"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629925935.501896/events.out.tfevents.1629925935.7e498afd5545.7645.155 +3 -0
- model-bin/finetune/base/log/1629926415.8598046/events.out.tfevents.1629926415.7e498afd5545.7645.157 +3 -0
- model-bin/finetune/base/log/1629926892.2024448/events.out.tfevents.1629926894.7e498afd5545.7645.159 +3 -0
- model-bin/finetune/base/log/1629927362.1056397/events.out.tfevents.1629927362.7e498afd5545.7645.161 +3 -0
- model-bin/finetune/base/log/1629927831.7978618/events.out.tfevents.1629927831.7e498afd5545.7645.163 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629925935.7e498afd5545.7645.154 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629926415.7e498afd5545.7645.156 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629926892.7e498afd5545.7645.158 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629927362.7e498afd5545.7645.160 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629927831.7e498afd5545.7645.162 +3 -0
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/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:a8bc3686dbe43b18d158600bbc73da971d3107e47c128d0290505faf633a2279
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/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 sha256:90f24c971fd277add21080d696ac93bf42778470903be4116a7d7279bc76d5ba
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/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 sha256:9cd302c43a1d5fefe9e23746f7da370464720b5481aac68cd427d03344fe3abc
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/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 sha256:366b40c92d99c2539fd4abe0fa5bd065289f2be7017a7878394de8391d6acb41
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/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:a32b27d201506de22e1468d6811196605d441be5f1ee480339d3343ad95f2cf3
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -208614,11 +208614,800 @@
|
|
| 208614 |
"eval_steps_per_second": 0.652,
|
| 208615 |
"eval_wer": 0.1846578526805172,
|
| 208616 |
"step": 85866
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 208617 |
}
|
| 208618 |
],
|
| 208619 |
"max_steps": 620000,
|
| 208620 |
"num_train_epochs": 5000,
|
| 208621 |
-
"total_flos": 2.
|
| 208622 |
"trial_name": null,
|
| 208623 |
"trial_params": null
|
| 208624 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 696.995983935743,
|
| 5 |
+
"global_step": 86488,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 208614 |
"eval_steps_per_second": 0.652,
|
| 208615 |
"eval_wer": 0.1846578526805172,
|
| 208616 |
"step": 85866
|
| 208617 |
+
},
|
| 208618 |
+
{
|
| 208619 |
+
"epoch": 692.03,
|
| 208620 |
+
"learning_rate": 8.629466882067852e-06,
|
| 208621 |
+
"loss": 0.3787,
|
| 208622 |
+
"step": 85870
|
| 208623 |
+
},
|
| 208624 |
+
{
|
| 208625 |
+
"epoch": 692.07,
|
| 208626 |
+
"learning_rate": 8.629386106623587e-06,
|
| 208627 |
+
"loss": 0.2704,
|
| 208628 |
+
"step": 85875
|
| 208629 |
+
},
|
| 208630 |
+
{
|
| 208631 |
+
"epoch": 692.11,
|
| 208632 |
+
"learning_rate": 8.629305331179322e-06,
|
| 208633 |
+
"loss": 0.2762,
|
| 208634 |
+
"step": 85880
|
| 208635 |
+
},
|
| 208636 |
+
{
|
| 208637 |
+
"epoch": 692.15,
|
| 208638 |
+
"learning_rate": 8.629224555735057e-06,
|
| 208639 |
+
"loss": 0.4332,
|
| 208640 |
+
"step": 85885
|
| 208641 |
+
},
|
| 208642 |
+
{
|
| 208643 |
+
"epoch": 692.19,
|
| 208644 |
+
"learning_rate": 8.629143780290792e-06,
|
| 208645 |
+
"loss": 1.0391,
|
| 208646 |
+
"step": 85890
|
| 208647 |
+
},
|
| 208648 |
+
{
|
| 208649 |
+
"epoch": 692.23,
|
| 208650 |
+
"learning_rate": 8.629063004846527e-06,
|
| 208651 |
+
"loss": 0.718,
|
| 208652 |
+
"step": 85895
|
| 208653 |
+
},
|
| 208654 |
+
{
|
| 208655 |
+
"epoch": 692.27,
|
| 208656 |
+
"learning_rate": 8.628982229402262e-06,
|
| 208657 |
+
"loss": 0.2963,
|
| 208658 |
+
"step": 85900
|
| 208659 |
+
},
|
| 208660 |
+
{
|
| 208661 |
+
"epoch": 692.31,
|
| 208662 |
+
"learning_rate": 8.628901453957997e-06,
|
| 208663 |
+
"loss": 0.3739,
|
| 208664 |
+
"step": 85905
|
| 208665 |
+
},
|
| 208666 |
+
{
|
| 208667 |
+
"epoch": 692.35,
|
| 208668 |
+
"learning_rate": 8.628820678513732e-06,
|
| 208669 |
+
"loss": 0.4223,
|
| 208670 |
+
"step": 85910
|
| 208671 |
+
},
|
| 208672 |
+
{
|
| 208673 |
+
"epoch": 692.39,
|
| 208674 |
+
"learning_rate": 8.628739903069468e-06,
|
| 208675 |
+
"loss": 0.8857,
|
| 208676 |
+
"step": 85915
|
| 208677 |
+
},
|
| 208678 |
+
{
|
| 208679 |
+
"epoch": 692.43,
|
| 208680 |
+
"learning_rate": 8.628659127625202e-06,
|
| 208681 |
+
"loss": 0.6445,
|
| 208682 |
+
"step": 85920
|
| 208683 |
+
},
|
| 208684 |
+
{
|
| 208685 |
+
"epoch": 692.47,
|
| 208686 |
+
"learning_rate": 8.628578352180938e-06,
|
| 208687 |
+
"loss": 0.3072,
|
| 208688 |
+
"step": 85925
|
| 208689 |
+
},
|
| 208690 |
+
{
|
| 208691 |
+
"epoch": 692.51,
|
| 208692 |
+
"learning_rate": 8.628497576736672e-06,
|
| 208693 |
+
"loss": 0.3156,
|
| 208694 |
+
"step": 85930
|
| 208695 |
+
},
|
| 208696 |
+
{
|
| 208697 |
+
"epoch": 692.55,
|
| 208698 |
+
"learning_rate": 8.628416801292408e-06,
|
| 208699 |
+
"loss": 0.4384,
|
| 208700 |
+
"step": 85935
|
| 208701 |
+
},
|
| 208702 |
+
{
|
| 208703 |
+
"epoch": 692.59,
|
| 208704 |
+
"learning_rate": 8.628336025848142e-06,
|
| 208705 |
+
"loss": 0.9746,
|
| 208706 |
+
"step": 85940
|
| 208707 |
+
},
|
| 208708 |
+
{
|
| 208709 |
+
"epoch": 692.63,
|
| 208710 |
+
"learning_rate": 8.628255250403878e-06,
|
| 208711 |
+
"loss": 0.6991,
|
| 208712 |
+
"step": 85945
|
| 208713 |
+
},
|
| 208714 |
+
{
|
| 208715 |
+
"epoch": 692.67,
|
| 208716 |
+
"learning_rate": 8.628174474959612e-06,
|
| 208717 |
+
"loss": 0.2924,
|
| 208718 |
+
"step": 85950
|
| 208719 |
+
},
|
| 208720 |
+
{
|
| 208721 |
+
"epoch": 692.71,
|
| 208722 |
+
"learning_rate": 8.628093699515348e-06,
|
| 208723 |
+
"loss": 0.271,
|
| 208724 |
+
"step": 85955
|
| 208725 |
+
},
|
| 208726 |
+
{
|
| 208727 |
+
"epoch": 692.76,
|
| 208728 |
+
"learning_rate": 8.628012924071082e-06,
|
| 208729 |
+
"loss": 0.4228,
|
| 208730 |
+
"step": 85960
|
| 208731 |
+
},
|
| 208732 |
+
{
|
| 208733 |
+
"epoch": 692.8,
|
| 208734 |
+
"learning_rate": 8.627932148626818e-06,
|
| 208735 |
+
"loss": 0.9741,
|
| 208736 |
+
"step": 85965
|
| 208737 |
+
},
|
| 208738 |
+
{
|
| 208739 |
+
"epoch": 692.84,
|
| 208740 |
+
"learning_rate": 8.627851373182552e-06,
|
| 208741 |
+
"loss": 0.5785,
|
| 208742 |
+
"step": 85970
|
| 208743 |
+
},
|
| 208744 |
+
{
|
| 208745 |
+
"epoch": 692.88,
|
| 208746 |
+
"learning_rate": 8.627770597738288e-06,
|
| 208747 |
+
"loss": 0.2988,
|
| 208748 |
+
"step": 85975
|
| 208749 |
+
},
|
| 208750 |
+
{
|
| 208751 |
+
"epoch": 692.92,
|
| 208752 |
+
"learning_rate": 8.627689822294024e-06,
|
| 208753 |
+
"loss": 0.3336,
|
| 208754 |
+
"step": 85980
|
| 208755 |
+
},
|
| 208756 |
+
{
|
| 208757 |
+
"epoch": 692.96,
|
| 208758 |
+
"learning_rate": 8.627609046849758e-06,
|
| 208759 |
+
"loss": 0.373,
|
| 208760 |
+
"step": 85985
|
| 208761 |
+
},
|
| 208762 |
+
{
|
| 208763 |
+
"epoch": 693.0,
|
| 208764 |
+
"learning_rate": 8.627528271405494e-06,
|
| 208765 |
+
"loss": 0.9426,
|
| 208766 |
+
"step": 85990
|
| 208767 |
+
},
|
| 208768 |
+
{
|
| 208769 |
+
"epoch": 693.0,
|
| 208770 |
+
"eval_loss": 0.33618369698524475,
|
| 208771 |
+
"eval_runtime": 45.8818,
|
| 208772 |
+
"eval_samples_per_second": 18.308,
|
| 208773 |
+
"eval_steps_per_second": 0.588,
|
| 208774 |
+
"eval_wer": 0.1844791889769401,
|
| 208775 |
+
"step": 85990
|
| 208776 |
+
},
|
| 208777 |
+
{
|
| 208778 |
+
"epoch": 687.04,
|
| 208779 |
+
"learning_rate": 8.627447495961228e-06,
|
| 208780 |
+
"loss": 0.345,
|
| 208781 |
+
"step": 85995
|
| 208782 |
+
},
|
| 208783 |
+
{
|
| 208784 |
+
"epoch": 687.08,
|
| 208785 |
+
"learning_rate": 8.627366720516964e-06,
|
| 208786 |
+
"loss": 0.3009,
|
| 208787 |
+
"step": 86000
|
| 208788 |
+
},
|
| 208789 |
+
{
|
| 208790 |
+
"epoch": 687.12,
|
| 208791 |
+
"learning_rate": 8.627285945072698e-06,
|
| 208792 |
+
"loss": 0.3208,
|
| 208793 |
+
"step": 86005
|
| 208794 |
+
},
|
| 208795 |
+
{
|
| 208796 |
+
"epoch": 687.16,
|
| 208797 |
+
"learning_rate": 8.627205169628434e-06,
|
| 208798 |
+
"loss": 0.5082,
|
| 208799 |
+
"step": 86010
|
| 208800 |
+
},
|
| 208801 |
+
{
|
| 208802 |
+
"epoch": 687.2,
|
| 208803 |
+
"learning_rate": 8.627124394184168e-06,
|
| 208804 |
+
"loss": 1.2031,
|
| 208805 |
+
"step": 86015
|
| 208806 |
+
},
|
| 208807 |
+
{
|
| 208808 |
+
"epoch": 687.24,
|
| 208809 |
+
"learning_rate": 8.627043618739904e-06,
|
| 208810 |
+
"loss": 0.3518,
|
| 208811 |
+
"step": 86020
|
| 208812 |
+
},
|
| 208813 |
+
{
|
| 208814 |
+
"epoch": 687.28,
|
| 208815 |
+
"learning_rate": 8.626962843295638e-06,
|
| 208816 |
+
"loss": 0.3394,
|
| 208817 |
+
"step": 86025
|
| 208818 |
+
},
|
| 208819 |
+
{
|
| 208820 |
+
"epoch": 687.32,
|
| 208821 |
+
"learning_rate": 8.626882067851374e-06,
|
| 208822 |
+
"loss": 0.3621,
|
| 208823 |
+
"step": 86030
|
| 208824 |
+
},
|
| 208825 |
+
{
|
| 208826 |
+
"epoch": 687.36,
|
| 208827 |
+
"learning_rate": 8.626801292407108e-06,
|
| 208828 |
+
"loss": 0.457,
|
| 208829 |
+
"step": 86035
|
| 208830 |
+
},
|
| 208831 |
+
{
|
| 208832 |
+
"epoch": 687.4,
|
| 208833 |
+
"learning_rate": 8.626720516962844e-06,
|
| 208834 |
+
"loss": 1.3182,
|
| 208835 |
+
"step": 86040
|
| 208836 |
+
},
|
| 208837 |
+
{
|
| 208838 |
+
"epoch": 687.44,
|
| 208839 |
+
"learning_rate": 8.62663974151858e-06,
|
| 208840 |
+
"loss": 0.3442,
|
| 208841 |
+
"step": 86045
|
| 208842 |
+
},
|
| 208843 |
+
{
|
| 208844 |
+
"epoch": 687.48,
|
| 208845 |
+
"learning_rate": 8.626558966074314e-06,
|
| 208846 |
+
"loss": 0.3368,
|
| 208847 |
+
"step": 86050
|
| 208848 |
+
},
|
| 208849 |
+
{
|
| 208850 |
+
"epoch": 687.52,
|
| 208851 |
+
"learning_rate": 8.62647819063005e-06,
|
| 208852 |
+
"loss": 0.3164,
|
| 208853 |
+
"step": 86055
|
| 208854 |
+
},
|
| 208855 |
+
{
|
| 208856 |
+
"epoch": 687.56,
|
| 208857 |
+
"learning_rate": 8.626397415185784e-06,
|
| 208858 |
+
"loss": 0.462,
|
| 208859 |
+
"step": 86060
|
| 208860 |
+
},
|
| 208861 |
+
{
|
| 208862 |
+
"epoch": 687.6,
|
| 208863 |
+
"learning_rate": 8.62631663974152e-06,
|
| 208864 |
+
"loss": 1.2451,
|
| 208865 |
+
"step": 86065
|
| 208866 |
+
},
|
| 208867 |
+
{
|
| 208868 |
+
"epoch": 687.64,
|
| 208869 |
+
"learning_rate": 8.626235864297254e-06,
|
| 208870 |
+
"loss": 0.3596,
|
| 208871 |
+
"step": 86070
|
| 208872 |
+
},
|
| 208873 |
+
{
|
| 208874 |
+
"epoch": 687.68,
|
| 208875 |
+
"learning_rate": 8.62615508885299e-06,
|
| 208876 |
+
"loss": 0.3069,
|
| 208877 |
+
"step": 86075
|
| 208878 |
+
},
|
| 208879 |
+
{
|
| 208880 |
+
"epoch": 687.72,
|
| 208881 |
+
"learning_rate": 8.626074313408724e-06,
|
| 208882 |
+
"loss": 0.3492,
|
| 208883 |
+
"step": 86080
|
| 208884 |
+
},
|
| 208885 |
+
{
|
| 208886 |
+
"epoch": 687.76,
|
| 208887 |
+
"learning_rate": 8.62599353796446e-06,
|
| 208888 |
+
"loss": 0.4954,
|
| 208889 |
+
"step": 86085
|
| 208890 |
+
},
|
| 208891 |
+
{
|
| 208892 |
+
"epoch": 687.8,
|
| 208893 |
+
"learning_rate": 8.625912762520194e-06,
|
| 208894 |
+
"loss": 1.1525,
|
| 208895 |
+
"step": 86090
|
| 208896 |
+
},
|
| 208897 |
+
{
|
| 208898 |
+
"epoch": 687.84,
|
| 208899 |
+
"learning_rate": 8.62583198707593e-06,
|
| 208900 |
+
"loss": 0.3123,
|
| 208901 |
+
"step": 86095
|
| 208902 |
+
},
|
| 208903 |
+
{
|
| 208904 |
+
"epoch": 687.88,
|
| 208905 |
+
"learning_rate": 8.625751211631664e-06,
|
| 208906 |
+
"loss": 0.2866,
|
| 208907 |
+
"step": 86100
|
| 208908 |
+
},
|
| 208909 |
+
{
|
| 208910 |
+
"epoch": 687.92,
|
| 208911 |
+
"learning_rate": 8.6256704361874e-06,
|
| 208912 |
+
"loss": 0.3291,
|
| 208913 |
+
"step": 86105
|
| 208914 |
+
},
|
| 208915 |
+
{
|
| 208916 |
+
"epoch": 687.96,
|
| 208917 |
+
"learning_rate": 8.625589660743134e-06,
|
| 208918 |
+
"loss": 0.6972,
|
| 208919 |
+
"step": 86110
|
| 208920 |
+
},
|
| 208921 |
+
{
|
| 208922 |
+
"epoch": 688.0,
|
| 208923 |
+
"learning_rate": 8.62550888529887e-06,
|
| 208924 |
+
"loss": 1.366,
|
| 208925 |
+
"step": 86115
|
| 208926 |
+
},
|
| 208927 |
+
{
|
| 208928 |
+
"epoch": 688.0,
|
| 208929 |
+
"eval_loss": 0.32431867718696594,
|
| 208930 |
+
"eval_runtime": 41.5952,
|
| 208931 |
+
"eval_samples_per_second": 20.171,
|
| 208932 |
+
"eval_steps_per_second": 0.649,
|
| 208933 |
+
"eval_wer": 0.18960425280567042,
|
| 208934 |
+
"step": 86115
|
| 208935 |
+
},
|
| 208936 |
+
{
|
| 208937 |
+
"epoch": 694.04,
|
| 208938 |
+
"learning_rate": 8.625428109854605e-06,
|
| 208939 |
+
"loss": 0.3579,
|
| 208940 |
+
"step": 86120
|
| 208941 |
+
},
|
| 208942 |
+
{
|
| 208943 |
+
"epoch": 694.08,
|
| 208944 |
+
"learning_rate": 8.62534733441034e-06,
|
| 208945 |
+
"loss": 0.2554,
|
| 208946 |
+
"step": 86125
|
| 208947 |
+
},
|
| 208948 |
+
{
|
| 208949 |
+
"epoch": 694.12,
|
| 208950 |
+
"learning_rate": 8.625266558966075e-06,
|
| 208951 |
+
"loss": 0.3368,
|
| 208952 |
+
"step": 86130
|
| 208953 |
+
},
|
| 208954 |
+
{
|
| 208955 |
+
"epoch": 694.16,
|
| 208956 |
+
"learning_rate": 8.62518578352181e-06,
|
| 208957 |
+
"loss": 0.3954,
|
| 208958 |
+
"step": 86135
|
| 208959 |
+
},
|
| 208960 |
+
{
|
| 208961 |
+
"epoch": 694.2,
|
| 208962 |
+
"learning_rate": 8.625105008077545e-06,
|
| 208963 |
+
"loss": 1.1405,
|
| 208964 |
+
"step": 86140
|
| 208965 |
+
},
|
| 208966 |
+
{
|
| 208967 |
+
"epoch": 694.24,
|
| 208968 |
+
"learning_rate": 8.62502423263328e-06,
|
| 208969 |
+
"loss": 0.3505,
|
| 208970 |
+
"step": 86145
|
| 208971 |
+
},
|
| 208972 |
+
{
|
| 208973 |
+
"epoch": 694.28,
|
| 208974 |
+
"learning_rate": 8.624943457189015e-06,
|
| 208975 |
+
"loss": 0.3632,
|
| 208976 |
+
"step": 86150
|
| 208977 |
+
},
|
| 208978 |
+
{
|
| 208979 |
+
"epoch": 694.32,
|
| 208980 |
+
"learning_rate": 8.62486268174475e-06,
|
| 208981 |
+
"loss": 0.2912,
|
| 208982 |
+
"step": 86155
|
| 208983 |
+
},
|
| 208984 |
+
{
|
| 208985 |
+
"epoch": 694.36,
|
| 208986 |
+
"learning_rate": 8.624781906300485e-06,
|
| 208987 |
+
"loss": 0.4371,
|
| 208988 |
+
"step": 86160
|
| 208989 |
+
},
|
| 208990 |
+
{
|
| 208991 |
+
"epoch": 694.4,
|
| 208992 |
+
"learning_rate": 8.62470113085622e-06,
|
| 208993 |
+
"loss": 1.0906,
|
| 208994 |
+
"step": 86165
|
| 208995 |
+
},
|
| 208996 |
+
{
|
| 208997 |
+
"epoch": 694.44,
|
| 208998 |
+
"learning_rate": 8.624620355411955e-06,
|
| 208999 |
+
"loss": 0.3488,
|
| 209000 |
+
"step": 86170
|
| 209001 |
+
},
|
| 209002 |
+
{
|
| 209003 |
+
"epoch": 694.48,
|
| 209004 |
+
"learning_rate": 8.62453957996769e-06,
|
| 209005 |
+
"loss": 0.281,
|
| 209006 |
+
"step": 86175
|
| 209007 |
+
},
|
| 209008 |
+
{
|
| 209009 |
+
"epoch": 694.52,
|
| 209010 |
+
"learning_rate": 8.624458804523425e-06,
|
| 209011 |
+
"loss": 0.3886,
|
| 209012 |
+
"step": 86180
|
| 209013 |
+
},
|
| 209014 |
+
{
|
| 209015 |
+
"epoch": 694.56,
|
| 209016 |
+
"learning_rate": 8.624378029079161e-06,
|
| 209017 |
+
"loss": 0.4976,
|
| 209018 |
+
"step": 86185
|
| 209019 |
+
},
|
| 209020 |
+
{
|
| 209021 |
+
"epoch": 694.6,
|
| 209022 |
+
"learning_rate": 8.624297253634895e-06,
|
| 209023 |
+
"loss": 1.2062,
|
| 209024 |
+
"step": 86190
|
| 209025 |
+
},
|
| 209026 |
+
{
|
| 209027 |
+
"epoch": 694.64,
|
| 209028 |
+
"learning_rate": 8.624216478190631e-06,
|
| 209029 |
+
"loss": 0.3087,
|
| 209030 |
+
"step": 86195
|
| 209031 |
+
},
|
| 209032 |
+
{
|
| 209033 |
+
"epoch": 694.68,
|
| 209034 |
+
"learning_rate": 8.624135702746365e-06,
|
| 209035 |
+
"loss": 0.3002,
|
| 209036 |
+
"step": 86200
|
| 209037 |
+
},
|
| 209038 |
+
{
|
| 209039 |
+
"epoch": 694.72,
|
| 209040 |
+
"learning_rate": 8.624054927302101e-06,
|
| 209041 |
+
"loss": 0.314,
|
| 209042 |
+
"step": 86205
|
| 209043 |
+
},
|
| 209044 |
+
{
|
| 209045 |
+
"epoch": 694.76,
|
| 209046 |
+
"learning_rate": 8.623974151857835e-06,
|
| 209047 |
+
"loss": 0.4838,
|
| 209048 |
+
"step": 86210
|
| 209049 |
+
},
|
| 209050 |
+
{
|
| 209051 |
+
"epoch": 694.8,
|
| 209052 |
+
"learning_rate": 8.623893376413571e-06,
|
| 209053 |
+
"loss": 1.1279,
|
| 209054 |
+
"step": 86215
|
| 209055 |
+
},
|
| 209056 |
+
{
|
| 209057 |
+
"epoch": 694.84,
|
| 209058 |
+
"learning_rate": 8.623812600969305e-06,
|
| 209059 |
+
"loss": 0.3075,
|
| 209060 |
+
"step": 86220
|
| 209061 |
+
},
|
| 209062 |
+
{
|
| 209063 |
+
"epoch": 694.88,
|
| 209064 |
+
"learning_rate": 8.62373182552504e-06,
|
| 209065 |
+
"loss": 0.3139,
|
| 209066 |
+
"step": 86225
|
| 209067 |
+
},
|
| 209068 |
+
{
|
| 209069 |
+
"epoch": 694.92,
|
| 209070 |
+
"learning_rate": 8.623651050080775e-06,
|
| 209071 |
+
"loss": 0.3913,
|
| 209072 |
+
"step": 86230
|
| 209073 |
+
},
|
| 209074 |
+
{
|
| 209075 |
+
"epoch": 694.96,
|
| 209076 |
+
"learning_rate": 8.62357027463651e-06,
|
| 209077 |
+
"loss": 0.4897,
|
| 209078 |
+
"step": 86235
|
| 209079 |
+
},
|
| 209080 |
+
{
|
| 209081 |
+
"epoch": 695.0,
|
| 209082 |
+
"eval_loss": 0.3379378020763397,
|
| 209083 |
+
"eval_runtime": 41.768,
|
| 209084 |
+
"eval_samples_per_second": 20.087,
|
| 209085 |
+
"eval_steps_per_second": 0.646,
|
| 209086 |
+
"eval_wer": 0.18313218183132182,
|
| 209087 |
+
"step": 86239
|
| 209088 |
+
},
|
| 209089 |
+
{
|
| 209090 |
+
"epoch": 689.01,
|
| 209091 |
+
"learning_rate": 8.623489499192245e-06,
|
| 209092 |
+
"loss": 0.28,
|
| 209093 |
+
"step": 86240
|
| 209094 |
+
},
|
| 209095 |
+
{
|
| 209096 |
+
"epoch": 689.05,
|
| 209097 |
+
"learning_rate": 8.62340872374798e-06,
|
| 209098 |
+
"loss": 0.5216,
|
| 209099 |
+
"step": 86245
|
| 209100 |
+
},
|
| 209101 |
+
{
|
| 209102 |
+
"epoch": 689.09,
|
| 209103 |
+
"learning_rate": 8.623327948303717e-06,
|
| 209104 |
+
"loss": 0.3201,
|
| 209105 |
+
"step": 86250
|
| 209106 |
+
},
|
| 209107 |
+
{
|
| 209108 |
+
"epoch": 689.13,
|
| 209109 |
+
"learning_rate": 8.62324717285945e-06,
|
| 209110 |
+
"loss": 0.342,
|
| 209111 |
+
"step": 86255
|
| 209112 |
+
},
|
| 209113 |
+
{
|
| 209114 |
+
"epoch": 689.17,
|
| 209115 |
+
"learning_rate": 8.623166397415187e-06,
|
| 209116 |
+
"loss": 0.5204,
|
| 209117 |
+
"step": 86260
|
| 209118 |
+
},
|
| 209119 |
+
{
|
| 209120 |
+
"epoch": 689.21,
|
| 209121 |
+
"learning_rate": 8.62308562197092e-06,
|
| 209122 |
+
"loss": 1.0585,
|
| 209123 |
+
"step": 86265
|
| 209124 |
+
},
|
| 209125 |
+
{
|
| 209126 |
+
"epoch": 689.25,
|
| 209127 |
+
"learning_rate": 8.623004846526657e-06,
|
| 209128 |
+
"loss": 0.33,
|
| 209129 |
+
"step": 86270
|
| 209130 |
+
},
|
| 209131 |
+
{
|
| 209132 |
+
"epoch": 689.29,
|
| 209133 |
+
"learning_rate": 8.62292407108239e-06,
|
| 209134 |
+
"loss": 0.3383,
|
| 209135 |
+
"step": 86275
|
| 209136 |
+
},
|
| 209137 |
+
{
|
| 209138 |
+
"epoch": 689.33,
|
| 209139 |
+
"learning_rate": 8.622843295638127e-06,
|
| 209140 |
+
"loss": 0.3156,
|
| 209141 |
+
"step": 86280
|
| 209142 |
+
},
|
| 209143 |
+
{
|
| 209144 |
+
"epoch": 689.37,
|
| 209145 |
+
"learning_rate": 8.62276252019386e-06,
|
| 209146 |
+
"loss": 0.5836,
|
| 209147 |
+
"step": 86285
|
| 209148 |
+
},
|
| 209149 |
+
{
|
| 209150 |
+
"epoch": 689.41,
|
| 209151 |
+
"learning_rate": 8.622681744749597e-06,
|
| 209152 |
+
"loss": 1.1502,
|
| 209153 |
+
"step": 86290
|
| 209154 |
+
},
|
| 209155 |
+
{
|
| 209156 |
+
"epoch": 689.45,
|
| 209157 |
+
"learning_rate": 8.62260096930533e-06,
|
| 209158 |
+
"loss": 0.3082,
|
| 209159 |
+
"step": 86295
|
| 209160 |
+
},
|
| 209161 |
+
{
|
| 209162 |
+
"epoch": 689.49,
|
| 209163 |
+
"learning_rate": 8.622520193861066e-06,
|
| 209164 |
+
"loss": 0.2888,
|
| 209165 |
+
"step": 86300
|
| 209166 |
+
},
|
| 209167 |
+
{
|
| 209168 |
+
"epoch": 689.53,
|
| 209169 |
+
"learning_rate": 8.622439418416802e-06,
|
| 209170 |
+
"loss": 0.3934,
|
| 209171 |
+
"step": 86305
|
| 209172 |
+
},
|
| 209173 |
+
{
|
| 209174 |
+
"epoch": 689.57,
|
| 209175 |
+
"learning_rate": 8.622358642972536e-06,
|
| 209176 |
+
"loss": 0.6293,
|
| 209177 |
+
"step": 86310
|
| 209178 |
+
},
|
| 209179 |
+
{
|
| 209180 |
+
"epoch": 689.61,
|
| 209181 |
+
"learning_rate": 8.622277867528272e-06,
|
| 209182 |
+
"loss": 1.0595,
|
| 209183 |
+
"step": 86315
|
| 209184 |
+
},
|
| 209185 |
+
{
|
| 209186 |
+
"epoch": 689.65,
|
| 209187 |
+
"learning_rate": 8.622197092084006e-06,
|
| 209188 |
+
"loss": 0.3427,
|
| 209189 |
+
"step": 86320
|
| 209190 |
+
},
|
| 209191 |
+
{
|
| 209192 |
+
"epoch": 689.69,
|
| 209193 |
+
"learning_rate": 8.622116316639742e-06,
|
| 209194 |
+
"loss": 0.3128,
|
| 209195 |
+
"step": 86325
|
| 209196 |
+
},
|
| 209197 |
+
{
|
| 209198 |
+
"epoch": 689.73,
|
| 209199 |
+
"learning_rate": 8.622035541195476e-06,
|
| 209200 |
+
"loss": 0.3138,
|
| 209201 |
+
"step": 86330
|
| 209202 |
+
},
|
| 209203 |
+
{
|
| 209204 |
+
"epoch": 689.77,
|
| 209205 |
+
"learning_rate": 8.621954765751212e-06,
|
| 209206 |
+
"loss": 0.4887,
|
| 209207 |
+
"step": 86335
|
| 209208 |
+
},
|
| 209209 |
+
{
|
| 209210 |
+
"epoch": 689.81,
|
| 209211 |
+
"learning_rate": 8.621873990306946e-06,
|
| 209212 |
+
"loss": 1.1533,
|
| 209213 |
+
"step": 86340
|
| 209214 |
+
},
|
| 209215 |
+
{
|
| 209216 |
+
"epoch": 689.85,
|
| 209217 |
+
"learning_rate": 8.621793214862682e-06,
|
| 209218 |
+
"loss": 0.3736,
|
| 209219 |
+
"step": 86345
|
| 209220 |
+
},
|
| 209221 |
+
{
|
| 209222 |
+
"epoch": 689.89,
|
| 209223 |
+
"learning_rate": 8.621712439418416e-06,
|
| 209224 |
+
"loss": 0.3076,
|
| 209225 |
+
"step": 86350
|
| 209226 |
+
},
|
| 209227 |
+
{
|
| 209228 |
+
"epoch": 689.93,
|
| 209229 |
+
"learning_rate": 8.621631663974152e-06,
|
| 209230 |
+
"loss": 0.3878,
|
| 209231 |
+
"step": 86355
|
| 209232 |
+
},
|
| 209233 |
+
{
|
| 209234 |
+
"epoch": 689.97,
|
| 209235 |
+
"learning_rate": 8.621550888529886e-06,
|
| 209236 |
+
"loss": 0.5588,
|
| 209237 |
+
"step": 86360
|
| 209238 |
+
},
|
| 209239 |
+
{
|
| 209240 |
+
"epoch": 690.0,
|
| 209241 |
+
"eval_loss": 0.4273076355457306,
|
| 209242 |
+
"eval_runtime": 43.4189,
|
| 209243 |
+
"eval_samples_per_second": 19.323,
|
| 209244 |
+
"eval_steps_per_second": 0.622,
|
| 209245 |
+
"eval_wer": 0.18854415274463007,
|
| 209246 |
+
"step": 86364
|
| 209247 |
+
},
|
| 209248 |
+
{
|
| 209249 |
+
"epoch": 696.01,
|
| 209250 |
+
"learning_rate": 8.621470113085622e-06,
|
| 209251 |
+
"loss": 0.3859,
|
| 209252 |
+
"step": 86365
|
| 209253 |
+
},
|
| 209254 |
+
{
|
| 209255 |
+
"epoch": 696.05,
|
| 209256 |
+
"learning_rate": 8.621389337641358e-06,
|
| 209257 |
+
"loss": 0.3438,
|
| 209258 |
+
"step": 86370
|
| 209259 |
+
},
|
| 209260 |
+
{
|
| 209261 |
+
"epoch": 696.09,
|
| 209262 |
+
"learning_rate": 8.621308562197092e-06,
|
| 209263 |
+
"loss": 0.2927,
|
| 209264 |
+
"step": 86375
|
| 209265 |
+
},
|
| 209266 |
+
{
|
| 209267 |
+
"epoch": 696.13,
|
| 209268 |
+
"learning_rate": 8.621227786752828e-06,
|
| 209269 |
+
"loss": 0.3259,
|
| 209270 |
+
"step": 86380
|
| 209271 |
+
},
|
| 209272 |
+
{
|
| 209273 |
+
"epoch": 696.17,
|
| 209274 |
+
"learning_rate": 8.621147011308562e-06,
|
| 209275 |
+
"loss": 0.5074,
|
| 209276 |
+
"step": 86385
|
| 209277 |
+
},
|
| 209278 |
+
{
|
| 209279 |
+
"epoch": 696.21,
|
| 209280 |
+
"learning_rate": 8.621066235864298e-06,
|
| 209281 |
+
"loss": 1.1853,
|
| 209282 |
+
"step": 86390
|
| 209283 |
+
},
|
| 209284 |
+
{
|
| 209285 |
+
"epoch": 696.25,
|
| 209286 |
+
"learning_rate": 8.620985460420032e-06,
|
| 209287 |
+
"loss": 0.3457,
|
| 209288 |
+
"step": 86395
|
| 209289 |
+
},
|
| 209290 |
+
{
|
| 209291 |
+
"epoch": 696.29,
|
| 209292 |
+
"learning_rate": 8.620904684975768e-06,
|
| 209293 |
+
"loss": 1.0157,
|
| 209294 |
+
"step": 86400
|
| 209295 |
+
},
|
| 209296 |
+
{
|
| 209297 |
+
"epoch": 696.33,
|
| 209298 |
+
"learning_rate": 8.620823909531502e-06,
|
| 209299 |
+
"loss": 0.375,
|
| 209300 |
+
"step": 86405
|
| 209301 |
+
},
|
| 209302 |
+
{
|
| 209303 |
+
"epoch": 696.37,
|
| 209304 |
+
"learning_rate": 8.620743134087238e-06,
|
| 209305 |
+
"loss": 0.4831,
|
| 209306 |
+
"step": 86410
|
| 209307 |
+
},
|
| 209308 |
+
{
|
| 209309 |
+
"epoch": 696.41,
|
| 209310 |
+
"learning_rate": 8.620662358642972e-06,
|
| 209311 |
+
"loss": 1.0284,
|
| 209312 |
+
"step": 86415
|
| 209313 |
+
},
|
| 209314 |
+
{
|
| 209315 |
+
"epoch": 696.45,
|
| 209316 |
+
"learning_rate": 8.620581583198708e-06,
|
| 209317 |
+
"loss": 0.3448,
|
| 209318 |
+
"step": 86420
|
| 209319 |
+
},
|
| 209320 |
+
{
|
| 209321 |
+
"epoch": 696.49,
|
| 209322 |
+
"learning_rate": 8.620500807754444e-06,
|
| 209323 |
+
"loss": 0.2996,
|
| 209324 |
+
"step": 86425
|
| 209325 |
+
},
|
| 209326 |
+
{
|
| 209327 |
+
"epoch": 696.53,
|
| 209328 |
+
"learning_rate": 8.620420032310178e-06,
|
| 209329 |
+
"loss": 0.3335,
|
| 209330 |
+
"step": 86430
|
| 209331 |
+
},
|
| 209332 |
+
{
|
| 209333 |
+
"epoch": 696.57,
|
| 209334 |
+
"learning_rate": 8.620339256865914e-06,
|
| 209335 |
+
"loss": 0.499,
|
| 209336 |
+
"step": 86435
|
| 209337 |
+
},
|
| 209338 |
+
{
|
| 209339 |
+
"epoch": 696.61,
|
| 209340 |
+
"learning_rate": 8.620258481421648e-06,
|
| 209341 |
+
"loss": 1.0738,
|
| 209342 |
+
"step": 86440
|
| 209343 |
+
},
|
| 209344 |
+
{
|
| 209345 |
+
"epoch": 696.65,
|
| 209346 |
+
"learning_rate": 8.620177705977384e-06,
|
| 209347 |
+
"loss": 0.3173,
|
| 209348 |
+
"step": 86445
|
| 209349 |
+
},
|
| 209350 |
+
{
|
| 209351 |
+
"epoch": 696.69,
|
| 209352 |
+
"learning_rate": 8.620096930533118e-06,
|
| 209353 |
+
"loss": 0.2994,
|
| 209354 |
+
"step": 86450
|
| 209355 |
+
},
|
| 209356 |
+
{
|
| 209357 |
+
"epoch": 696.73,
|
| 209358 |
+
"learning_rate": 8.620016155088854e-06,
|
| 209359 |
+
"loss": 0.3495,
|
| 209360 |
+
"step": 86455
|
| 209361 |
+
},
|
| 209362 |
+
{
|
| 209363 |
+
"epoch": 696.77,
|
| 209364 |
+
"learning_rate": 8.619935379644588e-06,
|
| 209365 |
+
"loss": 0.5834,
|
| 209366 |
+
"step": 86460
|
| 209367 |
+
},
|
| 209368 |
+
{
|
| 209369 |
+
"epoch": 696.81,
|
| 209370 |
+
"learning_rate": 8.619854604200324e-06,
|
| 209371 |
+
"loss": 1.2537,
|
| 209372 |
+
"step": 86465
|
| 209373 |
+
},
|
| 209374 |
+
{
|
| 209375 |
+
"epoch": 696.85,
|
| 209376 |
+
"learning_rate": 8.619773828756058e-06,
|
| 209377 |
+
"loss": 0.3,
|
| 209378 |
+
"step": 86470
|
| 209379 |
+
},
|
| 209380 |
+
{
|
| 209381 |
+
"epoch": 696.89,
|
| 209382 |
+
"learning_rate": 8.619693053311794e-06,
|
| 209383 |
+
"loss": 0.3382,
|
| 209384 |
+
"step": 86475
|
| 209385 |
+
},
|
| 209386 |
+
{
|
| 209387 |
+
"epoch": 696.93,
|
| 209388 |
+
"learning_rate": 8.61961227786753e-06,
|
| 209389 |
+
"loss": 0.3447,
|
| 209390 |
+
"step": 86480
|
| 209391 |
+
},
|
| 209392 |
+
{
|
| 209393 |
+
"epoch": 696.97,
|
| 209394 |
+
"learning_rate": 8.619531502423264e-06,
|
| 209395 |
+
"loss": 0.6205,
|
| 209396 |
+
"step": 86485
|
| 209397 |
+
},
|
| 209398 |
+
{
|
| 209399 |
+
"epoch": 697.0,
|
| 209400 |
+
"eval_loss": 0.3951117694377899,
|
| 209401 |
+
"eval_runtime": 41.9031,
|
| 209402 |
+
"eval_samples_per_second": 20.022,
|
| 209403 |
+
"eval_steps_per_second": 0.644,
|
| 209404 |
+
"eval_wer": 0.18736416098328712,
|
| 209405 |
+
"step": 86488
|
| 209406 |
}
|
| 209407 |
],
|
| 209408 |
"max_steps": 620000,
|
| 209409 |
"num_train_epochs": 5000,
|
| 209410 |
+
"total_flos": 2.434054730572205e+20,
|
| 209411 |
"trial_name": null,
|
| 209412 |
"trial_params": null
|
| 209413 |
}
|
model-bin/finetune/base/{checkpoint-85866 → checkpoint-86488}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629925935.501896/events.out.tfevents.1629925935.7e498afd5545.7645.155
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:10d35973eca77c60db745e002bb5a631b5bf7d5676ad910bc1835a11ffd27a01
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629926415.8598046/events.out.tfevents.1629926415.7e498afd5545.7645.157
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:438fd695c26b7b472fd72afe695ec74b4fb7aba6ff66ce7fe3cc518c6c69d2e3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629926892.2024448/events.out.tfevents.1629926894.7e498afd5545.7645.159
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:44dad85a7a60b622c5daf5bf80aefba6937bf41fb14e5460227e0a67a70901c0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629927362.1056397/events.out.tfevents.1629927362.7e498afd5545.7645.161
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:916c557b6636c8b3a4942acff7e70d28baa9a540ffe1b43049c06f5757066b51
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629927831.7978618/events.out.tfevents.1629927831.7e498afd5545.7645.163
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8c861c7ca7554ac3a20d873fed86fad3abc57cafab6550c4112487b57b2d45a2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629925935.7e498afd5545.7645.154
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:668fcc8b68cca7bef549e72b974308869eaeb3b0faa0be94f699d9c41cae17f6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629926415.7e498afd5545.7645.156
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a451b4bdf900516fc66d812eff3dc384b66ca340255f264acf1f18dc033c6dc6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629926892.7e498afd5545.7645.158
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:996bb611b82d79219d30f5e750ba4f0cf8dee9f09c4c493f667c145a1dbaa5f4
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629927362.7e498afd5545.7645.160
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:61720124e9f46598b64ca4b443547bc2f1268d14ad24a8c4405128bbf08c1eb6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629927831.7e498afd5545.7645.162
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:63c815b96007b09b21368cd671bfef0e48268901c1ad0c1e7c877d6741652bc7
|
| 3 |
+
size 8622
|