"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629950864.3685591/events.out.tfevents.1629950864.8e89bd551565.924.51 +3 -0
- model-bin/finetune/base/log/1629951305.726289/events.out.tfevents.1629951305.8e89bd551565.924.53 +3 -0
- model-bin/finetune/base/log/1629951830.0988564/events.out.tfevents.1629951830.8e89bd551565.924.55 +3 -0
- model-bin/finetune/base/log/1629952276.040209/events.out.tfevents.1629952276.8e89bd551565.924.57 +3 -0
- model-bin/finetune/base/log/1629952715.8959966/events.out.tfevents.1629952715.8e89bd551565.924.59 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629950864.8e89bd551565.924.50 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629951304.8e89bd551565.924.52 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629951830.8e89bd551565.924.54 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629952276.8e89bd551565.924.56 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629952715.8e89bd551565.924.58 +3 -0
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/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:afbc591eebb20b9dbb8dd3586e724b8f27e08539849f34515d060fcd4481234d
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/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 sha256:80b8220b9e51e764884c4b0efd0c522a8e61dfd178567a26b8d135892f3d7a61
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:45f2ebe8c52f444f5fbfd85934c90402e18a7a11cc12339a3165d5658075a12d
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/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 sha256:543a487eba208631d88196fdd30d0309f8032ae22b2ee45f6c8765542c0abfe1
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/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 sha256:2b985e4db4c9659bd4e2a9e08dc0bbf414c0dac0f248e47f20c3cbf6429b14ff
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -215739,11 +215739,806 @@
|
|
| 215739 |
"eval_steps_per_second": 0.682,
|
| 215740 |
"eval_wer": 0.18542114230683754,
|
| 215741 |
"step": 91468
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 215742 |
}
|
| 215743 |
],
|
| 215744 |
-
"max_steps":
|
| 215745 |
"num_train_epochs": 5000,
|
| 215746 |
-
"total_flos": 2.
|
| 215747 |
"trial_name": null,
|
| 215748 |
"trial_params": null
|
| 215749 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
+
"epoch": 735.9960159362549,
|
| 5 |
+
"global_step": 92092,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 215739 |
"eval_steps_per_second": 0.682,
|
| 215740 |
"eval_wer": 0.18542114230683754,
|
| 215741 |
"step": 91468
|
| 215742 |
+
},
|
| 215743 |
+
{
|
| 215744 |
+
"epoch": 731.02,
|
| 215745 |
+
"learning_rate": 8.55073717948718e-06,
|
| 215746 |
+
"loss": 0.299,
|
| 215747 |
+
"step": 91470
|
| 215748 |
+
},
|
| 215749 |
+
{
|
| 215750 |
+
"epoch": 731.06,
|
| 215751 |
+
"learning_rate": 8.550657051282051e-06,
|
| 215752 |
+
"loss": 0.3312,
|
| 215753 |
+
"step": 91475
|
| 215754 |
+
},
|
| 215755 |
+
{
|
| 215756 |
+
"epoch": 731.1,
|
| 215757 |
+
"learning_rate": 8.550576923076924e-06,
|
| 215758 |
+
"loss": 0.3024,
|
| 215759 |
+
"step": 91480
|
| 215760 |
+
},
|
| 215761 |
+
{
|
| 215762 |
+
"epoch": 731.14,
|
| 215763 |
+
"learning_rate": 8.550496794871796e-06,
|
| 215764 |
+
"loss": 0.3301,
|
| 215765 |
+
"step": 91485
|
| 215766 |
+
},
|
| 215767 |
+
{
|
| 215768 |
+
"epoch": 731.18,
|
| 215769 |
+
"learning_rate": 8.550416666666667e-06,
|
| 215770 |
+
"loss": 0.6771,
|
| 215771 |
+
"step": 91490
|
| 215772 |
+
},
|
| 215773 |
+
{
|
| 215774 |
+
"epoch": 731.22,
|
| 215775 |
+
"learning_rate": 8.550336538461538e-06,
|
| 215776 |
+
"loss": 1.0689,
|
| 215777 |
+
"step": 91495
|
| 215778 |
+
},
|
| 215779 |
+
{
|
| 215780 |
+
"epoch": 731.26,
|
| 215781 |
+
"learning_rate": 8.550256410256411e-06,
|
| 215782 |
+
"loss": 0.3782,
|
| 215783 |
+
"step": 91500
|
| 215784 |
+
},
|
| 215785 |
+
{
|
| 215786 |
+
"epoch": 731.3,
|
| 215787 |
+
"learning_rate": 8.550176282051283e-06,
|
| 215788 |
+
"loss": 0.3406,
|
| 215789 |
+
"step": 91505
|
| 215790 |
+
},
|
| 215791 |
+
{
|
| 215792 |
+
"epoch": 731.34,
|
| 215793 |
+
"learning_rate": 8.550096153846154e-06,
|
| 215794 |
+
"loss": 0.3023,
|
| 215795 |
+
"step": 91510
|
| 215796 |
+
},
|
| 215797 |
+
{
|
| 215798 |
+
"epoch": 731.38,
|
| 215799 |
+
"learning_rate": 8.550016025641026e-06,
|
| 215800 |
+
"loss": 0.5426,
|
| 215801 |
+
"step": 91515
|
| 215802 |
+
},
|
| 215803 |
+
{
|
| 215804 |
+
"epoch": 731.42,
|
| 215805 |
+
"learning_rate": 8.549935897435899e-06,
|
| 215806 |
+
"loss": 1.101,
|
| 215807 |
+
"step": 91520
|
| 215808 |
+
},
|
| 215809 |
+
{
|
| 215810 |
+
"epoch": 731.46,
|
| 215811 |
+
"learning_rate": 8.54985576923077e-06,
|
| 215812 |
+
"loss": 0.2966,
|
| 215813 |
+
"step": 91525
|
| 215814 |
+
},
|
| 215815 |
+
{
|
| 215816 |
+
"epoch": 731.5,
|
| 215817 |
+
"learning_rate": 8.549775641025641e-06,
|
| 215818 |
+
"loss": 0.2979,
|
| 215819 |
+
"step": 91530
|
| 215820 |
+
},
|
| 215821 |
+
{
|
| 215822 |
+
"epoch": 731.54,
|
| 215823 |
+
"learning_rate": 8.549695512820514e-06,
|
| 215824 |
+
"loss": 0.3577,
|
| 215825 |
+
"step": 91535
|
| 215826 |
+
},
|
| 215827 |
+
{
|
| 215828 |
+
"epoch": 731.58,
|
| 215829 |
+
"learning_rate": 8.549615384615386e-06,
|
| 215830 |
+
"loss": 0.6888,
|
| 215831 |
+
"step": 91540
|
| 215832 |
+
},
|
| 215833 |
+
{
|
| 215834 |
+
"epoch": 731.62,
|
| 215835 |
+
"learning_rate": 8.549535256410257e-06,
|
| 215836 |
+
"loss": 0.9692,
|
| 215837 |
+
"step": 91545
|
| 215838 |
+
},
|
| 215839 |
+
{
|
| 215840 |
+
"epoch": 731.66,
|
| 215841 |
+
"learning_rate": 8.549455128205128e-06,
|
| 215842 |
+
"loss": 0.2759,
|
| 215843 |
+
"step": 91550
|
| 215844 |
+
},
|
| 215845 |
+
{
|
| 215846 |
+
"epoch": 731.7,
|
| 215847 |
+
"learning_rate": 8.549375000000001e-06,
|
| 215848 |
+
"loss": 0.3488,
|
| 215849 |
+
"step": 91555
|
| 215850 |
+
},
|
| 215851 |
+
{
|
| 215852 |
+
"epoch": 731.74,
|
| 215853 |
+
"learning_rate": 8.549294871794873e-06,
|
| 215854 |
+
"loss": 0.4113,
|
| 215855 |
+
"step": 91560
|
| 215856 |
+
},
|
| 215857 |
+
{
|
| 215858 |
+
"epoch": 731.78,
|
| 215859 |
+
"learning_rate": 8.549214743589744e-06,
|
| 215860 |
+
"loss": 0.5813,
|
| 215861 |
+
"step": 91565
|
| 215862 |
+
},
|
| 215863 |
+
{
|
| 215864 |
+
"epoch": 731.82,
|
| 215865 |
+
"learning_rate": 8.549134615384616e-06,
|
| 215866 |
+
"loss": 1.0029,
|
| 215867 |
+
"step": 91570
|
| 215868 |
+
},
|
| 215869 |
+
{
|
| 215870 |
+
"epoch": 731.86,
|
| 215871 |
+
"learning_rate": 8.549054487179489e-06,
|
| 215872 |
+
"loss": 0.3106,
|
| 215873 |
+
"step": 91575
|
| 215874 |
+
},
|
| 215875 |
+
{
|
| 215876 |
+
"epoch": 731.9,
|
| 215877 |
+
"learning_rate": 8.54897435897436e-06,
|
| 215878 |
+
"loss": 0.2549,
|
| 215879 |
+
"step": 91580
|
| 215880 |
+
},
|
| 215881 |
+
{
|
| 215882 |
+
"epoch": 731.94,
|
| 215883 |
+
"learning_rate": 8.548894230769231e-06,
|
| 215884 |
+
"loss": 0.4328,
|
| 215885 |
+
"step": 91585
|
| 215886 |
+
},
|
| 215887 |
+
{
|
| 215888 |
+
"epoch": 731.98,
|
| 215889 |
+
"learning_rate": 8.548814102564104e-06,
|
| 215890 |
+
"loss": 0.7052,
|
| 215891 |
+
"step": 91590
|
| 215892 |
+
},
|
| 215893 |
+
{
|
| 215894 |
+
"epoch": 732.0,
|
| 215895 |
+
"eval_loss": 0.4339653551578522,
|
| 215896 |
+
"eval_runtime": 40.1887,
|
| 215897 |
+
"eval_samples_per_second": 20.951,
|
| 215898 |
+
"eval_steps_per_second": 0.672,
|
| 215899 |
+
"eval_wer": 0.19089456869009586,
|
| 215900 |
+
"step": 91593
|
| 215901 |
+
},
|
| 215902 |
+
{
|
| 215903 |
+
"epoch": 732.02,
|
| 215904 |
+
"learning_rate": 8.548733974358974e-06,
|
| 215905 |
+
"loss": 0.338,
|
| 215906 |
+
"step": 91595
|
| 215907 |
+
},
|
| 215908 |
+
{
|
| 215909 |
+
"epoch": 732.06,
|
| 215910 |
+
"learning_rate": 8.548653846153847e-06,
|
| 215911 |
+
"loss": 0.3145,
|
| 215912 |
+
"step": 91600
|
| 215913 |
+
},
|
| 215914 |
+
{
|
| 215915 |
+
"epoch": 732.1,
|
| 215916 |
+
"learning_rate": 8.548573717948718e-06,
|
| 215917 |
+
"loss": 0.3012,
|
| 215918 |
+
"step": 91605
|
| 215919 |
+
},
|
| 215920 |
+
{
|
| 215921 |
+
"epoch": 732.14,
|
| 215922 |
+
"learning_rate": 8.54849358974359e-06,
|
| 215923 |
+
"loss": 0.3511,
|
| 215924 |
+
"step": 91610
|
| 215925 |
+
},
|
| 215926 |
+
{
|
| 215927 |
+
"epoch": 732.18,
|
| 215928 |
+
"learning_rate": 8.548413461538461e-06,
|
| 215929 |
+
"loss": 0.6375,
|
| 215930 |
+
"step": 91615
|
| 215931 |
+
},
|
| 215932 |
+
{
|
| 215933 |
+
"epoch": 732.22,
|
| 215934 |
+
"learning_rate": 8.548333333333334e-06,
|
| 215935 |
+
"loss": 1.0038,
|
| 215936 |
+
"step": 91620
|
| 215937 |
+
},
|
| 215938 |
+
{
|
| 215939 |
+
"epoch": 732.26,
|
| 215940 |
+
"learning_rate": 8.548253205128206e-06,
|
| 215941 |
+
"loss": 0.3902,
|
| 215942 |
+
"step": 91625
|
| 215943 |
+
},
|
| 215944 |
+
{
|
| 215945 |
+
"epoch": 732.3,
|
| 215946 |
+
"learning_rate": 8.548173076923077e-06,
|
| 215947 |
+
"loss": 0.3743,
|
| 215948 |
+
"step": 91630
|
| 215949 |
+
},
|
| 215950 |
+
{
|
| 215951 |
+
"epoch": 732.34,
|
| 215952 |
+
"learning_rate": 8.54809294871795e-06,
|
| 215953 |
+
"loss": 0.3208,
|
| 215954 |
+
"step": 91635
|
| 215955 |
+
},
|
| 215956 |
+
{
|
| 215957 |
+
"epoch": 732.38,
|
| 215958 |
+
"learning_rate": 8.548012820512821e-06,
|
| 215959 |
+
"loss": 0.7151,
|
| 215960 |
+
"step": 91640
|
| 215961 |
+
},
|
| 215962 |
+
{
|
| 215963 |
+
"epoch": 732.42,
|
| 215964 |
+
"learning_rate": 8.547932692307693e-06,
|
| 215965 |
+
"loss": 1.1507,
|
| 215966 |
+
"step": 91645
|
| 215967 |
+
},
|
| 215968 |
+
{
|
| 215969 |
+
"epoch": 732.46,
|
| 215970 |
+
"learning_rate": 8.547852564102564e-06,
|
| 215971 |
+
"loss": 0.3511,
|
| 215972 |
+
"step": 91650
|
| 215973 |
+
},
|
| 215974 |
+
{
|
| 215975 |
+
"epoch": 732.5,
|
| 215976 |
+
"learning_rate": 8.547772435897437e-06,
|
| 215977 |
+
"loss": 0.3321,
|
| 215978 |
+
"step": 91655
|
| 215979 |
+
},
|
| 215980 |
+
{
|
| 215981 |
+
"epoch": 732.54,
|
| 215982 |
+
"learning_rate": 8.547692307692308e-06,
|
| 215983 |
+
"loss": 0.3649,
|
| 215984 |
+
"step": 91660
|
| 215985 |
+
},
|
| 215986 |
+
{
|
| 215987 |
+
"epoch": 732.58,
|
| 215988 |
+
"learning_rate": 8.54761217948718e-06,
|
| 215989 |
+
"loss": 0.638,
|
| 215990 |
+
"step": 91665
|
| 215991 |
+
},
|
| 215992 |
+
{
|
| 215993 |
+
"epoch": 732.62,
|
| 215994 |
+
"learning_rate": 8.547532051282051e-06,
|
| 215995 |
+
"loss": 1.1235,
|
| 215996 |
+
"step": 91670
|
| 215997 |
+
},
|
| 215998 |
+
{
|
| 215999 |
+
"epoch": 732.66,
|
| 216000 |
+
"learning_rate": 8.547451923076924e-06,
|
| 216001 |
+
"loss": 0.2796,
|
| 216002 |
+
"step": 91675
|
| 216003 |
+
},
|
| 216004 |
+
{
|
| 216005 |
+
"epoch": 732.7,
|
| 216006 |
+
"learning_rate": 8.547371794871796e-06,
|
| 216007 |
+
"loss": 0.2738,
|
| 216008 |
+
"step": 91680
|
| 216009 |
+
},
|
| 216010 |
+
{
|
| 216011 |
+
"epoch": 732.74,
|
| 216012 |
+
"learning_rate": 8.547291666666667e-06,
|
| 216013 |
+
"loss": 0.3643,
|
| 216014 |
+
"step": 91685
|
| 216015 |
+
},
|
| 216016 |
+
{
|
| 216017 |
+
"epoch": 732.78,
|
| 216018 |
+
"learning_rate": 8.54721153846154e-06,
|
| 216019 |
+
"loss": 0.592,
|
| 216020 |
+
"step": 91690
|
| 216021 |
+
},
|
| 216022 |
+
{
|
| 216023 |
+
"epoch": 732.82,
|
| 216024 |
+
"learning_rate": 8.547131410256411e-06,
|
| 216025 |
+
"loss": 1.0093,
|
| 216026 |
+
"step": 91695
|
| 216027 |
+
},
|
| 216028 |
+
{
|
| 216029 |
+
"epoch": 732.86,
|
| 216030 |
+
"learning_rate": 8.547051282051283e-06,
|
| 216031 |
+
"loss": 0.285,
|
| 216032 |
+
"step": 91700
|
| 216033 |
+
},
|
| 216034 |
+
{
|
| 216035 |
+
"epoch": 732.9,
|
| 216036 |
+
"learning_rate": 8.546971153846154e-06,
|
| 216037 |
+
"loss": 0.2928,
|
| 216038 |
+
"step": 91705
|
| 216039 |
+
},
|
| 216040 |
+
{
|
| 216041 |
+
"epoch": 732.94,
|
| 216042 |
+
"learning_rate": 8.546891025641027e-06,
|
| 216043 |
+
"loss": 0.3674,
|
| 216044 |
+
"step": 91710
|
| 216045 |
+
},
|
| 216046 |
+
{
|
| 216047 |
+
"epoch": 732.98,
|
| 216048 |
+
"learning_rate": 8.546810897435897e-06,
|
| 216049 |
+
"loss": 0.6795,
|
| 216050 |
+
"step": 91715
|
| 216051 |
+
},
|
| 216052 |
+
{
|
| 216053 |
+
"epoch": 733.0,
|
| 216054 |
+
"eval_loss": 0.3943796753883362,
|
| 216055 |
+
"eval_runtime": 40.8036,
|
| 216056 |
+
"eval_samples_per_second": 20.635,
|
| 216057 |
+
"eval_steps_per_second": 0.662,
|
| 216058 |
+
"eval_wer": 0.18691182857970176,
|
| 216059 |
+
"step": 91718
|
| 216060 |
+
},
|
| 216061 |
+
{
|
| 216062 |
+
"epoch": 739.02,
|
| 216063 |
+
"learning_rate": 8.54673076923077e-06,
|
| 216064 |
+
"loss": 0.3295,
|
| 216065 |
+
"step": 91720
|
| 216066 |
+
},
|
| 216067 |
+
{
|
| 216068 |
+
"epoch": 739.06,
|
| 216069 |
+
"learning_rate": 8.546650641025643e-06,
|
| 216070 |
+
"loss": 0.3028,
|
| 216071 |
+
"step": 91725
|
| 216072 |
+
},
|
| 216073 |
+
{
|
| 216074 |
+
"epoch": 739.1,
|
| 216075 |
+
"learning_rate": 8.546570512820514e-06,
|
| 216076 |
+
"loss": 0.3294,
|
| 216077 |
+
"step": 91730
|
| 216078 |
+
},
|
| 216079 |
+
{
|
| 216080 |
+
"epoch": 739.14,
|
| 216081 |
+
"learning_rate": 8.546490384615386e-06,
|
| 216082 |
+
"loss": 0.3243,
|
| 216083 |
+
"step": 91735
|
| 216084 |
+
},
|
| 216085 |
+
{
|
| 216086 |
+
"epoch": 739.18,
|
| 216087 |
+
"learning_rate": 8.546410256410257e-06,
|
| 216088 |
+
"loss": 0.67,
|
| 216089 |
+
"step": 91740
|
| 216090 |
+
},
|
| 216091 |
+
{
|
| 216092 |
+
"epoch": 739.22,
|
| 216093 |
+
"learning_rate": 8.54633012820513e-06,
|
| 216094 |
+
"loss": 0.8697,
|
| 216095 |
+
"step": 91745
|
| 216096 |
+
},
|
| 216097 |
+
{
|
| 216098 |
+
"epoch": 739.26,
|
| 216099 |
+
"learning_rate": 8.54625e-06,
|
| 216100 |
+
"loss": 0.3294,
|
| 216101 |
+
"step": 91750
|
| 216102 |
+
},
|
| 216103 |
+
{
|
| 216104 |
+
"epoch": 739.3,
|
| 216105 |
+
"learning_rate": 8.546169871794873e-06,
|
| 216106 |
+
"loss": 0.2796,
|
| 216107 |
+
"step": 91755
|
| 216108 |
+
},
|
| 216109 |
+
{
|
| 216110 |
+
"epoch": 739.34,
|
| 216111 |
+
"learning_rate": 8.546089743589744e-06,
|
| 216112 |
+
"loss": 0.3865,
|
| 216113 |
+
"step": 91760
|
| 216114 |
+
},
|
| 216115 |
+
{
|
| 216116 |
+
"epoch": 739.38,
|
| 216117 |
+
"learning_rate": 8.546009615384615e-06,
|
| 216118 |
+
"loss": 0.648,
|
| 216119 |
+
"step": 91765
|
| 216120 |
+
},
|
| 216121 |
+
{
|
| 216122 |
+
"epoch": 739.42,
|
| 216123 |
+
"learning_rate": 8.545929487179487e-06,
|
| 216124 |
+
"loss": 1.0979,
|
| 216125 |
+
"step": 91770
|
| 216126 |
+
},
|
| 216127 |
+
{
|
| 216128 |
+
"epoch": 739.46,
|
| 216129 |
+
"learning_rate": 8.54584935897436e-06,
|
| 216130 |
+
"loss": 0.2734,
|
| 216131 |
+
"step": 91775
|
| 216132 |
+
},
|
| 216133 |
+
{
|
| 216134 |
+
"epoch": 739.5,
|
| 216135 |
+
"learning_rate": 8.545769230769231e-06,
|
| 216136 |
+
"loss": 0.2806,
|
| 216137 |
+
"step": 91780
|
| 216138 |
+
},
|
| 216139 |
+
{
|
| 216140 |
+
"epoch": 739.54,
|
| 216141 |
+
"learning_rate": 8.545689102564103e-06,
|
| 216142 |
+
"loss": 0.3612,
|
| 216143 |
+
"step": 91785
|
| 216144 |
+
},
|
| 216145 |
+
{
|
| 216146 |
+
"epoch": 739.58,
|
| 216147 |
+
"learning_rate": 8.545608974358976e-06,
|
| 216148 |
+
"loss": 0.6085,
|
| 216149 |
+
"step": 91790
|
| 216150 |
+
},
|
| 216151 |
+
{
|
| 216152 |
+
"epoch": 739.62,
|
| 216153 |
+
"learning_rate": 8.545528846153847e-06,
|
| 216154 |
+
"loss": 0.9859,
|
| 216155 |
+
"step": 91795
|
| 216156 |
+
},
|
| 216157 |
+
{
|
| 216158 |
+
"epoch": 739.66,
|
| 216159 |
+
"learning_rate": 8.545448717948718e-06,
|
| 216160 |
+
"loss": 0.3682,
|
| 216161 |
+
"step": 91800
|
| 216162 |
+
},
|
| 216163 |
+
{
|
| 216164 |
+
"epoch": 739.7,
|
| 216165 |
+
"learning_rate": 8.54536858974359e-06,
|
| 216166 |
+
"loss": 0.299,
|
| 216167 |
+
"step": 91805
|
| 216168 |
+
},
|
| 216169 |
+
{
|
| 216170 |
+
"epoch": 739.74,
|
| 216171 |
+
"learning_rate": 8.545288461538463e-06,
|
| 216172 |
+
"loss": 0.341,
|
| 216173 |
+
"step": 91810
|
| 216174 |
+
},
|
| 216175 |
+
{
|
| 216176 |
+
"epoch": 739.78,
|
| 216177 |
+
"learning_rate": 8.545208333333334e-06,
|
| 216178 |
+
"loss": 0.6854,
|
| 216179 |
+
"step": 91815
|
| 216180 |
+
},
|
| 216181 |
+
{
|
| 216182 |
+
"epoch": 739.82,
|
| 216183 |
+
"learning_rate": 8.545128205128205e-06,
|
| 216184 |
+
"loss": 1.0646,
|
| 216185 |
+
"step": 91820
|
| 216186 |
+
},
|
| 216187 |
+
{
|
| 216188 |
+
"epoch": 739.86,
|
| 216189 |
+
"learning_rate": 8.545048076923079e-06,
|
| 216190 |
+
"loss": 0.3601,
|
| 216191 |
+
"step": 91825
|
| 216192 |
+
},
|
| 216193 |
+
{
|
| 216194 |
+
"epoch": 739.9,
|
| 216195 |
+
"learning_rate": 8.54496794871795e-06,
|
| 216196 |
+
"loss": 0.3256,
|
| 216197 |
+
"step": 91830
|
| 216198 |
+
},
|
| 216199 |
+
{
|
| 216200 |
+
"epoch": 739.94,
|
| 216201 |
+
"learning_rate": 8.544887820512821e-06,
|
| 216202 |
+
"loss": 0.5974,
|
| 216203 |
+
"step": 91835
|
| 216204 |
+
},
|
| 216205 |
+
{
|
| 216206 |
+
"epoch": 739.98,
|
| 216207 |
+
"learning_rate": 8.544807692307693e-06,
|
| 216208 |
+
"loss": 0.7661,
|
| 216209 |
+
"step": 91840
|
| 216210 |
+
},
|
| 216211 |
+
{
|
| 216212 |
+
"epoch": 740.0,
|
| 216213 |
+
"eval_loss": 0.3956696391105652,
|
| 216214 |
+
"eval_runtime": 40.4052,
|
| 216215 |
+
"eval_samples_per_second": 20.74,
|
| 216216 |
+
"eval_steps_per_second": 0.668,
|
| 216217 |
+
"eval_wer": 0.1869410310477714,
|
| 216218 |
+
"step": 91842
|
| 216219 |
+
},
|
| 216220 |
+
{
|
| 216221 |
+
"epoch": 734.02,
|
| 216222 |
+
"learning_rate": 8.544727564102566e-06,
|
| 216223 |
+
"loss": 0.333,
|
| 216224 |
+
"step": 91845
|
| 216225 |
+
},
|
| 216226 |
+
{
|
| 216227 |
+
"epoch": 734.06,
|
| 216228 |
+
"learning_rate": 8.544647435897437e-06,
|
| 216229 |
+
"loss": 0.2749,
|
| 216230 |
+
"step": 91850
|
| 216231 |
+
},
|
| 216232 |
+
{
|
| 216233 |
+
"epoch": 734.1,
|
| 216234 |
+
"learning_rate": 8.544567307692308e-06,
|
| 216235 |
+
"loss": 0.3329,
|
| 216236 |
+
"step": 91855
|
| 216237 |
+
},
|
| 216238 |
+
{
|
| 216239 |
+
"epoch": 734.14,
|
| 216240 |
+
"learning_rate": 8.54448717948718e-06,
|
| 216241 |
+
"loss": 0.3817,
|
| 216242 |
+
"step": 91860
|
| 216243 |
+
},
|
| 216244 |
+
{
|
| 216245 |
+
"epoch": 734.18,
|
| 216246 |
+
"learning_rate": 8.544407051282053e-06,
|
| 216247 |
+
"loss": 0.753,
|
| 216248 |
+
"step": 91865
|
| 216249 |
+
},
|
| 216250 |
+
{
|
| 216251 |
+
"epoch": 734.22,
|
| 216252 |
+
"learning_rate": 8.544326923076922e-06,
|
| 216253 |
+
"loss": 0.94,
|
| 216254 |
+
"step": 91870
|
| 216255 |
+
},
|
| 216256 |
+
{
|
| 216257 |
+
"epoch": 734.26,
|
| 216258 |
+
"learning_rate": 8.544246794871796e-06,
|
| 216259 |
+
"loss": 0.2708,
|
| 216260 |
+
"step": 91875
|
| 216261 |
+
},
|
| 216262 |
+
{
|
| 216263 |
+
"epoch": 734.3,
|
| 216264 |
+
"learning_rate": 8.544166666666669e-06,
|
| 216265 |
+
"loss": 0.3203,
|
| 216266 |
+
"step": 91880
|
| 216267 |
+
},
|
| 216268 |
+
{
|
| 216269 |
+
"epoch": 734.34,
|
| 216270 |
+
"learning_rate": 8.544086538461538e-06,
|
| 216271 |
+
"loss": 0.3598,
|
| 216272 |
+
"step": 91885
|
| 216273 |
+
},
|
| 216274 |
+
{
|
| 216275 |
+
"epoch": 734.38,
|
| 216276 |
+
"learning_rate": 8.544006410256411e-06,
|
| 216277 |
+
"loss": 0.7899,
|
| 216278 |
+
"step": 91890
|
| 216279 |
+
},
|
| 216280 |
+
{
|
| 216281 |
+
"epoch": 734.42,
|
| 216282 |
+
"learning_rate": 8.543926282051283e-06,
|
| 216283 |
+
"loss": 0.7525,
|
| 216284 |
+
"step": 91895
|
| 216285 |
+
},
|
| 216286 |
+
{
|
| 216287 |
+
"epoch": 734.46,
|
| 216288 |
+
"learning_rate": 8.543846153846154e-06,
|
| 216289 |
+
"loss": 0.3314,
|
| 216290 |
+
"step": 91900
|
| 216291 |
+
},
|
| 216292 |
+
{
|
| 216293 |
+
"epoch": 734.5,
|
| 216294 |
+
"learning_rate": 8.543766025641025e-06,
|
| 216295 |
+
"loss": 0.2971,
|
| 216296 |
+
"step": 91905
|
| 216297 |
+
},
|
| 216298 |
+
{
|
| 216299 |
+
"epoch": 734.54,
|
| 216300 |
+
"learning_rate": 8.543685897435898e-06,
|
| 216301 |
+
"loss": 0.4155,
|
| 216302 |
+
"step": 91910
|
| 216303 |
+
},
|
| 216304 |
+
{
|
| 216305 |
+
"epoch": 734.58,
|
| 216306 |
+
"learning_rate": 8.54360576923077e-06,
|
| 216307 |
+
"loss": 0.9399,
|
| 216308 |
+
"step": 91915
|
| 216309 |
+
},
|
| 216310 |
+
{
|
| 216311 |
+
"epoch": 734.62,
|
| 216312 |
+
"learning_rate": 8.543525641025641e-06,
|
| 216313 |
+
"loss": 0.843,
|
| 216314 |
+
"step": 91920
|
| 216315 |
+
},
|
| 216316 |
+
{
|
| 216317 |
+
"epoch": 734.66,
|
| 216318 |
+
"learning_rate": 8.543445512820514e-06,
|
| 216319 |
+
"loss": 0.2829,
|
| 216320 |
+
"step": 91925
|
| 216321 |
+
},
|
| 216322 |
+
{
|
| 216323 |
+
"epoch": 734.7,
|
| 216324 |
+
"learning_rate": 8.543365384615386e-06,
|
| 216325 |
+
"loss": 0.3364,
|
| 216326 |
+
"step": 91930
|
| 216327 |
+
},
|
| 216328 |
+
{
|
| 216329 |
+
"epoch": 734.74,
|
| 216330 |
+
"learning_rate": 8.543285256410257e-06,
|
| 216331 |
+
"loss": 0.3428,
|
| 216332 |
+
"step": 91935
|
| 216333 |
+
},
|
| 216334 |
+
{
|
| 216335 |
+
"epoch": 734.78,
|
| 216336 |
+
"learning_rate": 8.543205128205128e-06,
|
| 216337 |
+
"loss": 0.7198,
|
| 216338 |
+
"step": 91940
|
| 216339 |
+
},
|
| 216340 |
+
{
|
| 216341 |
+
"epoch": 734.82,
|
| 216342 |
+
"learning_rate": 8.543125000000001e-06,
|
| 216343 |
+
"loss": 0.8882,
|
| 216344 |
+
"step": 91945
|
| 216345 |
+
},
|
| 216346 |
+
{
|
| 216347 |
+
"epoch": 734.86,
|
| 216348 |
+
"learning_rate": 8.543044871794873e-06,
|
| 216349 |
+
"loss": 0.3002,
|
| 216350 |
+
"step": 91950
|
| 216351 |
+
},
|
| 216352 |
+
{
|
| 216353 |
+
"epoch": 734.9,
|
| 216354 |
+
"learning_rate": 8.542964743589744e-06,
|
| 216355 |
+
"loss": 0.3753,
|
| 216356 |
+
"step": 91955
|
| 216357 |
+
},
|
| 216358 |
+
{
|
| 216359 |
+
"epoch": 734.94,
|
| 216360 |
+
"learning_rate": 8.542884615384615e-06,
|
| 216361 |
+
"loss": 0.3742,
|
| 216362 |
+
"step": 91960
|
| 216363 |
+
},
|
| 216364 |
+
{
|
| 216365 |
+
"epoch": 734.98,
|
| 216366 |
+
"learning_rate": 8.542804487179488e-06,
|
| 216367 |
+
"loss": 0.8486,
|
| 216368 |
+
"step": 91965
|
| 216369 |
+
},
|
| 216370 |
+
{
|
| 216371 |
+
"epoch": 735.0,
|
| 216372 |
+
"eval_loss": 0.41339191794395447,
|
| 216373 |
+
"eval_runtime": 37.466,
|
| 216374 |
+
"eval_samples_per_second": 22.367,
|
| 216375 |
+
"eval_steps_per_second": 0.721,
|
| 216376 |
+
"eval_wer": 0.19123275799812234,
|
| 216377 |
+
"step": 91967
|
| 216378 |
+
},
|
| 216379 |
+
{
|
| 216380 |
+
"epoch": 735.02,
|
| 216381 |
+
"learning_rate": 8.54272435897436e-06,
|
| 216382 |
+
"loss": 0.3435,
|
| 216383 |
+
"step": 91970
|
| 216384 |
+
},
|
| 216385 |
+
{
|
| 216386 |
+
"epoch": 735.06,
|
| 216387 |
+
"learning_rate": 8.542644230769231e-06,
|
| 216388 |
+
"loss": 0.2895,
|
| 216389 |
+
"step": 91975
|
| 216390 |
+
},
|
| 216391 |
+
{
|
| 216392 |
+
"epoch": 735.1,
|
| 216393 |
+
"learning_rate": 8.542564102564104e-06,
|
| 216394 |
+
"loss": 0.2921,
|
| 216395 |
+
"step": 91980
|
| 216396 |
+
},
|
| 216397 |
+
{
|
| 216398 |
+
"epoch": 735.14,
|
| 216399 |
+
"learning_rate": 8.542483974358976e-06,
|
| 216400 |
+
"loss": 0.4499,
|
| 216401 |
+
"step": 91985
|
| 216402 |
+
},
|
| 216403 |
+
{
|
| 216404 |
+
"epoch": 735.18,
|
| 216405 |
+
"learning_rate": 8.542403846153847e-06,
|
| 216406 |
+
"loss": 0.7644,
|
| 216407 |
+
"step": 91990
|
| 216408 |
+
},
|
| 216409 |
+
{
|
| 216410 |
+
"epoch": 735.22,
|
| 216411 |
+
"learning_rate": 8.542323717948718e-06,
|
| 216412 |
+
"loss": 0.9578,
|
| 216413 |
+
"step": 91995
|
| 216414 |
+
},
|
| 216415 |
+
{
|
| 216416 |
+
"epoch": 735.26,
|
| 216417 |
+
"learning_rate": 8.542243589743591e-06,
|
| 216418 |
+
"loss": 0.3794,
|
| 216419 |
+
"step": 92000
|
| 216420 |
+
},
|
| 216421 |
+
{
|
| 216422 |
+
"epoch": 735.3,
|
| 216423 |
+
"learning_rate": 8.542163461538463e-06,
|
| 216424 |
+
"loss": 0.2991,
|
| 216425 |
+
"step": 92005
|
| 216426 |
+
},
|
| 216427 |
+
{
|
| 216428 |
+
"epoch": 735.34,
|
| 216429 |
+
"learning_rate": 8.542083333333334e-06,
|
| 216430 |
+
"loss": 0.4057,
|
| 216431 |
+
"step": 92010
|
| 216432 |
+
},
|
| 216433 |
+
{
|
| 216434 |
+
"epoch": 735.38,
|
| 216435 |
+
"learning_rate": 8.542003205128205e-06,
|
| 216436 |
+
"loss": 0.7372,
|
| 216437 |
+
"step": 92015
|
| 216438 |
+
},
|
| 216439 |
+
{
|
| 216440 |
+
"epoch": 735.42,
|
| 216441 |
+
"learning_rate": 8.541923076923078e-06,
|
| 216442 |
+
"loss": 1.0809,
|
| 216443 |
+
"step": 92020
|
| 216444 |
+
},
|
| 216445 |
+
{
|
| 216446 |
+
"epoch": 735.46,
|
| 216447 |
+
"learning_rate": 8.54184294871795e-06,
|
| 216448 |
+
"loss": 0.2875,
|
| 216449 |
+
"step": 92025
|
| 216450 |
+
},
|
| 216451 |
+
{
|
| 216452 |
+
"epoch": 735.5,
|
| 216453 |
+
"learning_rate": 8.541762820512821e-06,
|
| 216454 |
+
"loss": 0.4091,
|
| 216455 |
+
"step": 92030
|
| 216456 |
+
},
|
| 216457 |
+
{
|
| 216458 |
+
"epoch": 735.54,
|
| 216459 |
+
"learning_rate": 8.541682692307694e-06,
|
| 216460 |
+
"loss": 0.3895,
|
| 216461 |
+
"step": 92035
|
| 216462 |
+
},
|
| 216463 |
+
{
|
| 216464 |
+
"epoch": 735.58,
|
| 216465 |
+
"learning_rate": 8.541602564102564e-06,
|
| 216466 |
+
"loss": 0.6983,
|
| 216467 |
+
"step": 92040
|
| 216468 |
+
},
|
| 216469 |
+
{
|
| 216470 |
+
"epoch": 735.62,
|
| 216471 |
+
"learning_rate": 8.541522435897437e-06,
|
| 216472 |
+
"loss": 0.9221,
|
| 216473 |
+
"step": 92045
|
| 216474 |
+
},
|
| 216475 |
+
{
|
| 216476 |
+
"epoch": 735.66,
|
| 216477 |
+
"learning_rate": 8.541442307692308e-06,
|
| 216478 |
+
"loss": 0.315,
|
| 216479 |
+
"step": 92050
|
| 216480 |
+
},
|
| 216481 |
+
{
|
| 216482 |
+
"epoch": 735.7,
|
| 216483 |
+
"learning_rate": 8.54136217948718e-06,
|
| 216484 |
+
"loss": 0.3872,
|
| 216485 |
+
"step": 92055
|
| 216486 |
+
},
|
| 216487 |
+
{
|
| 216488 |
+
"epoch": 735.74,
|
| 216489 |
+
"learning_rate": 8.541282051282051e-06,
|
| 216490 |
+
"loss": 0.3626,
|
| 216491 |
+
"step": 92060
|
| 216492 |
+
},
|
| 216493 |
+
{
|
| 216494 |
+
"epoch": 735.78,
|
| 216495 |
+
"learning_rate": 8.541201923076924e-06,
|
| 216496 |
+
"loss": 0.7254,
|
| 216497 |
+
"step": 92065
|
| 216498 |
+
},
|
| 216499 |
+
{
|
| 216500 |
+
"epoch": 735.82,
|
| 216501 |
+
"learning_rate": 8.541121794871795e-06,
|
| 216502 |
+
"loss": 0.8808,
|
| 216503 |
+
"step": 92070
|
| 216504 |
+
},
|
| 216505 |
+
{
|
| 216506 |
+
"epoch": 735.86,
|
| 216507 |
+
"learning_rate": 8.541041666666667e-06,
|
| 216508 |
+
"loss": 0.252,
|
| 216509 |
+
"step": 92075
|
| 216510 |
+
},
|
| 216511 |
+
{
|
| 216512 |
+
"epoch": 735.9,
|
| 216513 |
+
"learning_rate": 8.54096153846154e-06,
|
| 216514 |
+
"loss": 0.2792,
|
| 216515 |
+
"step": 92080
|
| 216516 |
+
},
|
| 216517 |
+
{
|
| 216518 |
+
"epoch": 735.94,
|
| 216519 |
+
"learning_rate": 8.540881410256411e-06,
|
| 216520 |
+
"loss": 0.4183,
|
| 216521 |
+
"step": 92085
|
| 216522 |
+
},
|
| 216523 |
+
{
|
| 216524 |
+
"epoch": 735.98,
|
| 216525 |
+
"learning_rate": 8.540801282051283e-06,
|
| 216526 |
+
"loss": 0.7612,
|
| 216527 |
+
"step": 92090
|
| 216528 |
+
},
|
| 216529 |
+
{
|
| 216530 |
+
"epoch": 736.0,
|
| 216531 |
+
"eval_loss": 0.40517136454582214,
|
| 216532 |
+
"eval_runtime": 37.5938,
|
| 216533 |
+
"eval_samples_per_second": 22.291,
|
| 216534 |
+
"eval_steps_per_second": 0.718,
|
| 216535 |
+
"eval_wer": 0.1839959593044231,
|
| 216536 |
+
"step": 92092
|
| 216537 |
}
|
| 216538 |
],
|
| 216539 |
+
"max_steps": 625000,
|
| 216540 |
"num_train_epochs": 5000,
|
| 216541 |
+
"total_flos": 2.5915742366601983e+20,
|
| 216542 |
"trial_name": null,
|
| 216543 |
"trial_params": null
|
| 216544 |
}
|
model-bin/finetune/base/{checkpoint-91468 → checkpoint-92092}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629950864.3685591/events.out.tfevents.1629950864.8e89bd551565.924.51
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e02b1d93069034d0081e7cd833fc103132cb762da7eff3744c5cfa9384db05d4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629951305.726289/events.out.tfevents.1629951305.8e89bd551565.924.53
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d20b6feaaf7fb7c8ac26be761cde86c34d2993e1814d52c70c0c3031f7dffd7c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629951830.0988564/events.out.tfevents.1629951830.8e89bd551565.924.55
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1cf59e6d3bac47e6f474647f1a044e3148b9b3cdaf76ca8d4addd14f0eedd5b5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629952276.040209/events.out.tfevents.1629952276.8e89bd551565.924.57
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b7523b1993cf23a953be4f7285e045da9ca2154fd0a6eeb1375bb5f301efe525
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629952715.8959966/events.out.tfevents.1629952715.8e89bd551565.924.59
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9feed5cdcdcc61a77340ea32577b963f52446690359ad770fa43981e109fc4c8
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629950864.8e89bd551565.924.50
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd383743c347d749f150554d7880a63ea1184efb89393e9a1a17176363c753ca
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629951304.8e89bd551565.924.52
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4f152d33af2187b72f64b8ce6b9fb106e57851bdadfd4b5a35d737f334cdffd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629951830.8e89bd551565.924.54
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b510fbe19edc712064d2a9f9f5817ec99a524d187ee3abedbec7e23a70c6f0d9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629952276.8e89bd551565.924.56
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:19bd502aec7cf39b1ad3e800cfd35b45259c5f988340e575bc26e3f314ce5353
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629952715.8e89bd551565.924.58
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff7fa8b5cb92b9de57fe48c501e249424dcea2d094af99d2cd1a9b950921b1d2
|
| 3 |
+
size 8622
|