"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629962173.6222122/events.out.tfevents.1629962173.8e89bd551565.924.101 +3 -0
- model-bin/finetune/base/log/1629962621.2864316/events.out.tfevents.1629962621.8e89bd551565.924.103 +3 -0
- model-bin/finetune/base/log/1629963132.972444/events.out.tfevents.1629963132.8e89bd551565.924.105 +3 -0
- model-bin/finetune/base/log/1629963595.464257/events.out.tfevents.1629963595.8e89bd551565.924.107 +3 -0
- model-bin/finetune/base/log/1629964022.648677/events.out.tfevents.1629964022.8e89bd551565.924.109 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629962173.8e89bd551565.924.100 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629962620.8e89bd551565.924.102 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629963132.8e89bd551565.924.104 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629963595.8e89bd551565.924.106 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629964022.8e89bd551565.924.108 +3 -0
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/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 sha256:2659a34e917a75bf44ea9098c88f7ec69a641c2c56021fdcfeeeee6867230351
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/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 sha256:7a6ebd6b3f01e881ab9cc0e3de1da27d69fbcade5063e4daa9fc9f58b3cfcce5
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/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:c6036c6ef3bafb3cadfc12a8274b3ca9f3a75647660af17949bb2098c645a4d1
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/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 sha256:3ea1dcb8a64d00044fdd03ddccc110577c7eb81ecd8aad34e49e3d94f09fd63b
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/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 sha256:8f3535739f660ec3f51f187361d26affc8b8b369b384ed5f372ffd4fc2a923fa
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -219702,11 +219702,800 @@
|
|
| 219702 |
"eval_steps_per_second": 0.67,
|
| 219703 |
"eval_wer": 0.1897731364782139,
|
| 219704 |
"step": 94581
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 219705 |
}
|
| 219706 |
],
|
| 219707 |
"max_steps": 620000,
|
| 219708 |
"num_train_epochs": 5000,
|
| 219709 |
-
"total_flos": 2.
|
| 219710 |
"trial_name": null,
|
| 219711 |
"trial_params": null
|
| 219712 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 766.995983935743,
|
| 5 |
+
"global_step": 95203,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 219702 |
"eval_steps_per_second": 0.67,
|
| 219703 |
"eval_wer": 0.1897731364782139,
|
| 219704 |
"step": 94581
|
| 219705 |
+
},
|
| 219706 |
+
{
|
| 219707 |
+
"epoch": 756.03,
|
| 219708 |
+
"learning_rate": 8.50084935897436e-06,
|
| 219709 |
+
"loss": 0.3718,
|
| 219710 |
+
"step": 94585
|
| 219711 |
+
},
|
| 219712 |
+
{
|
| 219713 |
+
"epoch": 756.07,
|
| 219714 |
+
"learning_rate": 8.50076923076923e-06,
|
| 219715 |
+
"loss": 0.2754,
|
| 219716 |
+
"step": 94590
|
| 219717 |
+
},
|
| 219718 |
+
{
|
| 219719 |
+
"epoch": 756.11,
|
| 219720 |
+
"learning_rate": 8.500689102564104e-06,
|
| 219721 |
+
"loss": 0.2979,
|
| 219722 |
+
"step": 94595
|
| 219723 |
+
},
|
| 219724 |
+
{
|
| 219725 |
+
"epoch": 756.15,
|
| 219726 |
+
"learning_rate": 8.500608974358975e-06,
|
| 219727 |
+
"loss": 0.3518,
|
| 219728 |
+
"step": 94600
|
| 219729 |
+
},
|
| 219730 |
+
{
|
| 219731 |
+
"epoch": 756.19,
|
| 219732 |
+
"learning_rate": 8.500528846153847e-06,
|
| 219733 |
+
"loss": 0.9221,
|
| 219734 |
+
"step": 94605
|
| 219735 |
+
},
|
| 219736 |
+
{
|
| 219737 |
+
"epoch": 756.23,
|
| 219738 |
+
"learning_rate": 8.500448717948718e-06,
|
| 219739 |
+
"loss": 0.7317,
|
| 219740 |
+
"step": 94610
|
| 219741 |
+
},
|
| 219742 |
+
{
|
| 219743 |
+
"epoch": 756.27,
|
| 219744 |
+
"learning_rate": 8.500368589743591e-06,
|
| 219745 |
+
"loss": 0.3048,
|
| 219746 |
+
"step": 94615
|
| 219747 |
+
},
|
| 219748 |
+
{
|
| 219749 |
+
"epoch": 756.31,
|
| 219750 |
+
"learning_rate": 8.500288461538462e-06,
|
| 219751 |
+
"loss": 0.3322,
|
| 219752 |
+
"step": 94620
|
| 219753 |
+
},
|
| 219754 |
+
{
|
| 219755 |
+
"epoch": 756.35,
|
| 219756 |
+
"learning_rate": 8.500208333333334e-06,
|
| 219757 |
+
"loss": 0.4624,
|
| 219758 |
+
"step": 94625
|
| 219759 |
+
},
|
| 219760 |
+
{
|
| 219761 |
+
"epoch": 756.39,
|
| 219762 |
+
"learning_rate": 8.500128205128207e-06,
|
| 219763 |
+
"loss": 0.9054,
|
| 219764 |
+
"step": 94630
|
| 219765 |
+
},
|
| 219766 |
+
{
|
| 219767 |
+
"epoch": 756.43,
|
| 219768 |
+
"learning_rate": 8.500048076923078e-06,
|
| 219769 |
+
"loss": 0.6236,
|
| 219770 |
+
"step": 94635
|
| 219771 |
+
},
|
| 219772 |
+
{
|
| 219773 |
+
"epoch": 756.47,
|
| 219774 |
+
"learning_rate": 8.49996794871795e-06,
|
| 219775 |
+
"loss": 0.2952,
|
| 219776 |
+
"step": 94640
|
| 219777 |
+
},
|
| 219778 |
+
{
|
| 219779 |
+
"epoch": 756.51,
|
| 219780 |
+
"learning_rate": 8.499887820512821e-06,
|
| 219781 |
+
"loss": 0.2887,
|
| 219782 |
+
"step": 94645
|
| 219783 |
+
},
|
| 219784 |
+
{
|
| 219785 |
+
"epoch": 756.55,
|
| 219786 |
+
"learning_rate": 8.499807692307694e-06,
|
| 219787 |
+
"loss": 0.4188,
|
| 219788 |
+
"step": 94650
|
| 219789 |
+
},
|
| 219790 |
+
{
|
| 219791 |
+
"epoch": 756.59,
|
| 219792 |
+
"learning_rate": 8.499727564102565e-06,
|
| 219793 |
+
"loss": 0.9214,
|
| 219794 |
+
"step": 94655
|
| 219795 |
+
},
|
| 219796 |
+
{
|
| 219797 |
+
"epoch": 756.63,
|
| 219798 |
+
"learning_rate": 8.499647435897437e-06,
|
| 219799 |
+
"loss": 0.7252,
|
| 219800 |
+
"step": 94660
|
| 219801 |
+
},
|
| 219802 |
+
{
|
| 219803 |
+
"epoch": 756.67,
|
| 219804 |
+
"learning_rate": 8.499567307692308e-06,
|
| 219805 |
+
"loss": 0.2708,
|
| 219806 |
+
"step": 94665
|
| 219807 |
+
},
|
| 219808 |
+
{
|
| 219809 |
+
"epoch": 756.71,
|
| 219810 |
+
"learning_rate": 8.499487179487181e-06,
|
| 219811 |
+
"loss": 0.2798,
|
| 219812 |
+
"step": 94670
|
| 219813 |
+
},
|
| 219814 |
+
{
|
| 219815 |
+
"epoch": 756.75,
|
| 219816 |
+
"learning_rate": 8.49940705128205e-06,
|
| 219817 |
+
"loss": 0.3793,
|
| 219818 |
+
"step": 94675
|
| 219819 |
+
},
|
| 219820 |
+
{
|
| 219821 |
+
"epoch": 756.79,
|
| 219822 |
+
"learning_rate": 8.499326923076924e-06,
|
| 219823 |
+
"loss": 0.8439,
|
| 219824 |
+
"step": 94680
|
| 219825 |
+
},
|
| 219826 |
+
{
|
| 219827 |
+
"epoch": 756.83,
|
| 219828 |
+
"learning_rate": 8.499246794871797e-06,
|
| 219829 |
+
"loss": 0.5653,
|
| 219830 |
+
"step": 94685
|
| 219831 |
+
},
|
| 219832 |
+
{
|
| 219833 |
+
"epoch": 756.87,
|
| 219834 |
+
"learning_rate": 8.499166666666666e-06,
|
| 219835 |
+
"loss": 0.2796,
|
| 219836 |
+
"step": 94690
|
| 219837 |
+
},
|
| 219838 |
+
{
|
| 219839 |
+
"epoch": 756.91,
|
| 219840 |
+
"learning_rate": 8.49908653846154e-06,
|
| 219841 |
+
"loss": 0.2814,
|
| 219842 |
+
"step": 94695
|
| 219843 |
+
},
|
| 219844 |
+
{
|
| 219845 |
+
"epoch": 756.95,
|
| 219846 |
+
"learning_rate": 8.499006410256411e-06,
|
| 219847 |
+
"loss": 0.3786,
|
| 219848 |
+
"step": 94700
|
| 219849 |
+
},
|
| 219850 |
+
{
|
| 219851 |
+
"epoch": 756.99,
|
| 219852 |
+
"learning_rate": 8.498926282051282e-06,
|
| 219853 |
+
"loss": 0.9283,
|
| 219854 |
+
"step": 94705
|
| 219855 |
+
},
|
| 219856 |
+
{
|
| 219857 |
+
"epoch": 757.0,
|
| 219858 |
+
"eval_loss": 0.4585740864276886,
|
| 219859 |
+
"eval_runtime": 38.7312,
|
| 219860 |
+
"eval_samples_per_second": 21.662,
|
| 219861 |
+
"eval_steps_per_second": 0.697,
|
| 219862 |
+
"eval_wer": 0.1892521608417888,
|
| 219863 |
+
"step": 94706
|
| 219864 |
+
},
|
| 219865 |
+
{
|
| 219866 |
+
"epoch": 763.03,
|
| 219867 |
+
"learning_rate": 8.498846153846154e-06,
|
| 219868 |
+
"loss": 0.2911,
|
| 219869 |
+
"step": 94710
|
| 219870 |
+
},
|
| 219871 |
+
{
|
| 219872 |
+
"epoch": 763.07,
|
| 219873 |
+
"learning_rate": 8.498766025641027e-06,
|
| 219874 |
+
"loss": 0.3171,
|
| 219875 |
+
"step": 94715
|
| 219876 |
+
},
|
| 219877 |
+
{
|
| 219878 |
+
"epoch": 763.11,
|
| 219879 |
+
"learning_rate": 8.498685897435898e-06,
|
| 219880 |
+
"loss": 0.3134,
|
| 219881 |
+
"step": 94720
|
| 219882 |
+
},
|
| 219883 |
+
{
|
| 219884 |
+
"epoch": 763.15,
|
| 219885 |
+
"learning_rate": 8.49860576923077e-06,
|
| 219886 |
+
"loss": 0.361,
|
| 219887 |
+
"step": 94725
|
| 219888 |
+
},
|
| 219889 |
+
{
|
| 219890 |
+
"epoch": 763.19,
|
| 219891 |
+
"learning_rate": 8.498525641025642e-06,
|
| 219892 |
+
"loss": 0.9856,
|
| 219893 |
+
"step": 94730
|
| 219894 |
+
},
|
| 219895 |
+
{
|
| 219896 |
+
"epoch": 763.23,
|
| 219897 |
+
"learning_rate": 8.498445512820514e-06,
|
| 219898 |
+
"loss": 0.6757,
|
| 219899 |
+
"step": 94735
|
| 219900 |
+
},
|
| 219901 |
+
{
|
| 219902 |
+
"epoch": 763.27,
|
| 219903 |
+
"learning_rate": 8.498365384615385e-06,
|
| 219904 |
+
"loss": 0.3423,
|
| 219905 |
+
"step": 94740
|
| 219906 |
+
},
|
| 219907 |
+
{
|
| 219908 |
+
"epoch": 763.31,
|
| 219909 |
+
"learning_rate": 8.498285256410256e-06,
|
| 219910 |
+
"loss": 0.3902,
|
| 219911 |
+
"step": 94745
|
| 219912 |
+
},
|
| 219913 |
+
{
|
| 219914 |
+
"epoch": 763.35,
|
| 219915 |
+
"learning_rate": 8.49820512820513e-06,
|
| 219916 |
+
"loss": 0.3624,
|
| 219917 |
+
"step": 94750
|
| 219918 |
+
},
|
| 219919 |
+
{
|
| 219920 |
+
"epoch": 763.39,
|
| 219921 |
+
"learning_rate": 8.498125000000001e-06,
|
| 219922 |
+
"loss": 0.8491,
|
| 219923 |
+
"step": 94755
|
| 219924 |
+
},
|
| 219925 |
+
{
|
| 219926 |
+
"epoch": 763.43,
|
| 219927 |
+
"learning_rate": 8.498044871794872e-06,
|
| 219928 |
+
"loss": 0.7503,
|
| 219929 |
+
"step": 94760
|
| 219930 |
+
},
|
| 219931 |
+
{
|
| 219932 |
+
"epoch": 763.47,
|
| 219933 |
+
"learning_rate": 8.497964743589744e-06,
|
| 219934 |
+
"loss": 0.2657,
|
| 219935 |
+
"step": 94765
|
| 219936 |
+
},
|
| 219937 |
+
{
|
| 219938 |
+
"epoch": 763.51,
|
| 219939 |
+
"learning_rate": 8.497884615384617e-06,
|
| 219940 |
+
"loss": 0.4478,
|
| 219941 |
+
"step": 94770
|
| 219942 |
+
},
|
| 219943 |
+
{
|
| 219944 |
+
"epoch": 763.55,
|
| 219945 |
+
"learning_rate": 8.497804487179488e-06,
|
| 219946 |
+
"loss": 0.4349,
|
| 219947 |
+
"step": 94775
|
| 219948 |
+
},
|
| 219949 |
+
{
|
| 219950 |
+
"epoch": 763.59,
|
| 219951 |
+
"learning_rate": 8.49772435897436e-06,
|
| 219952 |
+
"loss": 0.8767,
|
| 219953 |
+
"step": 94780
|
| 219954 |
+
},
|
| 219955 |
+
{
|
| 219956 |
+
"epoch": 763.63,
|
| 219957 |
+
"learning_rate": 8.497644230769232e-06,
|
| 219958 |
+
"loss": 0.7005,
|
| 219959 |
+
"step": 94785
|
| 219960 |
+
},
|
| 219961 |
+
{
|
| 219962 |
+
"epoch": 763.67,
|
| 219963 |
+
"learning_rate": 8.497564102564104e-06,
|
| 219964 |
+
"loss": 0.2645,
|
| 219965 |
+
"step": 94790
|
| 219966 |
+
},
|
| 219967 |
+
{
|
| 219968 |
+
"epoch": 763.71,
|
| 219969 |
+
"learning_rate": 8.497483974358975e-06,
|
| 219970 |
+
"loss": 0.326,
|
| 219971 |
+
"step": 94795
|
| 219972 |
+
},
|
| 219973 |
+
{
|
| 219974 |
+
"epoch": 763.76,
|
| 219975 |
+
"learning_rate": 8.497403846153847e-06,
|
| 219976 |
+
"loss": 0.4294,
|
| 219977 |
+
"step": 94800
|
| 219978 |
+
},
|
| 219979 |
+
{
|
| 219980 |
+
"epoch": 763.8,
|
| 219981 |
+
"learning_rate": 8.49732371794872e-06,
|
| 219982 |
+
"loss": 0.9483,
|
| 219983 |
+
"step": 94805
|
| 219984 |
+
},
|
| 219985 |
+
{
|
| 219986 |
+
"epoch": 763.84,
|
| 219987 |
+
"learning_rate": 8.497243589743591e-06,
|
| 219988 |
+
"loss": 0.7171,
|
| 219989 |
+
"step": 94810
|
| 219990 |
+
},
|
| 219991 |
+
{
|
| 219992 |
+
"epoch": 763.88,
|
| 219993 |
+
"learning_rate": 8.497163461538462e-06,
|
| 219994 |
+
"loss": 0.2639,
|
| 219995 |
+
"step": 94815
|
| 219996 |
+
},
|
| 219997 |
+
{
|
| 219998 |
+
"epoch": 763.92,
|
| 219999 |
+
"learning_rate": 8.497083333333334e-06,
|
| 220000 |
+
"loss": 0.497,
|
| 220001 |
+
"step": 94820
|
| 220002 |
+
},
|
| 220003 |
+
{
|
| 220004 |
+
"epoch": 763.96,
|
| 220005 |
+
"learning_rate": 8.497003205128207e-06,
|
| 220006 |
+
"loss": 0.4221,
|
| 220007 |
+
"step": 94825
|
| 220008 |
+
},
|
| 220009 |
+
{
|
| 220010 |
+
"epoch": 764.0,
|
| 220011 |
+
"learning_rate": 8.496923076923078e-06,
|
| 220012 |
+
"loss": 0.9889,
|
| 220013 |
+
"step": 94830
|
| 220014 |
+
},
|
| 220015 |
+
{
|
| 220016 |
+
"epoch": 764.0,
|
| 220017 |
+
"eval_loss": 0.3748392164707184,
|
| 220018 |
+
"eval_runtime": 38.4646,
|
| 220019 |
+
"eval_samples_per_second": 21.786,
|
| 220020 |
+
"eval_steps_per_second": 0.702,
|
| 220021 |
+
"eval_wer": 0.18795899076565112,
|
| 220022 |
+
"step": 94830
|
| 220023 |
+
},
|
| 220024 |
+
{
|
| 220025 |
+
"epoch": 758.04,
|
| 220026 |
+
"learning_rate": 8.49684294871795e-06,
|
| 220027 |
+
"loss": 0.3387,
|
| 220028 |
+
"step": 94835
|
| 220029 |
+
},
|
| 220030 |
+
{
|
| 220031 |
+
"epoch": 758.08,
|
| 220032 |
+
"learning_rate": 8.496762820512822e-06,
|
| 220033 |
+
"loss": 0.3322,
|
| 220034 |
+
"step": 94840
|
| 220035 |
+
},
|
| 220036 |
+
{
|
| 220037 |
+
"epoch": 758.12,
|
| 220038 |
+
"learning_rate": 8.496682692307692e-06,
|
| 220039 |
+
"loss": 0.2927,
|
| 220040 |
+
"step": 94845
|
| 220041 |
+
},
|
| 220042 |
+
{
|
| 220043 |
+
"epoch": 758.16,
|
| 220044 |
+
"learning_rate": 8.496602564102565e-06,
|
| 220045 |
+
"loss": 0.4402,
|
| 220046 |
+
"step": 94850
|
| 220047 |
+
},
|
| 220048 |
+
{
|
| 220049 |
+
"epoch": 758.2,
|
| 220050 |
+
"learning_rate": 8.496522435897437e-06,
|
| 220051 |
+
"loss": 1.09,
|
| 220052 |
+
"step": 94855
|
| 220053 |
+
},
|
| 220054 |
+
{
|
| 220055 |
+
"epoch": 758.24,
|
| 220056 |
+
"learning_rate": 8.496442307692308e-06,
|
| 220057 |
+
"loss": 0.4025,
|
| 220058 |
+
"step": 94860
|
| 220059 |
+
},
|
| 220060 |
+
{
|
| 220061 |
+
"epoch": 758.28,
|
| 220062 |
+
"learning_rate": 8.49636217948718e-06,
|
| 220063 |
+
"loss": 0.3715,
|
| 220064 |
+
"step": 94865
|
| 220065 |
+
},
|
| 220066 |
+
{
|
| 220067 |
+
"epoch": 758.32,
|
| 220068 |
+
"learning_rate": 8.496282051282052e-06,
|
| 220069 |
+
"loss": 0.3208,
|
| 220070 |
+
"step": 94870
|
| 220071 |
+
},
|
| 220072 |
+
{
|
| 220073 |
+
"epoch": 758.36,
|
| 220074 |
+
"learning_rate": 8.496201923076924e-06,
|
| 220075 |
+
"loss": 0.4094,
|
| 220076 |
+
"step": 94875
|
| 220077 |
+
},
|
| 220078 |
+
{
|
| 220079 |
+
"epoch": 758.4,
|
| 220080 |
+
"learning_rate": 8.496121794871795e-06,
|
| 220081 |
+
"loss": 1.1571,
|
| 220082 |
+
"step": 94880
|
| 220083 |
+
},
|
| 220084 |
+
{
|
| 220085 |
+
"epoch": 758.44,
|
| 220086 |
+
"learning_rate": 8.496041666666668e-06,
|
| 220087 |
+
"loss": 0.3102,
|
| 220088 |
+
"step": 94885
|
| 220089 |
+
},
|
| 220090 |
+
{
|
| 220091 |
+
"epoch": 758.48,
|
| 220092 |
+
"learning_rate": 8.49596153846154e-06,
|
| 220093 |
+
"loss": 0.2724,
|
| 220094 |
+
"step": 94890
|
| 220095 |
+
},
|
| 220096 |
+
{
|
| 220097 |
+
"epoch": 758.52,
|
| 220098 |
+
"learning_rate": 8.49588141025641e-06,
|
| 220099 |
+
"loss": 0.3717,
|
| 220100 |
+
"step": 94895
|
| 220101 |
+
},
|
| 220102 |
+
{
|
| 220103 |
+
"epoch": 758.56,
|
| 220104 |
+
"learning_rate": 8.495801282051282e-06,
|
| 220105 |
+
"loss": 0.386,
|
| 220106 |
+
"step": 94900
|
| 220107 |
+
},
|
| 220108 |
+
{
|
| 220109 |
+
"epoch": 758.6,
|
| 220110 |
+
"learning_rate": 8.495721153846155e-06,
|
| 220111 |
+
"loss": 1.263,
|
| 220112 |
+
"step": 94905
|
| 220113 |
+
},
|
| 220114 |
+
{
|
| 220115 |
+
"epoch": 758.64,
|
| 220116 |
+
"learning_rate": 8.495641025641027e-06,
|
| 220117 |
+
"loss": 0.3325,
|
| 220118 |
+
"step": 94910
|
| 220119 |
+
},
|
| 220120 |
+
{
|
| 220121 |
+
"epoch": 758.68,
|
| 220122 |
+
"learning_rate": 8.495560897435898e-06,
|
| 220123 |
+
"loss": 0.3088,
|
| 220124 |
+
"step": 94915
|
| 220125 |
+
},
|
| 220126 |
+
{
|
| 220127 |
+
"epoch": 758.72,
|
| 220128 |
+
"learning_rate": 8.49548076923077e-06,
|
| 220129 |
+
"loss": 0.3919,
|
| 220130 |
+
"step": 94920
|
| 220131 |
+
},
|
| 220132 |
+
{
|
| 220133 |
+
"epoch": 758.76,
|
| 220134 |
+
"learning_rate": 8.495400641025642e-06,
|
| 220135 |
+
"loss": 0.4832,
|
| 220136 |
+
"step": 94925
|
| 220137 |
+
},
|
| 220138 |
+
{
|
| 220139 |
+
"epoch": 758.8,
|
| 220140 |
+
"learning_rate": 8.495320512820514e-06,
|
| 220141 |
+
"loss": 1.2671,
|
| 220142 |
+
"step": 94930
|
| 220143 |
+
},
|
| 220144 |
+
{
|
| 220145 |
+
"epoch": 758.84,
|
| 220146 |
+
"learning_rate": 8.495240384615385e-06,
|
| 220147 |
+
"loss": 0.3567,
|
| 220148 |
+
"step": 94935
|
| 220149 |
+
},
|
| 220150 |
+
{
|
| 220151 |
+
"epoch": 758.88,
|
| 220152 |
+
"learning_rate": 8.495160256410258e-06,
|
| 220153 |
+
"loss": 0.3045,
|
| 220154 |
+
"step": 94940
|
| 220155 |
+
},
|
| 220156 |
+
{
|
| 220157 |
+
"epoch": 758.92,
|
| 220158 |
+
"learning_rate": 8.49508012820513e-06,
|
| 220159 |
+
"loss": 0.379,
|
| 220160 |
+
"step": 94945
|
| 220161 |
+
},
|
| 220162 |
+
{
|
| 220163 |
+
"epoch": 758.96,
|
| 220164 |
+
"learning_rate": 8.495e-06,
|
| 220165 |
+
"loss": 0.5472,
|
| 220166 |
+
"step": 94950
|
| 220167 |
+
},
|
| 220168 |
+
{
|
| 220169 |
+
"epoch": 759.0,
|
| 220170 |
+
"learning_rate": 8.494919871794872e-06,
|
| 220171 |
+
"loss": 1.2067,
|
| 220172 |
+
"step": 94955
|
| 220173 |
+
},
|
| 220174 |
+
{
|
| 220175 |
+
"epoch": 759.0,
|
| 220176 |
+
"eval_loss": 0.3525693118572235,
|
| 220177 |
+
"eval_runtime": 40.2186,
|
| 220178 |
+
"eval_samples_per_second": 20.886,
|
| 220179 |
+
"eval_steps_per_second": 0.671,
|
| 220180 |
+
"eval_wer": 0.17930283224400873,
|
| 220181 |
+
"step": 94955
|
| 220182 |
+
},
|
| 220183 |
+
{
|
| 220184 |
+
"epoch": 765.04,
|
| 220185 |
+
"learning_rate": 8.494839743589745e-06,
|
| 220186 |
+
"loss": 0.3613,
|
| 220187 |
+
"step": 94960
|
| 220188 |
+
},
|
| 220189 |
+
{
|
| 220190 |
+
"epoch": 765.08,
|
| 220191 |
+
"learning_rate": 8.494759615384615e-06,
|
| 220192 |
+
"loss": 0.3132,
|
| 220193 |
+
"step": 94965
|
| 220194 |
+
},
|
| 220195 |
+
{
|
| 220196 |
+
"epoch": 765.12,
|
| 220197 |
+
"learning_rate": 8.494679487179488e-06,
|
| 220198 |
+
"loss": 0.3128,
|
| 220199 |
+
"step": 94970
|
| 220200 |
+
},
|
| 220201 |
+
{
|
| 220202 |
+
"epoch": 765.16,
|
| 220203 |
+
"learning_rate": 8.49459935897436e-06,
|
| 220204 |
+
"loss": 0.5602,
|
| 220205 |
+
"step": 94975
|
| 220206 |
+
},
|
| 220207 |
+
{
|
| 220208 |
+
"epoch": 765.2,
|
| 220209 |
+
"learning_rate": 8.49451923076923e-06,
|
| 220210 |
+
"loss": 1.2622,
|
| 220211 |
+
"step": 94980
|
| 220212 |
+
},
|
| 220213 |
+
{
|
| 220214 |
+
"epoch": 765.24,
|
| 220215 |
+
"learning_rate": 8.494439102564104e-06,
|
| 220216 |
+
"loss": 0.2838,
|
| 220217 |
+
"step": 94985
|
| 220218 |
+
},
|
| 220219 |
+
{
|
| 220220 |
+
"epoch": 765.28,
|
| 220221 |
+
"learning_rate": 8.494358974358975e-06,
|
| 220222 |
+
"loss": 0.3754,
|
| 220223 |
+
"step": 94990
|
| 220224 |
+
},
|
| 220225 |
+
{
|
| 220226 |
+
"epoch": 765.32,
|
| 220227 |
+
"learning_rate": 8.494278846153846e-06,
|
| 220228 |
+
"loss": 0.3182,
|
| 220229 |
+
"step": 94995
|
| 220230 |
+
},
|
| 220231 |
+
{
|
| 220232 |
+
"epoch": 765.36,
|
| 220233 |
+
"learning_rate": 8.494198717948718e-06,
|
| 220234 |
+
"loss": 0.4919,
|
| 220235 |
+
"step": 95000
|
| 220236 |
+
},
|
| 220237 |
+
{
|
| 220238 |
+
"epoch": 765.4,
|
| 220239 |
+
"learning_rate": 8.494118589743591e-06,
|
| 220240 |
+
"loss": 1.2478,
|
| 220241 |
+
"step": 95005
|
| 220242 |
+
},
|
| 220243 |
+
{
|
| 220244 |
+
"epoch": 765.44,
|
| 220245 |
+
"learning_rate": 8.494038461538462e-06,
|
| 220246 |
+
"loss": 0.3631,
|
| 220247 |
+
"step": 95010
|
| 220248 |
+
},
|
| 220249 |
+
{
|
| 220250 |
+
"epoch": 765.48,
|
| 220251 |
+
"learning_rate": 8.493958333333334e-06,
|
| 220252 |
+
"loss": 0.3297,
|
| 220253 |
+
"step": 95015
|
| 220254 |
+
},
|
| 220255 |
+
{
|
| 220256 |
+
"epoch": 765.52,
|
| 220257 |
+
"learning_rate": 8.493878205128205e-06,
|
| 220258 |
+
"loss": 0.2876,
|
| 220259 |
+
"step": 95020
|
| 220260 |
+
},
|
| 220261 |
+
{
|
| 220262 |
+
"epoch": 765.56,
|
| 220263 |
+
"learning_rate": 8.493798076923078e-06,
|
| 220264 |
+
"loss": 0.519,
|
| 220265 |
+
"step": 95025
|
| 220266 |
+
},
|
| 220267 |
+
{
|
| 220268 |
+
"epoch": 765.6,
|
| 220269 |
+
"learning_rate": 8.49371794871795e-06,
|
| 220270 |
+
"loss": 1.2066,
|
| 220271 |
+
"step": 95030
|
| 220272 |
+
},
|
| 220273 |
+
{
|
| 220274 |
+
"epoch": 765.65,
|
| 220275 |
+
"learning_rate": 8.49363782051282e-06,
|
| 220276 |
+
"loss": 0.2743,
|
| 220277 |
+
"step": 95035
|
| 220278 |
+
},
|
| 220279 |
+
{
|
| 220280 |
+
"epoch": 765.69,
|
| 220281 |
+
"learning_rate": 8.493557692307694e-06,
|
| 220282 |
+
"loss": 0.3329,
|
| 220283 |
+
"step": 95040
|
| 220284 |
+
},
|
| 220285 |
+
{
|
| 220286 |
+
"epoch": 765.73,
|
| 220287 |
+
"learning_rate": 8.493477564102565e-06,
|
| 220288 |
+
"loss": 0.28,
|
| 220289 |
+
"step": 95045
|
| 220290 |
+
},
|
| 220291 |
+
{
|
| 220292 |
+
"epoch": 765.77,
|
| 220293 |
+
"learning_rate": 8.493397435897436e-06,
|
| 220294 |
+
"loss": 0.3887,
|
| 220295 |
+
"step": 95050
|
| 220296 |
+
},
|
| 220297 |
+
{
|
| 220298 |
+
"epoch": 765.81,
|
| 220299 |
+
"learning_rate": 8.493317307692308e-06,
|
| 220300 |
+
"loss": 1.1578,
|
| 220301 |
+
"step": 95055
|
| 220302 |
+
},
|
| 220303 |
+
{
|
| 220304 |
+
"epoch": 765.85,
|
| 220305 |
+
"learning_rate": 8.493237179487181e-06,
|
| 220306 |
+
"loss": 0.3577,
|
| 220307 |
+
"step": 95060
|
| 220308 |
+
},
|
| 220309 |
+
{
|
| 220310 |
+
"epoch": 765.89,
|
| 220311 |
+
"learning_rate": 8.493157051282052e-06,
|
| 220312 |
+
"loss": 0.2978,
|
| 220313 |
+
"step": 95065
|
| 220314 |
+
},
|
| 220315 |
+
{
|
| 220316 |
+
"epoch": 765.93,
|
| 220317 |
+
"learning_rate": 8.493076923076924e-06,
|
| 220318 |
+
"loss": 0.3179,
|
| 220319 |
+
"step": 95070
|
| 220320 |
+
},
|
| 220321 |
+
{
|
| 220322 |
+
"epoch": 765.97,
|
| 220323 |
+
"learning_rate": 8.492996794871795e-06,
|
| 220324 |
+
"loss": 0.5798,
|
| 220325 |
+
"step": 95075
|
| 220326 |
+
},
|
| 220327 |
+
{
|
| 220328 |
+
"epoch": 766.0,
|
| 220329 |
+
"eval_loss": 0.37224629521369934,
|
| 220330 |
+
"eval_runtime": 38.0916,
|
| 220331 |
+
"eval_samples_per_second": 22.052,
|
| 220332 |
+
"eval_steps_per_second": 0.709,
|
| 220333 |
+
"eval_wer": 0.18887413309724066,
|
| 220334 |
+
"step": 95079
|
| 220335 |
+
},
|
| 220336 |
+
{
|
| 220337 |
+
"epoch": 766.01,
|
| 220338 |
+
"learning_rate": 8.492916666666668e-06,
|
| 220339 |
+
"loss": 0.3021,
|
| 220340 |
+
"step": 95080
|
| 220341 |
+
},
|
| 220342 |
+
{
|
| 220343 |
+
"epoch": 766.05,
|
| 220344 |
+
"learning_rate": 8.49283653846154e-06,
|
| 220345 |
+
"loss": 0.3712,
|
| 220346 |
+
"step": 95085
|
| 220347 |
+
},
|
| 220348 |
+
{
|
| 220349 |
+
"epoch": 766.09,
|
| 220350 |
+
"learning_rate": 8.49275641025641e-06,
|
| 220351 |
+
"loss": 0.3271,
|
| 220352 |
+
"step": 95090
|
| 220353 |
+
},
|
| 220354 |
+
{
|
| 220355 |
+
"epoch": 766.13,
|
| 220356 |
+
"learning_rate": 8.492676282051284e-06,
|
| 220357 |
+
"loss": 0.3178,
|
| 220358 |
+
"step": 95095
|
| 220359 |
+
},
|
| 220360 |
+
{
|
| 220361 |
+
"epoch": 766.17,
|
| 220362 |
+
"learning_rate": 8.492596153846155e-06,
|
| 220363 |
+
"loss": 0.4755,
|
| 220364 |
+
"step": 95100
|
| 220365 |
+
},
|
| 220366 |
+
{
|
| 220367 |
+
"epoch": 766.21,
|
| 220368 |
+
"learning_rate": 8.492516025641026e-06,
|
| 220369 |
+
"loss": 1.2469,
|
| 220370 |
+
"step": 95105
|
| 220371 |
+
},
|
| 220372 |
+
{
|
| 220373 |
+
"epoch": 766.25,
|
| 220374 |
+
"learning_rate": 8.492435897435898e-06,
|
| 220375 |
+
"loss": 0.2993,
|
| 220376 |
+
"step": 95110
|
| 220377 |
+
},
|
| 220378 |
+
{
|
| 220379 |
+
"epoch": 766.29,
|
| 220380 |
+
"learning_rate": 8.492355769230771e-06,
|
| 220381 |
+
"loss": 0.2966,
|
| 220382 |
+
"step": 95115
|
| 220383 |
+
},
|
| 220384 |
+
{
|
| 220385 |
+
"epoch": 766.33,
|
| 220386 |
+
"learning_rate": 8.49227564102564e-06,
|
| 220387 |
+
"loss": 0.2925,
|
| 220388 |
+
"step": 95120
|
| 220389 |
+
},
|
| 220390 |
+
{
|
| 220391 |
+
"epoch": 766.37,
|
| 220392 |
+
"learning_rate": 8.492195512820514e-06,
|
| 220393 |
+
"loss": 0.5868,
|
| 220394 |
+
"step": 95125
|
| 220395 |
+
},
|
| 220396 |
+
{
|
| 220397 |
+
"epoch": 766.41,
|
| 220398 |
+
"learning_rate": 8.492115384615385e-06,
|
| 220399 |
+
"loss": 1.0111,
|
| 220400 |
+
"step": 95130
|
| 220401 |
+
},
|
| 220402 |
+
{
|
| 220403 |
+
"epoch": 766.45,
|
| 220404 |
+
"learning_rate": 8.492035256410256e-06,
|
| 220405 |
+
"loss": 0.3314,
|
| 220406 |
+
"step": 95135
|
| 220407 |
+
},
|
| 220408 |
+
{
|
| 220409 |
+
"epoch": 766.49,
|
| 220410 |
+
"learning_rate": 8.49195512820513e-06,
|
| 220411 |
+
"loss": 0.2718,
|
| 220412 |
+
"step": 95140
|
| 220413 |
+
},
|
| 220414 |
+
{
|
| 220415 |
+
"epoch": 766.53,
|
| 220416 |
+
"learning_rate": 8.491875e-06,
|
| 220417 |
+
"loss": 0.3027,
|
| 220418 |
+
"step": 95145
|
| 220419 |
+
},
|
| 220420 |
+
{
|
| 220421 |
+
"epoch": 766.57,
|
| 220422 |
+
"learning_rate": 8.491794871794872e-06,
|
| 220423 |
+
"loss": 0.568,
|
| 220424 |
+
"step": 95150
|
| 220425 |
+
},
|
| 220426 |
+
{
|
| 220427 |
+
"epoch": 766.61,
|
| 220428 |
+
"learning_rate": 8.491714743589743e-06,
|
| 220429 |
+
"loss": 1.1562,
|
| 220430 |
+
"step": 95155
|
| 220431 |
+
},
|
| 220432 |
+
{
|
| 220433 |
+
"epoch": 766.65,
|
| 220434 |
+
"learning_rate": 8.491634615384617e-06,
|
| 220435 |
+
"loss": 0.3352,
|
| 220436 |
+
"step": 95160
|
| 220437 |
+
},
|
| 220438 |
+
{
|
| 220439 |
+
"epoch": 766.69,
|
| 220440 |
+
"learning_rate": 8.491554487179488e-06,
|
| 220441 |
+
"loss": 0.2736,
|
| 220442 |
+
"step": 95165
|
| 220443 |
+
},
|
| 220444 |
+
{
|
| 220445 |
+
"epoch": 766.73,
|
| 220446 |
+
"learning_rate": 8.49147435897436e-06,
|
| 220447 |
+
"loss": 0.3534,
|
| 220448 |
+
"step": 95170
|
| 220449 |
+
},
|
| 220450 |
+
{
|
| 220451 |
+
"epoch": 766.77,
|
| 220452 |
+
"learning_rate": 8.49139423076923e-06,
|
| 220453 |
+
"loss": 0.5592,
|
| 220454 |
+
"step": 95175
|
| 220455 |
+
},
|
| 220456 |
+
{
|
| 220457 |
+
"epoch": 766.81,
|
| 220458 |
+
"learning_rate": 8.491314102564104e-06,
|
| 220459 |
+
"loss": 1.0819,
|
| 220460 |
+
"step": 95180
|
| 220461 |
+
},
|
| 220462 |
+
{
|
| 220463 |
+
"epoch": 766.85,
|
| 220464 |
+
"learning_rate": 8.491233974358975e-06,
|
| 220465 |
+
"loss": 0.3304,
|
| 220466 |
+
"step": 95185
|
| 220467 |
+
},
|
| 220468 |
+
{
|
| 220469 |
+
"epoch": 766.89,
|
| 220470 |
+
"learning_rate": 8.491153846153846e-06,
|
| 220471 |
+
"loss": 0.3241,
|
| 220472 |
+
"step": 95190
|
| 220473 |
+
},
|
| 220474 |
+
{
|
| 220475 |
+
"epoch": 766.93,
|
| 220476 |
+
"learning_rate": 8.49107371794872e-06,
|
| 220477 |
+
"loss": 0.3405,
|
| 220478 |
+
"step": 95195
|
| 220479 |
+
},
|
| 220480 |
+
{
|
| 220481 |
+
"epoch": 766.97,
|
| 220482 |
+
"learning_rate": 8.49099358974359e-06,
|
| 220483 |
+
"loss": 0.6395,
|
| 220484 |
+
"step": 95200
|
| 220485 |
+
},
|
| 220486 |
+
{
|
| 220487 |
+
"epoch": 767.0,
|
| 220488 |
+
"eval_loss": 0.39154714345932007,
|
| 220489 |
+
"eval_runtime": 38.8655,
|
| 220490 |
+
"eval_samples_per_second": 21.613,
|
| 220491 |
+
"eval_steps_per_second": 0.695,
|
| 220492 |
+
"eval_wer": 0.18689181453921008,
|
| 220493 |
+
"step": 95203
|
| 220494 |
}
|
| 220495 |
],
|
| 220496 |
"max_steps": 620000,
|
| 220497 |
"num_train_epochs": 5000,
|
| 220498 |
+
"total_flos": 2.6790445779490485e+20,
|
| 220499 |
"trial_name": null,
|
| 220500 |
"trial_params": null
|
| 220501 |
}
|
model-bin/finetune/base/{checkpoint-94581 → checkpoint-95203}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629962173.6222122/events.out.tfevents.1629962173.8e89bd551565.924.101
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:39cd4f579bbf080652081ef19694b40a42fa0372047764a4d950361f270aa054
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629962621.2864316/events.out.tfevents.1629962621.8e89bd551565.924.103
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f6ca372b0c0d1b72e4cf4fde5f556bd803e6dce13c67b7b44c00d1e12d6b2b0f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629963132.972444/events.out.tfevents.1629963132.8e89bd551565.924.105
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3afe9f5372125de169cda98860f9697717806166440ad04405bd5547cb796df7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629963595.464257/events.out.tfevents.1629963595.8e89bd551565.924.107
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b7bc608f102a7225dee243d350de1a0c85171a2562f4c8855d5e2b5b2d8a860d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629964022.648677/events.out.tfevents.1629964022.8e89bd551565.924.109
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cfd05bc6b419e02b82ef5f6539f9cc30c00428babe7e8582fb9d38c7848c47be
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629962173.8e89bd551565.924.100
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:210b72792a374e97a675000d1c9ba210dd01276ce104c92f78bf534d78c21d6c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629962620.8e89bd551565.924.102
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ed0b27f757bd7a8966567994fa66dd5f3c7f0f38446141d5f024d592bb4b84a3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629963132.8e89bd551565.924.104
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:03494f7f2800abad921eb8b616190d68f77e8a6de8f56e18bdcf4f88cb47d088
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629963595.8e89bd551565.924.106
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:56e37a0d9375255f4baebd82c621ce878c1afd137deacc5c7408a93017149049
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629964022.8e89bd551565.924.108
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2166fe4ae1b6ca491cee8ed40813d06b9a08723cd361790e127f72b9e698b5a9
|
| 3 |
+
size 8622
|