"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630130593.3816462/events.out.tfevents.1630130593.86bb0ddabf9b.4092.41 +3 -0
- model-bin/finetune/base/log/1630130973.7216606/events.out.tfevents.1630130973.86bb0ddabf9b.4092.43 +3 -0
- model-bin/finetune/base/log/1630131357.8881845/events.out.tfevents.1630131357.86bb0ddabf9b.4092.45 +3 -0
- model-bin/finetune/base/log/1630131736.953837/events.out.tfevents.1630131736.86bb0ddabf9b.4092.47 +3 -0
- model-bin/finetune/base/log/1630132125.1505942/events.out.tfevents.1630132125.86bb0ddabf9b.4092.49 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630130593.86bb0ddabf9b.4092.40 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630130973.86bb0ddabf9b.4092.42 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630131357.86bb0ddabf9b.4092.44 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630131736.86bb0ddabf9b.4092.46 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630132125.86bb0ddabf9b.4092.48 +3 -0
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/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 sha256:092b700951279c90023bb11c835e0f5f54e15c88589d7fae37929377088a3adf
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/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:c7b657e06bef8e9094828e09219b103fa0568ba25f4c6a79a87c063cf5d83325
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c1c37a1cae1c9c87ffcb6d013494288d4405456b0beb4882c39271505234ed5f
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/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 sha256:3ebeea6250e483ecd40494e28df8405c19ff072bd9538b8863c97fc25c728f7f
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/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 sha256:3eefa8eb85bed9a407060ff02d546b477b03e6766a73c07eae1e5b9ca1444db3
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -252624,11 +252624,806 @@
|
|
| 252624 |
"eval_steps_per_second": 0.764,
|
| 252625 |
"eval_wer": 0.18568594312587258,
|
| 252626 |
"step": 120459
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 252627 |
}
|
| 252628 |
],
|
| 252629 |
-
"max_steps":
|
| 252630 |
"num_train_epochs": 5000,
|
| 252631 |
-
"total_flos": 3.
|
| 252632 |
"trial_name": null,
|
| 252633 |
"trial_params": null
|
| 252634 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 967.9960159362549,
|
| 5 |
+
"global_step": 121084,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 252624 |
"eval_steps_per_second": 0.764,
|
| 252625 |
"eval_wer": 0.18568594312587258,
|
| 252626 |
"step": 120459
|
| 252627 |
+
},
|
| 252628 |
+
{
|
| 252629 |
+
"epoch": 963.01,
|
| 252630 |
+
"learning_rate": 8.08633012820513e-06,
|
| 252631 |
+
"loss": 0.3063,
|
| 252632 |
+
"step": 120460
|
| 252633 |
+
},
|
| 252634 |
+
{
|
| 252635 |
+
"epoch": 963.05,
|
| 252636 |
+
"learning_rate": 8.08625e-06,
|
| 252637 |
+
"loss": 0.3186,
|
| 252638 |
+
"step": 120465
|
| 252639 |
+
},
|
| 252640 |
+
{
|
| 252641 |
+
"epoch": 963.09,
|
| 252642 |
+
"learning_rate": 8.086169871794872e-06,
|
| 252643 |
+
"loss": 0.2936,
|
| 252644 |
+
"step": 120470
|
| 252645 |
+
},
|
| 252646 |
+
{
|
| 252647 |
+
"epoch": 963.13,
|
| 252648 |
+
"learning_rate": 8.086089743589745e-06,
|
| 252649 |
+
"loss": 0.3874,
|
| 252650 |
+
"step": 120475
|
| 252651 |
+
},
|
| 252652 |
+
{
|
| 252653 |
+
"epoch": 963.17,
|
| 252654 |
+
"learning_rate": 8.086009615384616e-06,
|
| 252655 |
+
"loss": 0.538,
|
| 252656 |
+
"step": 120480
|
| 252657 |
+
},
|
| 252658 |
+
{
|
| 252659 |
+
"epoch": 963.21,
|
| 252660 |
+
"learning_rate": 8.085929487179488e-06,
|
| 252661 |
+
"loss": 1.1838,
|
| 252662 |
+
"step": 120485
|
| 252663 |
+
},
|
| 252664 |
+
{
|
| 252665 |
+
"epoch": 963.25,
|
| 252666 |
+
"learning_rate": 8.08584935897436e-06,
|
| 252667 |
+
"loss": 0.2738,
|
| 252668 |
+
"step": 120490
|
| 252669 |
+
},
|
| 252670 |
+
{
|
| 252671 |
+
"epoch": 963.29,
|
| 252672 |
+
"learning_rate": 8.085769230769232e-06,
|
| 252673 |
+
"loss": 0.2775,
|
| 252674 |
+
"step": 120495
|
| 252675 |
+
},
|
| 252676 |
+
{
|
| 252677 |
+
"epoch": 963.33,
|
| 252678 |
+
"learning_rate": 8.085689102564104e-06,
|
| 252679 |
+
"loss": 0.3505,
|
| 252680 |
+
"step": 120500
|
| 252681 |
+
},
|
| 252682 |
+
{
|
| 252683 |
+
"epoch": 963.37,
|
| 252684 |
+
"learning_rate": 8.085608974358975e-06,
|
| 252685 |
+
"loss": 0.5326,
|
| 252686 |
+
"step": 120505
|
| 252687 |
+
},
|
| 252688 |
+
{
|
| 252689 |
+
"epoch": 963.41,
|
| 252690 |
+
"learning_rate": 8.085528846153848e-06,
|
| 252691 |
+
"loss": 1.2263,
|
| 252692 |
+
"step": 120510
|
| 252693 |
+
},
|
| 252694 |
+
{
|
| 252695 |
+
"epoch": 963.45,
|
| 252696 |
+
"learning_rate": 8.085448717948718e-06,
|
| 252697 |
+
"loss": 0.2965,
|
| 252698 |
+
"step": 120515
|
| 252699 |
+
},
|
| 252700 |
+
{
|
| 252701 |
+
"epoch": 963.49,
|
| 252702 |
+
"learning_rate": 8.08536858974359e-06,
|
| 252703 |
+
"loss": 0.275,
|
| 252704 |
+
"step": 120520
|
| 252705 |
+
},
|
| 252706 |
+
{
|
| 252707 |
+
"epoch": 963.53,
|
| 252708 |
+
"learning_rate": 8.085288461538462e-06,
|
| 252709 |
+
"loss": 0.5078,
|
| 252710 |
+
"step": 120525
|
| 252711 |
+
},
|
| 252712 |
+
{
|
| 252713 |
+
"epoch": 963.57,
|
| 252714 |
+
"learning_rate": 8.085208333333333e-06,
|
| 252715 |
+
"loss": 0.5536,
|
| 252716 |
+
"step": 120530
|
| 252717 |
+
},
|
| 252718 |
+
{
|
| 252719 |
+
"epoch": 963.61,
|
| 252720 |
+
"learning_rate": 8.085128205128205e-06,
|
| 252721 |
+
"loss": 1.1345,
|
| 252722 |
+
"step": 120535
|
| 252723 |
+
},
|
| 252724 |
+
{
|
| 252725 |
+
"epoch": 963.65,
|
| 252726 |
+
"learning_rate": 8.085048076923078e-06,
|
| 252727 |
+
"loss": 0.3,
|
| 252728 |
+
"step": 120540
|
| 252729 |
+
},
|
| 252730 |
+
{
|
| 252731 |
+
"epoch": 963.69,
|
| 252732 |
+
"learning_rate": 8.08496794871795e-06,
|
| 252733 |
+
"loss": 0.2699,
|
| 252734 |
+
"step": 120545
|
| 252735 |
+
},
|
| 252736 |
+
{
|
| 252737 |
+
"epoch": 963.73,
|
| 252738 |
+
"learning_rate": 8.08488782051282e-06,
|
| 252739 |
+
"loss": 0.3185,
|
| 252740 |
+
"step": 120550
|
| 252741 |
+
},
|
| 252742 |
+
{
|
| 252743 |
+
"epoch": 963.77,
|
| 252744 |
+
"learning_rate": 8.084807692307694e-06,
|
| 252745 |
+
"loss": 0.5594,
|
| 252746 |
+
"step": 120555
|
| 252747 |
+
},
|
| 252748 |
+
{
|
| 252749 |
+
"epoch": 963.81,
|
| 252750 |
+
"learning_rate": 8.084727564102565e-06,
|
| 252751 |
+
"loss": 1.1854,
|
| 252752 |
+
"step": 120560
|
| 252753 |
+
},
|
| 252754 |
+
{
|
| 252755 |
+
"epoch": 963.85,
|
| 252756 |
+
"learning_rate": 8.084647435897436e-06,
|
| 252757 |
+
"loss": 0.3462,
|
| 252758 |
+
"step": 120565
|
| 252759 |
+
},
|
| 252760 |
+
{
|
| 252761 |
+
"epoch": 963.89,
|
| 252762 |
+
"learning_rate": 8.084567307692308e-06,
|
| 252763 |
+
"loss": 0.2817,
|
| 252764 |
+
"step": 120570
|
| 252765 |
+
},
|
| 252766 |
+
{
|
| 252767 |
+
"epoch": 963.93,
|
| 252768 |
+
"learning_rate": 8.08448717948718e-06,
|
| 252769 |
+
"loss": 0.3001,
|
| 252770 |
+
"step": 120575
|
| 252771 |
+
},
|
| 252772 |
+
{
|
| 252773 |
+
"epoch": 963.97,
|
| 252774 |
+
"learning_rate": 8.084407051282052e-06,
|
| 252775 |
+
"loss": 0.5627,
|
| 252776 |
+
"step": 120580
|
| 252777 |
+
},
|
| 252778 |
+
{
|
| 252779 |
+
"epoch": 964.0,
|
| 252780 |
+
"eval_loss": 0.3748638331890106,
|
| 252781 |
+
"eval_runtime": 35.9106,
|
| 252782 |
+
"eval_samples_per_second": 23.447,
|
| 252783 |
+
"eval_steps_per_second": 0.752,
|
| 252784 |
+
"eval_wer": 0.18153778365278694,
|
| 252785 |
+
"step": 120584
|
| 252786 |
+
},
|
| 252787 |
+
{
|
| 252788 |
+
"epoch": 964.01,
|
| 252789 |
+
"learning_rate": 8.084326923076923e-06,
|
| 252790 |
+
"loss": 0.4757,
|
| 252791 |
+
"step": 120585
|
| 252792 |
+
},
|
| 252793 |
+
{
|
| 252794 |
+
"epoch": 964.05,
|
| 252795 |
+
"learning_rate": 8.084246794871795e-06,
|
| 252796 |
+
"loss": 0.2449,
|
| 252797 |
+
"step": 120590
|
| 252798 |
+
},
|
| 252799 |
+
{
|
| 252800 |
+
"epoch": 964.09,
|
| 252801 |
+
"learning_rate": 8.084166666666668e-06,
|
| 252802 |
+
"loss": 0.2717,
|
| 252803 |
+
"step": 120595
|
| 252804 |
+
},
|
| 252805 |
+
{
|
| 252806 |
+
"epoch": 964.13,
|
| 252807 |
+
"learning_rate": 8.08408653846154e-06,
|
| 252808 |
+
"loss": 0.3517,
|
| 252809 |
+
"step": 120600
|
| 252810 |
+
},
|
| 252811 |
+
{
|
| 252812 |
+
"epoch": 964.17,
|
| 252813 |
+
"learning_rate": 8.08400641025641e-06,
|
| 252814 |
+
"loss": 0.5087,
|
| 252815 |
+
"step": 120605
|
| 252816 |
+
},
|
| 252817 |
+
{
|
| 252818 |
+
"epoch": 964.21,
|
| 252819 |
+
"learning_rate": 8.083926282051284e-06,
|
| 252820 |
+
"loss": 1.0103,
|
| 252821 |
+
"step": 120610
|
| 252822 |
+
},
|
| 252823 |
+
{
|
| 252824 |
+
"epoch": 964.25,
|
| 252825 |
+
"learning_rate": 8.083846153846155e-06,
|
| 252826 |
+
"loss": 0.3475,
|
| 252827 |
+
"step": 120615
|
| 252828 |
+
},
|
| 252829 |
+
{
|
| 252830 |
+
"epoch": 964.29,
|
| 252831 |
+
"learning_rate": 8.083766025641026e-06,
|
| 252832 |
+
"loss": 0.2822,
|
| 252833 |
+
"step": 120620
|
| 252834 |
+
},
|
| 252835 |
+
{
|
| 252836 |
+
"epoch": 964.33,
|
| 252837 |
+
"learning_rate": 8.083685897435898e-06,
|
| 252838 |
+
"loss": 0.2866,
|
| 252839 |
+
"step": 120625
|
| 252840 |
+
},
|
| 252841 |
+
{
|
| 252842 |
+
"epoch": 964.37,
|
| 252843 |
+
"learning_rate": 8.08360576923077e-06,
|
| 252844 |
+
"loss": 0.4979,
|
| 252845 |
+
"step": 120630
|
| 252846 |
+
},
|
| 252847 |
+
{
|
| 252848 |
+
"epoch": 964.41,
|
| 252849 |
+
"learning_rate": 8.08352564102564e-06,
|
| 252850 |
+
"loss": 1.1757,
|
| 252851 |
+
"step": 120635
|
| 252852 |
+
},
|
| 252853 |
+
{
|
| 252854 |
+
"epoch": 964.45,
|
| 252855 |
+
"learning_rate": 8.083445512820514e-06,
|
| 252856 |
+
"loss": 0.3024,
|
| 252857 |
+
"step": 120640
|
| 252858 |
+
},
|
| 252859 |
+
{
|
| 252860 |
+
"epoch": 964.49,
|
| 252861 |
+
"learning_rate": 8.083365384615385e-06,
|
| 252862 |
+
"loss": 0.2628,
|
| 252863 |
+
"step": 120645
|
| 252864 |
+
},
|
| 252865 |
+
{
|
| 252866 |
+
"epoch": 964.53,
|
| 252867 |
+
"learning_rate": 8.083285256410256e-06,
|
| 252868 |
+
"loss": 0.3263,
|
| 252869 |
+
"step": 120650
|
| 252870 |
+
},
|
| 252871 |
+
{
|
| 252872 |
+
"epoch": 964.57,
|
| 252873 |
+
"learning_rate": 8.08320512820513e-06,
|
| 252874 |
+
"loss": 0.5977,
|
| 252875 |
+
"step": 120655
|
| 252876 |
+
},
|
| 252877 |
+
{
|
| 252878 |
+
"epoch": 964.61,
|
| 252879 |
+
"learning_rate": 8.083125e-06,
|
| 252880 |
+
"loss": 1.1081,
|
| 252881 |
+
"step": 120660
|
| 252882 |
+
},
|
| 252883 |
+
{
|
| 252884 |
+
"epoch": 964.65,
|
| 252885 |
+
"learning_rate": 8.083044871794874e-06,
|
| 252886 |
+
"loss": 0.2912,
|
| 252887 |
+
"step": 120665
|
| 252888 |
+
},
|
| 252889 |
+
{
|
| 252890 |
+
"epoch": 964.69,
|
| 252891 |
+
"learning_rate": 8.082964743589743e-06,
|
| 252892 |
+
"loss": 0.2956,
|
| 252893 |
+
"step": 120670
|
| 252894 |
+
},
|
| 252895 |
+
{
|
| 252896 |
+
"epoch": 964.73,
|
| 252897 |
+
"learning_rate": 8.082884615384616e-06,
|
| 252898 |
+
"loss": 0.255,
|
| 252899 |
+
"step": 120675
|
| 252900 |
+
},
|
| 252901 |
+
{
|
| 252902 |
+
"epoch": 964.77,
|
| 252903 |
+
"learning_rate": 8.082804487179488e-06,
|
| 252904 |
+
"loss": 0.4588,
|
| 252905 |
+
"step": 120680
|
| 252906 |
+
},
|
| 252907 |
+
{
|
| 252908 |
+
"epoch": 964.81,
|
| 252909 |
+
"learning_rate": 8.082724358974359e-06,
|
| 252910 |
+
"loss": 1.0802,
|
| 252911 |
+
"step": 120685
|
| 252912 |
+
},
|
| 252913 |
+
{
|
| 252914 |
+
"epoch": 964.85,
|
| 252915 |
+
"learning_rate": 8.08264423076923e-06,
|
| 252916 |
+
"loss": 0.2736,
|
| 252917 |
+
"step": 120690
|
| 252918 |
+
},
|
| 252919 |
+
{
|
| 252920 |
+
"epoch": 964.89,
|
| 252921 |
+
"learning_rate": 8.082564102564104e-06,
|
| 252922 |
+
"loss": 0.2829,
|
| 252923 |
+
"step": 120695
|
| 252924 |
+
},
|
| 252925 |
+
{
|
| 252926 |
+
"epoch": 964.93,
|
| 252927 |
+
"learning_rate": 8.082483974358975e-06,
|
| 252928 |
+
"loss": 0.3132,
|
| 252929 |
+
"step": 120700
|
| 252930 |
+
},
|
| 252931 |
+
{
|
| 252932 |
+
"epoch": 964.97,
|
| 252933 |
+
"learning_rate": 8.082403846153846e-06,
|
| 252934 |
+
"loss": 0.5069,
|
| 252935 |
+
"step": 120705
|
| 252936 |
+
},
|
| 252937 |
+
{
|
| 252938 |
+
"epoch": 965.0,
|
| 252939 |
+
"eval_loss": 0.3655751347541809,
|
| 252940 |
+
"eval_runtime": 37.0544,
|
| 252941 |
+
"eval_samples_per_second": 22.723,
|
| 252942 |
+
"eval_steps_per_second": 0.729,
|
| 252943 |
+
"eval_wer": 0.18698390482855143,
|
| 252944 |
+
"step": 120709
|
| 252945 |
+
},
|
| 252946 |
+
{
|
| 252947 |
+
"epoch": 965.01,
|
| 252948 |
+
"learning_rate": 8.08232371794872e-06,
|
| 252949 |
+
"loss": 0.4841,
|
| 252950 |
+
"step": 120710
|
| 252951 |
+
},
|
| 252952 |
+
{
|
| 252953 |
+
"epoch": 965.05,
|
| 252954 |
+
"learning_rate": 8.08224358974359e-06,
|
| 252955 |
+
"loss": 0.2581,
|
| 252956 |
+
"step": 120715
|
| 252957 |
+
},
|
| 252958 |
+
{
|
| 252959 |
+
"epoch": 965.09,
|
| 252960 |
+
"learning_rate": 8.082163461538462e-06,
|
| 252961 |
+
"loss": 0.269,
|
| 252962 |
+
"step": 120720
|
| 252963 |
+
},
|
| 252964 |
+
{
|
| 252965 |
+
"epoch": 965.13,
|
| 252966 |
+
"learning_rate": 8.082083333333333e-06,
|
| 252967 |
+
"loss": 0.3394,
|
| 252968 |
+
"step": 120725
|
| 252969 |
+
},
|
| 252970 |
+
{
|
| 252971 |
+
"epoch": 965.17,
|
| 252972 |
+
"learning_rate": 8.082003205128206e-06,
|
| 252973 |
+
"loss": 0.5367,
|
| 252974 |
+
"step": 120730
|
| 252975 |
+
},
|
| 252976 |
+
{
|
| 252977 |
+
"epoch": 965.21,
|
| 252978 |
+
"learning_rate": 8.081923076923078e-06,
|
| 252979 |
+
"loss": 1.1198,
|
| 252980 |
+
"step": 120735
|
| 252981 |
+
},
|
| 252982 |
+
{
|
| 252983 |
+
"epoch": 965.25,
|
| 252984 |
+
"learning_rate": 8.08184294871795e-06,
|
| 252985 |
+
"loss": 0.2783,
|
| 252986 |
+
"step": 120740
|
| 252987 |
+
},
|
| 252988 |
+
{
|
| 252989 |
+
"epoch": 965.29,
|
| 252990 |
+
"learning_rate": 8.08176282051282e-06,
|
| 252991 |
+
"loss": 0.2792,
|
| 252992 |
+
"step": 120745
|
| 252993 |
+
},
|
| 252994 |
+
{
|
| 252995 |
+
"epoch": 965.33,
|
| 252996 |
+
"learning_rate": 8.081682692307694e-06,
|
| 252997 |
+
"loss": 0.2745,
|
| 252998 |
+
"step": 120750
|
| 252999 |
+
},
|
| 253000 |
+
{
|
| 253001 |
+
"epoch": 965.37,
|
| 253002 |
+
"learning_rate": 8.081602564102565e-06,
|
| 253003 |
+
"loss": 0.5469,
|
| 253004 |
+
"step": 120755
|
| 253005 |
+
},
|
| 253006 |
+
{
|
| 253007 |
+
"epoch": 965.41,
|
| 253008 |
+
"learning_rate": 8.081522435897436e-06,
|
| 253009 |
+
"loss": 1.0953,
|
| 253010 |
+
"step": 120760
|
| 253011 |
+
},
|
| 253012 |
+
{
|
| 253013 |
+
"epoch": 965.45,
|
| 253014 |
+
"learning_rate": 8.08144230769231e-06,
|
| 253015 |
+
"loss": 0.285,
|
| 253016 |
+
"step": 120765
|
| 253017 |
+
},
|
| 253018 |
+
{
|
| 253019 |
+
"epoch": 965.49,
|
| 253020 |
+
"learning_rate": 8.08136217948718e-06,
|
| 253021 |
+
"loss": 0.2642,
|
| 253022 |
+
"step": 120770
|
| 253023 |
+
},
|
| 253024 |
+
{
|
| 253025 |
+
"epoch": 965.53,
|
| 253026 |
+
"learning_rate": 8.081282051282052e-06,
|
| 253027 |
+
"loss": 0.3305,
|
| 253028 |
+
"step": 120775
|
| 253029 |
+
},
|
| 253030 |
+
{
|
| 253031 |
+
"epoch": 965.57,
|
| 253032 |
+
"learning_rate": 8.081201923076923e-06,
|
| 253033 |
+
"loss": 0.5227,
|
| 253034 |
+
"step": 120780
|
| 253035 |
+
},
|
| 253036 |
+
{
|
| 253037 |
+
"epoch": 965.61,
|
| 253038 |
+
"learning_rate": 8.081121794871796e-06,
|
| 253039 |
+
"loss": 1.1465,
|
| 253040 |
+
"step": 120785
|
| 253041 |
+
},
|
| 253042 |
+
{
|
| 253043 |
+
"epoch": 965.65,
|
| 253044 |
+
"learning_rate": 8.081041666666666e-06,
|
| 253045 |
+
"loss": 0.2476,
|
| 253046 |
+
"step": 120790
|
| 253047 |
+
},
|
| 253048 |
+
{
|
| 253049 |
+
"epoch": 965.69,
|
| 253050 |
+
"learning_rate": 8.08096153846154e-06,
|
| 253051 |
+
"loss": 0.2343,
|
| 253052 |
+
"step": 120795
|
| 253053 |
+
},
|
| 253054 |
+
{
|
| 253055 |
+
"epoch": 965.73,
|
| 253056 |
+
"learning_rate": 8.08088141025641e-06,
|
| 253057 |
+
"loss": 0.3151,
|
| 253058 |
+
"step": 120800
|
| 253059 |
+
},
|
| 253060 |
+
{
|
| 253061 |
+
"epoch": 965.77,
|
| 253062 |
+
"learning_rate": 8.080801282051282e-06,
|
| 253063 |
+
"loss": 0.5144,
|
| 253064 |
+
"step": 120805
|
| 253065 |
+
},
|
| 253066 |
+
{
|
| 253067 |
+
"epoch": 965.81,
|
| 253068 |
+
"learning_rate": 8.080721153846155e-06,
|
| 253069 |
+
"loss": 1.0767,
|
| 253070 |
+
"step": 120810
|
| 253071 |
+
},
|
| 253072 |
+
{
|
| 253073 |
+
"epoch": 965.85,
|
| 253074 |
+
"learning_rate": 8.080641025641026e-06,
|
| 253075 |
+
"loss": 0.3412,
|
| 253076 |
+
"step": 120815
|
| 253077 |
+
},
|
| 253078 |
+
{
|
| 253079 |
+
"epoch": 965.89,
|
| 253080 |
+
"learning_rate": 8.080560897435898e-06,
|
| 253081 |
+
"loss": 0.3241,
|
| 253082 |
+
"step": 120820
|
| 253083 |
+
},
|
| 253084 |
+
{
|
| 253085 |
+
"epoch": 965.93,
|
| 253086 |
+
"learning_rate": 8.080480769230769e-06,
|
| 253087 |
+
"loss": 0.363,
|
| 253088 |
+
"step": 120825
|
| 253089 |
+
},
|
| 253090 |
+
{
|
| 253091 |
+
"epoch": 965.97,
|
| 253092 |
+
"learning_rate": 8.080400641025642e-06,
|
| 253093 |
+
"loss": 0.5905,
|
| 253094 |
+
"step": 120830
|
| 253095 |
+
},
|
| 253096 |
+
{
|
| 253097 |
+
"epoch": 966.0,
|
| 253098 |
+
"eval_loss": 0.38227519392967224,
|
| 253099 |
+
"eval_runtime": 35.5959,
|
| 253100 |
+
"eval_samples_per_second": 23.654,
|
| 253101 |
+
"eval_steps_per_second": 0.759,
|
| 253102 |
+
"eval_wer": 0.18383645955451347,
|
| 253103 |
+
"step": 120834
|
| 253104 |
+
},
|
| 253105 |
+
{
|
| 253106 |
+
"epoch": 966.01,
|
| 253107 |
+
"learning_rate": 8.080320512820513e-06,
|
| 253108 |
+
"loss": 0.3745,
|
| 253109 |
+
"step": 120835
|
| 253110 |
+
},
|
| 253111 |
+
{
|
| 253112 |
+
"epoch": 966.05,
|
| 253113 |
+
"learning_rate": 8.080240384615385e-06,
|
| 253114 |
+
"loss": 0.2932,
|
| 253115 |
+
"step": 120840
|
| 253116 |
+
},
|
| 253117 |
+
{
|
| 253118 |
+
"epoch": 966.09,
|
| 253119 |
+
"learning_rate": 8.080160256410256e-06,
|
| 253120 |
+
"loss": 0.3262,
|
| 253121 |
+
"step": 120845
|
| 253122 |
+
},
|
| 253123 |
+
{
|
| 253124 |
+
"epoch": 966.13,
|
| 253125 |
+
"learning_rate": 8.08008012820513e-06,
|
| 253126 |
+
"loss": 0.3332,
|
| 253127 |
+
"step": 120850
|
| 253128 |
+
},
|
| 253129 |
+
{
|
| 253130 |
+
"epoch": 966.17,
|
| 253131 |
+
"learning_rate": 8.08e-06,
|
| 253132 |
+
"loss": 0.5078,
|
| 253133 |
+
"step": 120855
|
| 253134 |
+
},
|
| 253135 |
+
{
|
| 253136 |
+
"epoch": 966.21,
|
| 253137 |
+
"learning_rate": 8.079919871794872e-06,
|
| 253138 |
+
"loss": 1.1597,
|
| 253139 |
+
"step": 120860
|
| 253140 |
+
},
|
| 253141 |
+
{
|
| 253142 |
+
"epoch": 966.25,
|
| 253143 |
+
"learning_rate": 8.079839743589745e-06,
|
| 253144 |
+
"loss": 0.2896,
|
| 253145 |
+
"step": 120865
|
| 253146 |
+
},
|
| 253147 |
+
{
|
| 253148 |
+
"epoch": 966.29,
|
| 253149 |
+
"learning_rate": 8.079759615384616e-06,
|
| 253150 |
+
"loss": 0.2589,
|
| 253151 |
+
"step": 120870
|
| 253152 |
+
},
|
| 253153 |
+
{
|
| 253154 |
+
"epoch": 966.33,
|
| 253155 |
+
"learning_rate": 8.079679487179488e-06,
|
| 253156 |
+
"loss": 0.3202,
|
| 253157 |
+
"step": 120875
|
| 253158 |
+
},
|
| 253159 |
+
{
|
| 253160 |
+
"epoch": 966.37,
|
| 253161 |
+
"learning_rate": 8.079599358974359e-06,
|
| 253162 |
+
"loss": 0.4251,
|
| 253163 |
+
"step": 120880
|
| 253164 |
+
},
|
| 253165 |
+
{
|
| 253166 |
+
"epoch": 966.41,
|
| 253167 |
+
"learning_rate": 8.079519230769232e-06,
|
| 253168 |
+
"loss": 1.1363,
|
| 253169 |
+
"step": 120885
|
| 253170 |
+
},
|
| 253171 |
+
{
|
| 253172 |
+
"epoch": 966.45,
|
| 253173 |
+
"learning_rate": 8.079439102564103e-06,
|
| 253174 |
+
"loss": 0.3064,
|
| 253175 |
+
"step": 120890
|
| 253176 |
+
},
|
| 253177 |
+
{
|
| 253178 |
+
"epoch": 966.49,
|
| 253179 |
+
"learning_rate": 8.079358974358975e-06,
|
| 253180 |
+
"loss": 0.2289,
|
| 253181 |
+
"step": 120895
|
| 253182 |
+
},
|
| 253183 |
+
{
|
| 253184 |
+
"epoch": 966.53,
|
| 253185 |
+
"learning_rate": 8.079278846153846e-06,
|
| 253186 |
+
"loss": 0.3267,
|
| 253187 |
+
"step": 120900
|
| 253188 |
+
},
|
| 253189 |
+
{
|
| 253190 |
+
"epoch": 966.57,
|
| 253191 |
+
"learning_rate": 8.07919871794872e-06,
|
| 253192 |
+
"loss": 0.5204,
|
| 253193 |
+
"step": 120905
|
| 253194 |
+
},
|
| 253195 |
+
{
|
| 253196 |
+
"epoch": 966.61,
|
| 253197 |
+
"learning_rate": 8.07911858974359e-06,
|
| 253198 |
+
"loss": 1.2683,
|
| 253199 |
+
"step": 120910
|
| 253200 |
+
},
|
| 253201 |
+
{
|
| 253202 |
+
"epoch": 966.65,
|
| 253203 |
+
"learning_rate": 8.079038461538462e-06,
|
| 253204 |
+
"loss": 0.2852,
|
| 253205 |
+
"step": 120915
|
| 253206 |
+
},
|
| 253207 |
+
{
|
| 253208 |
+
"epoch": 966.69,
|
| 253209 |
+
"learning_rate": 8.078958333333335e-06,
|
| 253210 |
+
"loss": 0.2778,
|
| 253211 |
+
"step": 120920
|
| 253212 |
+
},
|
| 253213 |
+
{
|
| 253214 |
+
"epoch": 966.73,
|
| 253215 |
+
"learning_rate": 8.078878205128206e-06,
|
| 253216 |
+
"loss": 0.3523,
|
| 253217 |
+
"step": 120925
|
| 253218 |
+
},
|
| 253219 |
+
{
|
| 253220 |
+
"epoch": 966.77,
|
| 253221 |
+
"learning_rate": 8.078798076923078e-06,
|
| 253222 |
+
"loss": 0.5476,
|
| 253223 |
+
"step": 120930
|
| 253224 |
+
},
|
| 253225 |
+
{
|
| 253226 |
+
"epoch": 966.81,
|
| 253227 |
+
"learning_rate": 8.078717948717949e-06,
|
| 253228 |
+
"loss": 0.9729,
|
| 253229 |
+
"step": 120935
|
| 253230 |
+
},
|
| 253231 |
+
{
|
| 253232 |
+
"epoch": 966.85,
|
| 253233 |
+
"learning_rate": 8.078637820512822e-06,
|
| 253234 |
+
"loss": 0.5187,
|
| 253235 |
+
"step": 120940
|
| 253236 |
+
},
|
| 253237 |
+
{
|
| 253238 |
+
"epoch": 966.89,
|
| 253239 |
+
"learning_rate": 8.078557692307692e-06,
|
| 253240 |
+
"loss": 0.3767,
|
| 253241 |
+
"step": 120945
|
| 253242 |
+
},
|
| 253243 |
+
{
|
| 253244 |
+
"epoch": 966.93,
|
| 253245 |
+
"learning_rate": 8.078477564102565e-06,
|
| 253246 |
+
"loss": 0.3119,
|
| 253247 |
+
"step": 120950
|
| 253248 |
+
},
|
| 253249 |
+
{
|
| 253250 |
+
"epoch": 966.97,
|
| 253251 |
+
"learning_rate": 8.078397435897438e-06,
|
| 253252 |
+
"loss": 0.5822,
|
| 253253 |
+
"step": 120955
|
| 253254 |
+
},
|
| 253255 |
+
{
|
| 253256 |
+
"epoch": 967.0,
|
| 253257 |
+
"eval_loss": 0.39696982502937317,
|
| 253258 |
+
"eval_runtime": 35.5761,
|
| 253259 |
+
"eval_samples_per_second": 23.639,
|
| 253260 |
+
"eval_steps_per_second": 0.759,
|
| 253261 |
+
"eval_wer": 0.17967772790817452,
|
| 253262 |
+
"step": 120959
|
| 253263 |
+
},
|
| 253264 |
+
{
|
| 253265 |
+
"epoch": 967.01,
|
| 253266 |
+
"learning_rate": 8.078317307692308e-06,
|
| 253267 |
+
"loss": 0.3628,
|
| 253268 |
+
"step": 120960
|
| 253269 |
+
},
|
| 253270 |
+
{
|
| 253271 |
+
"epoch": 967.05,
|
| 253272 |
+
"learning_rate": 8.07823717948718e-06,
|
| 253273 |
+
"loss": 0.2908,
|
| 253274 |
+
"step": 120965
|
| 253275 |
+
},
|
| 253276 |
+
{
|
| 253277 |
+
"epoch": 967.09,
|
| 253278 |
+
"learning_rate": 8.078157051282052e-06,
|
| 253279 |
+
"loss": 0.2979,
|
| 253280 |
+
"step": 120970
|
| 253281 |
+
},
|
| 253282 |
+
{
|
| 253283 |
+
"epoch": 967.13,
|
| 253284 |
+
"learning_rate": 8.078076923076923e-06,
|
| 253285 |
+
"loss": 0.3033,
|
| 253286 |
+
"step": 120975
|
| 253287 |
+
},
|
| 253288 |
+
{
|
| 253289 |
+
"epoch": 967.17,
|
| 253290 |
+
"learning_rate": 8.077996794871795e-06,
|
| 253291 |
+
"loss": 0.6087,
|
| 253292 |
+
"step": 120980
|
| 253293 |
+
},
|
| 253294 |
+
{
|
| 253295 |
+
"epoch": 967.21,
|
| 253296 |
+
"learning_rate": 8.077916666666668e-06,
|
| 253297 |
+
"loss": 1.1057,
|
| 253298 |
+
"step": 120985
|
| 253299 |
+
},
|
| 253300 |
+
{
|
| 253301 |
+
"epoch": 967.25,
|
| 253302 |
+
"learning_rate": 8.077836538461539e-06,
|
| 253303 |
+
"loss": 0.2821,
|
| 253304 |
+
"step": 120990
|
| 253305 |
+
},
|
| 253306 |
+
{
|
| 253307 |
+
"epoch": 967.29,
|
| 253308 |
+
"learning_rate": 8.07775641025641e-06,
|
| 253309 |
+
"loss": 0.2461,
|
| 253310 |
+
"step": 120995
|
| 253311 |
+
},
|
| 253312 |
+
{
|
| 253313 |
+
"epoch": 967.33,
|
| 253314 |
+
"learning_rate": 8.077676282051282e-06,
|
| 253315 |
+
"loss": 0.5646,
|
| 253316 |
+
"step": 121000
|
| 253317 |
+
},
|
| 253318 |
+
{
|
| 253319 |
+
"epoch": 967.37,
|
| 253320 |
+
"learning_rate": 8.077596153846155e-06,
|
| 253321 |
+
"loss": 0.563,
|
| 253322 |
+
"step": 121005
|
| 253323 |
+
},
|
| 253324 |
+
{
|
| 253325 |
+
"epoch": 967.41,
|
| 253326 |
+
"learning_rate": 8.077516025641026e-06,
|
| 253327 |
+
"loss": 0.9909,
|
| 253328 |
+
"step": 121010
|
| 253329 |
+
},
|
| 253330 |
+
{
|
| 253331 |
+
"epoch": 967.45,
|
| 253332 |
+
"learning_rate": 8.077435897435898e-06,
|
| 253333 |
+
"loss": 0.2902,
|
| 253334 |
+
"step": 121015
|
| 253335 |
+
},
|
| 253336 |
+
{
|
| 253337 |
+
"epoch": 967.49,
|
| 253338 |
+
"learning_rate": 8.07735576923077e-06,
|
| 253339 |
+
"loss": 0.3619,
|
| 253340 |
+
"step": 121020
|
| 253341 |
+
},
|
| 253342 |
+
{
|
| 253343 |
+
"epoch": 967.53,
|
| 253344 |
+
"learning_rate": 8.077275641025642e-06,
|
| 253345 |
+
"loss": 0.3563,
|
| 253346 |
+
"step": 121025
|
| 253347 |
+
},
|
| 253348 |
+
{
|
| 253349 |
+
"epoch": 967.57,
|
| 253350 |
+
"learning_rate": 8.077195512820513e-06,
|
| 253351 |
+
"loss": 0.5795,
|
| 253352 |
+
"step": 121030
|
| 253353 |
+
},
|
| 253354 |
+
{
|
| 253355 |
+
"epoch": 967.61,
|
| 253356 |
+
"learning_rate": 8.077115384615385e-06,
|
| 253357 |
+
"loss": 1.1314,
|
| 253358 |
+
"step": 121035
|
| 253359 |
+
},
|
| 253360 |
+
{
|
| 253361 |
+
"epoch": 967.65,
|
| 253362 |
+
"learning_rate": 8.077035256410258e-06,
|
| 253363 |
+
"loss": 0.3236,
|
| 253364 |
+
"step": 121040
|
| 253365 |
+
},
|
| 253366 |
+
{
|
| 253367 |
+
"epoch": 967.69,
|
| 253368 |
+
"learning_rate": 8.076955128205129e-06,
|
| 253369 |
+
"loss": 0.2958,
|
| 253370 |
+
"step": 121045
|
| 253371 |
+
},
|
| 253372 |
+
{
|
| 253373 |
+
"epoch": 967.73,
|
| 253374 |
+
"learning_rate": 8.076875e-06,
|
| 253375 |
+
"loss": 0.2668,
|
| 253376 |
+
"step": 121050
|
| 253377 |
+
},
|
| 253378 |
+
{
|
| 253379 |
+
"epoch": 967.76,
|
| 253380 |
+
"learning_rate": 8.076794871794874e-06,
|
| 253381 |
+
"loss": 0.505,
|
| 253382 |
+
"step": 121055
|
| 253383 |
+
},
|
| 253384 |
+
{
|
| 253385 |
+
"epoch": 967.8,
|
| 253386 |
+
"learning_rate": 8.076714743589745e-06,
|
| 253387 |
+
"loss": 1.1167,
|
| 253388 |
+
"step": 121060
|
| 253389 |
+
},
|
| 253390 |
+
{
|
| 253391 |
+
"epoch": 967.84,
|
| 253392 |
+
"learning_rate": 8.076634615384616e-06,
|
| 253393 |
+
"loss": 0.2652,
|
| 253394 |
+
"step": 121065
|
| 253395 |
+
},
|
| 253396 |
+
{
|
| 253397 |
+
"epoch": 967.88,
|
| 253398 |
+
"learning_rate": 8.076554487179488e-06,
|
| 253399 |
+
"loss": 0.2603,
|
| 253400 |
+
"step": 121070
|
| 253401 |
+
},
|
| 253402 |
+
{
|
| 253403 |
+
"epoch": 967.92,
|
| 253404 |
+
"learning_rate": 8.07647435897436e-06,
|
| 253405 |
+
"loss": 0.3676,
|
| 253406 |
+
"step": 121075
|
| 253407 |
+
},
|
| 253408 |
+
{
|
| 253409 |
+
"epoch": 967.96,
|
| 253410 |
+
"learning_rate": 8.07639423076923e-06,
|
| 253411 |
+
"loss": 0.5594,
|
| 253412 |
+
"step": 121080
|
| 253413 |
+
},
|
| 253414 |
+
{
|
| 253415 |
+
"epoch": 968.0,
|
| 253416 |
+
"eval_loss": 0.4073833227157593,
|
| 253417 |
+
"eval_runtime": 35.3932,
|
| 253418 |
+
"eval_samples_per_second": 23.762,
|
| 253419 |
+
"eval_steps_per_second": 0.763,
|
| 253420 |
+
"eval_wer": 0.18734491315136476,
|
| 253421 |
+
"step": 121084
|
| 253422 |
}
|
| 253423 |
],
|
| 253424 |
+
"max_steps": 625000,
|
| 253425 |
"num_train_epochs": 5000,
|
| 253426 |
+
"total_flos": 3.407525538155879e+20,
|
| 253427 |
"trial_name": null,
|
| 253428 |
"trial_params": null
|
| 253429 |
}
|
model-bin/finetune/base/{checkpoint-120459 → checkpoint-121084}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630130593.3816462/events.out.tfevents.1630130593.86bb0ddabf9b.4092.41
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dc0e8a00a2ef51be7fb8787b9cb0133f9e18b8ad343f39d8f2c93b28efaac501
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630130973.7216606/events.out.tfevents.1630130973.86bb0ddabf9b.4092.43
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eeca7e6fd5d53b7cdd7d5863d9dbb703f942bec14f7ee8859b7efcd89baef2a5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630131357.8881845/events.out.tfevents.1630131357.86bb0ddabf9b.4092.45
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a1282b30922d1aa33658b6b050f893881ddbbc23b21a48855cc34fb6341277c3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630131736.953837/events.out.tfevents.1630131736.86bb0ddabf9b.4092.47
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8aba7cc3ad6a97807502e9d8fc7b0c7c70e15f4979a3ebc86f63d131e1e13e6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630132125.1505942/events.out.tfevents.1630132125.86bb0ddabf9b.4092.49
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3ffc4fb9d8b30d998c5540f0504c9aef5a93a60eaf845b88aa54b819568ab5e3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630130593.86bb0ddabf9b.4092.40
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3a2d0088ebc8ac48a43f684f224120bf5b960fc7ef754c5c8b2abd2d45fdca2c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630130973.86bb0ddabf9b.4092.42
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:475f78cc7c7aa32efa9485e9be0866cc8714e715d64fccdddea1d93322a47379
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630131357.86bb0ddabf9b.4092.44
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2d7e8e554fbf5a56b03696543205a58d3a7beefced2144c10c09b66c6e8d8f5c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630131736.86bb0ddabf9b.4092.46
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:37ea119a2e37164b37951258112f60443eabd98b0f8f91628b794e2ccb6c7056
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630132125.86bb0ddabf9b.4092.48
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49550e53dd6617feb12ed6a891c0dd77d7f2436ba592cf395b0b41130e854f9e
|
| 3 |
+
size 8622
|