"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629911360.5154393/events.out.tfevents.1629911360.7e498afd5545.7645.95 +3 -0
- model-bin/finetune/base/log/1629911829.7407227/events.out.tfevents.1629911829.7e498afd5545.7645.97 +3 -0
- model-bin/finetune/base/log/1629912375.8557265/events.out.tfevents.1629912375.7e498afd5545.7645.99 +3 -0
- model-bin/finetune/base/log/1629912862.109049/events.out.tfevents.1629912863.7e498afd5545.7645.101 +3 -0
- model-bin/finetune/base/log/1629913333.6385581/events.out.tfevents.1629913333.7e498afd5545.7645.103 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629911360.7e498afd5545.7645.94 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629911829.7e498afd5545.7645.96 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629912375.7e498afd5545.7645.98 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629912862.7e498afd5545.7645.100 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629913333.7e498afd5545.7645.102 +3 -0
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165393
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fe24d408e4fb7ffb990a164bc8a9e76ff50314e3ca2a1a098b92ef85c47b7458
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/pytorch_model.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 377909911
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c770f4b6b011e3ac322f175b730c7d5366331ac8000d2df659e0dc0c0f3d358f
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:269a9836b8326e173a7c1c2d5eab85d45ca91dee77bb0cb2730b8f3c60440756
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/scaler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 559
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:db292c316a79f456016322ef81169ebe4f14d1db899cb806a80f69948b3990ec
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/scheduler.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 623
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0da95b2d7c5c29c40af958513048d2caf3165525642fab25b978bf13bf2d5eb0
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -203862,11 +203862,806 @@
|
|
| 203862 |
"eval_steps_per_second": 0.646,
|
| 203863 |
"eval_wer": 0.1882754117390989,
|
| 203864 |
"step": 82134
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 203865 |
}
|
| 203866 |
],
|
| 203867 |
-
"max_steps":
|
| 203868 |
"num_train_epochs": 5000,
|
| 203869 |
-
"total_flos": 2.
|
| 203870 |
"trial_name": null,
|
| 203871 |
"trial_params": null
|
| 203872 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 662.0,
|
| 5 |
+
"global_step": 82758,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 203862 |
"eval_steps_per_second": 0.646,
|
| 203863 |
"eval_wer": 0.1882754117390989,
|
| 203864 |
"step": 82134
|
| 203865 |
+
},
|
| 203866 |
+
{
|
| 203867 |
+
"epoch": 657.01,
|
| 203868 |
+
"learning_rate": 8.689757673667206e-06,
|
| 203869 |
+
"loss": 0.3969,
|
| 203870 |
+
"step": 82135
|
| 203871 |
+
},
|
| 203872 |
+
{
|
| 203873 |
+
"epoch": 657.05,
|
| 203874 |
+
"learning_rate": 8.689676898222941e-06,
|
| 203875 |
+
"loss": 0.3168,
|
| 203876 |
+
"step": 82140
|
| 203877 |
+
},
|
| 203878 |
+
{
|
| 203879 |
+
"epoch": 657.09,
|
| 203880 |
+
"learning_rate": 8.689596122778676e-06,
|
| 203881 |
+
"loss": 0.3386,
|
| 203882 |
+
"step": 82145
|
| 203883 |
+
},
|
| 203884 |
+
{
|
| 203885 |
+
"epoch": 657.13,
|
| 203886 |
+
"learning_rate": 8.689515347334411e-06,
|
| 203887 |
+
"loss": 0.3562,
|
| 203888 |
+
"step": 82150
|
| 203889 |
+
},
|
| 203890 |
+
{
|
| 203891 |
+
"epoch": 657.17,
|
| 203892 |
+
"learning_rate": 8.689434571890146e-06,
|
| 203893 |
+
"loss": 0.5284,
|
| 203894 |
+
"step": 82155
|
| 203895 |
+
},
|
| 203896 |
+
{
|
| 203897 |
+
"epoch": 657.21,
|
| 203898 |
+
"learning_rate": 8.689353796445881e-06,
|
| 203899 |
+
"loss": 1.1899,
|
| 203900 |
+
"step": 82160
|
| 203901 |
+
},
|
| 203902 |
+
{
|
| 203903 |
+
"epoch": 657.25,
|
| 203904 |
+
"learning_rate": 8.689273021001616e-06,
|
| 203905 |
+
"loss": 0.2894,
|
| 203906 |
+
"step": 82165
|
| 203907 |
+
},
|
| 203908 |
+
{
|
| 203909 |
+
"epoch": 657.29,
|
| 203910 |
+
"learning_rate": 8.689192245557351e-06,
|
| 203911 |
+
"loss": 0.3797,
|
| 203912 |
+
"step": 82170
|
| 203913 |
+
},
|
| 203914 |
+
{
|
| 203915 |
+
"epoch": 657.33,
|
| 203916 |
+
"learning_rate": 8.689111470113086e-06,
|
| 203917 |
+
"loss": 0.4042,
|
| 203918 |
+
"step": 82175
|
| 203919 |
+
},
|
| 203920 |
+
{
|
| 203921 |
+
"epoch": 657.37,
|
| 203922 |
+
"learning_rate": 8.689030694668821e-06,
|
| 203923 |
+
"loss": 0.5138,
|
| 203924 |
+
"step": 82180
|
| 203925 |
+
},
|
| 203926 |
+
{
|
| 203927 |
+
"epoch": 657.41,
|
| 203928 |
+
"learning_rate": 8.688949919224555e-06,
|
| 203929 |
+
"loss": 1.1549,
|
| 203930 |
+
"step": 82185
|
| 203931 |
+
},
|
| 203932 |
+
{
|
| 203933 |
+
"epoch": 657.45,
|
| 203934 |
+
"learning_rate": 8.688869143780291e-06,
|
| 203935 |
+
"loss": 0.3237,
|
| 203936 |
+
"step": 82190
|
| 203937 |
+
},
|
| 203938 |
+
{
|
| 203939 |
+
"epoch": 657.49,
|
| 203940 |
+
"learning_rate": 8.688788368336027e-06,
|
| 203941 |
+
"loss": 0.3392,
|
| 203942 |
+
"step": 82195
|
| 203943 |
+
},
|
| 203944 |
+
{
|
| 203945 |
+
"epoch": 657.53,
|
| 203946 |
+
"learning_rate": 8.688707592891761e-06,
|
| 203947 |
+
"loss": 0.3629,
|
| 203948 |
+
"step": 82200
|
| 203949 |
+
},
|
| 203950 |
+
{
|
| 203951 |
+
"epoch": 657.57,
|
| 203952 |
+
"learning_rate": 8.688626817447497e-06,
|
| 203953 |
+
"loss": 0.4889,
|
| 203954 |
+
"step": 82205
|
| 203955 |
+
},
|
| 203956 |
+
{
|
| 203957 |
+
"epoch": 657.61,
|
| 203958 |
+
"learning_rate": 8.688546042003231e-06,
|
| 203959 |
+
"loss": 1.1361,
|
| 203960 |
+
"step": 82210
|
| 203961 |
+
},
|
| 203962 |
+
{
|
| 203963 |
+
"epoch": 657.65,
|
| 203964 |
+
"learning_rate": 8.688465266558967e-06,
|
| 203965 |
+
"loss": 0.2944,
|
| 203966 |
+
"step": 82215
|
| 203967 |
+
},
|
| 203968 |
+
{
|
| 203969 |
+
"epoch": 657.69,
|
| 203970 |
+
"learning_rate": 8.688384491114701e-06,
|
| 203971 |
+
"loss": 0.3102,
|
| 203972 |
+
"step": 82220
|
| 203973 |
+
},
|
| 203974 |
+
{
|
| 203975 |
+
"epoch": 657.73,
|
| 203976 |
+
"learning_rate": 8.688303715670437e-06,
|
| 203977 |
+
"loss": 0.3847,
|
| 203978 |
+
"step": 82225
|
| 203979 |
+
},
|
| 203980 |
+
{
|
| 203981 |
+
"epoch": 657.77,
|
| 203982 |
+
"learning_rate": 8.688222940226171e-06,
|
| 203983 |
+
"loss": 0.4707,
|
| 203984 |
+
"step": 82230
|
| 203985 |
+
},
|
| 203986 |
+
{
|
| 203987 |
+
"epoch": 657.81,
|
| 203988 |
+
"learning_rate": 8.688142164781907e-06,
|
| 203989 |
+
"loss": 1.212,
|
| 203990 |
+
"step": 82235
|
| 203991 |
+
},
|
| 203992 |
+
{
|
| 203993 |
+
"epoch": 657.85,
|
| 203994 |
+
"learning_rate": 8.688061389337641e-06,
|
| 203995 |
+
"loss": 0.3253,
|
| 203996 |
+
"step": 82240
|
| 203997 |
+
},
|
| 203998 |
+
{
|
| 203999 |
+
"epoch": 657.89,
|
| 204000 |
+
"learning_rate": 8.687980613893377e-06,
|
| 204001 |
+
"loss": 0.3342,
|
| 204002 |
+
"step": 82245
|
| 204003 |
+
},
|
| 204004 |
+
{
|
| 204005 |
+
"epoch": 657.93,
|
| 204006 |
+
"learning_rate": 8.687899838449111e-06,
|
| 204007 |
+
"loss": 0.3378,
|
| 204008 |
+
"step": 82250
|
| 204009 |
+
},
|
| 204010 |
+
{
|
| 204011 |
+
"epoch": 657.97,
|
| 204012 |
+
"learning_rate": 8.687819063004847e-06,
|
| 204013 |
+
"loss": 0.5694,
|
| 204014 |
+
"step": 82255
|
| 204015 |
+
},
|
| 204016 |
+
{
|
| 204017 |
+
"epoch": 658.0,
|
| 204018 |
+
"eval_loss": 0.3955002725124359,
|
| 204019 |
+
"eval_runtime": 44.0541,
|
| 204020 |
+
"eval_samples_per_second": 19.09,
|
| 204021 |
+
"eval_steps_per_second": 0.613,
|
| 204022 |
+
"eval_wer": 0.19089769122962308,
|
| 204023 |
+
"step": 82259
|
| 204024 |
+
},
|
| 204025 |
+
{
|
| 204026 |
+
"epoch": 663.01,
|
| 204027 |
+
"learning_rate": 8.687738287560583e-06,
|
| 204028 |
+
"loss": 0.4214,
|
| 204029 |
+
"step": 82260
|
| 204030 |
+
},
|
| 204031 |
+
{
|
| 204032 |
+
"epoch": 663.05,
|
| 204033 |
+
"learning_rate": 8.687657512116317e-06,
|
| 204034 |
+
"loss": 0.3219,
|
| 204035 |
+
"step": 82265
|
| 204036 |
+
},
|
| 204037 |
+
{
|
| 204038 |
+
"epoch": 663.09,
|
| 204039 |
+
"learning_rate": 8.687576736672053e-06,
|
| 204040 |
+
"loss": 0.3214,
|
| 204041 |
+
"step": 82270
|
| 204042 |
+
},
|
| 204043 |
+
{
|
| 204044 |
+
"epoch": 663.13,
|
| 204045 |
+
"learning_rate": 8.687495961227787e-06,
|
| 204046 |
+
"loss": 0.3444,
|
| 204047 |
+
"step": 82275
|
| 204048 |
+
},
|
| 204049 |
+
{
|
| 204050 |
+
"epoch": 663.17,
|
| 204051 |
+
"learning_rate": 8.687415185783523e-06,
|
| 204052 |
+
"loss": 0.5217,
|
| 204053 |
+
"step": 82280
|
| 204054 |
+
},
|
| 204055 |
+
{
|
| 204056 |
+
"epoch": 663.21,
|
| 204057 |
+
"learning_rate": 8.687334410339257e-06,
|
| 204058 |
+
"loss": 1.0719,
|
| 204059 |
+
"step": 82285
|
| 204060 |
+
},
|
| 204061 |
+
{
|
| 204062 |
+
"epoch": 663.25,
|
| 204063 |
+
"learning_rate": 8.687253634894993e-06,
|
| 204064 |
+
"loss": 0.3446,
|
| 204065 |
+
"step": 82290
|
| 204066 |
+
},
|
| 204067 |
+
{
|
| 204068 |
+
"epoch": 663.29,
|
| 204069 |
+
"learning_rate": 8.687172859450727e-06,
|
| 204070 |
+
"loss": 0.336,
|
| 204071 |
+
"step": 82295
|
| 204072 |
+
},
|
| 204073 |
+
{
|
| 204074 |
+
"epoch": 663.33,
|
| 204075 |
+
"learning_rate": 8.687092084006463e-06,
|
| 204076 |
+
"loss": 0.3395,
|
| 204077 |
+
"step": 82300
|
| 204078 |
+
},
|
| 204079 |
+
{
|
| 204080 |
+
"epoch": 663.37,
|
| 204081 |
+
"learning_rate": 8.687011308562197e-06,
|
| 204082 |
+
"loss": 0.6165,
|
| 204083 |
+
"step": 82305
|
| 204084 |
+
},
|
| 204085 |
+
{
|
| 204086 |
+
"epoch": 663.41,
|
| 204087 |
+
"learning_rate": 8.686930533117933e-06,
|
| 204088 |
+
"loss": 1.1607,
|
| 204089 |
+
"step": 82310
|
| 204090 |
+
},
|
| 204091 |
+
{
|
| 204092 |
+
"epoch": 663.45,
|
| 204093 |
+
"learning_rate": 8.686849757673667e-06,
|
| 204094 |
+
"loss": 0.3627,
|
| 204095 |
+
"step": 82315
|
| 204096 |
+
},
|
| 204097 |
+
{
|
| 204098 |
+
"epoch": 663.49,
|
| 204099 |
+
"learning_rate": 8.686768982229403e-06,
|
| 204100 |
+
"loss": 0.3126,
|
| 204101 |
+
"step": 82320
|
| 204102 |
+
},
|
| 204103 |
+
{
|
| 204104 |
+
"epoch": 663.53,
|
| 204105 |
+
"learning_rate": 8.686688206785139e-06,
|
| 204106 |
+
"loss": 0.3251,
|
| 204107 |
+
"step": 82325
|
| 204108 |
+
},
|
| 204109 |
+
{
|
| 204110 |
+
"epoch": 663.57,
|
| 204111 |
+
"learning_rate": 8.686607431340873e-06,
|
| 204112 |
+
"loss": 0.601,
|
| 204113 |
+
"step": 82330
|
| 204114 |
+
},
|
| 204115 |
+
{
|
| 204116 |
+
"epoch": 663.61,
|
| 204117 |
+
"learning_rate": 8.686526655896609e-06,
|
| 204118 |
+
"loss": 1.1649,
|
| 204119 |
+
"step": 82335
|
| 204120 |
+
},
|
| 204121 |
+
{
|
| 204122 |
+
"epoch": 663.65,
|
| 204123 |
+
"learning_rate": 8.686445880452343e-06,
|
| 204124 |
+
"loss": 0.3395,
|
| 204125 |
+
"step": 82340
|
| 204126 |
+
},
|
| 204127 |
+
{
|
| 204128 |
+
"epoch": 663.69,
|
| 204129 |
+
"learning_rate": 8.686365105008078e-06,
|
| 204130 |
+
"loss": 0.2699,
|
| 204131 |
+
"step": 82345
|
| 204132 |
+
},
|
| 204133 |
+
{
|
| 204134 |
+
"epoch": 663.73,
|
| 204135 |
+
"learning_rate": 8.686284329563813e-06,
|
| 204136 |
+
"loss": 0.3528,
|
| 204137 |
+
"step": 82350
|
| 204138 |
+
},
|
| 204139 |
+
{
|
| 204140 |
+
"epoch": 663.77,
|
| 204141 |
+
"learning_rate": 8.686203554119548e-06,
|
| 204142 |
+
"loss": 0.5987,
|
| 204143 |
+
"step": 82355
|
| 204144 |
+
},
|
| 204145 |
+
{
|
| 204146 |
+
"epoch": 663.81,
|
| 204147 |
+
"learning_rate": 8.686122778675283e-06,
|
| 204148 |
+
"loss": 1.0841,
|
| 204149 |
+
"step": 82360
|
| 204150 |
+
},
|
| 204151 |
+
{
|
| 204152 |
+
"epoch": 663.85,
|
| 204153 |
+
"learning_rate": 8.686042003231018e-06,
|
| 204154 |
+
"loss": 0.3186,
|
| 204155 |
+
"step": 82365
|
| 204156 |
+
},
|
| 204157 |
+
{
|
| 204158 |
+
"epoch": 663.9,
|
| 204159 |
+
"learning_rate": 8.685961227786753e-06,
|
| 204160 |
+
"loss": 0.3432,
|
| 204161 |
+
"step": 82370
|
| 204162 |
+
},
|
| 204163 |
+
{
|
| 204164 |
+
"epoch": 663.94,
|
| 204165 |
+
"learning_rate": 8.685880452342488e-06,
|
| 204166 |
+
"loss": 0.4133,
|
| 204167 |
+
"step": 82375
|
| 204168 |
+
},
|
| 204169 |
+
{
|
| 204170 |
+
"epoch": 663.98,
|
| 204171 |
+
"learning_rate": 8.685799676898223e-06,
|
| 204172 |
+
"loss": 0.7278,
|
| 204173 |
+
"step": 82380
|
| 204174 |
+
},
|
| 204175 |
+
{
|
| 204176 |
+
"epoch": 664.0,
|
| 204177 |
+
"eval_loss": 0.47679567337036133,
|
| 204178 |
+
"eval_runtime": 43.85,
|
| 204179 |
+
"eval_samples_per_second": 19.179,
|
| 204180 |
+
"eval_steps_per_second": 0.616,
|
| 204181 |
+
"eval_wer": 0.19013722756292653,
|
| 204182 |
+
"step": 82383
|
| 204183 |
+
},
|
| 204184 |
+
{
|
| 204185 |
+
"epoch": 659.02,
|
| 204186 |
+
"learning_rate": 8.685735056542812e-06,
|
| 204187 |
+
"loss": 0.3786,
|
| 204188 |
+
"step": 82385
|
| 204189 |
+
},
|
| 204190 |
+
{
|
| 204191 |
+
"epoch": 659.06,
|
| 204192 |
+
"learning_rate": 8.685654281098547e-06,
|
| 204193 |
+
"loss": 0.3614,
|
| 204194 |
+
"step": 82390
|
| 204195 |
+
},
|
| 204196 |
+
{
|
| 204197 |
+
"epoch": 659.1,
|
| 204198 |
+
"learning_rate": 8.685573505654282e-06,
|
| 204199 |
+
"loss": 0.3228,
|
| 204200 |
+
"step": 82395
|
| 204201 |
+
},
|
| 204202 |
+
{
|
| 204203 |
+
"epoch": 659.14,
|
| 204204 |
+
"learning_rate": 8.685492730210017e-06,
|
| 204205 |
+
"loss": 0.3479,
|
| 204206 |
+
"step": 82400
|
| 204207 |
+
},
|
| 204208 |
+
{
|
| 204209 |
+
"epoch": 659.18,
|
| 204210 |
+
"learning_rate": 8.685411954765752e-06,
|
| 204211 |
+
"loss": 0.645,
|
| 204212 |
+
"step": 82405
|
| 204213 |
+
},
|
| 204214 |
+
{
|
| 204215 |
+
"epoch": 659.22,
|
| 204216 |
+
"learning_rate": 8.685331179321487e-06,
|
| 204217 |
+
"loss": 1.0136,
|
| 204218 |
+
"step": 82410
|
| 204219 |
+
},
|
| 204220 |
+
{
|
| 204221 |
+
"epoch": 659.26,
|
| 204222 |
+
"learning_rate": 8.685250403877221e-06,
|
| 204223 |
+
"loss": 0.3144,
|
| 204224 |
+
"step": 82415
|
| 204225 |
+
},
|
| 204226 |
+
{
|
| 204227 |
+
"epoch": 659.3,
|
| 204228 |
+
"learning_rate": 8.685169628432957e-06,
|
| 204229 |
+
"loss": 0.2708,
|
| 204230 |
+
"step": 82420
|
| 204231 |
+
},
|
| 204232 |
+
{
|
| 204233 |
+
"epoch": 659.34,
|
| 204234 |
+
"learning_rate": 8.685088852988693e-06,
|
| 204235 |
+
"loss": 0.3647,
|
| 204236 |
+
"step": 82425
|
| 204237 |
+
},
|
| 204238 |
+
{
|
| 204239 |
+
"epoch": 659.38,
|
| 204240 |
+
"learning_rate": 8.685008077544427e-06,
|
| 204241 |
+
"loss": 0.6542,
|
| 204242 |
+
"step": 82430
|
| 204243 |
+
},
|
| 204244 |
+
{
|
| 204245 |
+
"epoch": 659.42,
|
| 204246 |
+
"learning_rate": 8.684927302100163e-06,
|
| 204247 |
+
"loss": 0.9257,
|
| 204248 |
+
"step": 82435
|
| 204249 |
+
},
|
| 204250 |
+
{
|
| 204251 |
+
"epoch": 659.46,
|
| 204252 |
+
"learning_rate": 8.684846526655897e-06,
|
| 204253 |
+
"loss": 0.3278,
|
| 204254 |
+
"step": 82440
|
| 204255 |
+
},
|
| 204256 |
+
{
|
| 204257 |
+
"epoch": 659.5,
|
| 204258 |
+
"learning_rate": 8.684765751211633e-06,
|
| 204259 |
+
"loss": 0.2851,
|
| 204260 |
+
"step": 82445
|
| 204261 |
+
},
|
| 204262 |
+
{
|
| 204263 |
+
"epoch": 659.54,
|
| 204264 |
+
"learning_rate": 8.684684975767367e-06,
|
| 204265 |
+
"loss": 0.3735,
|
| 204266 |
+
"step": 82450
|
| 204267 |
+
},
|
| 204268 |
+
{
|
| 204269 |
+
"epoch": 659.58,
|
| 204270 |
+
"learning_rate": 8.684604200323103e-06,
|
| 204271 |
+
"loss": 0.58,
|
| 204272 |
+
"step": 82455
|
| 204273 |
+
},
|
| 204274 |
+
{
|
| 204275 |
+
"epoch": 659.62,
|
| 204276 |
+
"learning_rate": 8.684523424878837e-06,
|
| 204277 |
+
"loss": 1.0125,
|
| 204278 |
+
"step": 82460
|
| 204279 |
+
},
|
| 204280 |
+
{
|
| 204281 |
+
"epoch": 659.66,
|
| 204282 |
+
"learning_rate": 8.684442649434573e-06,
|
| 204283 |
+
"loss": 0.3223,
|
| 204284 |
+
"step": 82465
|
| 204285 |
+
},
|
| 204286 |
+
{
|
| 204287 |
+
"epoch": 659.7,
|
| 204288 |
+
"learning_rate": 8.684361873990307e-06,
|
| 204289 |
+
"loss": 0.311,
|
| 204290 |
+
"step": 82470
|
| 204291 |
+
},
|
| 204292 |
+
{
|
| 204293 |
+
"epoch": 659.74,
|
| 204294 |
+
"learning_rate": 8.684281098546043e-06,
|
| 204295 |
+
"loss": 0.3833,
|
| 204296 |
+
"step": 82475
|
| 204297 |
+
},
|
| 204298 |
+
{
|
| 204299 |
+
"epoch": 659.78,
|
| 204300 |
+
"learning_rate": 8.684200323101777e-06,
|
| 204301 |
+
"loss": 0.5607,
|
| 204302 |
+
"step": 82480
|
| 204303 |
+
},
|
| 204304 |
+
{
|
| 204305 |
+
"epoch": 659.82,
|
| 204306 |
+
"learning_rate": 8.684119547657513e-06,
|
| 204307 |
+
"loss": 1.225,
|
| 204308 |
+
"step": 82485
|
| 204309 |
+
},
|
| 204310 |
+
{
|
| 204311 |
+
"epoch": 659.86,
|
| 204312 |
+
"learning_rate": 8.684038772213247e-06,
|
| 204313 |
+
"loss": 0.3396,
|
| 204314 |
+
"step": 82490
|
| 204315 |
+
},
|
| 204316 |
+
{
|
| 204317 |
+
"epoch": 659.9,
|
| 204318 |
+
"learning_rate": 8.683957996768983e-06,
|
| 204319 |
+
"loss": 0.2751,
|
| 204320 |
+
"step": 82495
|
| 204321 |
+
},
|
| 204322 |
+
{
|
| 204323 |
+
"epoch": 659.94,
|
| 204324 |
+
"learning_rate": 8.683877221324719e-06,
|
| 204325 |
+
"loss": 0.3637,
|
| 204326 |
+
"step": 82500
|
| 204327 |
+
},
|
| 204328 |
+
{
|
| 204329 |
+
"epoch": 659.98,
|
| 204330 |
+
"learning_rate": 8.683796445880453e-06,
|
| 204331 |
+
"loss": 0.6049,
|
| 204332 |
+
"step": 82505
|
| 204333 |
+
},
|
| 204334 |
+
{
|
| 204335 |
+
"epoch": 660.0,
|
| 204336 |
+
"eval_loss": 0.37412241101264954,
|
| 204337 |
+
"eval_runtime": 44.3983,
|
| 204338 |
+
"eval_samples_per_second": 18.92,
|
| 204339 |
+
"eval_steps_per_second": 0.608,
|
| 204340 |
+
"eval_wer": 0.19744901635800244,
|
| 204341 |
+
"step": 82508
|
| 204342 |
+
},
|
| 204343 |
+
{
|
| 204344 |
+
"epoch": 660.02,
|
| 204345 |
+
"learning_rate": 8.683715670436189e-06,
|
| 204346 |
+
"loss": 0.3228,
|
| 204347 |
+
"step": 82510
|
| 204348 |
+
},
|
| 204349 |
+
{
|
| 204350 |
+
"epoch": 660.06,
|
| 204351 |
+
"learning_rate": 8.683634894991923e-06,
|
| 204352 |
+
"loss": 0.2838,
|
| 204353 |
+
"step": 82515
|
| 204354 |
+
},
|
| 204355 |
+
{
|
| 204356 |
+
"epoch": 660.1,
|
| 204357 |
+
"learning_rate": 8.683554119547659e-06,
|
| 204358 |
+
"loss": 0.3297,
|
| 204359 |
+
"step": 82520
|
| 204360 |
+
},
|
| 204361 |
+
{
|
| 204362 |
+
"epoch": 660.14,
|
| 204363 |
+
"learning_rate": 8.683473344103393e-06,
|
| 204364 |
+
"loss": 0.3989,
|
| 204365 |
+
"step": 82525
|
| 204366 |
+
},
|
| 204367 |
+
{
|
| 204368 |
+
"epoch": 660.18,
|
| 204369 |
+
"learning_rate": 8.683392568659129e-06,
|
| 204370 |
+
"loss": 0.56,
|
| 204371 |
+
"step": 82530
|
| 204372 |
+
},
|
| 204373 |
+
{
|
| 204374 |
+
"epoch": 660.22,
|
| 204375 |
+
"learning_rate": 8.683311793214863e-06,
|
| 204376 |
+
"loss": 1.0621,
|
| 204377 |
+
"step": 82535
|
| 204378 |
+
},
|
| 204379 |
+
{
|
| 204380 |
+
"epoch": 660.26,
|
| 204381 |
+
"learning_rate": 8.683231017770599e-06,
|
| 204382 |
+
"loss": 0.2799,
|
| 204383 |
+
"step": 82540
|
| 204384 |
+
},
|
| 204385 |
+
{
|
| 204386 |
+
"epoch": 660.3,
|
| 204387 |
+
"learning_rate": 8.683150242326333e-06,
|
| 204388 |
+
"loss": 0.3602,
|
| 204389 |
+
"step": 82545
|
| 204390 |
+
},
|
| 204391 |
+
{
|
| 204392 |
+
"epoch": 660.34,
|
| 204393 |
+
"learning_rate": 8.683069466882069e-06,
|
| 204394 |
+
"loss": 0.2814,
|
| 204395 |
+
"step": 82550
|
| 204396 |
+
},
|
| 204397 |
+
{
|
| 204398 |
+
"epoch": 660.38,
|
| 204399 |
+
"learning_rate": 8.682988691437803e-06,
|
| 204400 |
+
"loss": 0.598,
|
| 204401 |
+
"step": 82555
|
| 204402 |
+
},
|
| 204403 |
+
{
|
| 204404 |
+
"epoch": 660.42,
|
| 204405 |
+
"learning_rate": 8.682907915993539e-06,
|
| 204406 |
+
"loss": 0.9839,
|
| 204407 |
+
"step": 82560
|
| 204408 |
+
},
|
| 204409 |
+
{
|
| 204410 |
+
"epoch": 660.46,
|
| 204411 |
+
"learning_rate": 8.682827140549274e-06,
|
| 204412 |
+
"loss": 0.2857,
|
| 204413 |
+
"step": 82565
|
| 204414 |
+
},
|
| 204415 |
+
{
|
| 204416 |
+
"epoch": 660.5,
|
| 204417 |
+
"learning_rate": 8.682746365105009e-06,
|
| 204418 |
+
"loss": 0.2945,
|
| 204419 |
+
"step": 82570
|
| 204420 |
+
},
|
| 204421 |
+
{
|
| 204422 |
+
"epoch": 660.54,
|
| 204423 |
+
"learning_rate": 8.682665589660744e-06,
|
| 204424 |
+
"loss": 0.3966,
|
| 204425 |
+
"step": 82575
|
| 204426 |
+
},
|
| 204427 |
+
{
|
| 204428 |
+
"epoch": 660.58,
|
| 204429 |
+
"learning_rate": 8.682584814216479e-06,
|
| 204430 |
+
"loss": 0.7018,
|
| 204431 |
+
"step": 82580
|
| 204432 |
+
},
|
| 204433 |
+
{
|
| 204434 |
+
"epoch": 660.62,
|
| 204435 |
+
"learning_rate": 8.682504038772214e-06,
|
| 204436 |
+
"loss": 1.0091,
|
| 204437 |
+
"step": 82585
|
| 204438 |
+
},
|
| 204439 |
+
{
|
| 204440 |
+
"epoch": 660.66,
|
| 204441 |
+
"learning_rate": 8.682423263327949e-06,
|
| 204442 |
+
"loss": 0.2892,
|
| 204443 |
+
"step": 82590
|
| 204444 |
+
},
|
| 204445 |
+
{
|
| 204446 |
+
"epoch": 660.7,
|
| 204447 |
+
"learning_rate": 8.682342487883684e-06,
|
| 204448 |
+
"loss": 0.3047,
|
| 204449 |
+
"step": 82595
|
| 204450 |
+
},
|
| 204451 |
+
{
|
| 204452 |
+
"epoch": 660.74,
|
| 204453 |
+
"learning_rate": 8.682261712439419e-06,
|
| 204454 |
+
"loss": 0.4108,
|
| 204455 |
+
"step": 82600
|
| 204456 |
+
},
|
| 204457 |
+
{
|
| 204458 |
+
"epoch": 660.78,
|
| 204459 |
+
"learning_rate": 8.682180936995154e-06,
|
| 204460 |
+
"loss": 0.6222,
|
| 204461 |
+
"step": 82605
|
| 204462 |
+
},
|
| 204463 |
+
{
|
| 204464 |
+
"epoch": 660.82,
|
| 204465 |
+
"learning_rate": 8.682100161550889e-06,
|
| 204466 |
+
"loss": 1.0505,
|
| 204467 |
+
"step": 82610
|
| 204468 |
+
},
|
| 204469 |
+
{
|
| 204470 |
+
"epoch": 660.86,
|
| 204471 |
+
"learning_rate": 8.682019386106624e-06,
|
| 204472 |
+
"loss": 0.2907,
|
| 204473 |
+
"step": 82615
|
| 204474 |
+
},
|
| 204475 |
+
{
|
| 204476 |
+
"epoch": 660.9,
|
| 204477 |
+
"learning_rate": 8.681938610662359e-06,
|
| 204478 |
+
"loss": 0.2489,
|
| 204479 |
+
"step": 82620
|
| 204480 |
+
},
|
| 204481 |
+
{
|
| 204482 |
+
"epoch": 660.94,
|
| 204483 |
+
"learning_rate": 8.681857835218094e-06,
|
| 204484 |
+
"loss": 0.3507,
|
| 204485 |
+
"step": 82625
|
| 204486 |
+
},
|
| 204487 |
+
{
|
| 204488 |
+
"epoch": 660.98,
|
| 204489 |
+
"learning_rate": 8.68177705977383e-06,
|
| 204490 |
+
"loss": 0.6509,
|
| 204491 |
+
"step": 82630
|
| 204492 |
+
},
|
| 204493 |
+
{
|
| 204494 |
+
"epoch": 661.0,
|
| 204495 |
+
"eval_loss": 0.4160799980163574,
|
| 204496 |
+
"eval_runtime": 42.35,
|
| 204497 |
+
"eval_samples_per_second": 19.858,
|
| 204498 |
+
"eval_steps_per_second": 0.638,
|
| 204499 |
+
"eval_wer": 0.19666399592104306,
|
| 204500 |
+
"step": 82633
|
| 204501 |
+
},
|
| 204502 |
+
{
|
| 204503 |
+
"epoch": 661.02,
|
| 204504 |
+
"learning_rate": 8.681696284329564e-06,
|
| 204505 |
+
"loss": 0.505,
|
| 204506 |
+
"step": 82635
|
| 204507 |
+
},
|
| 204508 |
+
{
|
| 204509 |
+
"epoch": 661.06,
|
| 204510 |
+
"learning_rate": 8.6816155088853e-06,
|
| 204511 |
+
"loss": 0.3202,
|
| 204512 |
+
"step": 82640
|
| 204513 |
+
},
|
| 204514 |
+
{
|
| 204515 |
+
"epoch": 661.1,
|
| 204516 |
+
"learning_rate": 8.681534733441034e-06,
|
| 204517 |
+
"loss": 0.3216,
|
| 204518 |
+
"step": 82645
|
| 204519 |
+
},
|
| 204520 |
+
{
|
| 204521 |
+
"epoch": 661.14,
|
| 204522 |
+
"learning_rate": 8.68145395799677e-06,
|
| 204523 |
+
"loss": 0.3955,
|
| 204524 |
+
"step": 82650
|
| 204525 |
+
},
|
| 204526 |
+
{
|
| 204527 |
+
"epoch": 661.18,
|
| 204528 |
+
"learning_rate": 8.681373182552504e-06,
|
| 204529 |
+
"loss": 0.6721,
|
| 204530 |
+
"step": 82655
|
| 204531 |
+
},
|
| 204532 |
+
{
|
| 204533 |
+
"epoch": 661.22,
|
| 204534 |
+
"learning_rate": 8.68129240710824e-06,
|
| 204535 |
+
"loss": 0.9869,
|
| 204536 |
+
"step": 82660
|
| 204537 |
+
},
|
| 204538 |
+
{
|
| 204539 |
+
"epoch": 661.26,
|
| 204540 |
+
"learning_rate": 8.681211631663974e-06,
|
| 204541 |
+
"loss": 0.37,
|
| 204542 |
+
"step": 82665
|
| 204543 |
+
},
|
| 204544 |
+
{
|
| 204545 |
+
"epoch": 661.3,
|
| 204546 |
+
"learning_rate": 8.68113085621971e-06,
|
| 204547 |
+
"loss": 0.2837,
|
| 204548 |
+
"step": 82670
|
| 204549 |
+
},
|
| 204550 |
+
{
|
| 204551 |
+
"epoch": 661.34,
|
| 204552 |
+
"learning_rate": 8.681050080775444e-06,
|
| 204553 |
+
"loss": 0.3437,
|
| 204554 |
+
"step": 82675
|
| 204555 |
+
},
|
| 204556 |
+
{
|
| 204557 |
+
"epoch": 661.38,
|
| 204558 |
+
"learning_rate": 8.68096930533118e-06,
|
| 204559 |
+
"loss": 0.7001,
|
| 204560 |
+
"step": 82680
|
| 204561 |
+
},
|
| 204562 |
+
{
|
| 204563 |
+
"epoch": 661.42,
|
| 204564 |
+
"learning_rate": 8.680888529886914e-06,
|
| 204565 |
+
"loss": 0.9983,
|
| 204566 |
+
"step": 82685
|
| 204567 |
+
},
|
| 204568 |
+
{
|
| 204569 |
+
"epoch": 661.46,
|
| 204570 |
+
"learning_rate": 8.68080775444265e-06,
|
| 204571 |
+
"loss": 0.4954,
|
| 204572 |
+
"step": 82690
|
| 204573 |
+
},
|
| 204574 |
+
{
|
| 204575 |
+
"epoch": 661.5,
|
| 204576 |
+
"learning_rate": 8.680726978998384e-06,
|
| 204577 |
+
"loss": 0.341,
|
| 204578 |
+
"step": 82695
|
| 204579 |
+
},
|
| 204580 |
+
{
|
| 204581 |
+
"epoch": 661.54,
|
| 204582 |
+
"learning_rate": 8.68064620355412e-06,
|
| 204583 |
+
"loss": 0.3555,
|
| 204584 |
+
"step": 82700
|
| 204585 |
+
},
|
| 204586 |
+
{
|
| 204587 |
+
"epoch": 661.58,
|
| 204588 |
+
"learning_rate": 8.680565428109856e-06,
|
| 204589 |
+
"loss": 0.7123,
|
| 204590 |
+
"step": 82705
|
| 204591 |
+
},
|
| 204592 |
+
{
|
| 204593 |
+
"epoch": 661.62,
|
| 204594 |
+
"learning_rate": 8.68048465266559e-06,
|
| 204595 |
+
"loss": 1.0405,
|
| 204596 |
+
"step": 82710
|
| 204597 |
+
},
|
| 204598 |
+
{
|
| 204599 |
+
"epoch": 661.66,
|
| 204600 |
+
"learning_rate": 8.680403877221326e-06,
|
| 204601 |
+
"loss": 0.3155,
|
| 204602 |
+
"step": 82715
|
| 204603 |
+
},
|
| 204604 |
+
{
|
| 204605 |
+
"epoch": 661.7,
|
| 204606 |
+
"learning_rate": 8.68032310177706e-06,
|
| 204607 |
+
"loss": 0.2971,
|
| 204608 |
+
"step": 82720
|
| 204609 |
+
},
|
| 204610 |
+
{
|
| 204611 |
+
"epoch": 661.74,
|
| 204612 |
+
"learning_rate": 8.680242326332796e-06,
|
| 204613 |
+
"loss": 0.4496,
|
| 204614 |
+
"step": 82725
|
| 204615 |
+
},
|
| 204616 |
+
{
|
| 204617 |
+
"epoch": 661.78,
|
| 204618 |
+
"learning_rate": 8.68016155088853e-06,
|
| 204619 |
+
"loss": 0.6356,
|
| 204620 |
+
"step": 82730
|
| 204621 |
+
},
|
| 204622 |
+
{
|
| 204623 |
+
"epoch": 661.82,
|
| 204624 |
+
"learning_rate": 8.680080775444266e-06,
|
| 204625 |
+
"loss": 1.0775,
|
| 204626 |
+
"step": 82735
|
| 204627 |
+
},
|
| 204628 |
+
{
|
| 204629 |
+
"epoch": 661.86,
|
| 204630 |
+
"learning_rate": 8.68e-06,
|
| 204631 |
+
"loss": 0.5475,
|
| 204632 |
+
"step": 82740
|
| 204633 |
+
},
|
| 204634 |
+
{
|
| 204635 |
+
"epoch": 661.9,
|
| 204636 |
+
"learning_rate": 8.679919224555736e-06,
|
| 204637 |
+
"loss": 0.2951,
|
| 204638 |
+
"step": 82745
|
| 204639 |
+
},
|
| 204640 |
+
{
|
| 204641 |
+
"epoch": 661.94,
|
| 204642 |
+
"learning_rate": 8.67983844911147e-06,
|
| 204643 |
+
"loss": 0.3912,
|
| 204644 |
+
"step": 82750
|
| 204645 |
+
},
|
| 204646 |
+
{
|
| 204647 |
+
"epoch": 661.98,
|
| 204648 |
+
"learning_rate": 8.679757673667206e-06,
|
| 204649 |
+
"loss": 0.6424,
|
| 204650 |
+
"step": 82755
|
| 204651 |
+
},
|
| 204652 |
+
{
|
| 204653 |
+
"epoch": 662.0,
|
| 204654 |
+
"eval_loss": 0.4037238359451294,
|
| 204655 |
+
"eval_runtime": 41.4212,
|
| 204656 |
+
"eval_samples_per_second": 20.304,
|
| 204657 |
+
"eval_steps_per_second": 0.652,
|
| 204658 |
+
"eval_wer": 0.18659354931813263,
|
| 204659 |
+
"step": 82758
|
| 204660 |
}
|
| 204661 |
],
|
| 204662 |
+
"max_steps": 625000,
|
| 204663 |
"num_train_epochs": 5000,
|
| 204664 |
+
"total_flos": 2.3289568086960737e+20,
|
| 204665 |
"trial_name": null,
|
| 204666 |
"trial_params": null
|
| 204667 |
}
|
model-bin/finetune/base/{checkpoint-82134 → checkpoint-82758}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629911360.5154393/events.out.tfevents.1629911360.7e498afd5545.7645.95
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2223313f19042abdf103cd948eff3e01590ee994ae7d9d316561692f121c2b19
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629911829.7407227/events.out.tfevents.1629911829.7e498afd5545.7645.97
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cf36c5a3bef24fdf888d6f18dd7293ffbab9d54925ca8605fd224ce2819a9883
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629912375.8557265/events.out.tfevents.1629912375.7e498afd5545.7645.99
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f5982527167f2eb761f566a40edb5026dd27e3bc2a328a3467537a88bfff8ebe
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629912862.109049/events.out.tfevents.1629912863.7e498afd5545.7645.101
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6a98d8eb673c08f8c73d63d204eed53152e07b83ab387bc968d4927abecb1d78
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629913333.6385581/events.out.tfevents.1629913333.7e498afd5545.7645.103
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f9650b8bfec733129a1ef1be9e0bfff38ae268d68ed025d78e6b9b4e29399d09
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629911360.7e498afd5545.7645.94
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:575adb2b57242b7b6aade64cc5cc39e4d296f907b620b3c1d092ef24f771c32d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629911829.7e498afd5545.7645.96
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:36006e797f00edfe6a1af2c001613ff5830d75503ff5bcdb310dc6a7bb9ed138
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629912375.7e498afd5545.7645.98
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9238bdd2229cc14fe659dd65fd2c7667d9c3df31e8c3e4770aa9e5daf97c7776
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629912862.7e498afd5545.7645.100
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:433192096413e3af5862869f29a54ae1127aa72436f1598b7425a047971f9d2b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629913333.7e498afd5545.7645.102
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8f9e08121c74367815ec30d01bd81d552a90baff7f2311a637b6cdd2ef525ac3
|
| 3 |
+
size 8622
|