"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629982277.5117593/events.out.tfevents.1629982277.8e89bd551565.924.191 +3 -0
- model-bin/finetune/base/log/1629982720.2748618/events.out.tfevents.1629982720.8e89bd551565.924.193 +3 -0
- model-bin/finetune/base/log/1629983163.766655/events.out.tfevents.1629983163.8e89bd551565.924.195 +3 -0
- model-bin/finetune/base/log/1629983583.3409672/events.out.tfevents.1629983583.8e89bd551565.924.197 +3 -0
- model-bin/finetune/base/log/1629984010.4668894/events.out.tfevents.1629984011.8e89bd551565.924.199 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629982277.8e89bd551565.924.190 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629982719.8e89bd551565.924.192 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629983162.8e89bd551565.924.194 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629983583.8e89bd551565.924.196 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629984010.8e89bd551565.924.198 +3 -0
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/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 sha256:106018a2f6d9b262414e2644ce43f843c8e696eca378d4249e34972449080519
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/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:beb041ebf9f3a4ede1423312f05f8532c405db1f5d083e5c547a31b72428b880
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/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:f57f37f9766dcd5f8e51d861073f14ae5f33d1d1afc938fd20d4c57dfff608a1
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/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 sha256:81eb64bd860b8d9ad973193c95bc88d4b976380774b6b6663ea8985193e5403e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/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 sha256:528c134b4c66f9c85e89201fe3a8a414fed74ad70bddf74b4097ed3f0bd77382
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -226827,11 +226827,800 @@
|
|
| 226827 |
"eval_steps_per_second": 0.705,
|
| 226828 |
"eval_wer": 0.18565555718690355,
|
| 226829 |
"step": 100181
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 226830 |
}
|
| 226831 |
],
|
| 226832 |
-
"max_steps":
|
| 226833 |
"num_train_epochs": 5000,
|
| 226834 |
-
"total_flos": 2.
|
| 226835 |
"trial_name": null,
|
| 226836 |
"trial_params": null
|
| 226837 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 806.0,
|
| 5 |
+
"global_step": 100803,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 226827 |
"eval_steps_per_second": 0.705,
|
| 226828 |
"eval_wer": 0.18565555718690355,
|
| 226829 |
"step": 100181
|
| 226830 |
+
},
|
| 226831 |
+
{
|
| 226832 |
+
"epoch": 807.03,
|
| 226833 |
+
"learning_rate": 8.411137820512821e-06,
|
| 226834 |
+
"loss": 0.3109,
|
| 226835 |
+
"step": 100185
|
| 226836 |
+
},
|
| 226837 |
+
{
|
| 226838 |
+
"epoch": 807.07,
|
| 226839 |
+
"learning_rate": 8.411057692307692e-06,
|
| 226840 |
+
"loss": 0.2535,
|
| 226841 |
+
"step": 100190
|
| 226842 |
+
},
|
| 226843 |
+
{
|
| 226844 |
+
"epoch": 807.11,
|
| 226845 |
+
"learning_rate": 8.410977564102565e-06,
|
| 226846 |
+
"loss": 0.3186,
|
| 226847 |
+
"step": 100195
|
| 226848 |
+
},
|
| 226849 |
+
{
|
| 226850 |
+
"epoch": 807.15,
|
| 226851 |
+
"learning_rate": 8.410897435897437e-06,
|
| 226852 |
+
"loss": 0.3538,
|
| 226853 |
+
"step": 100200
|
| 226854 |
+
},
|
| 226855 |
+
{
|
| 226856 |
+
"epoch": 807.19,
|
| 226857 |
+
"learning_rate": 8.410817307692308e-06,
|
| 226858 |
+
"loss": 0.8496,
|
| 226859 |
+
"step": 100205
|
| 226860 |
+
},
|
| 226861 |
+
{
|
| 226862 |
+
"epoch": 807.23,
|
| 226863 |
+
"learning_rate": 8.410737179487181e-06,
|
| 226864 |
+
"loss": 0.6444,
|
| 226865 |
+
"step": 100210
|
| 226866 |
+
},
|
| 226867 |
+
{
|
| 226868 |
+
"epoch": 807.27,
|
| 226869 |
+
"learning_rate": 8.410657051282053e-06,
|
| 226870 |
+
"loss": 0.2664,
|
| 226871 |
+
"step": 100215
|
| 226872 |
+
},
|
| 226873 |
+
{
|
| 226874 |
+
"epoch": 807.31,
|
| 226875 |
+
"learning_rate": 8.410576923076924e-06,
|
| 226876 |
+
"loss": 0.3318,
|
| 226877 |
+
"step": 100220
|
| 226878 |
+
},
|
| 226879 |
+
{
|
| 226880 |
+
"epoch": 807.35,
|
| 226881 |
+
"learning_rate": 8.410496794871795e-06,
|
| 226882 |
+
"loss": 0.4685,
|
| 226883 |
+
"step": 100225
|
| 226884 |
+
},
|
| 226885 |
+
{
|
| 226886 |
+
"epoch": 807.39,
|
| 226887 |
+
"learning_rate": 8.410416666666668e-06,
|
| 226888 |
+
"loss": 0.9129,
|
| 226889 |
+
"step": 100230
|
| 226890 |
+
},
|
| 226891 |
+
{
|
| 226892 |
+
"epoch": 807.43,
|
| 226893 |
+
"learning_rate": 8.410336538461538e-06,
|
| 226894 |
+
"loss": 0.7842,
|
| 226895 |
+
"step": 100235
|
| 226896 |
+
},
|
| 226897 |
+
{
|
| 226898 |
+
"epoch": 807.47,
|
| 226899 |
+
"learning_rate": 8.410256410256411e-06,
|
| 226900 |
+
"loss": 0.3964,
|
| 226901 |
+
"step": 100240
|
| 226902 |
+
},
|
| 226903 |
+
{
|
| 226904 |
+
"epoch": 807.51,
|
| 226905 |
+
"learning_rate": 8.410176282051282e-06,
|
| 226906 |
+
"loss": 0.3,
|
| 226907 |
+
"step": 100245
|
| 226908 |
+
},
|
| 226909 |
+
{
|
| 226910 |
+
"epoch": 807.55,
|
| 226911 |
+
"learning_rate": 8.410096153846154e-06,
|
| 226912 |
+
"loss": 0.4404,
|
| 226913 |
+
"step": 100250
|
| 226914 |
+
},
|
| 226915 |
+
{
|
| 226916 |
+
"epoch": 807.59,
|
| 226917 |
+
"learning_rate": 8.410016025641027e-06,
|
| 226918 |
+
"loss": 0.8669,
|
| 226919 |
+
"step": 100255
|
| 226920 |
+
},
|
| 226921 |
+
{
|
| 226922 |
+
"epoch": 807.63,
|
| 226923 |
+
"learning_rate": 8.409935897435898e-06,
|
| 226924 |
+
"loss": 0.676,
|
| 226925 |
+
"step": 100260
|
| 226926 |
+
},
|
| 226927 |
+
{
|
| 226928 |
+
"epoch": 807.67,
|
| 226929 |
+
"learning_rate": 8.40985576923077e-06,
|
| 226930 |
+
"loss": 0.3293,
|
| 226931 |
+
"step": 100265
|
| 226932 |
+
},
|
| 226933 |
+
{
|
| 226934 |
+
"epoch": 807.71,
|
| 226935 |
+
"learning_rate": 8.409775641025641e-06,
|
| 226936 |
+
"loss": 0.3637,
|
| 226937 |
+
"step": 100270
|
| 226938 |
+
},
|
| 226939 |
+
{
|
| 226940 |
+
"epoch": 807.76,
|
| 226941 |
+
"learning_rate": 8.409695512820514e-06,
|
| 226942 |
+
"loss": 0.5279,
|
| 226943 |
+
"step": 100275
|
| 226944 |
+
},
|
| 226945 |
+
{
|
| 226946 |
+
"epoch": 807.8,
|
| 226947 |
+
"learning_rate": 8.409615384615385e-06,
|
| 226948 |
+
"loss": 0.9272,
|
| 226949 |
+
"step": 100280
|
| 226950 |
+
},
|
| 226951 |
+
{
|
| 226952 |
+
"epoch": 807.84,
|
| 226953 |
+
"learning_rate": 8.409535256410257e-06,
|
| 226954 |
+
"loss": 0.7397,
|
| 226955 |
+
"step": 100285
|
| 226956 |
+
},
|
| 226957 |
+
{
|
| 226958 |
+
"epoch": 807.88,
|
| 226959 |
+
"learning_rate": 8.409455128205128e-06,
|
| 226960 |
+
"loss": 0.2691,
|
| 226961 |
+
"step": 100290
|
| 226962 |
+
},
|
| 226963 |
+
{
|
| 226964 |
+
"epoch": 807.92,
|
| 226965 |
+
"learning_rate": 8.409375000000001e-06,
|
| 226966 |
+
"loss": 0.3059,
|
| 226967 |
+
"step": 100295
|
| 226968 |
+
},
|
| 226969 |
+
{
|
| 226970 |
+
"epoch": 807.96,
|
| 226971 |
+
"learning_rate": 8.409294871794872e-06,
|
| 226972 |
+
"loss": 0.4232,
|
| 226973 |
+
"step": 100300
|
| 226974 |
+
},
|
| 226975 |
+
{
|
| 226976 |
+
"epoch": 808.0,
|
| 226977 |
+
"learning_rate": 8.409214743589744e-06,
|
| 226978 |
+
"loss": 1.1247,
|
| 226979 |
+
"step": 100305
|
| 226980 |
+
},
|
| 226981 |
+
{
|
| 226982 |
+
"epoch": 808.0,
|
| 226983 |
+
"eval_loss": 0.39919790625572205,
|
| 226984 |
+
"eval_runtime": 39.9523,
|
| 226985 |
+
"eval_samples_per_second": 20.85,
|
| 226986 |
+
"eval_steps_per_second": 0.676,
|
| 226987 |
+
"eval_wer": 0.1869192799070848,
|
| 226988 |
+
"step": 100305
|
| 226989 |
+
},
|
| 226990 |
+
{
|
| 226991 |
+
"epoch": 808.04,
|
| 226992 |
+
"learning_rate": 8.409134615384617e-06,
|
| 226993 |
+
"loss": 0.3744,
|
| 226994 |
+
"step": 100310
|
| 226995 |
+
},
|
| 226996 |
+
{
|
| 226997 |
+
"epoch": 808.08,
|
| 226998 |
+
"learning_rate": 8.409054487179488e-06,
|
| 226999 |
+
"loss": 0.279,
|
| 227000 |
+
"step": 100315
|
| 227001 |
+
},
|
| 227002 |
+
{
|
| 227003 |
+
"epoch": 808.12,
|
| 227004 |
+
"learning_rate": 8.40897435897436e-06,
|
| 227005 |
+
"loss": 0.3388,
|
| 227006 |
+
"step": 100320
|
| 227007 |
+
},
|
| 227008 |
+
{
|
| 227009 |
+
"epoch": 808.16,
|
| 227010 |
+
"learning_rate": 8.408894230769231e-06,
|
| 227011 |
+
"loss": 0.4341,
|
| 227012 |
+
"step": 100325
|
| 227013 |
+
},
|
| 227014 |
+
{
|
| 227015 |
+
"epoch": 808.2,
|
| 227016 |
+
"learning_rate": 8.408814102564104e-06,
|
| 227017 |
+
"loss": 1.1642,
|
| 227018 |
+
"step": 100330
|
| 227019 |
+
},
|
| 227020 |
+
{
|
| 227021 |
+
"epoch": 808.24,
|
| 227022 |
+
"learning_rate": 8.408733974358975e-06,
|
| 227023 |
+
"loss": 0.3418,
|
| 227024 |
+
"step": 100335
|
| 227025 |
+
},
|
| 227026 |
+
{
|
| 227027 |
+
"epoch": 808.28,
|
| 227028 |
+
"learning_rate": 8.408653846153847e-06,
|
| 227029 |
+
"loss": 0.3208,
|
| 227030 |
+
"step": 100340
|
| 227031 |
+
},
|
| 227032 |
+
{
|
| 227033 |
+
"epoch": 808.32,
|
| 227034 |
+
"learning_rate": 8.408573717948718e-06,
|
| 227035 |
+
"loss": 0.3099,
|
| 227036 |
+
"step": 100345
|
| 227037 |
+
},
|
| 227038 |
+
{
|
| 227039 |
+
"epoch": 808.36,
|
| 227040 |
+
"learning_rate": 8.408493589743591e-06,
|
| 227041 |
+
"loss": 0.5001,
|
| 227042 |
+
"step": 100350
|
| 227043 |
+
},
|
| 227044 |
+
{
|
| 227045 |
+
"epoch": 808.4,
|
| 227046 |
+
"learning_rate": 8.408413461538463e-06,
|
| 227047 |
+
"loss": 1.1834,
|
| 227048 |
+
"step": 100355
|
| 227049 |
+
},
|
| 227050 |
+
{
|
| 227051 |
+
"epoch": 808.44,
|
| 227052 |
+
"learning_rate": 8.408333333333334e-06,
|
| 227053 |
+
"loss": 0.3267,
|
| 227054 |
+
"step": 100360
|
| 227055 |
+
},
|
| 227056 |
+
{
|
| 227057 |
+
"epoch": 808.48,
|
| 227058 |
+
"learning_rate": 8.408253205128207e-06,
|
| 227059 |
+
"loss": 0.2782,
|
| 227060 |
+
"step": 100365
|
| 227061 |
+
},
|
| 227062 |
+
{
|
| 227063 |
+
"epoch": 808.52,
|
| 227064 |
+
"learning_rate": 8.408173076923077e-06,
|
| 227065 |
+
"loss": 0.2854,
|
| 227066 |
+
"step": 100370
|
| 227067 |
+
},
|
| 227068 |
+
{
|
| 227069 |
+
"epoch": 808.56,
|
| 227070 |
+
"learning_rate": 8.40809294871795e-06,
|
| 227071 |
+
"loss": 0.4589,
|
| 227072 |
+
"step": 100375
|
| 227073 |
+
},
|
| 227074 |
+
{
|
| 227075 |
+
"epoch": 808.6,
|
| 227076 |
+
"learning_rate": 8.408012820512821e-06,
|
| 227077 |
+
"loss": 1.0659,
|
| 227078 |
+
"step": 100380
|
| 227079 |
+
},
|
| 227080 |
+
{
|
| 227081 |
+
"epoch": 808.64,
|
| 227082 |
+
"learning_rate": 8.407932692307692e-06,
|
| 227083 |
+
"loss": 0.2958,
|
| 227084 |
+
"step": 100385
|
| 227085 |
+
},
|
| 227086 |
+
{
|
| 227087 |
+
"epoch": 808.68,
|
| 227088 |
+
"learning_rate": 8.407852564102564e-06,
|
| 227089 |
+
"loss": 0.2719,
|
| 227090 |
+
"step": 100390
|
| 227091 |
+
},
|
| 227092 |
+
{
|
| 227093 |
+
"epoch": 808.72,
|
| 227094 |
+
"learning_rate": 8.407772435897437e-06,
|
| 227095 |
+
"loss": 0.3458,
|
| 227096 |
+
"step": 100395
|
| 227097 |
+
},
|
| 227098 |
+
{
|
| 227099 |
+
"epoch": 808.76,
|
| 227100 |
+
"learning_rate": 8.407692307692308e-06,
|
| 227101 |
+
"loss": 0.4317,
|
| 227102 |
+
"step": 100400
|
| 227103 |
+
},
|
| 227104 |
+
{
|
| 227105 |
+
"epoch": 808.8,
|
| 227106 |
+
"learning_rate": 8.40761217948718e-06,
|
| 227107 |
+
"loss": 1.1277,
|
| 227108 |
+
"step": 100405
|
| 227109 |
+
},
|
| 227110 |
+
{
|
| 227111 |
+
"epoch": 808.84,
|
| 227112 |
+
"learning_rate": 8.407532051282053e-06,
|
| 227113 |
+
"loss": 0.365,
|
| 227114 |
+
"step": 100410
|
| 227115 |
+
},
|
| 227116 |
+
{
|
| 227117 |
+
"epoch": 808.88,
|
| 227118 |
+
"learning_rate": 8.407451923076924e-06,
|
| 227119 |
+
"loss": 0.2837,
|
| 227120 |
+
"step": 100415
|
| 227121 |
+
},
|
| 227122 |
+
{
|
| 227123 |
+
"epoch": 808.92,
|
| 227124 |
+
"learning_rate": 8.407371794871795e-06,
|
| 227125 |
+
"loss": 0.2807,
|
| 227126 |
+
"step": 100420
|
| 227127 |
+
},
|
| 227128 |
+
{
|
| 227129 |
+
"epoch": 808.96,
|
| 227130 |
+
"learning_rate": 8.407291666666667e-06,
|
| 227131 |
+
"loss": 0.4624,
|
| 227132 |
+
"step": 100425
|
| 227133 |
+
},
|
| 227134 |
+
{
|
| 227135 |
+
"epoch": 809.0,
|
| 227136 |
+
"eval_loss": 0.3385351896286011,
|
| 227137 |
+
"eval_runtime": 38.446,
|
| 227138 |
+
"eval_samples_per_second": 21.667,
|
| 227139 |
+
"eval_steps_per_second": 0.702,
|
| 227140 |
+
"eval_wer": 0.19216979779148144,
|
| 227141 |
+
"step": 100429
|
| 227142 |
+
},
|
| 227143 |
+
{
|
| 227144 |
+
"epoch": 809.01,
|
| 227145 |
+
"learning_rate": 8.40721153846154e-06,
|
| 227146 |
+
"loss": 0.5952,
|
| 227147 |
+
"step": 100430
|
| 227148 |
+
},
|
| 227149 |
+
{
|
| 227150 |
+
"epoch": 809.05,
|
| 227151 |
+
"learning_rate": 8.407131410256411e-06,
|
| 227152 |
+
"loss": 0.2565,
|
| 227153 |
+
"step": 100435
|
| 227154 |
+
},
|
| 227155 |
+
{
|
| 227156 |
+
"epoch": 809.09,
|
| 227157 |
+
"learning_rate": 8.407051282051282e-06,
|
| 227158 |
+
"loss": 0.2874,
|
| 227159 |
+
"step": 100440
|
| 227160 |
+
},
|
| 227161 |
+
{
|
| 227162 |
+
"epoch": 809.13,
|
| 227163 |
+
"learning_rate": 8.406971153846154e-06,
|
| 227164 |
+
"loss": 0.3954,
|
| 227165 |
+
"step": 100445
|
| 227166 |
+
},
|
| 227167 |
+
{
|
| 227168 |
+
"epoch": 809.17,
|
| 227169 |
+
"learning_rate": 8.406891025641027e-06,
|
| 227170 |
+
"loss": 0.5124,
|
| 227171 |
+
"step": 100450
|
| 227172 |
+
},
|
| 227173 |
+
{
|
| 227174 |
+
"epoch": 809.21,
|
| 227175 |
+
"learning_rate": 8.406810897435898e-06,
|
| 227176 |
+
"loss": 1.1811,
|
| 227177 |
+
"step": 100455
|
| 227178 |
+
},
|
| 227179 |
+
{
|
| 227180 |
+
"epoch": 809.25,
|
| 227181 |
+
"learning_rate": 8.40673076923077e-06,
|
| 227182 |
+
"loss": 0.2798,
|
| 227183 |
+
"step": 100460
|
| 227184 |
+
},
|
| 227185 |
+
{
|
| 227186 |
+
"epoch": 809.29,
|
| 227187 |
+
"learning_rate": 8.406650641025643e-06,
|
| 227188 |
+
"loss": 0.3528,
|
| 227189 |
+
"step": 100465
|
| 227190 |
+
},
|
| 227191 |
+
{
|
| 227192 |
+
"epoch": 809.33,
|
| 227193 |
+
"learning_rate": 8.406570512820514e-06,
|
| 227194 |
+
"loss": 0.3479,
|
| 227195 |
+
"step": 100470
|
| 227196 |
+
},
|
| 227197 |
+
{
|
| 227198 |
+
"epoch": 809.37,
|
| 227199 |
+
"learning_rate": 8.406490384615385e-06,
|
| 227200 |
+
"loss": 0.4995,
|
| 227201 |
+
"step": 100475
|
| 227202 |
+
},
|
| 227203 |
+
{
|
| 227204 |
+
"epoch": 809.41,
|
| 227205 |
+
"learning_rate": 8.406410256410257e-06,
|
| 227206 |
+
"loss": 1.3033,
|
| 227207 |
+
"step": 100480
|
| 227208 |
+
},
|
| 227209 |
+
{
|
| 227210 |
+
"epoch": 809.45,
|
| 227211 |
+
"learning_rate": 8.40633012820513e-06,
|
| 227212 |
+
"loss": 0.2764,
|
| 227213 |
+
"step": 100485
|
| 227214 |
+
},
|
| 227215 |
+
{
|
| 227216 |
+
"epoch": 809.49,
|
| 227217 |
+
"learning_rate": 8.406250000000001e-06,
|
| 227218 |
+
"loss": 0.3067,
|
| 227219 |
+
"step": 100490
|
| 227220 |
+
},
|
| 227221 |
+
{
|
| 227222 |
+
"epoch": 809.53,
|
| 227223 |
+
"learning_rate": 8.406169871794872e-06,
|
| 227224 |
+
"loss": 0.3278,
|
| 227225 |
+
"step": 100495
|
| 227226 |
+
},
|
| 227227 |
+
{
|
| 227228 |
+
"epoch": 809.57,
|
| 227229 |
+
"learning_rate": 8.406089743589744e-06,
|
| 227230 |
+
"loss": 0.606,
|
| 227231 |
+
"step": 100500
|
| 227232 |
+
},
|
| 227233 |
+
{
|
| 227234 |
+
"epoch": 809.61,
|
| 227235 |
+
"learning_rate": 8.406009615384617e-06,
|
| 227236 |
+
"loss": 0.9779,
|
| 227237 |
+
"step": 100505
|
| 227238 |
+
},
|
| 227239 |
+
{
|
| 227240 |
+
"epoch": 809.65,
|
| 227241 |
+
"learning_rate": 8.405929487179488e-06,
|
| 227242 |
+
"loss": 0.2627,
|
| 227243 |
+
"step": 100510
|
| 227244 |
+
},
|
| 227245 |
+
{
|
| 227246 |
+
"epoch": 809.69,
|
| 227247 |
+
"learning_rate": 8.40584935897436e-06,
|
| 227248 |
+
"loss": 0.3738,
|
| 227249 |
+
"step": 100515
|
| 227250 |
+
},
|
| 227251 |
+
{
|
| 227252 |
+
"epoch": 809.73,
|
| 227253 |
+
"learning_rate": 8.405769230769233e-06,
|
| 227254 |
+
"loss": 0.3222,
|
| 227255 |
+
"step": 100520
|
| 227256 |
+
},
|
| 227257 |
+
{
|
| 227258 |
+
"epoch": 809.77,
|
| 227259 |
+
"learning_rate": 8.405689102564102e-06,
|
| 227260 |
+
"loss": 0.5173,
|
| 227261 |
+
"step": 100525
|
| 227262 |
+
},
|
| 227263 |
+
{
|
| 227264 |
+
"epoch": 809.81,
|
| 227265 |
+
"learning_rate": 8.405608974358975e-06,
|
| 227266 |
+
"loss": 1.1401,
|
| 227267 |
+
"step": 100530
|
| 227268 |
+
},
|
| 227269 |
+
{
|
| 227270 |
+
"epoch": 809.85,
|
| 227271 |
+
"learning_rate": 8.405528846153847e-06,
|
| 227272 |
+
"loss": 0.3159,
|
| 227273 |
+
"step": 100535
|
| 227274 |
+
},
|
| 227275 |
+
{
|
| 227276 |
+
"epoch": 809.89,
|
| 227277 |
+
"learning_rate": 8.405448717948718e-06,
|
| 227278 |
+
"loss": 0.2589,
|
| 227279 |
+
"step": 100540
|
| 227280 |
+
},
|
| 227281 |
+
{
|
| 227282 |
+
"epoch": 809.93,
|
| 227283 |
+
"learning_rate": 8.40536858974359e-06,
|
| 227284 |
+
"loss": 0.3764,
|
| 227285 |
+
"step": 100545
|
| 227286 |
+
},
|
| 227287 |
+
{
|
| 227288 |
+
"epoch": 809.97,
|
| 227289 |
+
"learning_rate": 8.405288461538462e-06,
|
| 227290 |
+
"loss": 0.5405,
|
| 227291 |
+
"step": 100550
|
| 227292 |
+
},
|
| 227293 |
+
{
|
| 227294 |
+
"epoch": 810.0,
|
| 227295 |
+
"eval_loss": 0.3283374011516571,
|
| 227296 |
+
"eval_runtime": 37.22,
|
| 227297 |
+
"eval_samples_per_second": 22.407,
|
| 227298 |
+
"eval_steps_per_second": 0.725,
|
| 227299 |
+
"eval_wer": 0.17833504927195176,
|
| 227300 |
+
"step": 100553
|
| 227301 |
+
},
|
| 227302 |
+
{
|
| 227303 |
+
"epoch": 804.02,
|
| 227304 |
+
"learning_rate": 8.405208333333334e-06,
|
| 227305 |
+
"loss": 0.3528,
|
| 227306 |
+
"step": 100555
|
| 227307 |
+
},
|
| 227308 |
+
{
|
| 227309 |
+
"epoch": 804.06,
|
| 227310 |
+
"learning_rate": 8.405128205128205e-06,
|
| 227311 |
+
"loss": 0.2978,
|
| 227312 |
+
"step": 100560
|
| 227313 |
+
},
|
| 227314 |
+
{
|
| 227315 |
+
"epoch": 804.1,
|
| 227316 |
+
"learning_rate": 8.405048076923078e-06,
|
| 227317 |
+
"loss": 0.3136,
|
| 227318 |
+
"step": 100565
|
| 227319 |
+
},
|
| 227320 |
+
{
|
| 227321 |
+
"epoch": 804.14,
|
| 227322 |
+
"learning_rate": 8.40496794871795e-06,
|
| 227323 |
+
"loss": 0.3539,
|
| 227324 |
+
"step": 100570
|
| 227325 |
+
},
|
| 227326 |
+
{
|
| 227327 |
+
"epoch": 804.18,
|
| 227328 |
+
"learning_rate": 8.404887820512821e-06,
|
| 227329 |
+
"loss": 0.6009,
|
| 227330 |
+
"step": 100575
|
| 227331 |
+
},
|
| 227332 |
+
{
|
| 227333 |
+
"epoch": 804.22,
|
| 227334 |
+
"learning_rate": 8.404807692307692e-06,
|
| 227335 |
+
"loss": 1.057,
|
| 227336 |
+
"step": 100580
|
| 227337 |
+
},
|
| 227338 |
+
{
|
| 227339 |
+
"epoch": 804.26,
|
| 227340 |
+
"learning_rate": 8.404727564102565e-06,
|
| 227341 |
+
"loss": 0.3244,
|
| 227342 |
+
"step": 100585
|
| 227343 |
+
},
|
| 227344 |
+
{
|
| 227345 |
+
"epoch": 804.3,
|
| 227346 |
+
"learning_rate": 8.404647435897437e-06,
|
| 227347 |
+
"loss": 0.2644,
|
| 227348 |
+
"step": 100590
|
| 227349 |
+
},
|
| 227350 |
+
{
|
| 227351 |
+
"epoch": 804.34,
|
| 227352 |
+
"learning_rate": 8.404567307692308e-06,
|
| 227353 |
+
"loss": 0.3934,
|
| 227354 |
+
"step": 100595
|
| 227355 |
+
},
|
| 227356 |
+
{
|
| 227357 |
+
"epoch": 804.38,
|
| 227358 |
+
"learning_rate": 8.40448717948718e-06,
|
| 227359 |
+
"loss": 0.5217,
|
| 227360 |
+
"step": 100600
|
| 227361 |
+
},
|
| 227362 |
+
{
|
| 227363 |
+
"epoch": 804.42,
|
| 227364 |
+
"learning_rate": 8.404407051282052e-06,
|
| 227365 |
+
"loss": 1.0261,
|
| 227366 |
+
"step": 100605
|
| 227367 |
+
},
|
| 227368 |
+
{
|
| 227369 |
+
"epoch": 804.46,
|
| 227370 |
+
"learning_rate": 8.404326923076924e-06,
|
| 227371 |
+
"loss": 0.3029,
|
| 227372 |
+
"step": 100610
|
| 227373 |
+
},
|
| 227374 |
+
{
|
| 227375 |
+
"epoch": 804.5,
|
| 227376 |
+
"learning_rate": 8.404246794871795e-06,
|
| 227377 |
+
"loss": 0.3028,
|
| 227378 |
+
"step": 100615
|
| 227379 |
+
},
|
| 227380 |
+
{
|
| 227381 |
+
"epoch": 804.54,
|
| 227382 |
+
"learning_rate": 8.404166666666668e-06,
|
| 227383 |
+
"loss": 0.3211,
|
| 227384 |
+
"step": 100620
|
| 227385 |
+
},
|
| 227386 |
+
{
|
| 227387 |
+
"epoch": 804.58,
|
| 227388 |
+
"learning_rate": 8.40408653846154e-06,
|
| 227389 |
+
"loss": 0.6224,
|
| 227390 |
+
"step": 100625
|
| 227391 |
+
},
|
| 227392 |
+
{
|
| 227393 |
+
"epoch": 804.62,
|
| 227394 |
+
"learning_rate": 8.404006410256411e-06,
|
| 227395 |
+
"loss": 1.0374,
|
| 227396 |
+
"step": 100630
|
| 227397 |
+
},
|
| 227398 |
+
{
|
| 227399 |
+
"epoch": 804.66,
|
| 227400 |
+
"learning_rate": 8.403926282051282e-06,
|
| 227401 |
+
"loss": 0.3005,
|
| 227402 |
+
"step": 100635
|
| 227403 |
+
},
|
| 227404 |
+
{
|
| 227405 |
+
"epoch": 804.7,
|
| 227406 |
+
"learning_rate": 8.403846153846155e-06,
|
| 227407 |
+
"loss": 0.3301,
|
| 227408 |
+
"step": 100640
|
| 227409 |
+
},
|
| 227410 |
+
{
|
| 227411 |
+
"epoch": 804.74,
|
| 227412 |
+
"learning_rate": 8.403766025641025e-06,
|
| 227413 |
+
"loss": 0.3698,
|
| 227414 |
+
"step": 100645
|
| 227415 |
+
},
|
| 227416 |
+
{
|
| 227417 |
+
"epoch": 804.78,
|
| 227418 |
+
"learning_rate": 8.403685897435898e-06,
|
| 227419 |
+
"loss": 0.7145,
|
| 227420 |
+
"step": 100650
|
| 227421 |
+
},
|
| 227422 |
+
{
|
| 227423 |
+
"epoch": 804.82,
|
| 227424 |
+
"learning_rate": 8.403605769230771e-06,
|
| 227425 |
+
"loss": 1.0669,
|
| 227426 |
+
"step": 100655
|
| 227427 |
+
},
|
| 227428 |
+
{
|
| 227429 |
+
"epoch": 804.86,
|
| 227430 |
+
"learning_rate": 8.40352564102564e-06,
|
| 227431 |
+
"loss": 0.2803,
|
| 227432 |
+
"step": 100660
|
| 227433 |
+
},
|
| 227434 |
+
{
|
| 227435 |
+
"epoch": 804.9,
|
| 227436 |
+
"learning_rate": 8.403445512820514e-06,
|
| 227437 |
+
"loss": 0.2624,
|
| 227438 |
+
"step": 100665
|
| 227439 |
+
},
|
| 227440 |
+
{
|
| 227441 |
+
"epoch": 804.94,
|
| 227442 |
+
"learning_rate": 8.403365384615385e-06,
|
| 227443 |
+
"loss": 0.4437,
|
| 227444 |
+
"step": 100670
|
| 227445 |
+
},
|
| 227446 |
+
{
|
| 227447 |
+
"epoch": 804.98,
|
| 227448 |
+
"learning_rate": 8.403285256410257e-06,
|
| 227449 |
+
"loss": 0.5884,
|
| 227450 |
+
"step": 100675
|
| 227451 |
+
},
|
| 227452 |
+
{
|
| 227453 |
+
"epoch": 805.0,
|
| 227454 |
+
"eval_loss": 0.4828292727470398,
|
| 227455 |
+
"eval_runtime": 39.5012,
|
| 227456 |
+
"eval_samples_per_second": 21.088,
|
| 227457 |
+
"eval_steps_per_second": 0.684,
|
| 227458 |
+
"eval_wer": 0.18193024436226524,
|
| 227459 |
+
"step": 100678
|
| 227460 |
+
},
|
| 227461 |
+
{
|
| 227462 |
+
"epoch": 805.02,
|
| 227463 |
+
"learning_rate": 8.403205128205128e-06,
|
| 227464 |
+
"loss": 0.3817,
|
| 227465 |
+
"step": 100680
|
| 227466 |
+
},
|
| 227467 |
+
{
|
| 227468 |
+
"epoch": 805.06,
|
| 227469 |
+
"learning_rate": 8.403125000000001e-06,
|
| 227470 |
+
"loss": 0.2623,
|
| 227471 |
+
"step": 100685
|
| 227472 |
+
},
|
| 227473 |
+
{
|
| 227474 |
+
"epoch": 805.1,
|
| 227475 |
+
"learning_rate": 8.403044871794872e-06,
|
| 227476 |
+
"loss": 0.3245,
|
| 227477 |
+
"step": 100690
|
| 227478 |
+
},
|
| 227479 |
+
{
|
| 227480 |
+
"epoch": 805.14,
|
| 227481 |
+
"learning_rate": 8.402964743589744e-06,
|
| 227482 |
+
"loss": 0.3474,
|
| 227483 |
+
"step": 100695
|
| 227484 |
+
},
|
| 227485 |
+
{
|
| 227486 |
+
"epoch": 805.18,
|
| 227487 |
+
"learning_rate": 8.402884615384615e-06,
|
| 227488 |
+
"loss": 0.6056,
|
| 227489 |
+
"step": 100700
|
| 227490 |
+
},
|
| 227491 |
+
{
|
| 227492 |
+
"epoch": 805.22,
|
| 227493 |
+
"learning_rate": 8.402804487179488e-06,
|
| 227494 |
+
"loss": 1.093,
|
| 227495 |
+
"step": 100705
|
| 227496 |
+
},
|
| 227497 |
+
{
|
| 227498 |
+
"epoch": 805.26,
|
| 227499 |
+
"learning_rate": 8.40272435897436e-06,
|
| 227500 |
+
"loss": 0.3005,
|
| 227501 |
+
"step": 100710
|
| 227502 |
+
},
|
| 227503 |
+
{
|
| 227504 |
+
"epoch": 805.3,
|
| 227505 |
+
"learning_rate": 8.40264423076923e-06,
|
| 227506 |
+
"loss": 0.2933,
|
| 227507 |
+
"step": 100715
|
| 227508 |
+
},
|
| 227509 |
+
{
|
| 227510 |
+
"epoch": 805.34,
|
| 227511 |
+
"learning_rate": 8.402564102564104e-06,
|
| 227512 |
+
"loss": 0.3172,
|
| 227513 |
+
"step": 100720
|
| 227514 |
+
},
|
| 227515 |
+
{
|
| 227516 |
+
"epoch": 805.38,
|
| 227517 |
+
"learning_rate": 8.402483974358975e-06,
|
| 227518 |
+
"loss": 0.6234,
|
| 227519 |
+
"step": 100725
|
| 227520 |
+
},
|
| 227521 |
+
{
|
| 227522 |
+
"epoch": 805.42,
|
| 227523 |
+
"learning_rate": 8.402403846153847e-06,
|
| 227524 |
+
"loss": 0.9828,
|
| 227525 |
+
"step": 100730
|
| 227526 |
+
},
|
| 227527 |
+
{
|
| 227528 |
+
"epoch": 805.46,
|
| 227529 |
+
"learning_rate": 8.402323717948718e-06,
|
| 227530 |
+
"loss": 0.2691,
|
| 227531 |
+
"step": 100735
|
| 227532 |
+
},
|
| 227533 |
+
{
|
| 227534 |
+
"epoch": 805.5,
|
| 227535 |
+
"learning_rate": 8.402243589743591e-06,
|
| 227536 |
+
"loss": 0.2678,
|
| 227537 |
+
"step": 100740
|
| 227538 |
+
},
|
| 227539 |
+
{
|
| 227540 |
+
"epoch": 805.54,
|
| 227541 |
+
"learning_rate": 8.402163461538462e-06,
|
| 227542 |
+
"loss": 0.373,
|
| 227543 |
+
"step": 100745
|
| 227544 |
+
},
|
| 227545 |
+
{
|
| 227546 |
+
"epoch": 805.58,
|
| 227547 |
+
"learning_rate": 8.402083333333334e-06,
|
| 227548 |
+
"loss": 0.6393,
|
| 227549 |
+
"step": 100750
|
| 227550 |
+
},
|
| 227551 |
+
{
|
| 227552 |
+
"epoch": 805.62,
|
| 227553 |
+
"learning_rate": 8.402003205128207e-06,
|
| 227554 |
+
"loss": 1.05,
|
| 227555 |
+
"step": 100755
|
| 227556 |
+
},
|
| 227557 |
+
{
|
| 227558 |
+
"epoch": 805.66,
|
| 227559 |
+
"learning_rate": 8.401923076923078e-06,
|
| 227560 |
+
"loss": 0.2818,
|
| 227561 |
+
"step": 100760
|
| 227562 |
+
},
|
| 227563 |
+
{
|
| 227564 |
+
"epoch": 805.7,
|
| 227565 |
+
"learning_rate": 8.40184294871795e-06,
|
| 227566 |
+
"loss": 0.2837,
|
| 227567 |
+
"step": 100765
|
| 227568 |
+
},
|
| 227569 |
+
{
|
| 227570 |
+
"epoch": 805.74,
|
| 227571 |
+
"learning_rate": 8.40176282051282e-06,
|
| 227572 |
+
"loss": 0.3538,
|
| 227573 |
+
"step": 100770
|
| 227574 |
+
},
|
| 227575 |
+
{
|
| 227576 |
+
"epoch": 805.78,
|
| 227577 |
+
"learning_rate": 8.401682692307694e-06,
|
| 227578 |
+
"loss": 0.6882,
|
| 227579 |
+
"step": 100775
|
| 227580 |
+
},
|
| 227581 |
+
{
|
| 227582 |
+
"epoch": 805.82,
|
| 227583 |
+
"learning_rate": 8.401602564102565e-06,
|
| 227584 |
+
"loss": 1.0649,
|
| 227585 |
+
"step": 100780
|
| 227586 |
+
},
|
| 227587 |
+
{
|
| 227588 |
+
"epoch": 805.86,
|
| 227589 |
+
"learning_rate": 8.401522435897437e-06,
|
| 227590 |
+
"loss": 0.3007,
|
| 227591 |
+
"step": 100785
|
| 227592 |
+
},
|
| 227593 |
+
{
|
| 227594 |
+
"epoch": 805.9,
|
| 227595 |
+
"learning_rate": 8.401442307692308e-06,
|
| 227596 |
+
"loss": 0.2747,
|
| 227597 |
+
"step": 100790
|
| 227598 |
+
},
|
| 227599 |
+
{
|
| 227600 |
+
"epoch": 805.94,
|
| 227601 |
+
"learning_rate": 8.401362179487181e-06,
|
| 227602 |
+
"loss": 0.3714,
|
| 227603 |
+
"step": 100795
|
| 227604 |
+
},
|
| 227605 |
+
{
|
| 227606 |
+
"epoch": 805.98,
|
| 227607 |
+
"learning_rate": 8.40128205128205e-06,
|
| 227608 |
+
"loss": 0.6743,
|
| 227609 |
+
"step": 100800
|
| 227610 |
+
},
|
| 227611 |
+
{
|
| 227612 |
+
"epoch": 806.0,
|
| 227613 |
+
"eval_loss": 0.3387967646121979,
|
| 227614 |
+
"eval_runtime": 37.6542,
|
| 227615 |
+
"eval_samples_per_second": 22.122,
|
| 227616 |
+
"eval_steps_per_second": 0.717,
|
| 227617 |
+
"eval_wer": 0.1787529529672847,
|
| 227618 |
+
"step": 100803
|
| 227619 |
}
|
| 227620 |
],
|
| 227621 |
+
"max_steps": 625000,
|
| 227622 |
"num_train_epochs": 5000,
|
| 227623 |
+
"total_flos": 2.8369231696587882e+20,
|
| 227624 |
"trial_name": null,
|
| 227625 |
"trial_params": null
|
| 227626 |
}
|
model-bin/finetune/base/{checkpoint-100181 → checkpoint-100803}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629982277.5117593/events.out.tfevents.1629982277.8e89bd551565.924.191
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b8f8201e36b96465b01d5500153cb8f7545651bd21e8d14dfef8629e8ff8fbd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629982720.2748618/events.out.tfevents.1629982720.8e89bd551565.924.193
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f58eb466593607f88fbc20cd80971fd4b82adec595fd3ccf03f82caa8424b9f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629983163.766655/events.out.tfevents.1629983163.8e89bd551565.924.195
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1a71e7a584495a0eebdaf30b6769848bd435a4f478d64ffd1a9b5cb6fa0fea24
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629983583.3409672/events.out.tfevents.1629983583.8e89bd551565.924.197
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4517b96f5f4c51c074111033364ca0bf728abbd3357b45c2ce1d6d810e62d98c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629984010.4668894/events.out.tfevents.1629984011.8e89bd551565.924.199
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b01fd4b33222ec986f5a4ee71d255d13ce30cc7f7459ea3b889e8239120ea935
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629982277.8e89bd551565.924.190
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4b4a37efce3069dc789cfae4c31230b910592164552d4b4df6b8a53df5cee096
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629982719.8e89bd551565.924.192
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:72e357b4ada025982b57876e95f967f3fd0a3b06816a77870d840fa199fd2f94
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629983162.8e89bd551565.924.194
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9fc6544c86daf542a634d170faaded33b82d9bba7ec6abb61c5512309f6af999
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629983583.8e89bd551565.924.196
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d07eb5226ea0f1887217412cac12687e3dc2ed3481b49e86c537fd0ff67cc009
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629984010.8e89bd551565.924.198
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca6311d25649e2cb5b41310c198dcc5f849adc12ffdf0844d7a95cdecb20386f
|
| 3 |
+
size 8622
|