624113475177305,
|
| 1246 |
+
"grad_norm": 7.518368721008301,
|
| 1247 |
+
"learning_rate": 3.0015789473684214e-06,
|
| 1248 |
+
"loss": 0.1305,
|
| 1249 |
+
"step": 4300
|
| 1250 |
+
},
|
| 1251 |
+
{
|
| 1252 |
+
"epoch": 7.668439716312057,
|
| 1253 |
+
"grad_norm": 5.739889144897461,
|
| 1254 |
+
"learning_rate": 2.9884210526315793e-06,
|
| 1255 |
+
"loss": 0.1295,
|
| 1256 |
+
"step": 4325
|
| 1257 |
+
},
|
| 1258 |
+
{
|
| 1259 |
+
"epoch": 7.712765957446808,
|
| 1260 |
+
"grad_norm": 6.737969398498535,
|
| 1261 |
+
"learning_rate": 2.975263157894737e-06,
|
| 1262 |
+
"loss": 0.1256,
|
| 1263 |
+
"step": 4350
|
| 1264 |
+
},
|
| 1265 |
+
{
|
| 1266 |
+
"epoch": 7.75709219858156,
|
| 1267 |
+
"grad_norm": 5.012901782989502,
|
| 1268 |
+
"learning_rate": 2.962105263157895e-06,
|
| 1269 |
+
"loss": 0.1312,
|
| 1270 |
+
"step": 4375
|
| 1271 |
+
},
|
| 1272 |
+
{
|
| 1273 |
+
"epoch": 7.801418439716312,
|
| 1274 |
+
"grad_norm": 5.5256853103637695,
|
| 1275 |
+
"learning_rate": 2.948947368421053e-06,
|
| 1276 |
+
"loss": 0.1242,
|
| 1277 |
+
"step": 4400
|
| 1278 |
+
},
|
| 1279 |
+
{
|
| 1280 |
+
"epoch": 7.845744680851064,
|
| 1281 |
+
"grad_norm": 6.629995346069336,
|
| 1282 |
+
"learning_rate": 2.935789473684211e-06,
|
| 1283 |
+
"loss": 0.125,
|
| 1284 |
+
"step": 4425
|
| 1285 |
+
},
|
| 1286 |
+
{
|
| 1287 |
+
"epoch": 7.890070921985815,
|
| 1288 |
+
"grad_norm": 5.227272987365723,
|
| 1289 |
+
"learning_rate": 2.9226315789473684e-06,
|
| 1290 |
+
"loss": 0.1191,
|
| 1291 |
+
"step": 4450
|
| 1292 |
+
},
|
| 1293 |
+
{
|
| 1294 |
+
"epoch": 7.934397163120567,
|
| 1295 |
+
"grad_norm": 5.111964225769043,
|
| 1296 |
+
"learning_rate": 2.9094736842105263e-06,
|
| 1297 |
+
"loss": 0.1145,
|
| 1298 |
+
"step": 4475
|
| 1299 |
+
},
|
| 1300 |
+
{
|
| 1301 |
+
"epoch": 7.9787234042553195,
|
| 1302 |
+
"grad_norm": 5.537423610687256,
|
| 1303 |
+
"learning_rate": 2.8963157894736842e-06,
|
| 1304 |
+
"loss": 0.1278,
|
| 1305 |
+
"step": 4500
|
| 1306 |
+
},
|
| 1307 |
+
{
|
| 1308 |
+
"epoch": 8.02304964539007,
|
| 1309 |
+
"grad_norm": 4.478297710418701,
|
| 1310 |
+
"learning_rate": 2.883157894736842e-06,
|
| 1311 |
+
"loss": 0.1155,
|
| 1312 |
+
"step": 4525
|
| 1313 |
+
},
|
| 1314 |
+
{
|
| 1315 |
+
"epoch": 8.067375886524824,
|
| 1316 |
+
"grad_norm": 5.642357349395752,
|
| 1317 |
+
"learning_rate": 2.87e-06,
|
| 1318 |
+
"loss": 0.1102,
|
| 1319 |
+
"step": 4550
|
| 1320 |
+
},
|
| 1321 |
+
{
|
| 1322 |
+
"epoch": 8.111702127659575,
|
| 1323 |
+
"grad_norm": 5.228881359100342,
|
| 1324 |
+
"learning_rate": 2.856842105263158e-06,
|
| 1325 |
+
"loss": 0.1009,
|
| 1326 |
+
"step": 4575
|
| 1327 |
+
},
|
| 1328 |
+
{
|
| 1329 |
+
"epoch": 8.156028368794326,
|
| 1330 |
+
"grad_norm": 5.18090295791626,
|
| 1331 |
+
"learning_rate": 2.8436842105263163e-06,
|
| 1332 |
+
"loss": 0.0992,
|
| 1333 |
+
"step": 4600
|
| 1334 |
+
},
|
| 1335 |
+
{
|
| 1336 |
+
"epoch": 8.200354609929079,
|
| 1337 |
+
"grad_norm": 6.664114475250244,
|
| 1338 |
+
"learning_rate": 2.830526315789474e-06,
|
| 1339 |
+
"loss": 0.1092,
|
| 1340 |
+
"step": 4625
|
| 1341 |
+
},
|
| 1342 |
+
{
|
| 1343 |
+
"epoch": 8.24468085106383,
|
| 1344 |
+
"grad_norm": 5.435600280761719,
|
| 1345 |
+
"learning_rate": 2.817368421052632e-06,
|
| 1346 |
+
"loss": 0.1011,
|
| 1347 |
+
"step": 4650
|
| 1348 |
+
},
|
| 1349 |
+
{
|
| 1350 |
+
"epoch": 8.289007092198581,
|
| 1351 |
+
"grad_norm": 5.778509140014648,
|
| 1352 |
+
"learning_rate": 2.80421052631579e-06,
|
| 1353 |
+
"loss": 0.1025,
|
| 1354 |
+
"step": 4675
|
| 1355 |
+
},
|
| 1356 |
+
{
|
| 1357 |
+
"epoch": 8.333333333333334,
|
| 1358 |
+
"grad_norm": 5.977304458618164,
|
| 1359 |
+
"learning_rate": 2.7910526315789475e-06,
|
| 1360 |
+
"loss": 0.1021,
|
| 1361 |
+
"step": 4700
|
| 1362 |
+
},
|
| 1363 |
+
{
|
| 1364 |
+
"epoch": 8.377659574468085,
|
| 1365 |
+
"grad_norm": 5.274112701416016,
|
| 1366 |
+
"learning_rate": 2.7778947368421054e-06,
|
| 1367 |
+
"loss": 0.0935,
|
| 1368 |
+
"step": 4725
|
| 1369 |
+
},
|
| 1370 |
+
{
|
| 1371 |
+
"epoch": 8.421985815602836,
|
| 1372 |
+
"grad_norm": 5.418082237243652,
|
| 1373 |
+
"learning_rate": 2.7647368421052633e-06,
|
| 1374 |
+
"loss": 0.0965,
|
| 1375 |
+
"step": 4750
|
| 1376 |
+
},
|
| 1377 |
+
{
|
| 1378 |
+
"epoch": 8.46631205673759,
|
| 1379 |
+
"grad_norm": 5.248587131500244,
|
| 1380 |
+
"learning_rate": 2.7515789473684212e-06,
|
| 1381 |
+
"loss": 0.1014,
|
| 1382 |
+
"step": 4775
|
| 1383 |
+
},
|
| 1384 |
+
{
|
| 1385 |
+
"epoch": 8.51063829787234,
|
| 1386 |
+
"grad_norm": 5.329669952392578,
|
| 1387 |
+
"learning_rate": 2.738421052631579e-06,
|
| 1388 |
+
"loss": 0.1052,
|
| 1389 |
+
"step": 4800
|
| 1390 |
+
},
|
| 1391 |
+
{
|
| 1392 |
+
"epoch": 8.554964539007091,
|
| 1393 |
+
"grad_norm": 5.469305038452148,
|
| 1394 |
+
"learning_rate": 2.725263157894737e-06,
|
| 1395 |
+
"loss": 0.0988,
|
| 1396 |
+
"step": 4825
|
| 1397 |
+
},
|
| 1398 |
+
{
|
| 1399 |
+
"epoch": 8.599290780141844,
|
| 1400 |
+
"grad_norm": 4.839619159698486,
|
| 1401 |
+
"learning_rate": 2.712105263157895e-06,
|
| 1402 |
+
"loss": 0.1038,
|
| 1403 |
+
"step": 4850
|
| 1404 |
+
},
|
| 1405 |
+
{
|
| 1406 |
+
"epoch": 8.643617021276595,
|
| 1407 |
+
"grad_norm": 5.6988420486450195,
|
| 1408 |
+
"learning_rate": 2.6989473684210524e-06,
|
| 1409 |
+
"loss": 0.1005,
|
| 1410 |
+
"step": 4875
|
| 1411 |
+
},
|
| 1412 |
+
{
|
| 1413 |
+
"epoch": 8.687943262411348,
|
| 1414 |
+
"grad_norm": 6.122032165527344,
|
| 1415 |
+
"learning_rate": 2.6857894736842104e-06,
|
| 1416 |
+
"loss": 0.1071,
|
| 1417 |
+
"step": 4900
|
| 1418 |
+
},
|
| 1419 |
+
{
|
| 1420 |
+
"epoch": 8.7322695035461,
|
| 1421 |
+
"grad_norm": 5.00734806060791,
|
| 1422 |
+
"learning_rate": 2.672631578947369e-06,
|
| 1423 |
+
"loss": 0.1009,
|
| 1424 |
+
"step": 4925
|
| 1425 |
+
},
|
| 1426 |
+
{
|
| 1427 |
+
"epoch": 8.77659574468085,
|
| 1428 |
+
"grad_norm": 6.199928283691406,
|
| 1429 |
+
"learning_rate": 2.6594736842105266e-06,
|
| 1430 |
+
"loss": 0.1041,
|
| 1431 |
+
"step": 4950
|
| 1432 |
+
},
|
| 1433 |
+
{
|
| 1434 |
+
"epoch": 8.820921985815604,
|
| 1435 |
+
"grad_norm": 6.134685516357422,
|
| 1436 |
+
"learning_rate": 2.6463157894736845e-06,
|
| 1437 |
+
"loss": 0.1044,
|
| 1438 |
+
"step": 4975
|
| 1439 |
+
},
|
| 1440 |
+
{
|
| 1441 |
+
"epoch": 8.865248226950355,
|
| 1442 |
+
"grad_norm": 5.176562786102295,
|
| 1443 |
+
"learning_rate": 2.6331578947368424e-06,
|
| 1444 |
+
"loss": 0.1066,
|
| 1445 |
+
"step": 5000
|
| 1446 |
+
},
|
| 1447 |
+
{
|
| 1448 |
+
"epoch": 8.865248226950355,
|
| 1449 |
+
"eval_loss": 0.4846879839897156,
|
| 1450 |
+
"eval_runtime": 580.4904,
|
| 1451 |
+
"eval_samples_per_second": 6.712,
|
| 1452 |
+
"eval_steps_per_second": 0.839,
|
| 1453 |
+
"eval_wer": 37.96787634887283,
|
| 1454 |
+
"step": 5000
|
| 1455 |
+
},
|
| 1456 |
+
{
|
| 1457 |
+
"epoch": 8.909574468085106,
|
| 1458 |
+
"grad_norm": 5.835010051727295,
|
| 1459 |
+
"learning_rate": 2.6200000000000003e-06,
|
| 1460 |
+
"loss": 0.1001,
|
| 1461 |
+
"step": 5025
|
| 1462 |
+
},
|
| 1463 |
+
{
|
| 1464 |
+
"epoch": 8.953900709219859,
|
| 1465 |
+
"grad_norm": 6.407568454742432,
|
| 1466 |
+
"learning_rate": 2.6068421052631582e-06,
|
| 1467 |
+
"loss": 0.0965,
|
| 1468 |
+
"step": 5050
|
| 1469 |
+
},
|
| 1470 |
+
{
|
| 1471 |
+
"epoch": 8.99822695035461,
|
| 1472 |
+
"grad_norm": 6.197821617126465,
|
| 1473 |
+
"learning_rate": 2.593684210526316e-06,
|
| 1474 |
+
"loss": 0.1048,
|
| 1475 |
+
"step": 5075
|
| 1476 |
+
},
|
| 1477 |
+
{
|
| 1478 |
+
"epoch": 9.042553191489361,
|
| 1479 |
+
"grad_norm": 4.808340072631836,
|
| 1480 |
+
"learning_rate": 2.580526315789474e-06,
|
| 1481 |
+
"loss": 0.085,
|
| 1482 |
+
"step": 5100
|
| 1483 |
+
},
|
| 1484 |
+
{
|
| 1485 |
+
"epoch": 9.086879432624114,
|
| 1486 |
+
"grad_norm": 4.098535060882568,
|
| 1487 |
+
"learning_rate": 2.5673684210526315e-06,
|
| 1488 |
+
"loss": 0.0815,
|
| 1489 |
+
"step": 5125
|
| 1490 |
+
},
|
| 1491 |
+
{
|
| 1492 |
+
"epoch": 9.131205673758865,
|
| 1493 |
+
"grad_norm": 4.34876012802124,
|
| 1494 |
+
"learning_rate": 2.5542105263157894e-06,
|
| 1495 |
+
"loss": 0.0884,
|
| 1496 |
+
"step": 5150
|
| 1497 |
+
},
|
| 1498 |
+
{
|
| 1499 |
+
"epoch": 9.175531914893616,
|
| 1500 |
+
"grad_norm": 4.318136692047119,
|
| 1501 |
+
"learning_rate": 2.5410526315789474e-06,
|
| 1502 |
+
"loss": 0.0826,
|
| 1503 |
+
"step": 5175
|
| 1504 |
+
},
|
| 1505 |
+
{
|
| 1506 |
+
"epoch": 9.21985815602837,
|
| 1507 |
+
"grad_norm": 5.460968017578125,
|
| 1508 |
+
"learning_rate": 2.5278947368421053e-06,
|
| 1509 |
+
"loss": 0.0799,
|
| 1510 |
+
"step": 5200
|
| 1511 |
+
},
|
| 1512 |
+
{
|
| 1513 |
+
"epoch": 9.26418439716312,
|
| 1514 |
+
"grad_norm": 4.200242042541504,
|
| 1515 |
+
"learning_rate": 2.514736842105263e-06,
|
| 1516 |
+
"loss": 0.0835,
|
| 1517 |
+
"step": 5225
|
| 1518 |
+
},
|
| 1519 |
+
{
|
| 1520 |
+
"epoch": 9.308510638297872,
|
| 1521 |
+
"grad_norm": 5.984395503997803,
|
| 1522 |
+
"learning_rate": 2.501578947368421e-06,
|
| 1523 |
+
"loss": 0.0881,
|
| 1524 |
+
"step": 5250
|
| 1525 |
+
},
|
| 1526 |
+
{
|
| 1527 |
+
"epoch": 9.352836879432624,
|
| 1528 |
+
"grad_norm": 4.829773426055908,
|
| 1529 |
+
"learning_rate": 2.488421052631579e-06,
|
| 1530 |
+
"loss": 0.0786,
|
| 1531 |
+
"step": 5275
|
| 1532 |
+
},
|
| 1533 |
+
{
|
| 1534 |
+
"epoch": 9.397163120567376,
|
| 1535 |
+
"grad_norm": 3.996610403060913,
|
| 1536 |
+
"learning_rate": 2.475263157894737e-06,
|
| 1537 |
+
"loss": 0.0865,
|
| 1538 |
+
"step": 5300
|
| 1539 |
+
},
|
| 1540 |
+
{
|
| 1541 |
+
"epoch": 9.441489361702128,
|
| 1542 |
+
"grad_norm": 6.336328029632568,
|
| 1543 |
+
"learning_rate": 2.462105263157895e-06,
|
| 1544 |
+
"loss": 0.0802,
|
| 1545 |
+
"step": 5325
|
| 1546 |
+
},
|
| 1547 |
+
{
|
| 1548 |
+
"epoch": 9.48581560283688,
|
| 1549 |
+
"grad_norm": 5.054424285888672,
|
| 1550 |
+
"learning_rate": 2.448947368421053e-06,
|
| 1551 |
+
"loss": 0.0928,
|
| 1552 |
+
"step": 5350
|
| 1553 |
+
},
|
| 1554 |
+
{
|
| 1555 |
+
"epoch": 9.53014184397163,
|
| 1556 |
+
"grad_norm": 6.595405101776123,
|
| 1557 |
+
"learning_rate": 2.4357894736842106e-06,
|
| 1558 |
+
"loss": 0.0858,
|
| 1559 |
+
"step": 5375
|
| 1560 |
+
},
|
| 1561 |
+
{
|
| 1562 |
+
"epoch": 9.574468085106384,
|
| 1563 |
+
"grad_norm": 5.797497272491455,
|
| 1564 |
+
"learning_rate": 2.4226315789473685e-06,
|
| 1565 |
+
"loss": 0.0846,
|
| 1566 |
+
"step": 5400
|
| 1567 |
+
},
|
| 1568 |
+
{
|
| 1569 |
+
"epoch": 9.618794326241135,
|
| 1570 |
+
"grad_norm": 5.1372551918029785,
|
| 1571 |
+
"learning_rate": 2.4094736842105265e-06,
|
| 1572 |
+
"loss": 0.0789,
|
| 1573 |
+
"step": 5425
|
| 1574 |
+
},
|
| 1575 |
+
{
|
| 1576 |
+
"epoch": 9.663120567375886,
|
| 1577 |
+
"grad_norm": 6.663181304931641,
|
| 1578 |
+
"learning_rate": 2.3963157894736844e-06,
|
| 1579 |
+
"loss": 0.0977,
|
| 1580 |
+
"step": 5450
|
| 1581 |
+
},
|
| 1582 |
+
{
|
| 1583 |
+
"epoch": 9.707446808510639,
|
| 1584 |
+
"grad_norm": 4.910397529602051,
|
| 1585 |
+
"learning_rate": 2.3831578947368423e-06,
|
| 1586 |
+
"loss": 0.087,
|
| 1587 |
+
"step": 5475
|
| 1588 |
+
},
|
| 1589 |
+
{
|
| 1590 |
+
"epoch": 9.75177304964539,
|
| 1591 |
+
"grad_norm": 5.87327241897583,
|
| 1592 |
+
"learning_rate": 2.37e-06,
|
| 1593 |
+
"loss": 0.076,
|
| 1594 |
+
"step": 5500
|
| 1595 |
+
},
|
| 1596 |
+
{
|
| 1597 |
+
"epoch": 9.796099290780141,
|
| 1598 |
+
"grad_norm": 4.994716167449951,
|
| 1599 |
+
"learning_rate": 2.356842105263158e-06,
|
| 1600 |
+
"loss": 0.083,
|
| 1601 |
+
"step": 5525
|
| 1602 |
+
},
|
| 1603 |
+
{
|
| 1604 |
+
"epoch": 9.840425531914894,
|
| 1605 |
+
"grad_norm": 5.601754665374756,
|
| 1606 |
+
"learning_rate": 2.343684210526316e-06,
|
| 1607 |
+
"loss": 0.0819,
|
| 1608 |
+
"step": 5550
|
| 1609 |
+
},
|
| 1610 |
+
{
|
| 1611 |
+
"epoch": 9.884751773049645,
|
| 1612 |
+
"grad_norm": 5.767611980438232,
|
| 1613 |
+
"learning_rate": 2.330526315789474e-06,
|
| 1614 |
+
"loss": 0.0831,
|
| 1615 |
+
"step": 5575
|
| 1616 |
+
},
|
| 1617 |
+
{
|
| 1618 |
+
"epoch": 9.929078014184396,
|
| 1619 |
+
"grad_norm": 6.679659366607666,
|
| 1620 |
+
"learning_rate": 2.317368421052632e-06,
|
| 1621 |
+
"loss": 0.084,
|
| 1622 |
+
"step": 5600
|
| 1623 |
+
},
|
| 1624 |
+
{
|
| 1625 |
+
"epoch": 9.97340425531915,
|
| 1626 |
+
"grad_norm": 3.7785329818725586,
|
| 1627 |
+
"learning_rate": 2.3042105263157897e-06,
|
| 1628 |
+
"loss": 0.0894,
|
| 1629 |
+
"step": 5625
|
| 1630 |
+
},
|
| 1631 |
+
{
|
| 1632 |
+
"epoch": 10.0177304964539,
|
| 1633 |
+
"grad_norm": 4.858386516571045,
|
| 1634 |
+
"learning_rate": 2.2910526315789476e-06,
|
| 1635 |
+
"loss": 0.0817,
|
| 1636 |
+
"step": 5650
|
| 1637 |
+
},
|
| 1638 |
+
{
|
| 1639 |
+
"epoch": 10.062056737588652,
|
| 1640 |
+
"grad_norm": 3.9162485599517822,
|
| 1641 |
+
"learning_rate": 2.277894736842105e-06,
|
| 1642 |
+
"loss": 0.0669,
|
| 1643 |
+
"step": 5675
|
| 1644 |
+
},
|
| 1645 |
+
{
|
| 1646 |
+
"epoch": 10.106382978723405,
|
| 1647 |
+
"grad_norm": 5.168406963348389,
|
| 1648 |
+
"learning_rate": 2.2647368421052635e-06,
|
| 1649 |
+
"loss": 0.0677,
|
| 1650 |
+
"step": 5700
|
| 1651 |
+
},
|
| 1652 |
+
{
|
| 1653 |
+
"epoch": 10.150709219858156,
|
| 1654 |
+
"grad_norm": 3.850172758102417,
|
| 1655 |
+
"learning_rate": 2.2515789473684214e-06,
|
| 1656 |
+
"loss": 0.0687,
|
| 1657 |
+
"step": 5725
|
| 1658 |
+
},
|
| 1659 |
+
{
|
| 1660 |
+
"epoch": 10.195035460992909,
|
| 1661 |
+
"grad_norm": 3.854781150817871,
|
| 1662 |
+
"learning_rate": 2.2384210526315793e-06,
|
| 1663 |
+
"loss": 0.0734,
|
| 1664 |
+
"step": 5750
|
| 1665 |
+
},
|
| 1666 |
+
{
|
| 1667 |
+
"epoch": 10.23936170212766,
|
| 1668 |
+
"grad_norm": 3.807837724685669,
|
| 1669 |
+
"learning_rate": 2.225263157894737e-06,
|
| 1670 |
+
"loss": 0.0654,
|
| 1671 |
+
"step": 5775
|
| 1672 |
+
},
|
| 1673 |
+
{
|
| 1674 |
+
"epoch": 10.28368794326241,
|
| 1675 |
+
"grad_norm": 6.461479187011719,
|
| 1676 |
+
"learning_rate": 2.2121052631578947e-06,
|
| 1677 |
+
"loss": 0.0692,
|
| 1678 |
+
"step": 5800
|
| 1679 |
+
},
|
| 1680 |
+
{
|
| 1681 |
+
"epoch": 10.328014184397164,
|
| 1682 |
+
"grad_norm": 5.051649570465088,
|
| 1683 |
+
"learning_rate": 2.1989473684210526e-06,
|
| 1684 |
+
"loss": 0.0705,
|
| 1685 |
+
"step": 5825
|
| 1686 |
+
},
|
| 1687 |
+
{
|
| 1688 |
+
"epoch": 10.372340425531915,
|
| 1689 |
+
"grad_norm": 4.43517541885376,
|
| 1690 |
+
"learning_rate": 2.1857894736842105e-06,
|
| 1691 |
+
"loss": 0.0738,
|
| 1692 |
+
"step": 5850
|
| 1693 |
+
},
|
| 1694 |
+
{
|
| 1695 |
+
"epoch": 10.416666666666666,
|
| 1696 |
+
"grad_norm": 3.5894205570220947,
|
| 1697 |
+
"learning_rate": 2.172631578947369e-06,
|
| 1698 |
+
"loss": 0.0699,
|
| 1699 |
+
"step": 5875
|
| 1700 |
+
},
|
| 1701 |
+
{
|
| 1702 |
+
"epoch": 10.460992907801419,
|
| 1703 |
+
"grad_norm": 4.5283203125,
|
| 1704 |
+
"learning_rate": 2.1594736842105267e-06,
|
| 1705 |
+
"loss": 0.0688,
|
| 1706 |
+
"step": 5900
|
| 1707 |
+
},
|
| 1708 |
+
{
|
| 1709 |
+
"epoch": 10.50531914893617,
|
| 1710 |
+
"grad_norm": 3.9678380489349365,
|
| 1711 |
+
"learning_rate": 2.1463157894736842e-06,
|
| 1712 |
+
"loss": 0.0678,
|
| 1713 |
+
"step": 5925
|
| 1714 |
+
},
|
| 1715 |
+
{
|
| 1716 |
+
"epoch": 10.549645390070921,
|
| 1717 |
+
"grad_norm": 6.319568157196045,
|
| 1718 |
+
"learning_rate": 2.133157894736842e-06,
|
| 1719 |
+
"loss": 0.0716,
|
| 1720 |
+
"step": 5950
|
| 1721 |
+
},
|
| 1722 |
+
{
|
| 1723 |
+
"epoch": 10.593971631205674,
|
| 1724 |
+
"grad_norm": 4.849029541015625,
|
| 1725 |
+
"learning_rate": 2.12e-06,
|
| 1726 |
+
"loss": 0.0746,
|
| 1727 |
+
"step": 5975
|
| 1728 |
+
},
|
| 1729 |
+
{
|
| 1730 |
+
"epoch": 10.638297872340425,
|
| 1731 |
+
"grad_norm": 3.9395298957824707,
|
| 1732 |
+
"learning_rate": 2.106842105263158e-06,
|
| 1733 |
+
"loss": 0.0673,
|
| 1734 |
+
"step": 6000
|
| 1735 |
+
},
|
| 1736 |
+
{
|
| 1737 |
+
"epoch": 10.638297872340425,
|
| 1738 |
+
"eval_loss": 0.4972631335258484,
|
| 1739 |
+
"eval_runtime": 583.012,
|
| 1740 |
+
"eval_samples_per_second": 6.683,
|
| 1741 |
+
"eval_steps_per_second": 0.835,
|
| 1742 |
+
"eval_wer": 39.709391360091175,
|
| 1743 |
+
"step": 6000
|
| 1744 |
+
},
|
| 1745 |
+
{
|
| 1746 |
+
"epoch": 10.682624113475176,
|
| 1747 |
+
"grad_norm": 4.806758880615234,
|
| 1748 |
+
"learning_rate": 2.0936842105263163e-06,
|
| 1749 |
+
"loss": 0.069,
|
| 1750 |
+
"step": 6025
|
| 1751 |
+
},
|
| 1752 |
+
{
|
| 1753 |
+
"epoch": 10.72695035460993,
|
| 1754 |
+
"grad_norm": 5.572425842285156,
|
| 1755 |
+
"learning_rate": 2.0805263157894738e-06,
|
| 1756 |
+
"loss": 0.0699,
|
| 1757 |
+
"step": 6050
|
| 1758 |
+
},
|
| 1759 |
+
{
|
| 1760 |
+
"epoch": 10.77127659574468,
|
| 1761 |
+
"grad_norm": 4.8004302978515625,
|
| 1762 |
+
"learning_rate": 2.0673684210526317e-06,
|
| 1763 |
+
"loss": 0.0712,
|
| 1764 |
+
"step": 6075
|
| 1765 |
+
},
|
| 1766 |
+
{
|
| 1767 |
+
"epoch": 10.815602836879432,
|
| 1768 |
+
"grad_norm": 4.476444721221924,
|
| 1769 |
+
"learning_rate": 2.0547368421052633e-06,
|
| 1770 |
+
"loss": 0.0746,
|
| 1771 |
+
"step": 6100
|
| 1772 |
+
},
|
| 1773 |
+
{
|
| 1774 |
+
"epoch": 10.859929078014185,
|
| 1775 |
+
"grad_norm": 4.727671146392822,
|
| 1776 |
+
"learning_rate": 2.0415789473684213e-06,
|
| 1777 |
+
"loss": 0.073,
|
| 1778 |
+
"step": 6125
|
| 1779 |
+
},
|
| 1780 |
+
{
|
| 1781 |
+
"epoch": 10.904255319148936,
|
| 1782 |
+
"grad_norm": 4.828220367431641,
|
| 1783 |
+
"learning_rate": 2.028421052631579e-06,
|
| 1784 |
+
"loss": 0.0698,
|
| 1785 |
+
"step": 6150
|
| 1786 |
+
},
|
| 1787 |
+
{
|
| 1788 |
+
"epoch": 10.948581560283689,
|
| 1789 |
+
"grad_norm": 5.050329685211182,
|
| 1790 |
+
"learning_rate": 2.015263157894737e-06,
|
| 1791 |
+
"loss": 0.0705,
|
| 1792 |
+
"step": 6175
|
| 1793 |
+
},
|
| 1794 |
+
{
|
| 1795 |
+
"epoch": 10.99290780141844,
|
| 1796 |
+
"grad_norm": 4.282689094543457,
|
| 1797 |
+
"learning_rate": 2.002105263157895e-06,
|
| 1798 |
+
"loss": 0.0718,
|
| 1799 |
+
"step": 6200
|
| 1800 |
+
},
|
| 1801 |
+
{
|
| 1802 |
+
"epoch": 11.037234042553191,
|
| 1803 |
+
"grad_norm": 4.424275875091553,
|
| 1804 |
+
"learning_rate": 1.988947368421053e-06,
|
| 1805 |
+
"loss": 0.0636,
|
| 1806 |
+
"step": 6225
|
| 1807 |
+
},
|
| 1808 |
+
{
|
| 1809 |
+
"epoch": 11.081560283687944,
|
| 1810 |
+
"grad_norm": 3.237255573272705,
|
| 1811 |
+
"learning_rate": 1.975789473684211e-06,
|
| 1812 |
+
"loss": 0.0544,
|
| 1813 |
+
"step": 6250
|
| 1814 |
+
},
|
| 1815 |
+
{
|
| 1816 |
+
"epoch": 11.125886524822695,
|
| 1817 |
+
"grad_norm": 3.363708972930908,
|
| 1818 |
+
"learning_rate": 1.9626315789473683e-06,
|
| 1819 |
+
"loss": 0.0582,
|
| 1820 |
+
"step": 6275
|
| 1821 |
+
},
|
| 1822 |
+
{
|
| 1823 |
+
"epoch": 11.170212765957446,
|
| 1824 |
+
"grad_norm": 4.743597030639648,
|
| 1825 |
+
"learning_rate": 1.949473684210526e-06,
|
| 1826 |
+
"loss": 0.0567,
|
| 1827 |
+
"step": 6300
|
| 1828 |
+
},
|
| 1829 |
+
{
|
| 1830 |
+
"epoch": 11.214539007092199,
|
| 1831 |
+
"grad_norm": 4.212203502655029,
|
| 1832 |
+
"learning_rate": 1.9363157894736845e-06,
|
| 1833 |
+
"loss": 0.0643,
|
| 1834 |
+
"step": 6325
|
| 1835 |
+
},
|
| 1836 |
+
{
|
| 1837 |
+
"epoch": 11.25886524822695,
|
| 1838 |
+
"grad_norm": 3.580488443374634,
|
| 1839 |
+
"learning_rate": 1.9231578947368424e-06,
|
| 1840 |
+
"loss": 0.057,
|
| 1841 |
+
"step": 6350
|
| 1842 |
+
},
|
| 1843 |
+
{
|
| 1844 |
+
"epoch": 11.303191489361701,
|
| 1845 |
+
"grad_norm": 3.409921646118164,
|
| 1846 |
+
"learning_rate": 1.9100000000000003e-06,
|
| 1847 |
+
"loss": 0.0567,
|
| 1848 |
+
"step": 6375
|
| 1849 |
+
},
|
| 1850 |
+
{
|
| 1851 |
+
"epoch": 11.347517730496454,
|
| 1852 |
+
"grad_norm": 3.3070523738861084,
|
| 1853 |
+
"learning_rate": 1.896842105263158e-06,
|
| 1854 |
+
"loss": 0.0562,
|
| 1855 |
+
"step": 6400
|
| 1856 |
+
},
|
| 1857 |
+
{
|
| 1858 |
+
"epoch": 11.391843971631205,
|
| 1859 |
+
"grad_norm": 4.054013252258301,
|
| 1860 |
+
"learning_rate": 1.883684210526316e-06,
|
| 1861 |
+
"loss": 0.0645,
|
| 1862 |
+
"step": 6425
|
| 1863 |
+
},
|
| 1864 |
+
{
|
| 1865 |
+
"epoch": 11.436170212765958,
|
| 1866 |
+
"grad_norm": 3.5053253173828125,
|
| 1867 |
+
"learning_rate": 1.8705263157894737e-06,
|
| 1868 |
+
"loss": 0.0632,
|
| 1869 |
+
"step": 6450
|
| 1870 |
+
},
|
| 1871 |
+
{
|
| 1872 |
+
"epoch": 11.48049645390071,
|
| 1873 |
+
"grad_norm": 3.654541492462158,
|
| 1874 |
+
"learning_rate": 1.8573684210526316e-06,
|
| 1875 |
+
"loss": 0.0586,
|
| 1876 |
+
"step": 6475
|
| 1877 |
+
},
|
| 1878 |
+
{
|
| 1879 |
+
"epoch": 11.52482269503546,
|
| 1880 |
+
"grad_norm": 4.121072769165039,
|
| 1881 |
+
"learning_rate": 1.8442105263157897e-06,
|
| 1882 |
+
"loss": 0.061,
|
| 1883 |
+
"step": 6500
|
| 1884 |
+
},
|
| 1885 |
+
{
|
| 1886 |
+
"epoch": 11.569148936170214,
|
| 1887 |
+
"grad_norm": 4.159468173980713,
|
| 1888 |
+
"learning_rate": 1.8310526315789476e-06,
|
| 1889 |
+
"loss": 0.0611,
|
| 1890 |
+
"step": 6525
|
| 1891 |
+
},
|
| 1892 |
+
{
|
| 1893 |
+
"epoch": 11.613475177304965,
|
| 1894 |
+
"grad_norm": 4.2946672439575195,
|
| 1895 |
+
"learning_rate": 1.8178947368421055e-06,
|
| 1896 |
+
"loss": 0.0609,
|
| 1897 |
+
"step": 6550
|
| 1898 |
+
},
|
| 1899 |
+
{
|
| 1900 |
+
"epoch": 11.657801418439716,
|
| 1901 |
+
"grad_norm": 3.857961416244507,
|
| 1902 |
+
"learning_rate": 1.8047368421052632e-06,
|
| 1903 |
+
"loss": 0.0613,
|
| 1904 |
+
"step": 6575
|
| 1905 |
+
},
|
| 1906 |
+
{
|
| 1907 |
+
"epoch": 11.702127659574469,
|
| 1908 |
+
"grad_norm": 3.1902592182159424,
|
| 1909 |
+
"learning_rate": 1.7915789473684211e-06,
|
| 1910 |
+
"loss": 0.0607,
|
| 1911 |
+
"step": 6600
|
| 1912 |
+
},
|
| 1913 |
+
{
|
| 1914 |
+
"epoch": 11.74645390070922,
|
| 1915 |
+
"grad_norm": 5.016479015350342,
|
| 1916 |
+
"learning_rate": 1.778421052631579e-06,
|
| 1917 |
+
"loss": 0.0586,
|
| 1918 |
+
"step": 6625
|
| 1919 |
+
},
|
| 1920 |
+
{
|
| 1921 |
+
"epoch": 11.790780141843971,
|
| 1922 |
+
"grad_norm": 4.0299601554870605,
|
| 1923 |
+
"learning_rate": 1.7652631578947371e-06,
|
| 1924 |
+
"loss": 0.059,
|
| 1925 |
+
"step": 6650
|
| 1926 |
+
},
|
| 1927 |
+
{
|
| 1928 |
+
"epoch": 11.835106382978724,
|
| 1929 |
+
"grad_norm": 4.405561447143555,
|
| 1930 |
+
"learning_rate": 1.752105263157895e-06,
|
| 1931 |
+
"loss": 0.0623,
|
| 1932 |
+
"step": 6675
|
| 1933 |
+
},
|
| 1934 |
+
{
|
| 1935 |
+
"epoch": 11.879432624113475,
|
| 1936 |
+
"grad_norm": 3.684788465499878,
|
| 1937 |
+
"learning_rate": 1.7389473684210527e-06,
|
| 1938 |
+
"loss": 0.0614,
|
| 1939 |
+
"step": 6700
|
| 1940 |
+
},
|
| 1941 |
+
{
|
| 1942 |
+
"epoch": 11.923758865248226,
|
| 1943 |
+
"grad_norm": 4.335251331329346,
|
| 1944 |
+
"learning_rate": 1.7257894736842107e-06,
|
| 1945 |
+
"loss": 0.0581,
|
| 1946 |
+
"step": 6725
|
| 1947 |
+
},
|
| 1948 |
+
{
|
| 1949 |
+
"epoch": 11.96808510638298,
|
| 1950 |
+
"grad_norm": 4.67876672744751,
|
| 1951 |
+
"learning_rate": 1.7126315789473686e-06,
|
| 1952 |
+
"loss": 0.0569,
|
| 1953 |
+
"step": 6750
|
| 1954 |
+
},
|
| 1955 |
+
{
|
| 1956 |
+
"epoch": 12.01241134751773,
|
| 1957 |
+
"grad_norm": 5.14631462097168,
|
| 1958 |
+
"learning_rate": 1.6994736842105265e-06,
|
| 1959 |
+
"loss": 0.0548,
|
| 1960 |
+
"step": 6775
|
| 1961 |
+
},
|
| 1962 |
+
{
|
| 1963 |
+
"epoch": 12.056737588652481,
|
| 1964 |
+
"grad_norm": 3.651719331741333,
|
| 1965 |
+
"learning_rate": 1.6863157894736842e-06,
|
| 1966 |
+
"loss": 0.0514,
|
| 1967 |
+
"step": 6800
|
| 1968 |
+
},
|
| 1969 |
+
{
|
| 1970 |
+
"epoch": 12.101063829787234,
|
| 1971 |
+
"grad_norm": 3.455418109893799,
|
| 1972 |
+
"learning_rate": 1.6731578947368423e-06,
|
| 1973 |
+
"loss": 0.0534,
|
| 1974 |
+
"step": 6825
|
| 1975 |
+
},
|
| 1976 |
+
{
|
| 1977 |
+
"epoch": 12.145390070921986,
|
| 1978 |
+
"grad_norm": 3.9486734867095947,
|
| 1979 |
+
"learning_rate": 1.6600000000000002e-06,
|
| 1980 |
+
"loss": 0.0495,
|
| 1981 |
+
"step": 6850
|
| 1982 |
+
},
|
| 1983 |
+
{
|
| 1984 |
+
"epoch": 12.189716312056738,
|
| 1985 |
+
"grad_norm": 2.9897313117980957,
|
| 1986 |
+
"learning_rate": 1.6468421052631581e-06,
|
| 1987 |
+
"loss": 0.051,
|
| 1988 |
+
"step": 6875
|
| 1989 |
+
},
|
| 1990 |
+
{
|
| 1991 |
+
"epoch": 12.23404255319149,
|
| 1992 |
+
"grad_norm": 4.208747863769531,
|
| 1993 |
+
"learning_rate": 1.633684210526316e-06,
|
| 1994 |
+
"loss": 0.0486,
|
| 1995 |
+
"step": 6900
|
| 1996 |
+
},
|
| 1997 |
+
{
|
| 1998 |
+
"epoch": 12.27836879432624,
|
| 1999 |
+
"grad_norm": 3.3527841567993164,
|
| 2000 |
+
"learning_rate": 1.6205263157894737e-06,
|
| 2001 |
+
"loss": 0.0524,
|
| 2002 |
+
"step": 6925
|
| 2003 |
+
},
|
| 2004 |
+
{
|
| 2005 |
+
"epoch": 12.322695035460994,
|
| 2006 |
+
"grad_norm": 3.6749916076660156,
|
| 2007 |
+
"learning_rate": 1.6073684210526316e-06,
|
| 2008 |
+
"loss": 0.0577,
|
| 2009 |
+
"step": 6950
|
| 2010 |
+
},
|
| 2011 |
+
{
|
| 2012 |
+
"epoch": 12.367021276595745,
|
| 2013 |
+
"grad_norm": 3.266439437866211,
|
| 2014 |
+
"learning_rate": 1.5942105263157895e-06,
|
| 2015 |
+
"loss": 0.0491,
|
| 2016 |
+
"step": 6975
|
| 2017 |
+
},
|
| 2018 |
+
{
|
| 2019 |
+
"epoch": 12.411347517730496,
|
| 2020 |
+
"grad_norm": 3.9574387073516846,
|
| 2021 |
+
"learning_rate": 1.5810526315789477e-06,
|
| 2022 |
+
"loss": 0.0537,
|
| 2023 |
+
"step": 7000
|
| 2024 |
+
},
|
| 2025 |
+
{
|
| 2026 |
+
"epoch": 12.411347517730496,
|
| 2027 |
+
"eval_loss": 0.5095303654670715,
|
| 2028 |
+
"eval_runtime": 586.2939,
|
| 2029 |
+
"eval_samples_per_second": 6.645,
|
| 2030 |
+
"eval_steps_per_second": 0.831,
|
| 2031 |
+
"eval_wer": 41.29064425371274,
|
| 2032 |
+
"step": 7000
|
| 2033 |
+
},
|
| 2034 |
+
{
|
| 2035 |
+
"epoch": 12.455673758865249,
|
| 2036 |
+
"grad_norm": 4.798894882202148,
|
| 2037 |
+
"learning_rate": 1.5678947368421056e-06,
|
| 2038 |
+
"loss": 0.0533,
|
| 2039 |
+
"step": 7025
|
| 2040 |
+
},
|
| 2041 |
+
{
|
| 2042 |
+
"epoch": 12.5,
|
| 2043 |
+
"grad_norm": 3.0210700035095215,
|
| 2044 |
+
"learning_rate": 1.5547368421052633e-06,
|
| 2045 |
+
"loss": 0.0483,
|
| 2046 |
+
"step": 7050
|
| 2047 |
+
},
|
| 2048 |
+
{
|
| 2049 |
+
"epoch": 12.544326241134751,
|
| 2050 |
+
"grad_norm": 4.632834434509277,
|
| 2051 |
+
"learning_rate": 1.5415789473684212e-06,
|
| 2052 |
+
"loss": 0.0509,
|
| 2053 |
+
"step": 7075
|
| 2054 |
+
},
|
| 2055 |
+
{
|
| 2056 |
+
"epoch": 12.588652482269504,
|
| 2057 |
+
"grad_norm": 4.397753715515137,
|
| 2058 |
+
"learning_rate": 1.528421052631579e-06,
|
| 2059 |
+
"loss": 0.0515,
|
| 2060 |
+
"step": 7100
|
| 2061 |
+
},
|
| 2062 |
+
{
|
| 2063 |
+
"epoch": 12.632978723404255,
|
| 2064 |
+
"grad_norm": 2.9680283069610596,
|
| 2065 |
+
"learning_rate": 1.5152631578947368e-06,
|
| 2066 |
+
"loss": 0.0475,
|
| 2067 |
+
"step": 7125
|
| 2068 |
+
},
|
| 2069 |
+
{
|
| 2070 |
+
"epoch": 12.677304964539006,
|
| 2071 |
+
"grad_norm": 3.9441206455230713,
|
| 2072 |
+
"learning_rate": 1.5021052631578947e-06,
|
| 2073 |
+
"loss": 0.055,
|
| 2074 |
+
"step": 7150
|
| 2075 |
+
},
|
| 2076 |
+
{
|
| 2077 |
+
"epoch": 12.72163120567376,
|
| 2078 |
+
"grad_norm": 3.183037519454956,
|
| 2079 |
+
"learning_rate": 1.4889473684210528e-06,
|
| 2080 |
+
"loss": 0.0522,
|
| 2081 |
+
"step": 7175
|
| 2082 |
+
},
|
| 2083 |
+
{
|
| 2084 |
+
"epoch": 12.76595744680851,
|
| 2085 |
+
"grad_norm": 3.4659500122070312,
|
| 2086 |
+
"learning_rate": 1.4757894736842107e-06,
|
| 2087 |
+
"loss": 0.0477,
|
| 2088 |
+
"step": 7200
|
| 2089 |
+
},
|
| 2090 |
+
{
|
| 2091 |
+
"epoch": 12.810283687943262,
|
| 2092 |
+
"grad_norm": 3.1689703464508057,
|
| 2093 |
+
"learning_rate": 1.4626315789473686e-06,
|
| 2094 |
+
"loss": 0.0512,
|
| 2095 |
+
"step": 7225
|
| 2096 |
+
},
|
| 2097 |
+
{
|
| 2098 |
+
"epoch": 12.854609929078014,
|
| 2099 |
+
"grad_norm": 5.009653568267822,
|
| 2100 |
+
"learning_rate": 1.4494736842105263e-06,
|
| 2101 |
+
"loss": 0.0487,
|
| 2102 |
+
"step": 7250
|
| 2103 |
+
},
|
| 2104 |
+
{
|
| 2105 |
+
"epoch": 12.898936170212766,
|
| 2106 |
+
"grad_norm": 3.3407084941864014,
|
| 2107 |
+
"learning_rate": 1.4363157894736842e-06,
|
| 2108 |
+
"loss": 0.0489,
|
| 2109 |
+
"step": 7275
|
| 2110 |
+
},
|
| 2111 |
+
{
|
| 2112 |
+
"epoch": 12.943262411347519,
|
| 2113 |
+
"grad_norm": 4.140749454498291,
|
| 2114 |
+
"learning_rate": 1.4231578947368421e-06,
|
| 2115 |
+
"loss": 0.047,
|
| 2116 |
+
"step": 7300
|
| 2117 |
+
},
|
| 2118 |
+
{
|
| 2119 |
+
"epoch": 12.98758865248227,
|
| 2120 |
+
"grad_norm": 3.0766468048095703,
|
| 2121 |
+
"learning_rate": 1.41e-06,
|
| 2122 |
+
"loss": 0.052,
|
| 2123 |
+
"step": 7325
|
| 2124 |
+
},
|
| 2125 |
+
{
|
| 2126 |
+
"epoch": 13.03191489361702,
|
| 2127 |
+
"grad_norm": 3.058790683746338,
|
| 2128 |
+
"learning_rate": 1.3968421052631582e-06,
|
| 2129 |
+
"loss": 0.0445,
|
| 2130 |
+
"step": 7350
|
| 2131 |
+
},
|
| 2132 |
+
{
|
| 2133 |
+
"epoch": 13.076241134751774,
|
| 2134 |
+
"grad_norm": 2.4315567016601562,
|
| 2135 |
+
"learning_rate": 1.3836842105263159e-06,
|
| 2136 |
+
"loss": 0.0447,
|
| 2137 |
+
"step": 7375
|
| 2138 |
+
},
|
| 2139 |
+
{
|
| 2140 |
+
"epoch": 13.120567375886525,
|
| 2141 |
+
"grad_norm": 2.692753314971924,
|
| 2142 |
+
"learning_rate": 1.3705263157894738e-06,
|
| 2143 |
+
"loss": 0.0398,
|
| 2144 |
+
"step": 7400
|
| 2145 |
+
},
|
| 2146 |
+
{
|
| 2147 |
+
"epoch": 13.164893617021276,
|
| 2148 |
+
"grad_norm": 3.2242069244384766,
|
| 2149 |
+
"learning_rate": 1.3573684210526317e-06,
|
| 2150 |
+
"loss": 0.0425,
|
| 2151 |
+
"step": 7425
|
| 2152 |
+
},
|
| 2153 |
+
{
|
| 2154 |
+
"epoch": 13.209219858156029,
|
| 2155 |
+
"grad_norm": 3.640981674194336,
|
| 2156 |
+
"learning_rate": 1.3442105263157896e-06,
|
| 2157 |
+
"loss": 0.043,
|
| 2158 |
+
"step": 7450
|
| 2159 |
+
},
|
| 2160 |
+
{
|
| 2161 |
+
"epoch": 13.25354609929078,
|
| 2162 |
+
"grad_norm": 2.966660261154175,
|
| 2163 |
+
"learning_rate": 1.3310526315789473e-06,
|
| 2164 |
+
"loss": 0.0422,
|
| 2165 |
+
"step": 7475
|
| 2166 |
+
},
|
| 2167 |
+
{
|
| 2168 |
+
"epoch": 13.297872340425531,
|
| 2169 |
+
"grad_norm": 2.7896602153778076,
|
| 2170 |
+
"learning_rate": 1.3178947368421054e-06,
|
| 2171 |
+
"loss": 0.044,
|
| 2172 |
+
"step": 7500
|
| 2173 |
+
},
|
| 2174 |
+
{
|
| 2175 |
+
"epoch": 13.342198581560284,
|
| 2176 |
+
"grad_norm": 3.0664894580841064,
|
| 2177 |
+
"learning_rate": 1.3047368421052633e-06,
|
| 2178 |
+
"loss": 0.0432,
|
| 2179 |
+
"step": 7525
|
| 2180 |
+
},
|
| 2181 |
+
{
|
| 2182 |
+
"epoch": 13.386524822695035,
|
| 2183 |
+
"grad_norm": 2.7736198902130127,
|
| 2184 |
+
"learning_rate": 1.2915789473684212e-06,
|
| 2185 |
+
"loss": 0.0416,
|
| 2186 |
+
"step": 7550
|
| 2187 |
+
},
|
| 2188 |
+
{
|
| 2189 |
+
"epoch": 13.430851063829786,
|
| 2190 |
+
"grad_norm": 3.5016989707946777,
|
| 2191 |
+
"learning_rate": 1.2784210526315791e-06,
|
| 2192 |
+
"loss": 0.0481,
|
| 2193 |
+
"step": 7575
|
| 2194 |
+
},
|
| 2195 |
+
{
|
| 2196 |
+
"epoch": 13.47517730496454,
|
| 2197 |
+
"grad_norm": 3.0631349086761475,
|
| 2198 |
+
"learning_rate": 1.2652631578947368e-06,
|
| 2199 |
+
"loss": 0.0441,
|
| 2200 |
+
"step": 7600
|
| 2201 |
+
},
|
| 2202 |
+
{
|
| 2203 |
+
"epoch": 13.51950354609929,
|
| 2204 |
+
"grad_norm": 3.7912166118621826,
|
| 2205 |
+
"learning_rate": 1.2521052631578948e-06,
|
| 2206 |
+
"loss": 0.047,
|
| 2207 |
+
"step": 7625
|
| 2208 |
+
},
|
| 2209 |
+
{
|
| 2210 |
+
"epoch": 13.563829787234042,
|
| 2211 |
+
"grad_norm": 3.7112090587615967,
|
| 2212 |
+
"learning_rate": 1.2389473684210527e-06,
|
| 2213 |
+
"loss": 0.0442,
|
| 2214 |
+
"step": 7650
|
| 2215 |
+
},
|
| 2216 |
+
{
|
| 2217 |
+
"epoch": 13.608156028368795,
|
| 2218 |
+
"grad_norm": 4.523186206817627,
|
| 2219 |
+
"learning_rate": 1.2257894736842106e-06,
|
| 2220 |
+
"loss": 0.0458,
|
| 2221 |
+
"step": 7675
|
| 2222 |
+
},
|
| 2223 |
+
{
|
| 2224 |
+
"epoch": 13.652482269503546,
|
| 2225 |
+
"grad_norm": 3.0612213611602783,
|
| 2226 |
+
"learning_rate": 1.2126315789473685e-06,
|
| 2227 |
+
"loss": 0.0461,
|
| 2228 |
+
"step": 7700
|
| 2229 |
+
},
|
| 2230 |
+
{
|
| 2231 |
+
"epoch": 13.696808510638299,
|
| 2232 |
+
"grad_norm": 2.902688980102539,
|
| 2233 |
+
"learning_rate": 1.1994736842105264e-06,
|
| 2234 |
+
"loss": 0.0446,
|
| 2235 |
+
"step": 7725
|
| 2236 |
+
},
|
| 2237 |
+
{
|
| 2238 |
+
"epoch": 13.74113475177305,
|
| 2239 |
+
"grad_norm": 2.876624822616577,
|
| 2240 |
+
"learning_rate": 1.1863157894736843e-06,
|
| 2241 |
+
"loss": 0.0433,
|
| 2242 |
+
"step": 7750
|
| 2243 |
+
},
|
| 2244 |
+
{
|
| 2245 |
+
"epoch": 13.785460992907801,
|
| 2246 |
+
"grad_norm": 3.696685314178467,
|
| 2247 |
+
"learning_rate": 1.1731578947368422e-06,
|
| 2248 |
+
"loss": 0.0464,
|
| 2249 |
+
"step": 7775
|
| 2250 |
+
},
|
| 2251 |
+
{
|
| 2252 |
+
"epoch": 13.829787234042554,
|
| 2253 |
+
"grad_norm": 3.2120165824890137,
|
| 2254 |
+
"learning_rate": 1.1600000000000001e-06,
|
| 2255 |
+
"loss": 0.0444,
|
| 2256 |
+
"step": 7800
|
| 2257 |
+
},
|
| 2258 |
+
{
|
| 2259 |
+
"epoch": 13.874113475177305,
|
| 2260 |
+
"grad_norm": 3.735292434692383,
|
| 2261 |
+
"learning_rate": 1.146842105263158e-06,
|
| 2262 |
+
"loss": 0.0417,
|
| 2263 |
+
"step": 7825
|
| 2264 |
+
},
|
| 2265 |
+
{
|
| 2266 |
+
"epoch": 13.918439716312056,
|
| 2267 |
+
"grad_norm": 3.8104641437530518,
|
| 2268 |
+
"learning_rate": 1.133684210526316e-06,
|
| 2269 |
+
"loss": 0.045,
|
| 2270 |
+
"step": 7850
|
| 2271 |
+
},
|
| 2272 |
+
{
|
| 2273 |
+
"epoch": 13.962765957446809,
|
| 2274 |
+
"grad_norm": 3.321183919906616,
|
| 2275 |
+
"learning_rate": 1.1205263157894736e-06,
|
| 2276 |
+
"loss": 0.046,
|
| 2277 |
+
"step": 7875
|
| 2278 |
+
},
|
| 2279 |
+
{
|
| 2280 |
+
"epoch": 14.00709219858156,
|
| 2281 |
+
"grad_norm": 3.0110223293304443,
|
| 2282 |
+
"learning_rate": 1.1073684210526318e-06,
|
| 2283 |
+
"loss": 0.0452,
|
| 2284 |
+
"step": 7900
|
| 2285 |
+
},
|
| 2286 |
+
{
|
| 2287 |
+
"epoch": 14.051418439716311,
|
| 2288 |
+
"grad_norm": 2.797724485397339,
|
| 2289 |
+
"learning_rate": 1.0942105263157895e-06,
|
| 2290 |
+
"loss": 0.0384,
|
| 2291 |
+
"step": 7925
|
| 2292 |
+
},
|
| 2293 |
+
{
|
| 2294 |
+
"epoch": 14.095744680851064,
|
| 2295 |
+
"grad_norm": 2.8559882640838623,
|
| 2296 |
+
"learning_rate": 1.0810526315789474e-06,
|
| 2297 |
+
"loss": 0.039,
|
| 2298 |
+
"step": 7950
|
| 2299 |
+
},
|
| 2300 |
+
{
|
| 2301 |
+
"epoch": 14.140070921985815,
|
| 2302 |
+
"grad_norm": 3.0210611820220947,
|
| 2303 |
+
"learning_rate": 1.0678947368421055e-06,
|
| 2304 |
+
"loss": 0.0359,
|
| 2305 |
+
"step": 7975
|
| 2306 |
+
},
|
| 2307 |
+
{
|
| 2308 |
+
"epoch": 14.184397163120567,
|
| 2309 |
+
"grad_norm": 3.4683313369750977,
|
| 2310 |
+
"learning_rate": 1.0547368421052632e-06,
|
| 2311 |
+
"loss": 0.0393,
|
| 2312 |
+
"step": 8000
|
| 2313 |
+
},
|
| 2314 |
+
{
|
| 2315 |
+
"epoch": 14.184397163120567,
|
| 2316 |
+
"eval_loss": 0.5176098942756653,
|
| 2317 |
+
"eval_runtime": 588.4879,
|
| 2318 |
+
"eval_samples_per_second": 6.62,
|
| 2319 |
+
"eval_steps_per_second": 0.828,
|
| 2320 |
+
"eval_wer": 41.376117383097686,
|
| 2321 |
+
"step": 8000
|
| 2322 |
+
},
|
| 2323 |
+
{
|
| 2324 |
+
"epoch": 14.22872340425532,
|
| 2325 |
+
"grad_norm": 3.2651422023773193,
|
| 2326 |
+
"learning_rate": 1.041578947368421e-06,
|
| 2327 |
+
"loss": 0.039,
|
| 2328 |
+
"step": 8025
|
| 2329 |
+
},
|
| 2330 |
+
{
|
| 2331 |
+
"epoch": 14.27304964539007,
|
| 2332 |
+
"grad_norm": 3.2940969467163086,
|
| 2333 |
+
"learning_rate": 1.028421052631579e-06,
|
| 2334 |
+
"loss": 0.0393,
|
| 2335 |
+
"step": 8050
|
| 2336 |
+
},
|
| 2337 |
+
{
|
| 2338 |
+
"epoch": 14.317375886524824,
|
| 2339 |
+
"grad_norm": 3.090914011001587,
|
| 2340 |
+
"learning_rate": 1.015263157894737e-06,
|
| 2341 |
+
"loss": 0.039,
|
| 2342 |
+
"step": 8075
|
| 2343 |
+
},
|
| 2344 |
+
{
|
| 2345 |
+
"epoch": 14.361702127659575,
|
| 2346 |
+
"grad_norm": 3.464435577392578,
|
| 2347 |
+
"learning_rate": 1.0021052631578948e-06,
|
| 2348 |
+
"loss": 0.0395,
|
| 2349 |
+
"step": 8100
|
| 2350 |
+
},
|
| 2351 |
+
{
|
| 2352 |
+
"epoch": 14.406028368794326,
|
| 2353 |
+
"grad_norm": 2.341763734817505,
|
| 2354 |
+
"learning_rate": 9.889473684210527e-07,
|
| 2355 |
+
"loss": 0.0392,
|
| 2356 |
+
"step": 8125
|
| 2357 |
+
},
|
| 2358 |
+
{
|
| 2359 |
+
"epoch": 14.450354609929079,
|
| 2360 |
+
"grad_norm": 2.7853071689605713,
|
| 2361 |
+
"learning_rate": 9.757894736842106e-07,
|
| 2362 |
+
"loss": 0.0399,
|
| 2363 |
+
"step": 8150
|
| 2364 |
+
},
|
| 2365 |
+
{
|
| 2366 |
+
"epoch": 14.49468085106383,
|
| 2367 |
+
"grad_norm": 3.5469071865081787,
|
| 2368 |
+
"learning_rate": 9.626315789473685e-07,
|
| 2369 |
+
"loss": 0.0404,
|
| 2370 |
+
"step": 8175
|
| 2371 |
+
},
|
| 2372 |
+
{
|
| 2373 |
+
"epoch": 14.539007092198581,
|
| 2374 |
+
"grad_norm": 3.5632236003875732,
|
| 2375 |
+
"learning_rate": 9.494736842105263e-07,
|
| 2376 |
+
"loss": 0.0409,
|
| 2377 |
+
"step": 8200
|
| 2378 |
+
},
|
| 2379 |
+
{
|
| 2380 |
+
"epoch": 14.583333333333334,
|
| 2381 |
+
"grad_norm": 2.9529061317443848,
|
| 2382 |
+
"learning_rate": 9.363157894736844e-07,
|
| 2383 |
+
"loss": 0.0411,
|
| 2384 |
+
"step": 8225
|
| 2385 |
+
},
|
| 2386 |
+
{
|
| 2387 |
+
"epoch": 14.627659574468085,
|
| 2388 |
+
"grad_norm": 2.856344223022461,
|
| 2389 |
+
"learning_rate": 9.231578947368422e-07,
|
| 2390 |
+
"loss": 0.0384,
|
| 2391 |
+
"step": 8250
|
| 2392 |
+
},
|
| 2393 |
+
{
|
| 2394 |
+
"epoch": 14.671985815602836,
|
| 2395 |
+
"grad_norm": 3.559720039367676,
|
| 2396 |
+
"learning_rate": 9.100000000000001e-07,
|
| 2397 |
+
"loss": 0.039,
|
| 2398 |
+
"step": 8275
|
| 2399 |
+
},
|
| 2400 |
+
{
|
| 2401 |
+
"epoch": 14.71631205673759,
|
| 2402 |
+
"grad_norm": 3.8412675857543945,
|
| 2403 |
+
"learning_rate": 8.968421052631579e-07,
|
| 2404 |
+
"loss": 0.0361,
|
| 2405 |
+
"step": 8300
|
| 2406 |
+
},
|
| 2407 |
+
{
|
| 2408 |
+
"epoch": 14.76063829787234,
|
| 2409 |
+
"grad_norm": 3.8791191577911377,
|
| 2410 |
+
"learning_rate": 8.836842105263159e-07,
|
| 2411 |
+
"loss": 0.0357,
|
| 2412 |
+
"step": 8325
|
| 2413 |
+
},
|
| 2414 |
+
{
|
| 2415 |
+
"epoch": 14.804964539007091,
|
| 2416 |
+
"grad_norm": 4.187379837036133,
|
| 2417 |
+
"learning_rate": 8.705263157894737e-07,
|
| 2418 |
+
"loss": 0.0428,
|
| 2419 |
+
"step": 8350
|
| 2420 |
+
},
|
| 2421 |
+
{
|
| 2422 |
+
"epoch": 14.849290780141844,
|
| 2423 |
+
"grad_norm": 4.432793617248535,
|
| 2424 |
+
"learning_rate": 8.573684210526316e-07,
|
| 2425 |
+
"loss": 0.0375,
|
| 2426 |
+
"step": 8375
|
| 2427 |
+
},
|
| 2428 |
+
{
|
| 2429 |
+
"epoch": 14.893617021276595,
|
| 2430 |
+
"grad_norm": 3.823516368865967,
|
| 2431 |
+
"learning_rate": 8.442105263157896e-07,
|
| 2432 |
+
"loss": 0.041,
|
| 2433 |
+
"step": 8400
|
| 2434 |
+
},
|
| 2435 |
+
{
|
| 2436 |
+
"epoch": 14.937943262411348,
|
| 2437 |
+
"grad_norm": 3.2699050903320312,
|
| 2438 |
+
"learning_rate": 8.310526315789474e-07,
|
| 2439 |
+
"loss": 0.0399,
|
| 2440 |
+
"step": 8425
|
| 2441 |
+
},
|
| 2442 |
+
{
|
| 2443 |
+
"epoch": 14.9822695035461,
|
| 2444 |
+
"grad_norm": 2.571930408477783,
|
| 2445 |
+
"learning_rate": 8.178947368421053e-07,
|
| 2446 |
+
"loss": 0.0405,
|
| 2447 |
+
"step": 8450
|
| 2448 |
+
},
|
| 2449 |
+
{
|
| 2450 |
+
"epoch": 15.02659574468085,
|
| 2451 |
+
"grad_norm": 2.3387529850006104,
|
| 2452 |
+
"learning_rate": 8.047368421052632e-07,
|
| 2453 |
+
"loss": 0.0367,
|
| 2454 |
+
"step": 8475
|
| 2455 |
+
},
|
| 2456 |
+
{
|
| 2457 |
+
"epoch": 15.070921985815604,
|
| 2458 |
+
"grad_norm": 3.301847219467163,
|
| 2459 |
+
"learning_rate": 7.915789473684212e-07,
|
| 2460 |
+
"loss": 0.0347,
|
| 2461 |
+
"step": 8500
|
| 2462 |
+
},
|
| 2463 |
+
{
|
| 2464 |
+
"epoch": 15.115248226950355,
|
| 2465 |
+
"grad_norm": 3.649311065673828,
|
| 2466 |
+
"learning_rate": 7.78421052631579e-07,
|
| 2467 |
+
"loss": 0.0368,
|
| 2468 |
+
"step": 8525
|
| 2469 |
+
},
|
| 2470 |
+
{
|
| 2471 |
+
"epoch": 15.159574468085106,
|
| 2472 |
+
"grad_norm": 2.7183964252471924,
|
| 2473 |
+
"learning_rate": 7.652631578947369e-07,
|
| 2474 |
+
"loss": 0.0357,
|
| 2475 |
+
"step": 8550
|
| 2476 |
+
},
|
| 2477 |
+
{
|
| 2478 |
+
"epoch": 15.203900709219859,
|
| 2479 |
+
"grad_norm": 4.076670169830322,
|
| 2480 |
+
"learning_rate": 7.521052631578949e-07,
|
| 2481 |
+
"loss": 0.0345,
|
| 2482 |
+
"step": 8575
|
| 2483 |
+
},
|
| 2484 |
+
{
|
| 2485 |
+
"epoch": 15.24822695035461,
|
| 2486 |
+
"grad_norm": 2.5695323944091797,
|
| 2487 |
+
"learning_rate": 7.389473684210527e-07,
|
| 2488 |
+
"loss": 0.0332,
|
| 2489 |
+
"step": 8600
|
| 2490 |
+
},
|
| 2491 |
+
{
|
| 2492 |
+
"epoch": 15.292553191489361,
|
| 2493 |
+
"grad_norm": 3.0496561527252197,
|
| 2494 |
+
"learning_rate": 7.257894736842106e-07,
|
| 2495 |
+
"loss": 0.0352,
|
| 2496 |
+
"step": 8625
|
| 2497 |
+
},
|
| 2498 |
+
{
|
| 2499 |
+
"epoch": 15.336879432624114,
|
| 2500 |
+
"grad_norm": 2.6376793384552,
|
| 2501 |
+
"learning_rate": 7.126315789473685e-07,
|
| 2502 |
+
"loss": 0.0381,
|
| 2503 |
+
"step": 8650
|
| 2504 |
+
},
|
| 2505 |
+
{
|
| 2506 |
+
"epoch": 15.381205673758865,
|
| 2507 |
+
"grad_norm": 2.617739200592041,
|
| 2508 |
+
"learning_rate": 6.994736842105264e-07,
|
| 2509 |
+
"loss": 0.0382,
|
| 2510 |
+
"step": 8675
|
| 2511 |
+
},
|
| 2512 |
+
{
|
| 2513 |
+
"epoch": 15.425531914893616,
|
| 2514 |
+
"grad_norm": 1.8115471601486206,
|
| 2515 |
+
"learning_rate": 6.863157894736842e-07,
|
| 2516 |
+
"loss": 0.0358,
|
| 2517 |
+
"step": 8700
|
| 2518 |
+
},
|
| 2519 |
+
{
|
| 2520 |
+
"epoch": 15.46985815602837,
|
| 2521 |
+
"grad_norm": 2.169344902038574,
|
| 2522 |
+
"learning_rate": 6.731578947368421e-07,
|
| 2523 |
+
"loss": 0.0336,
|
| 2524 |
+
"step": 8725
|
| 2525 |
+
},
|
| 2526 |
+
{
|
| 2527 |
+
"epoch": 15.51418439716312,
|
| 2528 |
+
"grad_norm": 2.6424083709716797,
|
| 2529 |
+
"learning_rate": 6.6e-07,
|
| 2530 |
+
"loss": 0.0358,
|
| 2531 |
+
"step": 8750
|
| 2532 |
+
},
|
| 2533 |
+
{
|
| 2534 |
+
"epoch": 15.558510638297872,
|
| 2535 |
+
"grad_norm": 3.1048264503479004,
|
| 2536 |
+
"learning_rate": 6.468421052631579e-07,
|
| 2537 |
+
"loss": 0.0351,
|
| 2538 |
+
"step": 8775
|
| 2539 |
+
},
|
| 2540 |
+
{
|
| 2541 |
+
"epoch": 15.602836879432624,
|
| 2542 |
+
"grad_norm": 2.3110456466674805,
|
| 2543 |
+
"learning_rate": 6.336842105263157e-07,
|
| 2544 |
+
"loss": 0.0338,
|
| 2545 |
+
"step": 8800
|
| 2546 |
+
},
|
| 2547 |
+
{
|
| 2548 |
+
"epoch": 15.647163120567376,
|
| 2549 |
+
"grad_norm": 3.729184150695801,
|
| 2550 |
+
"learning_rate": 6.205263157894738e-07,
|
| 2551 |
+
"loss": 0.0355,
|
| 2552 |
+
"step": 8825
|
| 2553 |
+
},
|
| 2554 |
+
{
|
| 2555 |
+
"epoch": 15.691489361702128,
|
| 2556 |
+
"grad_norm": 2.4826672077178955,
|
| 2557 |
+
"learning_rate": 6.073684210526317e-07,
|
| 2558 |
+
"loss": 0.0314,
|
| 2559 |
+
"step": 8850
|
| 2560 |
+
},
|
| 2561 |
+
{
|
| 2562 |
+
"epoch": 15.73581560283688,
|
| 2563 |
+
"grad_norm": 2.23388934135437,
|
| 2564 |
+
"learning_rate": 5.942105263157895e-07,
|
| 2565 |
+
"loss": 0.0369,
|
| 2566 |
+
"step": 8875
|
| 2567 |
+
},
|
| 2568 |
+
{
|
| 2569 |
+
"epoch": 15.78014184397163,
|
| 2570 |
+
"grad_norm": 2.6081252098083496,
|
| 2571 |
+
"learning_rate": 5.810526315789474e-07,
|
| 2572 |
+
"loss": 0.0335,
|
| 2573 |
+
"step": 8900
|
| 2574 |
+
},
|
| 2575 |
+
{
|
| 2576 |
+
"epoch": 15.824468085106384,
|
| 2577 |
+
"grad_norm": 2.6870105266571045,
|
| 2578 |
+
"learning_rate": 5.678947368421053e-07,
|
| 2579 |
+
"loss": 0.0355,
|
| 2580 |
+
"step": 8925
|
| 2581 |
+
},
|
| 2582 |
+
{
|
| 2583 |
+
"epoch": 15.868794326241135,
|
| 2584 |
+
"grad_norm": 2.663280725479126,
|
| 2585 |
+
"learning_rate": 5.547368421052632e-07,
|
| 2586 |
+
"loss": 0.0386,
|
| 2587 |
+
"step": 8950
|
| 2588 |
+
},
|
| 2589 |
+
{
|
| 2590 |
+
"epoch": 15.913120567375886,
|
| 2591 |
+
"grad_norm": 2.3229103088378906,
|
| 2592 |
+
"learning_rate": 5.415789473684211e-07,
|
| 2593 |
+
"loss": 0.0343,
|
| 2594 |
+
"step": 8975
|
| 2595 |
+
},
|
| 2596 |
+
{
|
| 2597 |
+
"epoch": 15.957446808510639,
|
| 2598 |
+
"grad_norm": 2.0813558101654053,
|
| 2599 |
+
"learning_rate": 5.284210526315789e-07,
|
| 2600 |
+
"loss": 0.0354,
|
| 2601 |
+
"step": 9000
|
| 2602 |
+
},
|
| 2603 |
+
{
|
| 2604 |
+
"epoch": 15.957446808510639,
|
| 2605 |
+
"eval_loss": 0.525884747505188,
|
| 2606 |
+
"eval_runtime": 589.6497,
|
| 2607 |
+
"eval_samples_per_second": 6.607,
|
| 2608 |
+
"eval_steps_per_second": 0.826,
|
| 2609 |
+
"eval_wer": 42.95380889632822,
|
| 2610 |
+
"step": 9000
|
| 2611 |
+
},
|
| 2612 |
+
{
|
| 2613 |
+
"epoch": 16.00177304964539,
|
| 2614 |
+
"grad_norm": 2.0741195678710938,
|
| 2615 |
+
"learning_rate": 5.152631578947369e-07,
|
| 2616 |
+
"loss": 0.0357,
|
| 2617 |
+
"step": 9025
|
| 2618 |
+
},
|
| 2619 |
+
{
|
| 2620 |
+
"epoch": 16.04609929078014,
|
| 2621 |
+
"grad_norm": 2.2553582191467285,
|
| 2622 |
+
"learning_rate": 5.021052631578948e-07,
|
| 2623 |
+
"loss": 0.0348,
|
| 2624 |
+
"step": 9050
|
| 2625 |
+
},
|
| 2626 |
+
{
|
| 2627 |
+
"epoch": 16.090425531914892,
|
| 2628 |
+
"grad_norm": 2.8926563262939453,
|
| 2629 |
+
"learning_rate": 4.889473684210526e-07,
|
| 2630 |
+
"loss": 0.0329,
|
| 2631 |
+
"step": 9075
|
| 2632 |
+
},
|
| 2633 |
+
{
|
| 2634 |
+
"epoch": 16.134751773049647,
|
| 2635 |
+
"grad_norm": 2.0093374252319336,
|
| 2636 |
+
"learning_rate": 4.757894736842106e-07,
|
| 2637 |
+
"loss": 0.0316,
|
| 2638 |
+
"step": 9100
|
| 2639 |
+
},
|
| 2640 |
+
{
|
| 2641 |
+
"epoch": 16.179078014184398,
|
| 2642 |
+
"grad_norm": 2.607196569442749,
|
| 2643 |
+
"learning_rate": 4.626315789473684e-07,
|
| 2644 |
+
"loss": 0.0343,
|
| 2645 |
+
"step": 9125
|
| 2646 |
+
},
|
| 2647 |
+
{
|
| 2648 |
+
"epoch": 16.22340425531915,
|
| 2649 |
+
"grad_norm": 3.0515787601470947,
|
| 2650 |
+
"learning_rate": 4.4947368421052637e-07,
|
| 2651 |
+
"loss": 0.0337,
|
| 2652 |
+
"step": 9150
|
| 2653 |
+
},
|
| 2654 |
+
{
|
| 2655 |
+
"epoch": 16.2677304964539,
|
| 2656 |
+
"grad_norm": 2.6530652046203613,
|
| 2657 |
+
"learning_rate": 4.363157894736843e-07,
|
| 2658 |
+
"loss": 0.0328,
|
| 2659 |
+
"step": 9175
|
| 2660 |
+
},
|
| 2661 |
+
{
|
| 2662 |
+
"epoch": 16.31205673758865,
|
| 2663 |
+
"grad_norm": 2.286144733428955,
|
| 2664 |
+
"learning_rate": 4.2315789473684214e-07,
|
| 2665 |
+
"loss": 0.0314,
|
| 2666 |
+
"step": 9200
|
| 2667 |
+
},
|
| 2668 |
+
{
|
| 2669 |
+
"epoch": 16.356382978723403,
|
| 2670 |
+
"grad_norm": 1.9126091003417969,
|
| 2671 |
+
"learning_rate": 4.1000000000000004e-07,
|
| 2672 |
+
"loss": 0.0337,
|
| 2673 |
+
"step": 9225
|
| 2674 |
+
},
|
| 2675 |
+
{
|
| 2676 |
+
"epoch": 16.400709219858157,
|
| 2677 |
+
"grad_norm": 2.4638893604278564,
|
| 2678 |
+
"learning_rate": 3.968421052631579e-07,
|
| 2679 |
+
"loss": 0.0326,
|
| 2680 |
+
"step": 9250
|
| 2681 |
+
},
|
| 2682 |
+
{
|
| 2683 |
+
"epoch": 16.44503546099291,
|
| 2684 |
+
"grad_norm": 1.9740127325057983,
|
| 2685 |
+
"learning_rate": 3.836842105263158e-07,
|
| 2686 |
+
"loss": 0.0316,
|
| 2687 |
+
"step": 9275
|
| 2688 |
+
},
|
| 2689 |
+
{
|
| 2690 |
+
"epoch": 16.48936170212766,
|
| 2691 |
+
"grad_norm": 2.466771125793457,
|
| 2692 |
+
"learning_rate": 3.7052631578947377e-07,
|
| 2693 |
+
"loss": 0.0323,
|
| 2694 |
+
"step": 9300
|
| 2695 |
+
},
|
| 2696 |
+
{
|
| 2697 |
+
"epoch": 16.53368794326241,
|
| 2698 |
+
"grad_norm": 3.461355686187744,
|
| 2699 |
+
"learning_rate": 3.573684210526316e-07,
|
| 2700 |
+
"loss": 0.0329,
|
| 2701 |
+
"step": 9325
|
| 2702 |
+
},
|
| 2703 |
+
{
|
| 2704 |
+
"epoch": 16.578014184397162,
|
| 2705 |
+
"grad_norm": 4.0049662590026855,
|
| 2706 |
+
"learning_rate": 3.4421052631578954e-07,
|
| 2707 |
+
"loss": 0.0307,
|
| 2708 |
+
"step": 9350
|
| 2709 |
+
},
|
| 2710 |
+
{
|
| 2711 |
+
"epoch": 16.622340425531917,
|
| 2712 |
+
"grad_norm": 3.518848419189453,
|
| 2713 |
+
"learning_rate": 3.310526315789474e-07,
|
| 2714 |
+
"loss": 0.0303,
|
| 2715 |
+
"step": 9375
|
| 2716 |
+
},
|
| 2717 |
+
{
|
| 2718 |
+
"epoch": 16.666666666666668,
|
| 2719 |
+
"grad_norm": 2.83296799659729,
|
| 2720 |
+
"learning_rate": 3.178947368421053e-07,
|
| 2721 |
+
"loss": 0.0357,
|
| 2722 |
+
"step": 9400
|
| 2723 |
+
},
|
| 2724 |
+
{
|
| 2725 |
+
"epoch": 16.71099290780142,
|
| 2726 |
+
"grad_norm": 2.5147087574005127,
|
| 2727 |
+
"learning_rate": 3.0473684210526316e-07,
|
| 2728 |
+
"loss": 0.0328,
|
| 2729 |
+
"step": 9425
|
| 2730 |
+
},
|
| 2731 |
+
{
|
| 2732 |
+
"epoch": 16.75531914893617,
|
| 2733 |
+
"grad_norm": 3.1541314125061035,
|
| 2734 |
+
"learning_rate": 2.9157894736842107e-07,
|
| 2735 |
+
"loss": 0.0344,
|
| 2736 |
+
"step": 9450
|
| 2737 |
+
},
|
| 2738 |
+
{
|
| 2739 |
+
"epoch": 16.79964539007092,
|
| 2740 |
+
"grad_norm": 2.6284327507019043,
|
| 2741 |
+
"learning_rate": 2.78421052631579e-07,
|
| 2742 |
+
"loss": 0.0317,
|
| 2743 |
+
"step": 9475
|
| 2744 |
+
},
|
| 2745 |
+
{
|
| 2746 |
+
"epoch": 16.843971631205672,
|
| 2747 |
+
"grad_norm": 2.6868457794189453,
|
| 2748 |
+
"learning_rate": 2.6526315789473684e-07,
|
| 2749 |
+
"loss": 0.0336,
|
| 2750 |
+
"step": 9500
|
| 2751 |
+
},
|
| 2752 |
+
{
|
| 2753 |
+
"epoch": 16.888297872340427,
|
| 2754 |
+
"grad_norm": 1.9752613306045532,
|
| 2755 |
+
"learning_rate": 2.5210526315789474e-07,
|
| 2756 |
+
"loss": 0.0321,
|
| 2757 |
+
"step": 9525
|
| 2758 |
+
},
|
| 2759 |
+
{
|
| 2760 |
+
"epoch": 16.93262411347518,
|
| 2761 |
+
"grad_norm": 2.544431447982788,
|
| 2762 |
+
"learning_rate": 2.3894736842105265e-07,
|
| 2763 |
+
"loss": 0.0334,
|
| 2764 |
+
"step": 9550
|
| 2765 |
+
},
|
| 2766 |
+
{
|
| 2767 |
+
"epoch": 16.97695035460993,
|
| 2768 |
+
"grad_norm": 3.398198366165161,
|
| 2769 |
+
"learning_rate": 2.2578947368421054e-07,
|
| 2770 |
+
"loss": 0.0342,
|
| 2771 |
+
"step": 9575
|
| 2772 |
+
},
|
| 2773 |
+
{
|
| 2774 |
+
"epoch": 17.02127659574468,
|
| 2775 |
+
"grad_norm": 2.5984408855438232,
|
| 2776 |
+
"learning_rate": 2.1263157894736842e-07,
|
| 2777 |
+
"loss": 0.0341,
|
| 2778 |
+
"step": 9600
|
| 2779 |
+
},
|
| 2780 |
+
{
|
| 2781 |
+
"epoch": 17.06560283687943,
|
| 2782 |
+
"grad_norm": 2.4616684913635254,
|
| 2783 |
+
"learning_rate": 1.9947368421052633e-07,
|
| 2784 |
+
"loss": 0.0297,
|
| 2785 |
+
"step": 9625
|
| 2786 |
+
},
|
| 2787 |
+
{
|
| 2788 |
+
"epoch": 17.109929078014183,
|
| 2789 |
+
"grad_norm": 1.7951653003692627,
|
| 2790 |
+
"learning_rate": 1.8631578947368424e-07,
|
| 2791 |
+
"loss": 0.0315,
|
| 2792 |
+
"step": 9650
|
| 2793 |
+
},
|
| 2794 |
+
{
|
| 2795 |
+
"epoch": 17.154255319148938,
|
| 2796 |
+
"grad_norm": 2.757528066635132,
|
| 2797 |
+
"learning_rate": 1.7315789473684212e-07,
|
| 2798 |
+
"loss": 0.0315,
|
| 2799 |
+
"step": 9675
|
| 2800 |
+
},
|
| 2801 |
+
{
|
| 2802 |
+
"epoch": 17.19858156028369,
|
| 2803 |
+
"grad_norm": 2.636103868484497,
|
| 2804 |
+
"learning_rate": 1.6e-07,
|
| 2805 |
+
"loss": 0.034,
|
| 2806 |
+
"step": 9700
|
| 2807 |
+
},
|
| 2808 |
+
{
|
| 2809 |
+
"epoch": 17.24290780141844,
|
| 2810 |
+
"grad_norm": 2.6068308353424072,
|
| 2811 |
+
"learning_rate": 1.468421052631579e-07,
|
| 2812 |
+
"loss": 0.0318,
|
| 2813 |
+
"step": 9725
|
| 2814 |
+
},
|
| 2815 |
+
{
|
| 2816 |
+
"epoch": 17.28723404255319,
|
| 2817 |
+
"grad_norm": 2.8780813217163086,
|
| 2818 |
+
"learning_rate": 1.3368421052631582e-07,
|
| 2819 |
+
"loss": 0.0291,
|
| 2820 |
+
"step": 9750
|
| 2821 |
+
},
|
| 2822 |
+
{
|
| 2823 |
+
"epoch": 17.331560283687942,
|
| 2824 |
+
"grad_norm": 2.056938409805298,
|
| 2825 |
+
"learning_rate": 1.205263157894737e-07,
|
| 2826 |
+
"loss": 0.0312,
|
| 2827 |
+
"step": 9775
|
| 2828 |
+
},
|
| 2829 |
+
{
|
| 2830 |
+
"epoch": 17.375886524822697,
|
| 2831 |
+
"grad_norm": 1.919609785079956,
|
| 2832 |
+
"learning_rate": 1.0736842105263159e-07,
|
| 2833 |
+
"loss": 0.0301,
|
| 2834 |
+
"step": 9800
|
| 2835 |
+
},
|
| 2836 |
+
{
|
| 2837 |
+
"epoch": 17.420212765957448,
|
| 2838 |
+
"grad_norm": 2.8991692066192627,
|
| 2839 |
+
"learning_rate": 9.421052631578948e-08,
|
| 2840 |
+
"loss": 0.0325,
|
| 2841 |
+
"step": 9825
|
| 2842 |
+
},
|
| 2843 |
+
{
|
| 2844 |
+
"epoch": 17.4645390070922,
|
| 2845 |
+
"grad_norm": 3.554067611694336,
|
| 2846 |
+
"learning_rate": 8.105263157894738e-08,
|
| 2847 |
+
"loss": 0.0325,
|
| 2848 |
+
"step": 9850
|
| 2849 |
+
},
|
| 2850 |
+
{
|
| 2851 |
+
"epoch": 17.50886524822695,
|
| 2852 |
+
"grad_norm": 2.3162522315979004,
|
| 2853 |
+
"learning_rate": 6.789473684210528e-08,
|
| 2854 |
+
"loss": 0.0295,
|
| 2855 |
+
"step": 9875
|
| 2856 |
+
},
|
| 2857 |
+
{
|
| 2858 |
+
"epoch": 17.5531914893617,
|
| 2859 |
+
"grad_norm": 1.9810761213302612,
|
| 2860 |
+
"learning_rate": 5.473684210526316e-08,
|
| 2861 |
+
"loss": 0.0316,
|
| 2862 |
+
"step": 9900
|
| 2863 |
+
},
|
| 2864 |
+
{
|
| 2865 |
+
"epoch": 17.597517730496453,
|
| 2866 |
+
"grad_norm": 2.424508571624756,
|
| 2867 |
+
"learning_rate": 4.1578947368421054e-08,
|
| 2868 |
+
"loss": 0.0299,
|
| 2869 |
+
"step": 9925
|
| 2870 |
+
},
|
| 2871 |
+
{
|
| 2872 |
+
"epoch": 17.641843971631207,
|
| 2873 |
+
"grad_norm": 3.3512706756591797,
|
| 2874 |
+
"learning_rate": 2.842105263157895e-08,
|
| 2875 |
+
"loss": 0.0316,
|
| 2876 |
+
"step": 9950
|
| 2877 |
+
},
|
| 2878 |
+
{
|
| 2879 |
+
"epoch": 17.68617021276596,
|
| 2880 |
+
"grad_norm": 2.7042930126190186,
|
| 2881 |
+
"learning_rate": 1.5263157894736843e-08,
|
| 2882 |
+
"loss": 0.0308,
|
| 2883 |
+
"step": 9975
|
| 2884 |
+
},
|
| 2885 |
+
{
|
| 2886 |
+
"epoch": 17.73049645390071,
|
| 2887 |
+
"grad_norm": 2.975961685180664,
|
| 2888 |
+
"learning_rate": 2.105263157894737e-09,
|
| 2889 |
+
"loss": 0.0317,
|
| 2890 |
+
"step": 10000
|
| 2891 |
+
},
|
| 2892 |
+
{
|
| 2893 |
+
"epoch": 17.73049645390071,
|
| 2894 |
+
"eval_loss": 0.5293700695037842,
|
| 2895 |
+
"eval_runtime": 590.1218,
|
| 2896 |
+
"eval_samples_per_second": 6.602,
|
| 2897 |
+
"eval_steps_per_second": 0.825,
|
| 2898 |
+
"eval_wer": 44.56711421346914,
|
| 2899 |
+
"step": 10000
|
| 2900 |
+
},
|
| 2901 |
+
{
|
| 2902 |
+
"epoch": 17.73049645390071,
|
| 2903 |
+
"step": 10000,
|
| 2904 |
+
"total_flos": 3.93818457710592e+18,
|
| 2905 |
+
"train_loss": 0.1984061565220356,
|
| 2906 |
+
"train_runtime": 10260.2689,
|
| 2907 |
+
"train_samples_per_second": 15.594,
|
| 2908 |
+
"train_steps_per_second": 0.975
|
| 2909 |
+
},
|
| 2910 |
+
{
|
| 2911 |
+
"epoch": 17.73049645390071,
|
| 2912 |
+
"eval_loss": 0.4848020076751709,
|
| 2913 |
+
"eval_runtime": 614.3795,
|
| 2914 |
+
"eval_samples_per_second": 6.341,
|
| 2915 |
+
"eval_steps_per_second": 0.793,
|
| 2916 |
+
"eval_wer": 37.98212187043698,
|
| 2917 |
+
"step": 10000
|
| 2918 |
+
}
|
| 2919 |
+
],
|
| 2920 |
+
"logging_steps": 25,
|
| 2921 |
+
"max_steps": 10000,
|
| 2922 |
+
"num_input_tokens_seen": 0,
|
| 2923 |
+
"num_train_epochs": 18,
|
| 2924 |
+
"save_steps": 1000,
|
| 2925 |
+
"stateful_callbacks": {
|
| 2926 |
+
"TrainerControl": {
|
| 2927 |
+
"args": {
|
| 2928 |
+
"should_epoch_stop": false,
|
| 2929 |
+
"should_evaluate": false,
|
| 2930 |
+
"should_log": false,
|
| 2931 |
+
"should_save": true,
|
| 2932 |
+
"should_training_stop": true
|
| 2933 |
+
},
|
| 2934 |
+
"attributes": {}
|
| 2935 |
+
}
|
| 2936 |
+
},
|
| 2937 |
+
"total_flos": 3.93818457710592e+18,
|
| 2938 |
+
"train_batch_size": 16,
|
| 2939 |
+
"trial_name": null,
|
| 2940 |
+
"trial_params": null
|
| 2941 |
+
}
|