"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129674}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129674}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129674}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129674}/trainer_state.json +636 -6
- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/training_args.bin +0 -0
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129799}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129799}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129799}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-124947 → checkpoint-129799}/trainer_state.json +0 -0
- model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630158854.6730864/events.out.tfevents.1630158854.86bb0ddabf9b.4092.181 +3 -0
- model-bin/finetune/base/log/1630159245.4578865/events.out.tfevents.1630159245.86bb0ddabf9b.4092.183 +3 -0
- model-bin/finetune/base/log/1630159646.2298772/events.out.tfevents.1630159646.86bb0ddabf9b.4092.185 +3 -0
- model-bin/finetune/base/log/1630160047.7808826/events.out.tfevents.1630160047.86bb0ddabf9b.4092.187 +3 -0
- model-bin/finetune/base/log/1630160440.877656/events.out.tfevents.1630160440.86bb0ddabf9b.4092.189 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630158854.86bb0ddabf9b.4092.180 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630159245.86bb0ddabf9b.4092.182 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630159646.86bb0ddabf9b.4092.184 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630160047.86bb0ddabf9b.4092.186 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630160440.86bb0ddabf9b.4092.188 +3 -0
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/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 sha256:2ab0ea0f33fc607a3d20c20fb5943d8f60aa638ab2c397abfa1add25fb550342
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129674}/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 sha256:38eb82634b3b5e6d8ef274e02321900b21a2f83bc07066b3483ebf2ee80700c0
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129674}/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:f554dfcdada4394fc7e8bf4ff181a820cba3bd5ee38a9705118bfd13f34d5c3c
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129674}/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 sha256:4dedda88080e30bde75dd07d24e3957d07cf65d683f3fcca5da54a4d262734ab
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/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:d49936a1e9f29b7739dd609185c56e6246ae02806e09e5605ecbe08b50f9022b
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129674}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -263718,11 +263718,641 @@
|
|
| 263718 |
"eval_steps_per_second": 0.721,
|
| 263719 |
"eval_wer": 0.1812391430225825,
|
| 263720 |
"step": 129176
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 263721 |
}
|
| 263722 |
],
|
| 263723 |
-
"max_steps":
|
| 263724 |
"num_train_epochs": 5000,
|
| 263725 |
-
"total_flos": 3.
|
| 263726 |
"trial_name": null,
|
| 263727 |
"trial_params": null
|
| 263728 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.17105075053609722,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-129674",
|
| 4 |
+
"epoch": 1037.0,
|
| 5 |
+
"global_step": 129674,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 263718 |
"eval_steps_per_second": 0.721,
|
| 263719 |
"eval_wer": 0.1812391430225825,
|
| 263720 |
"step": 129176
|
| 263721 |
+
},
|
| 263722 |
+
{
|
| 263723 |
+
"epoch": 1033.03,
|
| 263724 |
+
"learning_rate": 7.946634615384616e-06,
|
| 263725 |
+
"loss": 0.336,
|
| 263726 |
+
"step": 129180
|
| 263727 |
+
},
|
| 263728 |
+
{
|
| 263729 |
+
"epoch": 1033.07,
|
| 263730 |
+
"learning_rate": 7.946554487179487e-06,
|
| 263731 |
+
"loss": 0.2629,
|
| 263732 |
+
"step": 129185
|
| 263733 |
+
},
|
| 263734 |
+
{
|
| 263735 |
+
"epoch": 1033.11,
|
| 263736 |
+
"learning_rate": 7.94647435897436e-06,
|
| 263737 |
+
"loss": 0.2842,
|
| 263738 |
+
"step": 129190
|
| 263739 |
+
},
|
| 263740 |
+
{
|
| 263741 |
+
"epoch": 1033.15,
|
| 263742 |
+
"learning_rate": 7.946394230769232e-06,
|
| 263743 |
+
"loss": 0.368,
|
| 263744 |
+
"step": 129195
|
| 263745 |
+
},
|
| 263746 |
+
{
|
| 263747 |
+
"epoch": 1033.19,
|
| 263748 |
+
"learning_rate": 7.946314102564103e-06,
|
| 263749 |
+
"loss": 0.8138,
|
| 263750 |
+
"step": 129200
|
| 263751 |
+
},
|
| 263752 |
+
{
|
| 263753 |
+
"epoch": 1033.23,
|
| 263754 |
+
"learning_rate": 7.946233974358974e-06,
|
| 263755 |
+
"loss": 0.6127,
|
| 263756 |
+
"step": 129205
|
| 263757 |
+
},
|
| 263758 |
+
{
|
| 263759 |
+
"epoch": 1033.27,
|
| 263760 |
+
"learning_rate": 7.946153846153847e-06,
|
| 263761 |
+
"loss": 0.2991,
|
| 263762 |
+
"step": 129210
|
| 263763 |
+
},
|
| 263764 |
+
{
|
| 263765 |
+
"epoch": 1033.31,
|
| 263766 |
+
"learning_rate": 7.946073717948719e-06,
|
| 263767 |
+
"loss": 0.268,
|
| 263768 |
+
"step": 129215
|
| 263769 |
+
},
|
| 263770 |
+
{
|
| 263771 |
+
"epoch": 1033.35,
|
| 263772 |
+
"learning_rate": 7.94599358974359e-06,
|
| 263773 |
+
"loss": 0.3597,
|
| 263774 |
+
"step": 129220
|
| 263775 |
+
},
|
| 263776 |
+
{
|
| 263777 |
+
"epoch": 1033.39,
|
| 263778 |
+
"learning_rate": 7.945913461538463e-06,
|
| 263779 |
+
"loss": 0.9515,
|
| 263780 |
+
"step": 129225
|
| 263781 |
+
},
|
| 263782 |
+
{
|
| 263783 |
+
"epoch": 1033.43,
|
| 263784 |
+
"learning_rate": 7.945833333333334e-06,
|
| 263785 |
+
"loss": 0.6679,
|
| 263786 |
+
"step": 129230
|
| 263787 |
+
},
|
| 263788 |
+
{
|
| 263789 |
+
"epoch": 1033.47,
|
| 263790 |
+
"learning_rate": 7.945753205128206e-06,
|
| 263791 |
+
"loss": 0.2819,
|
| 263792 |
+
"step": 129235
|
| 263793 |
+
},
|
| 263794 |
+
{
|
| 263795 |
+
"epoch": 1033.51,
|
| 263796 |
+
"learning_rate": 7.945673076923077e-06,
|
| 263797 |
+
"loss": 0.2869,
|
| 263798 |
+
"step": 129240
|
| 263799 |
+
},
|
| 263800 |
+
{
|
| 263801 |
+
"epoch": 1033.55,
|
| 263802 |
+
"learning_rate": 7.94559294871795e-06,
|
| 263803 |
+
"loss": 0.3977,
|
| 263804 |
+
"step": 129245
|
| 263805 |
+
},
|
| 263806 |
+
{
|
| 263807 |
+
"epoch": 1033.59,
|
| 263808 |
+
"learning_rate": 7.94551282051282e-06,
|
| 263809 |
+
"loss": 0.9283,
|
| 263810 |
+
"step": 129250
|
| 263811 |
+
},
|
| 263812 |
+
{
|
| 263813 |
+
"epoch": 1033.63,
|
| 263814 |
+
"learning_rate": 7.945432692307693e-06,
|
| 263815 |
+
"loss": 0.6327,
|
| 263816 |
+
"step": 129255
|
| 263817 |
+
},
|
| 263818 |
+
{
|
| 263819 |
+
"epoch": 1033.67,
|
| 263820 |
+
"learning_rate": 7.945352564102566e-06,
|
| 263821 |
+
"loss": 0.2767,
|
| 263822 |
+
"step": 129260
|
| 263823 |
+
},
|
| 263824 |
+
{
|
| 263825 |
+
"epoch": 1033.71,
|
| 263826 |
+
"learning_rate": 7.945272435897436e-06,
|
| 263827 |
+
"loss": 0.3074,
|
| 263828 |
+
"step": 129265
|
| 263829 |
+
},
|
| 263830 |
+
{
|
| 263831 |
+
"epoch": 1033.75,
|
| 263832 |
+
"learning_rate": 7.945192307692309e-06,
|
| 263833 |
+
"loss": 0.3771,
|
| 263834 |
+
"step": 129270
|
| 263835 |
+
},
|
| 263836 |
+
{
|
| 263837 |
+
"epoch": 1033.79,
|
| 263838 |
+
"learning_rate": 7.94511217948718e-06,
|
| 263839 |
+
"loss": 0.9088,
|
| 263840 |
+
"step": 129275
|
| 263841 |
+
},
|
| 263842 |
+
{
|
| 263843 |
+
"epoch": 1033.83,
|
| 263844 |
+
"learning_rate": 7.945032051282051e-06,
|
| 263845 |
+
"loss": 0.5685,
|
| 263846 |
+
"step": 129280
|
| 263847 |
+
},
|
| 263848 |
+
{
|
| 263849 |
+
"epoch": 1033.87,
|
| 263850 |
+
"learning_rate": 7.944951923076923e-06,
|
| 263851 |
+
"loss": 0.465,
|
| 263852 |
+
"step": 129285
|
| 263853 |
+
},
|
| 263854 |
+
{
|
| 263855 |
+
"epoch": 1033.91,
|
| 263856 |
+
"learning_rate": 7.944871794871796e-06,
|
| 263857 |
+
"loss": 0.2975,
|
| 263858 |
+
"step": 129290
|
| 263859 |
+
},
|
| 263860 |
+
{
|
| 263861 |
+
"epoch": 1033.95,
|
| 263862 |
+
"learning_rate": 7.944791666666667e-06,
|
| 263863 |
+
"loss": 0.4209,
|
| 263864 |
+
"step": 129295
|
| 263865 |
+
},
|
| 263866 |
+
{
|
| 263867 |
+
"epoch": 1033.99,
|
| 263868 |
+
"learning_rate": 7.944711538461539e-06,
|
| 263869 |
+
"loss": 1.091,
|
| 263870 |
+
"step": 129300
|
| 263871 |
+
},
|
| 263872 |
+
{
|
| 263873 |
+
"epoch": 1034.0,
|
| 263874 |
+
"eval_loss": 0.4637022018432617,
|
| 263875 |
+
"eval_runtime": 37.1126,
|
| 263876 |
+
"eval_samples_per_second": 22.526,
|
| 263877 |
+
"eval_steps_per_second": 0.728,
|
| 263878 |
+
"eval_wer": 0.17635815628905388,
|
| 263879 |
+
"step": 129301
|
| 263880 |
+
},
|
| 263881 |
+
{
|
| 263882 |
+
"epoch": 1042.03,
|
| 263883 |
+
"learning_rate": 7.94463141025641e-06,
|
| 263884 |
+
"loss": 0.3302,
|
| 263885 |
+
"step": 129305
|
| 263886 |
+
},
|
| 263887 |
+
{
|
| 263888 |
+
"epoch": 1042.07,
|
| 263889 |
+
"learning_rate": 7.944551282051283e-06,
|
| 263890 |
+
"loss": 0.2612,
|
| 263891 |
+
"step": 129310
|
| 263892 |
+
},
|
| 263893 |
+
{
|
| 263894 |
+
"epoch": 1042.11,
|
| 263895 |
+
"learning_rate": 7.944471153846154e-06,
|
| 263896 |
+
"loss": 0.298,
|
| 263897 |
+
"step": 129315
|
| 263898 |
+
},
|
| 263899 |
+
{
|
| 263900 |
+
"epoch": 1042.15,
|
| 263901 |
+
"learning_rate": 7.944391025641026e-06,
|
| 263902 |
+
"loss": 0.4075,
|
| 263903 |
+
"step": 129320
|
| 263904 |
+
},
|
| 263905 |
+
{
|
| 263906 |
+
"epoch": 1042.19,
|
| 263907 |
+
"learning_rate": 7.944310897435899e-06,
|
| 263908 |
+
"loss": 0.9383,
|
| 263909 |
+
"step": 129325
|
| 263910 |
+
},
|
| 263911 |
+
{
|
| 263912 |
+
"epoch": 1042.23,
|
| 263913 |
+
"learning_rate": 7.94423076923077e-06,
|
| 263914 |
+
"loss": 0.6966,
|
| 263915 |
+
"step": 129330
|
| 263916 |
+
},
|
| 263917 |
+
{
|
| 263918 |
+
"epoch": 1042.27,
|
| 263919 |
+
"learning_rate": 7.944150641025641e-06,
|
| 263920 |
+
"loss": 0.2463,
|
| 263921 |
+
"step": 129335
|
| 263922 |
+
},
|
| 263923 |
+
{
|
| 263924 |
+
"epoch": 1042.31,
|
| 263925 |
+
"learning_rate": 7.944070512820513e-06,
|
| 263926 |
+
"loss": 0.3064,
|
| 263927 |
+
"step": 129340
|
| 263928 |
+
},
|
| 263929 |
+
{
|
| 263930 |
+
"epoch": 1042.35,
|
| 263931 |
+
"learning_rate": 7.943990384615386e-06,
|
| 263932 |
+
"loss": 0.4184,
|
| 263933 |
+
"step": 129345
|
| 263934 |
+
},
|
| 263935 |
+
{
|
| 263936 |
+
"epoch": 1042.39,
|
| 263937 |
+
"learning_rate": 7.943910256410257e-06,
|
| 263938 |
+
"loss": 0.9273,
|
| 263939 |
+
"step": 129350
|
| 263940 |
+
},
|
| 263941 |
+
{
|
| 263942 |
+
"epoch": 1042.43,
|
| 263943 |
+
"learning_rate": 7.943830128205129e-06,
|
| 263944 |
+
"loss": 0.6881,
|
| 263945 |
+
"step": 129355
|
| 263946 |
+
},
|
| 263947 |
+
{
|
| 263948 |
+
"epoch": 1042.47,
|
| 263949 |
+
"learning_rate": 7.943750000000002e-06,
|
| 263950 |
+
"loss": 0.2911,
|
| 263951 |
+
"step": 129360
|
| 263952 |
+
},
|
| 263953 |
+
{
|
| 263954 |
+
"epoch": 1042.51,
|
| 263955 |
+
"learning_rate": 7.943669871794873e-06,
|
| 263956 |
+
"loss": 0.2801,
|
| 263957 |
+
"step": 129365
|
| 263958 |
+
},
|
| 263959 |
+
{
|
| 263960 |
+
"epoch": 1042.55,
|
| 263961 |
+
"learning_rate": 7.943589743589744e-06,
|
| 263962 |
+
"loss": 0.3756,
|
| 263963 |
+
"step": 129370
|
| 263964 |
+
},
|
| 263965 |
+
{
|
| 263966 |
+
"epoch": 1042.59,
|
| 263967 |
+
"learning_rate": 7.943509615384616e-06,
|
| 263968 |
+
"loss": 0.9498,
|
| 263969 |
+
"step": 129375
|
| 263970 |
+
},
|
| 263971 |
+
{
|
| 263972 |
+
"epoch": 1042.63,
|
| 263973 |
+
"learning_rate": 7.943429487179489e-06,
|
| 263974 |
+
"loss": 0.694,
|
| 263975 |
+
"step": 129380
|
| 263976 |
+
},
|
| 263977 |
+
{
|
| 263978 |
+
"epoch": 1042.67,
|
| 263979 |
+
"learning_rate": 7.94334935897436e-06,
|
| 263980 |
+
"loss": 0.2574,
|
| 263981 |
+
"step": 129385
|
| 263982 |
+
},
|
| 263983 |
+
{
|
| 263984 |
+
"epoch": 1042.71,
|
| 263985 |
+
"learning_rate": 7.943269230769232e-06,
|
| 263986 |
+
"loss": 0.2722,
|
| 263987 |
+
"step": 129390
|
| 263988 |
+
},
|
| 263989 |
+
{
|
| 263990 |
+
"epoch": 1042.76,
|
| 263991 |
+
"learning_rate": 7.943189102564103e-06,
|
| 263992 |
+
"loss": 0.3887,
|
| 263993 |
+
"step": 129395
|
| 263994 |
+
},
|
| 263995 |
+
{
|
| 263996 |
+
"epoch": 1042.8,
|
| 263997 |
+
"learning_rate": 7.943108974358976e-06,
|
| 263998 |
+
"loss": 0.8722,
|
| 263999 |
+
"step": 129400
|
| 264000 |
+
},
|
| 264001 |
+
{
|
| 264002 |
+
"epoch": 1042.84,
|
| 264003 |
+
"learning_rate": 7.943028846153846e-06,
|
| 264004 |
+
"loss": 0.6307,
|
| 264005 |
+
"step": 129405
|
| 264006 |
+
},
|
| 264007 |
+
{
|
| 264008 |
+
"epoch": 1042.88,
|
| 264009 |
+
"learning_rate": 7.942948717948719e-06,
|
| 264010 |
+
"loss": 0.3165,
|
| 264011 |
+
"step": 129410
|
| 264012 |
+
},
|
| 264013 |
+
{
|
| 264014 |
+
"epoch": 1042.92,
|
| 264015 |
+
"learning_rate": 7.942868589743592e-06,
|
| 264016 |
+
"loss": 0.2951,
|
| 264017 |
+
"step": 129415
|
| 264018 |
+
},
|
| 264019 |
+
{
|
| 264020 |
+
"epoch": 1042.96,
|
| 264021 |
+
"learning_rate": 7.942788461538461e-06,
|
| 264022 |
+
"loss": 0.4033,
|
| 264023 |
+
"step": 129420
|
| 264024 |
+
},
|
| 264025 |
+
{
|
| 264026 |
+
"epoch": 1043.0,
|
| 264027 |
+
"learning_rate": 7.942708333333334e-06,
|
| 264028 |
+
"loss": 0.9835,
|
| 264029 |
+
"step": 129425
|
| 264030 |
+
},
|
| 264031 |
+
{
|
| 264032 |
+
"epoch": 1043.0,
|
| 264033 |
+
"eval_loss": 0.34297874569892883,
|
| 264034 |
+
"eval_runtime": 36.6288,
|
| 264035 |
+
"eval_samples_per_second": 22.824,
|
| 264036 |
+
"eval_steps_per_second": 0.737,
|
| 264037 |
+
"eval_wer": 0.18068257863038145,
|
| 264038 |
+
"step": 129425
|
| 264039 |
+
},
|
| 264040 |
+
{
|
| 264041 |
+
"epoch": 1043.04,
|
| 264042 |
+
"learning_rate": 7.942628205128206e-06,
|
| 264043 |
+
"loss": 0.3216,
|
| 264044 |
+
"step": 129430
|
| 264045 |
+
},
|
| 264046 |
+
{
|
| 264047 |
+
"epoch": 1043.08,
|
| 264048 |
+
"learning_rate": 7.942548076923077e-06,
|
| 264049 |
+
"loss": 0.2541,
|
| 264050 |
+
"step": 129435
|
| 264051 |
+
},
|
| 264052 |
+
{
|
| 264053 |
+
"epoch": 1043.12,
|
| 264054 |
+
"learning_rate": 7.942467948717948e-06,
|
| 264055 |
+
"loss": 0.3287,
|
| 264056 |
+
"step": 129440
|
| 264057 |
+
},
|
| 264058 |
+
{
|
| 264059 |
+
"epoch": 1043.16,
|
| 264060 |
+
"learning_rate": 7.942387820512822e-06,
|
| 264061 |
+
"loss": 0.4313,
|
| 264062 |
+
"step": 129445
|
| 264063 |
+
},
|
| 264064 |
+
{
|
| 264065 |
+
"epoch": 1043.2,
|
| 264066 |
+
"learning_rate": 7.942307692307693e-06,
|
| 264067 |
+
"loss": 1.3484,
|
| 264068 |
+
"step": 129450
|
| 264069 |
+
},
|
| 264070 |
+
{
|
| 264071 |
+
"epoch": 1043.24,
|
| 264072 |
+
"learning_rate": 7.942227564102564e-06,
|
| 264073 |
+
"loss": 0.3509,
|
| 264074 |
+
"step": 129455
|
| 264075 |
+
},
|
| 264076 |
+
{
|
| 264077 |
+
"epoch": 1043.28,
|
| 264078 |
+
"learning_rate": 7.942147435897437e-06,
|
| 264079 |
+
"loss": 0.2928,
|
| 264080 |
+
"step": 129460
|
| 264081 |
+
},
|
| 264082 |
+
{
|
| 264083 |
+
"epoch": 1043.32,
|
| 264084 |
+
"learning_rate": 7.942067307692309e-06,
|
| 264085 |
+
"loss": 0.3163,
|
| 264086 |
+
"step": 129465
|
| 264087 |
+
},
|
| 264088 |
+
{
|
| 264089 |
+
"epoch": 1043.36,
|
| 264090 |
+
"learning_rate": 7.94198717948718e-06,
|
| 264091 |
+
"loss": 0.3945,
|
| 264092 |
+
"step": 129470
|
| 264093 |
+
},
|
| 264094 |
+
{
|
| 264095 |
+
"epoch": 1043.4,
|
| 264096 |
+
"learning_rate": 7.941923076923077e-06,
|
| 264097 |
+
"loss": 2.4296,
|
| 264098 |
+
"step": 129475
|
| 264099 |
+
},
|
| 264100 |
+
{
|
| 264101 |
+
"epoch": 1043.44,
|
| 264102 |
+
"learning_rate": 7.941842948717949e-06,
|
| 264103 |
+
"loss": 0.3489,
|
| 264104 |
+
"step": 129480
|
| 264105 |
+
},
|
| 264106 |
+
{
|
| 264107 |
+
"epoch": 1043.48,
|
| 264108 |
+
"learning_rate": 7.941762820512822e-06,
|
| 264109 |
+
"loss": 0.2532,
|
| 264110 |
+
"step": 129485
|
| 264111 |
+
},
|
| 264112 |
+
{
|
| 264113 |
+
"epoch": 1043.52,
|
| 264114 |
+
"learning_rate": 7.941682692307693e-06,
|
| 264115 |
+
"loss": 0.3105,
|
| 264116 |
+
"step": 129490
|
| 264117 |
+
},
|
| 264118 |
+
{
|
| 264119 |
+
"epoch": 1043.56,
|
| 264120 |
+
"learning_rate": 7.941602564102565e-06,
|
| 264121 |
+
"loss": 0.4227,
|
| 264122 |
+
"step": 129495
|
| 264123 |
+
},
|
| 264124 |
+
{
|
| 264125 |
+
"epoch": 1043.6,
|
| 264126 |
+
"learning_rate": 7.941522435897436e-06,
|
| 264127 |
+
"loss": 1.0799,
|
| 264128 |
+
"step": 129500
|
| 264129 |
+
},
|
| 264130 |
+
{
|
| 264131 |
+
"epoch": 1043.65,
|
| 264132 |
+
"learning_rate": 7.941442307692309e-06,
|
| 264133 |
+
"loss": 0.3115,
|
| 264134 |
+
"step": 129505
|
| 264135 |
+
},
|
| 264136 |
+
{
|
| 264137 |
+
"epoch": 1043.69,
|
| 264138 |
+
"learning_rate": 7.94136217948718e-06,
|
| 264139 |
+
"loss": 0.2782,
|
| 264140 |
+
"step": 129510
|
| 264141 |
+
},
|
| 264142 |
+
{
|
| 264143 |
+
"epoch": 1043.73,
|
| 264144 |
+
"learning_rate": 7.941282051282052e-06,
|
| 264145 |
+
"loss": 0.3032,
|
| 264146 |
+
"step": 129515
|
| 264147 |
+
},
|
| 264148 |
+
{
|
| 264149 |
+
"epoch": 1043.77,
|
| 264150 |
+
"learning_rate": 7.941201923076923e-06,
|
| 264151 |
+
"loss": 0.417,
|
| 264152 |
+
"step": 129520
|
| 264153 |
+
},
|
| 264154 |
+
{
|
| 264155 |
+
"epoch": 1043.81,
|
| 264156 |
+
"learning_rate": 7.941121794871796e-06,
|
| 264157 |
+
"loss": 1.0323,
|
| 264158 |
+
"step": 129525
|
| 264159 |
+
},
|
| 264160 |
+
{
|
| 264161 |
+
"epoch": 1043.85,
|
| 264162 |
+
"learning_rate": 7.941041666666667e-06,
|
| 264163 |
+
"loss": 0.3683,
|
| 264164 |
+
"step": 129530
|
| 264165 |
+
},
|
| 264166 |
+
{
|
| 264167 |
+
"epoch": 1043.89,
|
| 264168 |
+
"learning_rate": 7.940961538461539e-06,
|
| 264169 |
+
"loss": 0.262,
|
| 264170 |
+
"step": 129535
|
| 264171 |
+
},
|
| 264172 |
+
{
|
| 264173 |
+
"epoch": 1043.93,
|
| 264174 |
+
"learning_rate": 7.940881410256412e-06,
|
| 264175 |
+
"loss": 0.2751,
|
| 264176 |
+
"step": 129540
|
| 264177 |
+
},
|
| 264178 |
+
{
|
| 264179 |
+
"epoch": 1043.97,
|
| 264180 |
+
"learning_rate": 7.940801282051283e-06,
|
| 264181 |
+
"loss": 0.6299,
|
| 264182 |
+
"step": 129545
|
| 264183 |
+
},
|
| 264184 |
+
{
|
| 264185 |
+
"epoch": 1044.0,
|
| 264186 |
+
"eval_loss": 0.3327125012874603,
|
| 264187 |
+
"eval_runtime": 37.5621,
|
| 264188 |
+
"eval_samples_per_second": 22.256,
|
| 264189 |
+
"eval_steps_per_second": 0.719,
|
| 264190 |
+
"eval_wer": 0.17799092107189926,
|
| 264191 |
+
"step": 129549
|
| 264192 |
+
},
|
| 264193 |
+
{
|
| 264194 |
+
"epoch": 1036.01,
|
| 264195 |
+
"learning_rate": 7.940721153846155e-06,
|
| 264196 |
+
"loss": 0.3303,
|
| 264197 |
+
"step": 129550
|
| 264198 |
+
},
|
| 264199 |
+
{
|
| 264200 |
+
"epoch": 1036.05,
|
| 264201 |
+
"learning_rate": 7.940641025641026e-06,
|
| 264202 |
+
"loss": 0.4137,
|
| 264203 |
+
"step": 129555
|
| 264204 |
+
},
|
| 264205 |
+
{
|
| 264206 |
+
"epoch": 1036.09,
|
| 264207 |
+
"learning_rate": 7.940560897435899e-06,
|
| 264208 |
+
"loss": 0.274,
|
| 264209 |
+
"step": 129560
|
| 264210 |
+
},
|
| 264211 |
+
{
|
| 264212 |
+
"epoch": 1036.13,
|
| 264213 |
+
"learning_rate": 7.940480769230769e-06,
|
| 264214 |
+
"loss": 0.2856,
|
| 264215 |
+
"step": 129565
|
| 264216 |
+
},
|
| 264217 |
+
{
|
| 264218 |
+
"epoch": 1036.17,
|
| 264219 |
+
"learning_rate": 7.940400641025642e-06,
|
| 264220 |
+
"loss": 0.5237,
|
| 264221 |
+
"step": 129570
|
| 264222 |
+
},
|
| 264223 |
+
{
|
| 264224 |
+
"epoch": 1036.21,
|
| 264225 |
+
"learning_rate": 7.940320512820513e-06,
|
| 264226 |
+
"loss": 1.1421,
|
| 264227 |
+
"step": 129575
|
| 264228 |
+
},
|
| 264229 |
+
{
|
| 264230 |
+
"epoch": 1036.25,
|
| 264231 |
+
"learning_rate": 7.940240384615384e-06,
|
| 264232 |
+
"loss": 0.299,
|
| 264233 |
+
"step": 129580
|
| 264234 |
+
},
|
| 264235 |
+
{
|
| 264236 |
+
"epoch": 1036.29,
|
| 264237 |
+
"learning_rate": 7.940160256410257e-06,
|
| 264238 |
+
"loss": 0.264,
|
| 264239 |
+
"step": 129585
|
| 264240 |
+
},
|
| 264241 |
+
{
|
| 264242 |
+
"epoch": 1036.33,
|
| 264243 |
+
"learning_rate": 7.940080128205129e-06,
|
| 264244 |
+
"loss": 0.3278,
|
| 264245 |
+
"step": 129590
|
| 264246 |
+
},
|
| 264247 |
+
{
|
| 264248 |
+
"epoch": 1036.37,
|
| 264249 |
+
"learning_rate": 7.94e-06,
|
| 264250 |
+
"loss": 0.4604,
|
| 264251 |
+
"step": 129595
|
| 264252 |
+
},
|
| 264253 |
+
{
|
| 264254 |
+
"epoch": 1036.41,
|
| 264255 |
+
"learning_rate": 7.939919871794872e-06,
|
| 264256 |
+
"loss": 1.1322,
|
| 264257 |
+
"step": 129600
|
| 264258 |
+
},
|
| 264259 |
+
{
|
| 264260 |
+
"epoch": 1036.45,
|
| 264261 |
+
"learning_rate": 7.939839743589745e-06,
|
| 264262 |
+
"loss": 0.3,
|
| 264263 |
+
"step": 129605
|
| 264264 |
+
},
|
| 264265 |
+
{
|
| 264266 |
+
"epoch": 1036.49,
|
| 264267 |
+
"learning_rate": 7.939759615384616e-06,
|
| 264268 |
+
"loss": 0.2407,
|
| 264269 |
+
"step": 129610
|
| 264270 |
+
},
|
| 264271 |
+
{
|
| 264272 |
+
"epoch": 1036.53,
|
| 264273 |
+
"learning_rate": 7.939679487179487e-06,
|
| 264274 |
+
"loss": 0.321,
|
| 264275 |
+
"step": 129615
|
| 264276 |
+
},
|
| 264277 |
+
{
|
| 264278 |
+
"epoch": 1036.57,
|
| 264279 |
+
"learning_rate": 7.939599358974359e-06,
|
| 264280 |
+
"loss": 0.5329,
|
| 264281 |
+
"step": 129620
|
| 264282 |
+
},
|
| 264283 |
+
{
|
| 264284 |
+
"epoch": 1036.61,
|
| 264285 |
+
"learning_rate": 7.939519230769232e-06,
|
| 264286 |
+
"loss": 1.1307,
|
| 264287 |
+
"step": 129625
|
| 264288 |
+
},
|
| 264289 |
+
{
|
| 264290 |
+
"epoch": 1036.65,
|
| 264291 |
+
"learning_rate": 7.939439102564103e-06,
|
| 264292 |
+
"loss": 0.2848,
|
| 264293 |
+
"step": 129630
|
| 264294 |
+
},
|
| 264295 |
+
{
|
| 264296 |
+
"epoch": 1036.69,
|
| 264297 |
+
"learning_rate": 7.939358974358974e-06,
|
| 264298 |
+
"loss": 0.2602,
|
| 264299 |
+
"step": 129635
|
| 264300 |
+
},
|
| 264301 |
+
{
|
| 264302 |
+
"epoch": 1036.73,
|
| 264303 |
+
"learning_rate": 7.939278846153848e-06,
|
| 264304 |
+
"loss": 0.3013,
|
| 264305 |
+
"step": 129640
|
| 264306 |
+
},
|
| 264307 |
+
{
|
| 264308 |
+
"epoch": 1036.77,
|
| 264309 |
+
"learning_rate": 7.939198717948719e-06,
|
| 264310 |
+
"loss": 0.5769,
|
| 264311 |
+
"step": 129645
|
| 264312 |
+
},
|
| 264313 |
+
{
|
| 264314 |
+
"epoch": 1036.81,
|
| 264315 |
+
"learning_rate": 7.93911858974359e-06,
|
| 264316 |
+
"loss": 1.0882,
|
| 264317 |
+
"step": 129650
|
| 264318 |
+
},
|
| 264319 |
+
{
|
| 264320 |
+
"epoch": 1036.85,
|
| 264321 |
+
"learning_rate": 7.939038461538462e-06,
|
| 264322 |
+
"loss": 0.3345,
|
| 264323 |
+
"step": 129655
|
| 264324 |
+
},
|
| 264325 |
+
{
|
| 264326 |
+
"epoch": 1036.89,
|
| 264327 |
+
"learning_rate": 7.938958333333335e-06,
|
| 264328 |
+
"loss": 0.2642,
|
| 264329 |
+
"step": 129660
|
| 264330 |
+
},
|
| 264331 |
+
{
|
| 264332 |
+
"epoch": 1036.93,
|
| 264333 |
+
"learning_rate": 7.938878205128206e-06,
|
| 264334 |
+
"loss": 0.3457,
|
| 264335 |
+
"step": 129665
|
| 264336 |
+
},
|
| 264337 |
+
{
|
| 264338 |
+
"epoch": 1036.97,
|
| 264339 |
+
"learning_rate": 7.938798076923077e-06,
|
| 264340 |
+
"loss": 0.5265,
|
| 264341 |
+
"step": 129670
|
| 264342 |
+
},
|
| 264343 |
+
{
|
| 264344 |
+
"epoch": 1037.0,
|
| 264345 |
+
"eval_loss": 0.3877243101596832,
|
| 264346 |
+
"eval_runtime": 37.2117,
|
| 264347 |
+
"eval_samples_per_second": 22.466,
|
| 264348 |
+
"eval_steps_per_second": 0.726,
|
| 264349 |
+
"eval_wer": 0.17105075053609722,
|
| 264350 |
+
"step": 129674
|
| 264351 |
}
|
| 264352 |
],
|
| 264353 |
+
"max_steps": 625000,
|
| 264354 |
"num_train_epochs": 5000,
|
| 264355 |
+
"total_flos": 3.649092774457167e+20,
|
| 264356 |
"trial_name": null,
|
| 264357 |
"trial_params": null
|
| 264358 |
}
|
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129674}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/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:b625c058e4faca45be25b4f12ffa5ddb492254fee24e780dca1a0e944a4b460b
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129799}/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 sha256:767309603379ab5670f4f913d0c69c5817193c15825f5f5fa9d0d740c2dbd52b
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129799}/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:b8675d42f073ebf4d2004c004c2b56930892249657b03f4a1470c321cd24e807
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129799}/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 sha256:4ff59c52ad1b3ee36995b7074f5bc48f42c02150ebca0f8c790d967cdd695e1f
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/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 sha256:9105cc09ce9425afcec55b8ea22dba1e47c016dafac983246ad6d83808a7b2d0
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-124947 → checkpoint-129799}/trainer_state.json
RENAMED
|
The diff for this file is too large to render.
See raw diff
|
|
|
model-bin/finetune/base/{checkpoint-129176 → checkpoint-129799}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630158854.6730864/events.out.tfevents.1630158854.86bb0ddabf9b.4092.181
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3f9f7edefd30f2998ab5be51fba7e3c971ee4a79d67dd29b51855a03d2137a88
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630159245.4578865/events.out.tfevents.1630159245.86bb0ddabf9b.4092.183
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:41c6bd5a666dcf5afb167e59b3331a392cc2699e08f560498e657da19de22fd6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630159646.2298772/events.out.tfevents.1630159646.86bb0ddabf9b.4092.185
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d99f1a96386120c47549539de278c8dfe08c7edd3a687ae77088b0384ed1db52
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630160047.7808826/events.out.tfevents.1630160047.86bb0ddabf9b.4092.187
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4fafbe74dd3353bf4d917a57be7768c4bf91187451df2fadd0daa473f2e1369f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630160440.877656/events.out.tfevents.1630160440.86bb0ddabf9b.4092.189
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b322509b84387e6762d6dd940e6eb4e7acfde9ca8a858919483cf234aa7d9ad9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630158854.86bb0ddabf9b.4092.180
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3651233c640c186c1fac90a0265cff2003bf90be78089c5f6b810b72af085047
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630159245.86bb0ddabf9b.4092.182
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd22089d259b7b5666e60cbc124c62ec2c87a00284eec6ac676570803f994fc6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630159646.86bb0ddabf9b.4092.184
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c68fef167381493c2613769b6a8689347e17efd137d293a4bf558f5b567d918c
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630160047.86bb0ddabf9b.4092.186
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5790c687a9f613ac7a3f9329aab87a396bce12f7bed6534a3203fa3a4c77c15d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630160440.86bb0ddabf9b.4092.188
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ada3e6d9778c54ec500ad4cce1ff33956ea7a6da8a659f9315a220d45c23682
|
| 3 |
+
size 8622
|