"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629984472.6452265/events.out.tfevents.1629984472.8e89bd551565.924.201 +3 -0
- model-bin/finetune/base/log/1629984899.856262/events.out.tfevents.1629984899.8e89bd551565.924.203 +3 -0
- model-bin/finetune/base/log/1629985415.1269803/events.out.tfevents.1629985415.8e89bd551565.924.205 +3 -0
- model-bin/finetune/base/log/1629985858.0043113/events.out.tfevents.1629985858.8e89bd551565.924.207 +3 -0
- model-bin/finetune/base/log/1629986302.622504/events.out.tfevents.1629986302.8e89bd551565.924.209 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629984472.8e89bd551565.924.200 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629984899.8e89bd551565.924.202 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629985415.8e89bd551565.924.204 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629985858.8e89bd551565.924.206 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629986302.8e89bd551565.924.208 +3 -0
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/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:d3a771d7edb6aa41faa0a6a21bd195173b670dd0b09e6dab38bd0ac7661300d2
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/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:bd7c9466469e8561bf618808a7c2f005c58e031e02d6b5d7a5e9f99b36077bc8
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/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:c7ff32f0345d9842236266d835234b3d344bc6da9a81f9f41b6fa61f493e6a98
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/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 sha256:949c8be174ac18f6208fa3845f0a25711ace273e30219ad5f9c9828daebad6e1
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/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:d80c4826293eb78c5c01d3872b7e293297ecba83a8f1aec804a5cc40cd45e424
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/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,
|
|
@@ -227616,11 +227616,806 @@
|
|
| 227616 |
"eval_steps_per_second": 0.717,
|
| 227617 |
"eval_wer": 0.1787529529672847,
|
| 227618 |
"step": 100803
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 227619 |
}
|
| 227620 |
],
|
| 227621 |
-
"max_steps":
|
| 227622 |
"num_train_epochs": 5000,
|
| 227623 |
-
"total_flos": 2.
|
| 227624 |
"trial_name": null,
|
| 227625 |
"trial_params": null
|
| 227626 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 816.995983935743,
|
| 5 |
+
"global_step": 101426,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 227616 |
"eval_steps_per_second": 0.717,
|
| 227617 |
"eval_wer": 0.1787529529672847,
|
| 227618 |
"step": 100803
|
| 227619 |
+
},
|
| 227620 |
+
{
|
| 227621 |
+
"epoch": 812.02,
|
| 227622 |
+
"learning_rate": 8.401201923076924e-06,
|
| 227623 |
+
"loss": 0.3378,
|
| 227624 |
+
"step": 100805
|
| 227625 |
+
},
|
| 227626 |
+
{
|
| 227627 |
+
"epoch": 812.06,
|
| 227628 |
+
"learning_rate": 8.401121794871797e-06,
|
| 227629 |
+
"loss": 0.2857,
|
| 227630 |
+
"step": 100810
|
| 227631 |
+
},
|
| 227632 |
+
{
|
| 227633 |
+
"epoch": 812.1,
|
| 227634 |
+
"learning_rate": 8.401041666666666e-06,
|
| 227635 |
+
"loss": 0.2505,
|
| 227636 |
+
"step": 100815
|
| 227637 |
+
},
|
| 227638 |
+
{
|
| 227639 |
+
"epoch": 812.14,
|
| 227640 |
+
"learning_rate": 8.40096153846154e-06,
|
| 227641 |
+
"loss": 0.3367,
|
| 227642 |
+
"step": 100820
|
| 227643 |
+
},
|
| 227644 |
+
{
|
| 227645 |
+
"epoch": 812.18,
|
| 227646 |
+
"learning_rate": 8.400881410256411e-06,
|
| 227647 |
+
"loss": 0.6967,
|
| 227648 |
+
"step": 100825
|
| 227649 |
+
},
|
| 227650 |
+
{
|
| 227651 |
+
"epoch": 812.22,
|
| 227652 |
+
"learning_rate": 8.400801282051282e-06,
|
| 227653 |
+
"loss": 0.9902,
|
| 227654 |
+
"step": 100830
|
| 227655 |
+
},
|
| 227656 |
+
{
|
| 227657 |
+
"epoch": 812.26,
|
| 227658 |
+
"learning_rate": 8.400721153846154e-06,
|
| 227659 |
+
"loss": 0.2897,
|
| 227660 |
+
"step": 100835
|
| 227661 |
+
},
|
| 227662 |
+
{
|
| 227663 |
+
"epoch": 812.3,
|
| 227664 |
+
"learning_rate": 8.400641025641027e-06,
|
| 227665 |
+
"loss": 0.3277,
|
| 227666 |
+
"step": 100840
|
| 227667 |
+
},
|
| 227668 |
+
{
|
| 227669 |
+
"epoch": 812.34,
|
| 227670 |
+
"learning_rate": 8.400560897435898e-06,
|
| 227671 |
+
"loss": 0.3201,
|
| 227672 |
+
"step": 100845
|
| 227673 |
+
},
|
| 227674 |
+
{
|
| 227675 |
+
"epoch": 812.38,
|
| 227676 |
+
"learning_rate": 8.40048076923077e-06,
|
| 227677 |
+
"loss": 0.5825,
|
| 227678 |
+
"step": 100850
|
| 227679 |
+
},
|
| 227680 |
+
{
|
| 227681 |
+
"epoch": 812.42,
|
| 227682 |
+
"learning_rate": 8.400400641025642e-06,
|
| 227683 |
+
"loss": 1.1866,
|
| 227684 |
+
"step": 100855
|
| 227685 |
+
},
|
| 227686 |
+
{
|
| 227687 |
+
"epoch": 812.46,
|
| 227688 |
+
"learning_rate": 8.400320512820514e-06,
|
| 227689 |
+
"loss": 0.2778,
|
| 227690 |
+
"step": 100860
|
| 227691 |
+
},
|
| 227692 |
+
{
|
| 227693 |
+
"epoch": 812.5,
|
| 227694 |
+
"learning_rate": 8.400240384615385e-06,
|
| 227695 |
+
"loss": 0.3438,
|
| 227696 |
+
"step": 100865
|
| 227697 |
+
},
|
| 227698 |
+
{
|
| 227699 |
+
"epoch": 812.54,
|
| 227700 |
+
"learning_rate": 8.400160256410256e-06,
|
| 227701 |
+
"loss": 0.3764,
|
| 227702 |
+
"step": 100870
|
| 227703 |
+
},
|
| 227704 |
+
{
|
| 227705 |
+
"epoch": 812.58,
|
| 227706 |
+
"learning_rate": 8.40008012820513e-06,
|
| 227707 |
+
"loss": 0.6869,
|
| 227708 |
+
"step": 100875
|
| 227709 |
+
},
|
| 227710 |
+
{
|
| 227711 |
+
"epoch": 812.62,
|
| 227712 |
+
"learning_rate": 8.400000000000001e-06,
|
| 227713 |
+
"loss": 0.9914,
|
| 227714 |
+
"step": 100880
|
| 227715 |
+
},
|
| 227716 |
+
{
|
| 227717 |
+
"epoch": 812.66,
|
| 227718 |
+
"learning_rate": 8.399919871794872e-06,
|
| 227719 |
+
"loss": 0.2822,
|
| 227720 |
+
"step": 100885
|
| 227721 |
+
},
|
| 227722 |
+
{
|
| 227723 |
+
"epoch": 812.7,
|
| 227724 |
+
"learning_rate": 8.399839743589744e-06,
|
| 227725 |
+
"loss": 0.2664,
|
| 227726 |
+
"step": 100890
|
| 227727 |
+
},
|
| 227728 |
+
{
|
| 227729 |
+
"epoch": 812.74,
|
| 227730 |
+
"learning_rate": 8.399759615384617e-06,
|
| 227731 |
+
"loss": 0.3402,
|
| 227732 |
+
"step": 100895
|
| 227733 |
+
},
|
| 227734 |
+
{
|
| 227735 |
+
"epoch": 812.78,
|
| 227736 |
+
"learning_rate": 8.399679487179488e-06,
|
| 227737 |
+
"loss": 0.5972,
|
| 227738 |
+
"step": 100900
|
| 227739 |
+
},
|
| 227740 |
+
{
|
| 227741 |
+
"epoch": 812.82,
|
| 227742 |
+
"learning_rate": 8.39959935897436e-06,
|
| 227743 |
+
"loss": 1.0647,
|
| 227744 |
+
"step": 100905
|
| 227745 |
+
},
|
| 227746 |
+
{
|
| 227747 |
+
"epoch": 812.86,
|
| 227748 |
+
"learning_rate": 8.399519230769232e-06,
|
| 227749 |
+
"loss": 0.2687,
|
| 227750 |
+
"step": 100910
|
| 227751 |
+
},
|
| 227752 |
+
{
|
| 227753 |
+
"epoch": 812.9,
|
| 227754 |
+
"learning_rate": 8.399439102564104e-06,
|
| 227755 |
+
"loss": 0.2978,
|
| 227756 |
+
"step": 100915
|
| 227757 |
+
},
|
| 227758 |
+
{
|
| 227759 |
+
"epoch": 812.94,
|
| 227760 |
+
"learning_rate": 8.399358974358975e-06,
|
| 227761 |
+
"loss": 0.3943,
|
| 227762 |
+
"step": 100920
|
| 227763 |
+
},
|
| 227764 |
+
{
|
| 227765 |
+
"epoch": 812.98,
|
| 227766 |
+
"learning_rate": 8.399278846153847e-06,
|
| 227767 |
+
"loss": 0.8063,
|
| 227768 |
+
"step": 100925
|
| 227769 |
+
},
|
| 227770 |
+
{
|
| 227771 |
+
"epoch": 813.0,
|
| 227772 |
+
"eval_loss": 0.38394448161125183,
|
| 227773 |
+
"eval_runtime": 39.8467,
|
| 227774 |
+
"eval_samples_per_second": 20.905,
|
| 227775 |
+
"eval_steps_per_second": 0.678,
|
| 227776 |
+
"eval_wer": 0.1810313210591946,
|
| 227777 |
+
"step": 100927
|
| 227778 |
+
},
|
| 227779 |
+
{
|
| 227780 |
+
"epoch": 807.02,
|
| 227781 |
+
"learning_rate": 8.39919871794872e-06,
|
| 227782 |
+
"loss": 0.3239,
|
| 227783 |
+
"step": 100930
|
| 227784 |
+
},
|
| 227785 |
+
{
|
| 227786 |
+
"epoch": 807.06,
|
| 227787 |
+
"learning_rate": 8.39911858974359e-06,
|
| 227788 |
+
"loss": 0.4577,
|
| 227789 |
+
"step": 100935
|
| 227790 |
+
},
|
| 227791 |
+
{
|
| 227792 |
+
"epoch": 807.1,
|
| 227793 |
+
"learning_rate": 8.399038461538462e-06,
|
| 227794 |
+
"loss": 0.3141,
|
| 227795 |
+
"step": 100940
|
| 227796 |
+
},
|
| 227797 |
+
{
|
| 227798 |
+
"epoch": 807.14,
|
| 227799 |
+
"learning_rate": 8.398958333333334e-06,
|
| 227800 |
+
"loss": 0.3403,
|
| 227801 |
+
"step": 100945
|
| 227802 |
+
},
|
| 227803 |
+
{
|
| 227804 |
+
"epoch": 807.18,
|
| 227805 |
+
"learning_rate": 8.398878205128207e-06,
|
| 227806 |
+
"loss": 0.7388,
|
| 227807 |
+
"step": 100950
|
| 227808 |
+
},
|
| 227809 |
+
{
|
| 227810 |
+
"epoch": 807.22,
|
| 227811 |
+
"learning_rate": 8.398798076923078e-06,
|
| 227812 |
+
"loss": 0.8915,
|
| 227813 |
+
"step": 100955
|
| 227814 |
+
},
|
| 227815 |
+
{
|
| 227816 |
+
"epoch": 807.26,
|
| 227817 |
+
"learning_rate": 8.39871794871795e-06,
|
| 227818 |
+
"loss": 0.2946,
|
| 227819 |
+
"step": 100960
|
| 227820 |
+
},
|
| 227821 |
+
{
|
| 227822 |
+
"epoch": 807.3,
|
| 227823 |
+
"learning_rate": 8.398637820512822e-06,
|
| 227824 |
+
"loss": 0.269,
|
| 227825 |
+
"step": 100965
|
| 227826 |
+
},
|
| 227827 |
+
{
|
| 227828 |
+
"epoch": 807.34,
|
| 227829 |
+
"learning_rate": 8.398557692307692e-06,
|
| 227830 |
+
"loss": 0.3845,
|
| 227831 |
+
"step": 100970
|
| 227832 |
+
},
|
| 227833 |
+
{
|
| 227834 |
+
"epoch": 807.38,
|
| 227835 |
+
"learning_rate": 8.398477564102565e-06,
|
| 227836 |
+
"loss": 0.7681,
|
| 227837 |
+
"step": 100975
|
| 227838 |
+
},
|
| 227839 |
+
{
|
| 227840 |
+
"epoch": 807.42,
|
| 227841 |
+
"learning_rate": 8.398397435897437e-06,
|
| 227842 |
+
"loss": 0.8968,
|
| 227843 |
+
"step": 100980
|
| 227844 |
+
},
|
| 227845 |
+
{
|
| 227846 |
+
"epoch": 807.46,
|
| 227847 |
+
"learning_rate": 8.398317307692308e-06,
|
| 227848 |
+
"loss": 0.3255,
|
| 227849 |
+
"step": 100985
|
| 227850 |
+
},
|
| 227851 |
+
{
|
| 227852 |
+
"epoch": 807.5,
|
| 227853 |
+
"learning_rate": 8.39823717948718e-06,
|
| 227854 |
+
"loss": 0.3212,
|
| 227855 |
+
"step": 100990
|
| 227856 |
+
},
|
| 227857 |
+
{
|
| 227858 |
+
"epoch": 807.54,
|
| 227859 |
+
"learning_rate": 8.398157051282052e-06,
|
| 227860 |
+
"loss": 0.3626,
|
| 227861 |
+
"step": 100995
|
| 227862 |
+
},
|
| 227863 |
+
{
|
| 227864 |
+
"epoch": 807.58,
|
| 227865 |
+
"learning_rate": 8.398076923076924e-06,
|
| 227866 |
+
"loss": 1.031,
|
| 227867 |
+
"step": 101000
|
| 227868 |
+
},
|
| 227869 |
+
{
|
| 227870 |
+
"epoch": 807.62,
|
| 227871 |
+
"learning_rate": 8.397996794871795e-06,
|
| 227872 |
+
"loss": 0.8177,
|
| 227873 |
+
"step": 101005
|
| 227874 |
+
},
|
| 227875 |
+
{
|
| 227876 |
+
"epoch": 807.66,
|
| 227877 |
+
"learning_rate": 8.397916666666668e-06,
|
| 227878 |
+
"loss": 0.2561,
|
| 227879 |
+
"step": 101010
|
| 227880 |
+
},
|
| 227881 |
+
{
|
| 227882 |
+
"epoch": 807.7,
|
| 227883 |
+
"learning_rate": 8.39783653846154e-06,
|
| 227884 |
+
"loss": 0.2728,
|
| 227885 |
+
"step": 101015
|
| 227886 |
+
},
|
| 227887 |
+
{
|
| 227888 |
+
"epoch": 807.74,
|
| 227889 |
+
"learning_rate": 8.39775641025641e-06,
|
| 227890 |
+
"loss": 0.3881,
|
| 227891 |
+
"step": 101020
|
| 227892 |
+
},
|
| 227893 |
+
{
|
| 227894 |
+
"epoch": 807.78,
|
| 227895 |
+
"learning_rate": 8.397676282051282e-06,
|
| 227896 |
+
"loss": 0.6644,
|
| 227897 |
+
"step": 101025
|
| 227898 |
+
},
|
| 227899 |
+
{
|
| 227900 |
+
"epoch": 807.82,
|
| 227901 |
+
"learning_rate": 8.397596153846155e-06,
|
| 227902 |
+
"loss": 0.8532,
|
| 227903 |
+
"step": 101030
|
| 227904 |
+
},
|
| 227905 |
+
{
|
| 227906 |
+
"epoch": 807.86,
|
| 227907 |
+
"learning_rate": 8.397516025641027e-06,
|
| 227908 |
+
"loss": 0.314,
|
| 227909 |
+
"step": 101035
|
| 227910 |
+
},
|
| 227911 |
+
{
|
| 227912 |
+
"epoch": 807.9,
|
| 227913 |
+
"learning_rate": 8.397435897435898e-06,
|
| 227914 |
+
"loss": 0.2902,
|
| 227915 |
+
"step": 101040
|
| 227916 |
+
},
|
| 227917 |
+
{
|
| 227918 |
+
"epoch": 807.94,
|
| 227919 |
+
"learning_rate": 8.39735576923077e-06,
|
| 227920 |
+
"loss": 0.4595,
|
| 227921 |
+
"step": 101045
|
| 227922 |
+
},
|
| 227923 |
+
{
|
| 227924 |
+
"epoch": 807.98,
|
| 227925 |
+
"learning_rate": 8.397275641025642e-06,
|
| 227926 |
+
"loss": 0.7156,
|
| 227927 |
+
"step": 101050
|
| 227928 |
+
},
|
| 227929 |
+
{
|
| 227930 |
+
"epoch": 808.0,
|
| 227931 |
+
"eval_loss": 0.3595258593559265,
|
| 227932 |
+
"eval_runtime": 40.7337,
|
| 227933 |
+
"eval_samples_per_second": 20.45,
|
| 227934 |
+
"eval_steps_per_second": 0.663,
|
| 227935 |
+
"eval_wer": 0.17914092802758225,
|
| 227936 |
+
"step": 101052
|
| 227937 |
+
},
|
| 227938 |
+
{
|
| 227939 |
+
"epoch": 808.02,
|
| 227940 |
+
"learning_rate": 8.397195512820514e-06,
|
| 227941 |
+
"loss": 0.3318,
|
| 227942 |
+
"step": 101055
|
| 227943 |
+
},
|
| 227944 |
+
{
|
| 227945 |
+
"epoch": 808.06,
|
| 227946 |
+
"learning_rate": 8.397115384615385e-06,
|
| 227947 |
+
"loss": 0.3463,
|
| 227948 |
+
"step": 101060
|
| 227949 |
+
},
|
| 227950 |
+
{
|
| 227951 |
+
"epoch": 808.1,
|
| 227952 |
+
"learning_rate": 8.397035256410258e-06,
|
| 227953 |
+
"loss": 0.2749,
|
| 227954 |
+
"step": 101065
|
| 227955 |
+
},
|
| 227956 |
+
{
|
| 227957 |
+
"epoch": 808.14,
|
| 227958 |
+
"learning_rate": 8.39695512820513e-06,
|
| 227959 |
+
"loss": 0.3859,
|
| 227960 |
+
"step": 101070
|
| 227961 |
+
},
|
| 227962 |
+
{
|
| 227963 |
+
"epoch": 808.18,
|
| 227964 |
+
"learning_rate": 8.396875e-06,
|
| 227965 |
+
"loss": 1.0331,
|
| 227966 |
+
"step": 101075
|
| 227967 |
+
},
|
| 227968 |
+
{
|
| 227969 |
+
"epoch": 808.22,
|
| 227970 |
+
"learning_rate": 8.396794871794872e-06,
|
| 227971 |
+
"loss": 0.8274,
|
| 227972 |
+
"step": 101080
|
| 227973 |
+
},
|
| 227974 |
+
{
|
| 227975 |
+
"epoch": 808.26,
|
| 227976 |
+
"learning_rate": 8.396714743589745e-06,
|
| 227977 |
+
"loss": 0.3095,
|
| 227978 |
+
"step": 101085
|
| 227979 |
+
},
|
| 227980 |
+
{
|
| 227981 |
+
"epoch": 808.3,
|
| 227982 |
+
"learning_rate": 8.396634615384615e-06,
|
| 227983 |
+
"loss": 0.3409,
|
| 227984 |
+
"step": 101090
|
| 227985 |
+
},
|
| 227986 |
+
{
|
| 227987 |
+
"epoch": 808.34,
|
| 227988 |
+
"learning_rate": 8.396554487179488e-06,
|
| 227989 |
+
"loss": 0.3863,
|
| 227990 |
+
"step": 101095
|
| 227991 |
+
},
|
| 227992 |
+
{
|
| 227993 |
+
"epoch": 808.38,
|
| 227994 |
+
"learning_rate": 8.39647435897436e-06,
|
| 227995 |
+
"loss": 0.78,
|
| 227996 |
+
"step": 101100
|
| 227997 |
+
},
|
| 227998 |
+
{
|
| 227999 |
+
"epoch": 808.42,
|
| 228000 |
+
"learning_rate": 8.39639423076923e-06,
|
| 228001 |
+
"loss": 0.9809,
|
| 228002 |
+
"step": 101105
|
| 228003 |
+
},
|
| 228004 |
+
{
|
| 228005 |
+
"epoch": 808.46,
|
| 228006 |
+
"learning_rate": 8.396314102564104e-06,
|
| 228007 |
+
"loss": 0.3009,
|
| 228008 |
+
"step": 101110
|
| 228009 |
+
},
|
| 228010 |
+
{
|
| 228011 |
+
"epoch": 808.5,
|
| 228012 |
+
"learning_rate": 8.396233974358975e-06,
|
| 228013 |
+
"loss": 0.2904,
|
| 228014 |
+
"step": 101115
|
| 228015 |
+
},
|
| 228016 |
+
{
|
| 228017 |
+
"epoch": 808.54,
|
| 228018 |
+
"learning_rate": 8.396153846153846e-06,
|
| 228019 |
+
"loss": 0.3599,
|
| 228020 |
+
"step": 101120
|
| 228021 |
+
},
|
| 228022 |
+
{
|
| 228023 |
+
"epoch": 808.58,
|
| 228024 |
+
"learning_rate": 8.396073717948718e-06,
|
| 228025 |
+
"loss": 0.9045,
|
| 228026 |
+
"step": 101125
|
| 228027 |
+
},
|
| 228028 |
+
{
|
| 228029 |
+
"epoch": 808.62,
|
| 228030 |
+
"learning_rate": 8.39599358974359e-06,
|
| 228031 |
+
"loss": 1.0318,
|
| 228032 |
+
"step": 101130
|
| 228033 |
+
},
|
| 228034 |
+
{
|
| 228035 |
+
"epoch": 808.66,
|
| 228036 |
+
"learning_rate": 8.395913461538462e-06,
|
| 228037 |
+
"loss": 0.2434,
|
| 228038 |
+
"step": 101135
|
| 228039 |
+
},
|
| 228040 |
+
{
|
| 228041 |
+
"epoch": 808.7,
|
| 228042 |
+
"learning_rate": 8.395833333333334e-06,
|
| 228043 |
+
"loss": 0.3177,
|
| 228044 |
+
"step": 101140
|
| 228045 |
+
},
|
| 228046 |
+
{
|
| 228047 |
+
"epoch": 808.74,
|
| 228048 |
+
"learning_rate": 8.395753205128205e-06,
|
| 228049 |
+
"loss": 0.3773,
|
| 228050 |
+
"step": 101145
|
| 228051 |
+
},
|
| 228052 |
+
{
|
| 228053 |
+
"epoch": 808.78,
|
| 228054 |
+
"learning_rate": 8.395673076923078e-06,
|
| 228055 |
+
"loss": 0.8804,
|
| 228056 |
+
"step": 101150
|
| 228057 |
+
},
|
| 228058 |
+
{
|
| 228059 |
+
"epoch": 808.82,
|
| 228060 |
+
"learning_rate": 8.39559294871795e-06,
|
| 228061 |
+
"loss": 0.9089,
|
| 228062 |
+
"step": 101155
|
| 228063 |
+
},
|
| 228064 |
+
{
|
| 228065 |
+
"epoch": 808.86,
|
| 228066 |
+
"learning_rate": 8.39551282051282e-06,
|
| 228067 |
+
"loss": 0.2774,
|
| 228068 |
+
"step": 101160
|
| 228069 |
+
},
|
| 228070 |
+
{
|
| 228071 |
+
"epoch": 808.9,
|
| 228072 |
+
"learning_rate": 8.395432692307694e-06,
|
| 228073 |
+
"loss": 0.3395,
|
| 228074 |
+
"step": 101165
|
| 228075 |
+
},
|
| 228076 |
+
{
|
| 228077 |
+
"epoch": 808.94,
|
| 228078 |
+
"learning_rate": 8.395352564102565e-06,
|
| 228079 |
+
"loss": 0.4058,
|
| 228080 |
+
"step": 101170
|
| 228081 |
+
},
|
| 228082 |
+
{
|
| 228083 |
+
"epoch": 808.98,
|
| 228084 |
+
"learning_rate": 8.395272435897436e-06,
|
| 228085 |
+
"loss": 0.7291,
|
| 228086 |
+
"step": 101175
|
| 228087 |
+
},
|
| 228088 |
+
{
|
| 228089 |
+
"epoch": 809.0,
|
| 228090 |
+
"eval_loss": 0.4380325973033905,
|
| 228091 |
+
"eval_runtime": 39.7922,
|
| 228092 |
+
"eval_samples_per_second": 21.11,
|
| 228093 |
+
"eval_steps_per_second": 0.679,
|
| 228094 |
+
"eval_wer": 0.18656987295825772,
|
| 228095 |
+
"step": 101177
|
| 228096 |
+
},
|
| 228097 |
+
{
|
| 228098 |
+
"epoch": 809.02,
|
| 228099 |
+
"learning_rate": 8.395192307692308e-06,
|
| 228100 |
+
"loss": 0.3575,
|
| 228101 |
+
"step": 101180
|
| 228102 |
+
},
|
| 228103 |
+
{
|
| 228104 |
+
"epoch": 809.06,
|
| 228105 |
+
"learning_rate": 8.395112179487181e-06,
|
| 228106 |
+
"loss": 0.2703,
|
| 228107 |
+
"step": 101185
|
| 228108 |
+
},
|
| 228109 |
+
{
|
| 228110 |
+
"epoch": 809.1,
|
| 228111 |
+
"learning_rate": 8.395032051282052e-06,
|
| 228112 |
+
"loss": 0.2827,
|
| 228113 |
+
"step": 101190
|
| 228114 |
+
},
|
| 228115 |
+
{
|
| 228116 |
+
"epoch": 809.14,
|
| 228117 |
+
"learning_rate": 8.394951923076924e-06,
|
| 228118 |
+
"loss": 0.364,
|
| 228119 |
+
"step": 101195
|
| 228120 |
+
},
|
| 228121 |
+
{
|
| 228122 |
+
"epoch": 809.18,
|
| 228123 |
+
"learning_rate": 8.394871794871795e-06,
|
| 228124 |
+
"loss": 0.7247,
|
| 228125 |
+
"step": 101200
|
| 228126 |
+
},
|
| 228127 |
+
{
|
| 228128 |
+
"epoch": 809.22,
|
| 228129 |
+
"learning_rate": 8.394791666666668e-06,
|
| 228130 |
+
"loss": 0.8795,
|
| 228131 |
+
"step": 101205
|
| 228132 |
+
},
|
| 228133 |
+
{
|
| 228134 |
+
"epoch": 809.26,
|
| 228135 |
+
"learning_rate": 8.39471153846154e-06,
|
| 228136 |
+
"loss": 0.2718,
|
| 228137 |
+
"step": 101210
|
| 228138 |
+
},
|
| 228139 |
+
{
|
| 228140 |
+
"epoch": 809.3,
|
| 228141 |
+
"learning_rate": 8.39463141025641e-06,
|
| 228142 |
+
"loss": 0.2908,
|
| 228143 |
+
"step": 101215
|
| 228144 |
+
},
|
| 228145 |
+
{
|
| 228146 |
+
"epoch": 809.34,
|
| 228147 |
+
"learning_rate": 8.394551282051284e-06,
|
| 228148 |
+
"loss": 0.372,
|
| 228149 |
+
"step": 101220
|
| 228150 |
+
},
|
| 228151 |
+
{
|
| 228152 |
+
"epoch": 809.38,
|
| 228153 |
+
"learning_rate": 8.394471153846155e-06,
|
| 228154 |
+
"loss": 0.7974,
|
| 228155 |
+
"step": 101225
|
| 228156 |
+
},
|
| 228157 |
+
{
|
| 228158 |
+
"epoch": 809.42,
|
| 228159 |
+
"learning_rate": 8.394391025641026e-06,
|
| 228160 |
+
"loss": 0.9957,
|
| 228161 |
+
"step": 101230
|
| 228162 |
+
},
|
| 228163 |
+
{
|
| 228164 |
+
"epoch": 809.46,
|
| 228165 |
+
"learning_rate": 8.394310897435898e-06,
|
| 228166 |
+
"loss": 0.3685,
|
| 228167 |
+
"step": 101235
|
| 228168 |
+
},
|
| 228169 |
+
{
|
| 228170 |
+
"epoch": 809.5,
|
| 228171 |
+
"learning_rate": 8.394230769230771e-06,
|
| 228172 |
+
"loss": 0.3201,
|
| 228173 |
+
"step": 101240
|
| 228174 |
+
},
|
| 228175 |
+
{
|
| 228176 |
+
"epoch": 809.54,
|
| 228177 |
+
"learning_rate": 8.39415064102564e-06,
|
| 228178 |
+
"loss": 0.3592,
|
| 228179 |
+
"step": 101245
|
| 228180 |
+
},
|
| 228181 |
+
{
|
| 228182 |
+
"epoch": 809.58,
|
| 228183 |
+
"learning_rate": 8.394070512820514e-06,
|
| 228184 |
+
"loss": 0.7414,
|
| 228185 |
+
"step": 101250
|
| 228186 |
+
},
|
| 228187 |
+
{
|
| 228188 |
+
"epoch": 809.62,
|
| 228189 |
+
"learning_rate": 8.393990384615387e-06,
|
| 228190 |
+
"loss": 0.8616,
|
| 228191 |
+
"step": 101255
|
| 228192 |
+
},
|
| 228193 |
+
{
|
| 228194 |
+
"epoch": 809.66,
|
| 228195 |
+
"learning_rate": 8.393910256410256e-06,
|
| 228196 |
+
"loss": 0.2793,
|
| 228197 |
+
"step": 101260
|
| 228198 |
+
},
|
| 228199 |
+
{
|
| 228200 |
+
"epoch": 809.7,
|
| 228201 |
+
"learning_rate": 8.39383012820513e-06,
|
| 228202 |
+
"loss": 0.3023,
|
| 228203 |
+
"step": 101265
|
| 228204 |
+
},
|
| 228205 |
+
{
|
| 228206 |
+
"epoch": 809.74,
|
| 228207 |
+
"learning_rate": 8.39375e-06,
|
| 228208 |
+
"loss": 0.3786,
|
| 228209 |
+
"step": 101270
|
| 228210 |
+
},
|
| 228211 |
+
{
|
| 228212 |
+
"epoch": 809.78,
|
| 228213 |
+
"learning_rate": 8.393669871794872e-06,
|
| 228214 |
+
"loss": 0.6854,
|
| 228215 |
+
"step": 101275
|
| 228216 |
+
},
|
| 228217 |
+
{
|
| 228218 |
+
"epoch": 809.82,
|
| 228219 |
+
"learning_rate": 8.393589743589743e-06,
|
| 228220 |
+
"loss": 0.8627,
|
| 228221 |
+
"step": 101280
|
| 228222 |
+
},
|
| 228223 |
+
{
|
| 228224 |
+
"epoch": 809.86,
|
| 228225 |
+
"learning_rate": 8.393509615384617e-06,
|
| 228226 |
+
"loss": 0.315,
|
| 228227 |
+
"step": 101285
|
| 228228 |
+
},
|
| 228229 |
+
{
|
| 228230 |
+
"epoch": 809.9,
|
| 228231 |
+
"learning_rate": 8.393429487179488e-06,
|
| 228232 |
+
"loss": 0.3104,
|
| 228233 |
+
"step": 101290
|
| 228234 |
+
},
|
| 228235 |
+
{
|
| 228236 |
+
"epoch": 809.94,
|
| 228237 |
+
"learning_rate": 8.39334935897436e-06,
|
| 228238 |
+
"loss": 0.3336,
|
| 228239 |
+
"step": 101295
|
| 228240 |
+
},
|
| 228241 |
+
{
|
| 228242 |
+
"epoch": 809.98,
|
| 228243 |
+
"learning_rate": 8.39326923076923e-06,
|
| 228244 |
+
"loss": 0.6299,
|
| 228245 |
+
"step": 101300
|
| 228246 |
+
},
|
| 228247 |
+
{
|
| 228248 |
+
"epoch": 810.0,
|
| 228249 |
+
"eval_loss": 0.3927081227302551,
|
| 228250 |
+
"eval_runtime": 37.1456,
|
| 228251 |
+
"eval_samples_per_second": 22.614,
|
| 228252 |
+
"eval_steps_per_second": 0.727,
|
| 228253 |
+
"eval_wer": 0.18203601510310777,
|
| 228254 |
+
"step": 101302
|
| 228255 |
+
},
|
| 228256 |
+
{
|
| 228257 |
+
"epoch": 816.02,
|
| 228258 |
+
"learning_rate": 8.393189102564104e-06,
|
| 228259 |
+
"loss": 0.3374,
|
| 228260 |
+
"step": 101305
|
| 228261 |
+
},
|
| 228262 |
+
{
|
| 228263 |
+
"epoch": 816.06,
|
| 228264 |
+
"learning_rate": 8.393108974358975e-06,
|
| 228265 |
+
"loss": 0.2778,
|
| 228266 |
+
"step": 101310
|
| 228267 |
+
},
|
| 228268 |
+
{
|
| 228269 |
+
"epoch": 816.1,
|
| 228270 |
+
"learning_rate": 8.393028846153846e-06,
|
| 228271 |
+
"loss": 0.2851,
|
| 228272 |
+
"step": 101315
|
| 228273 |
+
},
|
| 228274 |
+
{
|
| 228275 |
+
"epoch": 816.14,
|
| 228276 |
+
"learning_rate": 8.39294871794872e-06,
|
| 228277 |
+
"loss": 0.4203,
|
| 228278 |
+
"step": 101320
|
| 228279 |
+
},
|
| 228280 |
+
{
|
| 228281 |
+
"epoch": 816.18,
|
| 228282 |
+
"learning_rate": 8.39286858974359e-06,
|
| 228283 |
+
"loss": 0.7721,
|
| 228284 |
+
"step": 101325
|
| 228285 |
+
},
|
| 228286 |
+
{
|
| 228287 |
+
"epoch": 816.22,
|
| 228288 |
+
"learning_rate": 8.392788461538462e-06,
|
| 228289 |
+
"loss": 0.9135,
|
| 228290 |
+
"step": 101330
|
| 228291 |
+
},
|
| 228292 |
+
{
|
| 228293 |
+
"epoch": 816.27,
|
| 228294 |
+
"learning_rate": 8.392708333333333e-06,
|
| 228295 |
+
"loss": 0.2625,
|
| 228296 |
+
"step": 101335
|
| 228297 |
+
},
|
| 228298 |
+
{
|
| 228299 |
+
"epoch": 816.31,
|
| 228300 |
+
"learning_rate": 8.392628205128207e-06,
|
| 228301 |
+
"loss": 0.2985,
|
| 228302 |
+
"step": 101340
|
| 228303 |
+
},
|
| 228304 |
+
{
|
| 228305 |
+
"epoch": 816.35,
|
| 228306 |
+
"learning_rate": 8.392548076923078e-06,
|
| 228307 |
+
"loss": 0.3627,
|
| 228308 |
+
"step": 101345
|
| 228309 |
+
},
|
| 228310 |
+
{
|
| 228311 |
+
"epoch": 816.39,
|
| 228312 |
+
"learning_rate": 8.39246794871795e-06,
|
| 228313 |
+
"loss": 0.6898,
|
| 228314 |
+
"step": 101350
|
| 228315 |
+
},
|
| 228316 |
+
{
|
| 228317 |
+
"epoch": 816.43,
|
| 228318 |
+
"learning_rate": 8.392387820512822e-06,
|
| 228319 |
+
"loss": 0.701,
|
| 228320 |
+
"step": 101355
|
| 228321 |
+
},
|
| 228322 |
+
{
|
| 228323 |
+
"epoch": 816.47,
|
| 228324 |
+
"learning_rate": 8.392307692307694e-06,
|
| 228325 |
+
"loss": 0.2535,
|
| 228326 |
+
"step": 101360
|
| 228327 |
+
},
|
| 228328 |
+
{
|
| 228329 |
+
"epoch": 816.51,
|
| 228330 |
+
"learning_rate": 8.392227564102565e-06,
|
| 228331 |
+
"loss": 0.3312,
|
| 228332 |
+
"step": 101365
|
| 228333 |
+
},
|
| 228334 |
+
{
|
| 228335 |
+
"epoch": 816.55,
|
| 228336 |
+
"learning_rate": 8.392147435897436e-06,
|
| 228337 |
+
"loss": 0.3892,
|
| 228338 |
+
"step": 101370
|
| 228339 |
+
},
|
| 228340 |
+
{
|
| 228341 |
+
"epoch": 816.59,
|
| 228342 |
+
"learning_rate": 8.39206730769231e-06,
|
| 228343 |
+
"loss": 0.7533,
|
| 228344 |
+
"step": 101375
|
| 228345 |
+
},
|
| 228346 |
+
{
|
| 228347 |
+
"epoch": 816.63,
|
| 228348 |
+
"learning_rate": 8.391987179487179e-06,
|
| 228349 |
+
"loss": 0.8716,
|
| 228350 |
+
"step": 101380
|
| 228351 |
+
},
|
| 228352 |
+
{
|
| 228353 |
+
"epoch": 816.67,
|
| 228354 |
+
"learning_rate": 8.391907051282052e-06,
|
| 228355 |
+
"loss": 0.2673,
|
| 228356 |
+
"step": 101385
|
| 228357 |
+
},
|
| 228358 |
+
{
|
| 228359 |
+
"epoch": 816.71,
|
| 228360 |
+
"learning_rate": 8.391826923076924e-06,
|
| 228361 |
+
"loss": 0.3044,
|
| 228362 |
+
"step": 101390
|
| 228363 |
+
},
|
| 228364 |
+
{
|
| 228365 |
+
"epoch": 816.75,
|
| 228366 |
+
"learning_rate": 8.391746794871795e-06,
|
| 228367 |
+
"loss": 0.3774,
|
| 228368 |
+
"step": 101395
|
| 228369 |
+
},
|
| 228370 |
+
{
|
| 228371 |
+
"epoch": 816.79,
|
| 228372 |
+
"learning_rate": 8.391666666666666e-06,
|
| 228373 |
+
"loss": 0.8407,
|
| 228374 |
+
"step": 101400
|
| 228375 |
+
},
|
| 228376 |
+
{
|
| 228377 |
+
"epoch": 816.83,
|
| 228378 |
+
"learning_rate": 8.39158653846154e-06,
|
| 228379 |
+
"loss": 0.902,
|
| 228380 |
+
"step": 101405
|
| 228381 |
+
},
|
| 228382 |
+
{
|
| 228383 |
+
"epoch": 816.87,
|
| 228384 |
+
"learning_rate": 8.39150641025641e-06,
|
| 228385 |
+
"loss": 0.3134,
|
| 228386 |
+
"step": 101410
|
| 228387 |
+
},
|
| 228388 |
+
{
|
| 228389 |
+
"epoch": 816.91,
|
| 228390 |
+
"learning_rate": 8.391426282051282e-06,
|
| 228391 |
+
"loss": 0.339,
|
| 228392 |
+
"step": 101415
|
| 228393 |
+
},
|
| 228394 |
+
{
|
| 228395 |
+
"epoch": 816.95,
|
| 228396 |
+
"learning_rate": 8.391346153846155e-06,
|
| 228397 |
+
"loss": 0.3602,
|
| 228398 |
+
"step": 101420
|
| 228399 |
+
},
|
| 228400 |
+
{
|
| 228401 |
+
"epoch": 816.99,
|
| 228402 |
+
"learning_rate": 8.391266025641026e-06,
|
| 228403 |
+
"loss": 0.8532,
|
| 228404 |
+
"step": 101425
|
| 228405 |
+
},
|
| 228406 |
+
{
|
| 228407 |
+
"epoch": 817.0,
|
| 228408 |
+
"eval_loss": 0.3548583388328552,
|
| 228409 |
+
"eval_runtime": 38.6846,
|
| 228410 |
+
"eval_samples_per_second": 21.714,
|
| 228411 |
+
"eval_steps_per_second": 0.698,
|
| 228412 |
+
"eval_wer": 0.18432907466213722,
|
| 228413 |
+
"step": 101426
|
| 228414 |
}
|
| 228415 |
],
|
| 228416 |
+
"max_steps": 620000,
|
| 228417 |
"num_train_epochs": 5000,
|
| 228418 |
+
"total_flos": 2.854393645809328e+20,
|
| 228419 |
"trial_name": null,
|
| 228420 |
"trial_params": null
|
| 228421 |
}
|
model-bin/finetune/base/{checkpoint-100803 → checkpoint-101426}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629984472.6452265/events.out.tfevents.1629984472.8e89bd551565.924.201
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:96ed4f80086a0b3273b5fc65d0ba9b025d0ca6b64165c0ab21810cfa183771ec
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629984899.856262/events.out.tfevents.1629984899.8e89bd551565.924.203
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ce6fa4d82c579b605ea891ef5f92931abafb1954a9d98c7feeab039e30a70433
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629985415.1269803/events.out.tfevents.1629985415.8e89bd551565.924.205
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f7905a13ce3bdcfa26775e02207b71c803946fff7a1c01388ed8e380d091074
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629985858.0043113/events.out.tfevents.1629985858.8e89bd551565.924.207
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:184e73e004f5732de722d42a4846f9c3e9583c5d9a87079f4efd4f31c50d94a5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629986302.622504/events.out.tfevents.1629986302.8e89bd551565.924.209
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f60551831ea8f163beb39b343df3ca3025cbc7be9fa240bddf052408d233f73c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629984472.8e89bd551565.924.200
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7ef1e1cc3058b85f4741ff5dfe03035c3d3ef72b7e0aeb385fa8cb46e77862cc
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629984899.8e89bd551565.924.202
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0e4b553374d32ca4ded8ee20db7cdeec601c2348fd86979c10214d4081ee8214
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629985415.8e89bd551565.924.204
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9198f407d76e58ef0e1dc46cd22e032fd4ee9f5f2346fc5638d6ed4ffe9cb182
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629985858.8e89bd551565.924.206
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4d80ec2afdcb9794e5caceddd799295ef002303928d0e635172c7b576cb1a8fd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629986302.8e89bd551565.924.208
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f0c070947758af9312d9010ba4d0c6207044426af608431181342a33887ab661
|
| 3 |
+
size 8622
|