"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629749097.2738104/events.out.tfevents.1629749097.74272264b15c.932.213 +3 -0
- model-bin/finetune/base/log/1629749718.7735815/events.out.tfevents.1629749718.74272264b15c.932.215 +3 -0
- model-bin/finetune/base/log/1629750364.5758934/events.out.tfevents.1629750364.74272264b15c.932.217 +3 -0
- model-bin/finetune/base/log/1629751114.6670787/events.out.tfevents.1629751114.74272264b15c.932.219 +3 -0
- model-bin/finetune/base/log/1629751765.2580316/events.out.tfevents.1629751765.74272264b15c.932.221 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629749097.74272264b15c.932.212 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629749718.74272264b15c.932.214 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629750364.74272264b15c.932.216 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629751114.74272264b15c.932.218 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629751765.74272264b15c.932.220 +3 -0
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:44b00d4f58219b4d318df10c240cc6442e9022f96f2d6f17ab28ad2f843e3d57
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/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:ccd0de1e9539b726c0716fdb3a195e2d50475b80cff249324c9c6298512f047c
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/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 sha256:6e05861741a7b5f0d8dde4975b7b4ab4ba0db65702615da6c6340b05b2596ea5
|
| 3 |
+
size 14439
|
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/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 sha256:83aaac26ff07408c2a26c36f15750d57617e8958aad261fe178da94ede075a9a
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/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 sha256:8b0222eb9c2e85654b83c859b959cbaa02d9327f4d42f81bbeea8e5e0ff98e3a
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -164757,11 +164757,800 @@
|
|
| 164757 |
"eval_steps_per_second": 0.688,
|
| 164758 |
"eval_wer": 0.2009889470622455,
|
| 164759 |
"step": 51395
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 164760 |
}
|
| 164761 |
],
|
| 164762 |
"max_steps": 620000,
|
| 164763 |
"num_train_epochs": 5000,
|
| 164764 |
-
"total_flos": 1.
|
| 164765 |
"trial_name": null,
|
| 164766 |
"trial_params": null
|
| 164767 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.19748327029386092,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-46666",
|
| 4 |
+
"epoch": 418.99598393574297,
|
| 5 |
+
"global_step": 52015,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 164757 |
"eval_steps_per_second": 0.688,
|
| 164758 |
"eval_wer": 0.2009889470622455,
|
| 164759 |
"step": 51395
|
| 164760 |
+
},
|
| 164761 |
+
{
|
| 164762 |
+
"epoch": 414.04,
|
| 164763 |
+
"learning_rate": 9.19261217948718e-06,
|
| 164764 |
+
"loss": 0.353,
|
| 164765 |
+
"step": 51400
|
| 164766 |
+
},
|
| 164767 |
+
{
|
| 164768 |
+
"epoch": 414.08,
|
| 164769 |
+
"learning_rate": 9.192532051282051e-06,
|
| 164770 |
+
"loss": 0.2921,
|
| 164771 |
+
"step": 51405
|
| 164772 |
+
},
|
| 164773 |
+
{
|
| 164774 |
+
"epoch": 414.12,
|
| 164775 |
+
"learning_rate": 9.192451923076924e-06,
|
| 164776 |
+
"loss": 0.3673,
|
| 164777 |
+
"step": 51410
|
| 164778 |
+
},
|
| 164779 |
+
{
|
| 164780 |
+
"epoch": 414.16,
|
| 164781 |
+
"learning_rate": 9.192371794871795e-06,
|
| 164782 |
+
"loss": 0.5239,
|
| 164783 |
+
"step": 51415
|
| 164784 |
+
},
|
| 164785 |
+
{
|
| 164786 |
+
"epoch": 414.2,
|
| 164787 |
+
"learning_rate": 9.192291666666667e-06,
|
| 164788 |
+
"loss": 1.3389,
|
| 164789 |
+
"step": 51420
|
| 164790 |
+
},
|
| 164791 |
+
{
|
| 164792 |
+
"epoch": 414.24,
|
| 164793 |
+
"learning_rate": 9.19221153846154e-06,
|
| 164794 |
+
"loss": 0.3955,
|
| 164795 |
+
"step": 51425
|
| 164796 |
+
},
|
| 164797 |
+
{
|
| 164798 |
+
"epoch": 414.28,
|
| 164799 |
+
"learning_rate": 9.192131410256411e-06,
|
| 164800 |
+
"loss": 0.3084,
|
| 164801 |
+
"step": 51430
|
| 164802 |
+
},
|
| 164803 |
+
{
|
| 164804 |
+
"epoch": 414.32,
|
| 164805 |
+
"learning_rate": 9.192051282051282e-06,
|
| 164806 |
+
"loss": 0.4483,
|
| 164807 |
+
"step": 51435
|
| 164808 |
+
},
|
| 164809 |
+
{
|
| 164810 |
+
"epoch": 414.36,
|
| 164811 |
+
"learning_rate": 9.191971153846154e-06,
|
| 164812 |
+
"loss": 0.5474,
|
| 164813 |
+
"step": 51440
|
| 164814 |
+
},
|
| 164815 |
+
{
|
| 164816 |
+
"epoch": 414.4,
|
| 164817 |
+
"learning_rate": 9.191891025641027e-06,
|
| 164818 |
+
"loss": 1.3813,
|
| 164819 |
+
"step": 51445
|
| 164820 |
+
},
|
| 164821 |
+
{
|
| 164822 |
+
"epoch": 414.44,
|
| 164823 |
+
"learning_rate": 9.191810897435898e-06,
|
| 164824 |
+
"loss": 0.3601,
|
| 164825 |
+
"step": 51450
|
| 164826 |
+
},
|
| 164827 |
+
{
|
| 164828 |
+
"epoch": 414.48,
|
| 164829 |
+
"learning_rate": 9.19173076923077e-06,
|
| 164830 |
+
"loss": 0.3359,
|
| 164831 |
+
"step": 51455
|
| 164832 |
+
},
|
| 164833 |
+
{
|
| 164834 |
+
"epoch": 414.52,
|
| 164835 |
+
"learning_rate": 9.191650641025643e-06,
|
| 164836 |
+
"loss": 0.5013,
|
| 164837 |
+
"step": 51460
|
| 164838 |
+
},
|
| 164839 |
+
{
|
| 164840 |
+
"epoch": 414.56,
|
| 164841 |
+
"learning_rate": 9.191570512820514e-06,
|
| 164842 |
+
"loss": 0.5967,
|
| 164843 |
+
"step": 51465
|
| 164844 |
+
},
|
| 164845 |
+
{
|
| 164846 |
+
"epoch": 414.6,
|
| 164847 |
+
"learning_rate": 9.191490384615385e-06,
|
| 164848 |
+
"loss": 1.2045,
|
| 164849 |
+
"step": 51470
|
| 164850 |
+
},
|
| 164851 |
+
{
|
| 164852 |
+
"epoch": 414.64,
|
| 164853 |
+
"learning_rate": 9.191410256410257e-06,
|
| 164854 |
+
"loss": 0.3618,
|
| 164855 |
+
"step": 51475
|
| 164856 |
+
},
|
| 164857 |
+
{
|
| 164858 |
+
"epoch": 414.68,
|
| 164859 |
+
"learning_rate": 9.19133012820513e-06,
|
| 164860 |
+
"loss": 0.3224,
|
| 164861 |
+
"step": 51480
|
| 164862 |
+
},
|
| 164863 |
+
{
|
| 164864 |
+
"epoch": 414.72,
|
| 164865 |
+
"learning_rate": 9.191250000000001e-06,
|
| 164866 |
+
"loss": 0.4466,
|
| 164867 |
+
"step": 51485
|
| 164868 |
+
},
|
| 164869 |
+
{
|
| 164870 |
+
"epoch": 414.76,
|
| 164871 |
+
"learning_rate": 9.191169871794873e-06,
|
| 164872 |
+
"loss": 0.5253,
|
| 164873 |
+
"step": 51490
|
| 164874 |
+
},
|
| 164875 |
+
{
|
| 164876 |
+
"epoch": 414.8,
|
| 164877 |
+
"learning_rate": 9.191089743589744e-06,
|
| 164878 |
+
"loss": 1.3537,
|
| 164879 |
+
"step": 51495
|
| 164880 |
+
},
|
| 164881 |
+
{
|
| 164882 |
+
"epoch": 414.84,
|
| 164883 |
+
"learning_rate": 9.191009615384617e-06,
|
| 164884 |
+
"loss": 0.412,
|
| 164885 |
+
"step": 51500
|
| 164886 |
+
},
|
| 164887 |
+
{
|
| 164888 |
+
"epoch": 414.88,
|
| 164889 |
+
"learning_rate": 9.190929487179487e-06,
|
| 164890 |
+
"loss": 0.4065,
|
| 164891 |
+
"step": 51505
|
| 164892 |
+
},
|
| 164893 |
+
{
|
| 164894 |
+
"epoch": 414.92,
|
| 164895 |
+
"learning_rate": 9.19084935897436e-06,
|
| 164896 |
+
"loss": 0.3398,
|
| 164897 |
+
"step": 51510
|
| 164898 |
+
},
|
| 164899 |
+
{
|
| 164900 |
+
"epoch": 414.96,
|
| 164901 |
+
"learning_rate": 9.190769230769233e-06,
|
| 164902 |
+
"loss": 0.5099,
|
| 164903 |
+
"step": 51515
|
| 164904 |
+
},
|
| 164905 |
+
{
|
| 164906 |
+
"epoch": 415.0,
|
| 164907 |
+
"eval_loss": 0.4426712095737457,
|
| 164908 |
+
"eval_runtime": 38.7111,
|
| 164909 |
+
"eval_samples_per_second": 21.699,
|
| 164910 |
+
"eval_steps_per_second": 0.697,
|
| 164911 |
+
"eval_wer": 0.20031044423091138,
|
| 164912 |
+
"step": 51519
|
| 164913 |
+
},
|
| 164914 |
+
{
|
| 164915 |
+
"epoch": 415.01,
|
| 164916 |
+
"learning_rate": 9.190689102564102e-06,
|
| 164917 |
+
"loss": 0.4553,
|
| 164918 |
+
"step": 51520
|
| 164919 |
+
},
|
| 164920 |
+
{
|
| 164921 |
+
"epoch": 415.05,
|
| 164922 |
+
"learning_rate": 9.190608974358975e-06,
|
| 164923 |
+
"loss": 0.3547,
|
| 164924 |
+
"step": 51525
|
| 164925 |
+
},
|
| 164926 |
+
{
|
| 164927 |
+
"epoch": 415.09,
|
| 164928 |
+
"learning_rate": 9.190528846153847e-06,
|
| 164929 |
+
"loss": 0.3651,
|
| 164930 |
+
"step": 51530
|
| 164931 |
+
},
|
| 164932 |
+
{
|
| 164933 |
+
"epoch": 415.13,
|
| 164934 |
+
"learning_rate": 9.190448717948718e-06,
|
| 164935 |
+
"loss": 0.4613,
|
| 164936 |
+
"step": 51535
|
| 164937 |
+
},
|
| 164938 |
+
{
|
| 164939 |
+
"epoch": 415.17,
|
| 164940 |
+
"learning_rate": 9.19036858974359e-06,
|
| 164941 |
+
"loss": 0.6468,
|
| 164942 |
+
"step": 51540
|
| 164943 |
+
},
|
| 164944 |
+
{
|
| 164945 |
+
"epoch": 415.21,
|
| 164946 |
+
"learning_rate": 9.190288461538463e-06,
|
| 164947 |
+
"loss": 1.1471,
|
| 164948 |
+
"step": 51545
|
| 164949 |
+
},
|
| 164950 |
+
{
|
| 164951 |
+
"epoch": 415.25,
|
| 164952 |
+
"learning_rate": 9.190208333333334e-06,
|
| 164953 |
+
"loss": 0.4193,
|
| 164954 |
+
"step": 51550
|
| 164955 |
+
},
|
| 164956 |
+
{
|
| 164957 |
+
"epoch": 415.29,
|
| 164958 |
+
"learning_rate": 9.190128205128205e-06,
|
| 164959 |
+
"loss": 0.3993,
|
| 164960 |
+
"step": 51555
|
| 164961 |
+
},
|
| 164962 |
+
{
|
| 164963 |
+
"epoch": 415.33,
|
| 164964 |
+
"learning_rate": 9.190048076923078e-06,
|
| 164965 |
+
"loss": 0.3883,
|
| 164966 |
+
"step": 51560
|
| 164967 |
+
},
|
| 164968 |
+
{
|
| 164969 |
+
"epoch": 415.37,
|
| 164970 |
+
"learning_rate": 9.18996794871795e-06,
|
| 164971 |
+
"loss": 0.6178,
|
| 164972 |
+
"step": 51565
|
| 164973 |
+
},
|
| 164974 |
+
{
|
| 164975 |
+
"epoch": 415.41,
|
| 164976 |
+
"learning_rate": 9.189887820512821e-06,
|
| 164977 |
+
"loss": 1.2703,
|
| 164978 |
+
"step": 51570
|
| 164979 |
+
},
|
| 164980 |
+
{
|
| 164981 |
+
"epoch": 415.45,
|
| 164982 |
+
"learning_rate": 9.189807692307692e-06,
|
| 164983 |
+
"loss": 0.3288,
|
| 164984 |
+
"step": 51575
|
| 164985 |
+
},
|
| 164986 |
+
{
|
| 164987 |
+
"epoch": 415.49,
|
| 164988 |
+
"learning_rate": 9.189727564102565e-06,
|
| 164989 |
+
"loss": 0.3124,
|
| 164990 |
+
"step": 51580
|
| 164991 |
+
},
|
| 164992 |
+
{
|
| 164993 |
+
"epoch": 415.53,
|
| 164994 |
+
"learning_rate": 9.189647435897437e-06,
|
| 164995 |
+
"loss": 0.3634,
|
| 164996 |
+
"step": 51585
|
| 164997 |
+
},
|
| 164998 |
+
{
|
| 164999 |
+
"epoch": 415.57,
|
| 165000 |
+
"learning_rate": 9.189567307692308e-06,
|
| 165001 |
+
"loss": 0.6493,
|
| 165002 |
+
"step": 51590
|
| 165003 |
+
},
|
| 165004 |
+
{
|
| 165005 |
+
"epoch": 415.61,
|
| 165006 |
+
"learning_rate": 9.18948717948718e-06,
|
| 165007 |
+
"loss": 1.2041,
|
| 165008 |
+
"step": 51595
|
| 165009 |
+
},
|
| 165010 |
+
{
|
| 165011 |
+
"epoch": 415.65,
|
| 165012 |
+
"learning_rate": 9.189407051282053e-06,
|
| 165013 |
+
"loss": 0.3486,
|
| 165014 |
+
"step": 51600
|
| 165015 |
+
},
|
| 165016 |
+
{
|
| 165017 |
+
"epoch": 415.69,
|
| 165018 |
+
"learning_rate": 9.189326923076924e-06,
|
| 165019 |
+
"loss": 0.3547,
|
| 165020 |
+
"step": 51605
|
| 165021 |
+
},
|
| 165022 |
+
{
|
| 165023 |
+
"epoch": 415.73,
|
| 165024 |
+
"learning_rate": 9.189246794871795e-06,
|
| 165025 |
+
"loss": 0.3555,
|
| 165026 |
+
"step": 51610
|
| 165027 |
+
},
|
| 165028 |
+
{
|
| 165029 |
+
"epoch": 415.77,
|
| 165030 |
+
"learning_rate": 9.189166666666668e-06,
|
| 165031 |
+
"loss": 0.6093,
|
| 165032 |
+
"step": 51615
|
| 165033 |
+
},
|
| 165034 |
+
{
|
| 165035 |
+
"epoch": 415.81,
|
| 165036 |
+
"learning_rate": 9.18908653846154e-06,
|
| 165037 |
+
"loss": 1.2668,
|
| 165038 |
+
"step": 51620
|
| 165039 |
+
},
|
| 165040 |
+
{
|
| 165041 |
+
"epoch": 415.85,
|
| 165042 |
+
"learning_rate": 9.189006410256411e-06,
|
| 165043 |
+
"loss": 0.3427,
|
| 165044 |
+
"step": 51625
|
| 165045 |
+
},
|
| 165046 |
+
{
|
| 165047 |
+
"epoch": 415.89,
|
| 165048 |
+
"learning_rate": 9.188926282051282e-06,
|
| 165049 |
+
"loss": 0.3484,
|
| 165050 |
+
"step": 51630
|
| 165051 |
+
},
|
| 165052 |
+
{
|
| 165053 |
+
"epoch": 415.93,
|
| 165054 |
+
"learning_rate": 9.188846153846155e-06,
|
| 165055 |
+
"loss": 0.4014,
|
| 165056 |
+
"step": 51635
|
| 165057 |
+
},
|
| 165058 |
+
{
|
| 165059 |
+
"epoch": 415.97,
|
| 165060 |
+
"learning_rate": 9.188766025641027e-06,
|
| 165061 |
+
"loss": 0.6165,
|
| 165062 |
+
"step": 51640
|
| 165063 |
+
},
|
| 165064 |
+
{
|
| 165065 |
+
"epoch": 416.0,
|
| 165066 |
+
"eval_loss": 0.43282684683799744,
|
| 165067 |
+
"eval_runtime": 39.9024,
|
| 165068 |
+
"eval_samples_per_second": 21.026,
|
| 165069 |
+
"eval_steps_per_second": 0.677,
|
| 165070 |
+
"eval_wer": 0.20908092273892348,
|
| 165071 |
+
"step": 51643
|
| 165072 |
+
},
|
| 165073 |
+
{
|
| 165074 |
+
"epoch": 416.02,
|
| 165075 |
+
"learning_rate": 9.188685897435898e-06,
|
| 165076 |
+
"loss": 0.4721,
|
| 165077 |
+
"step": 51645
|
| 165078 |
+
},
|
| 165079 |
+
{
|
| 165080 |
+
"epoch": 416.06,
|
| 165081 |
+
"learning_rate": 9.18860576923077e-06,
|
| 165082 |
+
"loss": 0.3121,
|
| 165083 |
+
"step": 51650
|
| 165084 |
+
},
|
| 165085 |
+
{
|
| 165086 |
+
"epoch": 416.1,
|
| 165087 |
+
"learning_rate": 9.188525641025643e-06,
|
| 165088 |
+
"loss": 0.2956,
|
| 165089 |
+
"step": 51655
|
| 165090 |
+
},
|
| 165091 |
+
{
|
| 165092 |
+
"epoch": 416.14,
|
| 165093 |
+
"learning_rate": 9.188445512820514e-06,
|
| 165094 |
+
"loss": 0.4189,
|
| 165095 |
+
"step": 51660
|
| 165096 |
+
},
|
| 165097 |
+
{
|
| 165098 |
+
"epoch": 416.18,
|
| 165099 |
+
"learning_rate": 9.188365384615385e-06,
|
| 165100 |
+
"loss": 0.7809,
|
| 165101 |
+
"step": 51665
|
| 165102 |
+
},
|
| 165103 |
+
{
|
| 165104 |
+
"epoch": 416.22,
|
| 165105 |
+
"learning_rate": 9.188285256410258e-06,
|
| 165106 |
+
"loss": 1.1376,
|
| 165107 |
+
"step": 51670
|
| 165108 |
+
},
|
| 165109 |
+
{
|
| 165110 |
+
"epoch": 416.26,
|
| 165111 |
+
"learning_rate": 9.188205128205128e-06,
|
| 165112 |
+
"loss": 0.2887,
|
| 165113 |
+
"step": 51675
|
| 165114 |
+
},
|
| 165115 |
+
{
|
| 165116 |
+
"epoch": 416.3,
|
| 165117 |
+
"learning_rate": 9.188125000000001e-06,
|
| 165118 |
+
"loss": 0.3829,
|
| 165119 |
+
"step": 51680
|
| 165120 |
+
},
|
| 165121 |
+
{
|
| 165122 |
+
"epoch": 416.34,
|
| 165123 |
+
"learning_rate": 9.188044871794872e-06,
|
| 165124 |
+
"loss": 0.4552,
|
| 165125 |
+
"step": 51685
|
| 165126 |
+
},
|
| 165127 |
+
{
|
| 165128 |
+
"epoch": 416.38,
|
| 165129 |
+
"learning_rate": 9.187964743589744e-06,
|
| 165130 |
+
"loss": 0.869,
|
| 165131 |
+
"step": 51690
|
| 165132 |
+
},
|
| 165133 |
+
{
|
| 165134 |
+
"epoch": 416.42,
|
| 165135 |
+
"learning_rate": 9.187884615384615e-06,
|
| 165136 |
+
"loss": 1.0696,
|
| 165137 |
+
"step": 51695
|
| 165138 |
+
},
|
| 165139 |
+
{
|
| 165140 |
+
"epoch": 416.46,
|
| 165141 |
+
"learning_rate": 9.187804487179488e-06,
|
| 165142 |
+
"loss": 0.3194,
|
| 165143 |
+
"step": 51700
|
| 165144 |
+
},
|
| 165145 |
+
{
|
| 165146 |
+
"epoch": 416.5,
|
| 165147 |
+
"learning_rate": 9.18772435897436e-06,
|
| 165148 |
+
"loss": 0.2884,
|
| 165149 |
+
"step": 51705
|
| 165150 |
+
},
|
| 165151 |
+
{
|
| 165152 |
+
"epoch": 416.54,
|
| 165153 |
+
"learning_rate": 9.187644230769231e-06,
|
| 165154 |
+
"loss": 0.4256,
|
| 165155 |
+
"step": 51710
|
| 165156 |
+
},
|
| 165157 |
+
{
|
| 165158 |
+
"epoch": 416.58,
|
| 165159 |
+
"learning_rate": 9.187564102564104e-06,
|
| 165160 |
+
"loss": 0.6619,
|
| 165161 |
+
"step": 51715
|
| 165162 |
+
},
|
| 165163 |
+
{
|
| 165164 |
+
"epoch": 416.62,
|
| 165165 |
+
"learning_rate": 9.187483974358975e-06,
|
| 165166 |
+
"loss": 1.0305,
|
| 165167 |
+
"step": 51720
|
| 165168 |
+
},
|
| 165169 |
+
{
|
| 165170 |
+
"epoch": 416.66,
|
| 165171 |
+
"learning_rate": 9.187403846153847e-06,
|
| 165172 |
+
"loss": 0.3894,
|
| 165173 |
+
"step": 51725
|
| 165174 |
+
},
|
| 165175 |
+
{
|
| 165176 |
+
"epoch": 416.7,
|
| 165177 |
+
"learning_rate": 9.187323717948718e-06,
|
| 165178 |
+
"loss": 0.429,
|
| 165179 |
+
"step": 51730
|
| 165180 |
+
},
|
| 165181 |
+
{
|
| 165182 |
+
"epoch": 416.74,
|
| 165183 |
+
"learning_rate": 9.187243589743591e-06,
|
| 165184 |
+
"loss": 0.395,
|
| 165185 |
+
"step": 51735
|
| 165186 |
+
},
|
| 165187 |
+
{
|
| 165188 |
+
"epoch": 416.78,
|
| 165189 |
+
"learning_rate": 9.187163461538462e-06,
|
| 165190 |
+
"loss": 0.6476,
|
| 165191 |
+
"step": 51740
|
| 165192 |
+
},
|
| 165193 |
+
{
|
| 165194 |
+
"epoch": 416.82,
|
| 165195 |
+
"learning_rate": 9.187083333333334e-06,
|
| 165196 |
+
"loss": 1.0216,
|
| 165197 |
+
"step": 51745
|
| 165198 |
+
},
|
| 165199 |
+
{
|
| 165200 |
+
"epoch": 416.86,
|
| 165201 |
+
"learning_rate": 9.187003205128205e-06,
|
| 165202 |
+
"loss": 0.3681,
|
| 165203 |
+
"step": 51750
|
| 165204 |
+
},
|
| 165205 |
+
{
|
| 165206 |
+
"epoch": 416.9,
|
| 165207 |
+
"learning_rate": 9.186923076923078e-06,
|
| 165208 |
+
"loss": 0.4025,
|
| 165209 |
+
"step": 51755
|
| 165210 |
+
},
|
| 165211 |
+
{
|
| 165212 |
+
"epoch": 416.94,
|
| 165213 |
+
"learning_rate": 9.18684294871795e-06,
|
| 165214 |
+
"loss": 0.4326,
|
| 165215 |
+
"step": 51760
|
| 165216 |
+
},
|
| 165217 |
+
{
|
| 165218 |
+
"epoch": 416.98,
|
| 165219 |
+
"learning_rate": 9.186762820512821e-06,
|
| 165220 |
+
"loss": 0.669,
|
| 165221 |
+
"step": 51765
|
| 165222 |
+
},
|
| 165223 |
+
{
|
| 165224 |
+
"epoch": 417.0,
|
| 165225 |
+
"eval_loss": 0.47711676359176636,
|
| 165226 |
+
"eval_runtime": 40.1533,
|
| 165227 |
+
"eval_samples_per_second": 20.895,
|
| 165228 |
+
"eval_steps_per_second": 0.672,
|
| 165229 |
+
"eval_wer": 0.20052234474753336,
|
| 165230 |
+
"step": 51767
|
| 165231 |
+
},
|
| 165232 |
+
{
|
| 165233 |
+
"epoch": 417.02,
|
| 165234 |
+
"learning_rate": 9.186682692307694e-06,
|
| 165235 |
+
"loss": 0.3275,
|
| 165236 |
+
"step": 51770
|
| 165237 |
+
},
|
| 165238 |
+
{
|
| 165239 |
+
"epoch": 417.06,
|
| 165240 |
+
"learning_rate": 9.186602564102565e-06,
|
| 165241 |
+
"loss": 0.4032,
|
| 165242 |
+
"step": 51775
|
| 165243 |
+
},
|
| 165244 |
+
{
|
| 165245 |
+
"epoch": 417.1,
|
| 165246 |
+
"learning_rate": 9.186522435897437e-06,
|
| 165247 |
+
"loss": 0.2821,
|
| 165248 |
+
"step": 51780
|
| 165249 |
+
},
|
| 165250 |
+
{
|
| 165251 |
+
"epoch": 417.15,
|
| 165252 |
+
"learning_rate": 9.186442307692308e-06,
|
| 165253 |
+
"loss": 0.4831,
|
| 165254 |
+
"step": 51785
|
| 165255 |
+
},
|
| 165256 |
+
{
|
| 165257 |
+
"epoch": 417.19,
|
| 165258 |
+
"learning_rate": 9.186362179487181e-06,
|
| 165259 |
+
"loss": 0.9399,
|
| 165260 |
+
"step": 51790
|
| 165261 |
+
},
|
| 165262 |
+
{
|
| 165263 |
+
"epoch": 417.23,
|
| 165264 |
+
"learning_rate": 9.186282051282052e-06,
|
| 165265 |
+
"loss": 0.9047,
|
| 165266 |
+
"step": 51795
|
| 165267 |
+
},
|
| 165268 |
+
{
|
| 165269 |
+
"epoch": 417.27,
|
| 165270 |
+
"learning_rate": 9.186201923076924e-06,
|
| 165271 |
+
"loss": 0.3493,
|
| 165272 |
+
"step": 51800
|
| 165273 |
+
},
|
| 165274 |
+
{
|
| 165275 |
+
"epoch": 417.31,
|
| 165276 |
+
"learning_rate": 9.186121794871795e-06,
|
| 165277 |
+
"loss": 0.4136,
|
| 165278 |
+
"step": 51805
|
| 165279 |
+
},
|
| 165280 |
+
{
|
| 165281 |
+
"epoch": 417.35,
|
| 165282 |
+
"learning_rate": 9.186041666666668e-06,
|
| 165283 |
+
"loss": 0.3883,
|
| 165284 |
+
"step": 51810
|
| 165285 |
+
},
|
| 165286 |
+
{
|
| 165287 |
+
"epoch": 417.39,
|
| 165288 |
+
"learning_rate": 9.18596153846154e-06,
|
| 165289 |
+
"loss": 0.7535,
|
| 165290 |
+
"step": 51815
|
| 165291 |
+
},
|
| 165292 |
+
{
|
| 165293 |
+
"epoch": 417.43,
|
| 165294 |
+
"learning_rate": 9.185881410256411e-06,
|
| 165295 |
+
"loss": 0.9066,
|
| 165296 |
+
"step": 51820
|
| 165297 |
+
},
|
| 165298 |
+
{
|
| 165299 |
+
"epoch": 417.47,
|
| 165300 |
+
"learning_rate": 9.185801282051284e-06,
|
| 165301 |
+
"loss": 0.3561,
|
| 165302 |
+
"step": 51825
|
| 165303 |
+
},
|
| 165304 |
+
{
|
| 165305 |
+
"epoch": 417.51,
|
| 165306 |
+
"learning_rate": 9.185721153846154e-06,
|
| 165307 |
+
"loss": 0.4193,
|
| 165308 |
+
"step": 51830
|
| 165309 |
+
},
|
| 165310 |
+
{
|
| 165311 |
+
"epoch": 417.55,
|
| 165312 |
+
"learning_rate": 9.185641025641027e-06,
|
| 165313 |
+
"loss": 0.3924,
|
| 165314 |
+
"step": 51835
|
| 165315 |
+
},
|
| 165316 |
+
{
|
| 165317 |
+
"epoch": 417.59,
|
| 165318 |
+
"learning_rate": 9.185560897435898e-06,
|
| 165319 |
+
"loss": 0.7668,
|
| 165320 |
+
"step": 51840
|
| 165321 |
+
},
|
| 165322 |
+
{
|
| 165323 |
+
"epoch": 417.63,
|
| 165324 |
+
"learning_rate": 9.18548076923077e-06,
|
| 165325 |
+
"loss": 0.967,
|
| 165326 |
+
"step": 51845
|
| 165327 |
+
},
|
| 165328 |
+
{
|
| 165329 |
+
"epoch": 417.67,
|
| 165330 |
+
"learning_rate": 9.18540064102564e-06,
|
| 165331 |
+
"loss": 0.3295,
|
| 165332 |
+
"step": 51850
|
| 165333 |
+
},
|
| 165334 |
+
{
|
| 165335 |
+
"epoch": 417.71,
|
| 165336 |
+
"learning_rate": 9.185320512820514e-06,
|
| 165337 |
+
"loss": 0.3158,
|
| 165338 |
+
"step": 51855
|
| 165339 |
+
},
|
| 165340 |
+
{
|
| 165341 |
+
"epoch": 417.75,
|
| 165342 |
+
"learning_rate": 9.185240384615385e-06,
|
| 165343 |
+
"loss": 0.4142,
|
| 165344 |
+
"step": 51860
|
| 165345 |
+
},
|
| 165346 |
+
{
|
| 165347 |
+
"epoch": 417.79,
|
| 165348 |
+
"learning_rate": 9.185160256410257e-06,
|
| 165349 |
+
"loss": 0.9217,
|
| 165350 |
+
"step": 51865
|
| 165351 |
+
},
|
| 165352 |
+
{
|
| 165353 |
+
"epoch": 417.83,
|
| 165354 |
+
"learning_rate": 9.18508012820513e-06,
|
| 165355 |
+
"loss": 0.9746,
|
| 165356 |
+
"step": 51870
|
| 165357 |
+
},
|
| 165358 |
+
{
|
| 165359 |
+
"epoch": 417.87,
|
| 165360 |
+
"learning_rate": 9.185000000000001e-06,
|
| 165361 |
+
"loss": 0.2906,
|
| 165362 |
+
"step": 51875
|
| 165363 |
+
},
|
| 165364 |
+
{
|
| 165365 |
+
"epoch": 417.91,
|
| 165366 |
+
"learning_rate": 9.184919871794872e-06,
|
| 165367 |
+
"loss": 0.3185,
|
| 165368 |
+
"step": 51880
|
| 165369 |
+
},
|
| 165370 |
+
{
|
| 165371 |
+
"epoch": 417.95,
|
| 165372 |
+
"learning_rate": 9.184839743589744e-06,
|
| 165373 |
+
"loss": 0.389,
|
| 165374 |
+
"step": 51885
|
| 165375 |
+
},
|
| 165376 |
+
{
|
| 165377 |
+
"epoch": 417.99,
|
| 165378 |
+
"learning_rate": 9.184759615384617e-06,
|
| 165379 |
+
"loss": 0.9404,
|
| 165380 |
+
"step": 51890
|
| 165381 |
+
},
|
| 165382 |
+
{
|
| 165383 |
+
"epoch": 418.0,
|
| 165384 |
+
"eval_loss": 0.4447157382965088,
|
| 165385 |
+
"eval_runtime": 37.3454,
|
| 165386 |
+
"eval_samples_per_second": 22.466,
|
| 165387 |
+
"eval_steps_per_second": 0.723,
|
| 165388 |
+
"eval_wer": 0.2077396021699819,
|
| 165389 |
+
"step": 51891
|
| 165390 |
+
},
|
| 165391 |
+
{
|
| 165392 |
+
"epoch": 418.03,
|
| 165393 |
+
"learning_rate": 9.184679487179488e-06,
|
| 165394 |
+
"loss": 0.348,
|
| 165395 |
+
"step": 51895
|
| 165396 |
+
},
|
| 165397 |
+
{
|
| 165398 |
+
"epoch": 418.07,
|
| 165399 |
+
"learning_rate": 9.18459935897436e-06,
|
| 165400 |
+
"loss": 0.3307,
|
| 165401 |
+
"step": 51900
|
| 165402 |
+
},
|
| 165403 |
+
{
|
| 165404 |
+
"epoch": 418.11,
|
| 165405 |
+
"learning_rate": 9.18451923076923e-06,
|
| 165406 |
+
"loss": 0.345,
|
| 165407 |
+
"step": 51905
|
| 165408 |
+
},
|
| 165409 |
+
{
|
| 165410 |
+
"epoch": 418.15,
|
| 165411 |
+
"learning_rate": 9.184439102564104e-06,
|
| 165412 |
+
"loss": 0.4252,
|
| 165413 |
+
"step": 51910
|
| 165414 |
+
},
|
| 165415 |
+
{
|
| 165416 |
+
"epoch": 418.19,
|
| 165417 |
+
"learning_rate": 9.184358974358975e-06,
|
| 165418 |
+
"loss": 1.1198,
|
| 165419 |
+
"step": 51915
|
| 165420 |
+
},
|
| 165421 |
+
{
|
| 165422 |
+
"epoch": 418.23,
|
| 165423 |
+
"learning_rate": 9.184278846153847e-06,
|
| 165424 |
+
"loss": 0.9071,
|
| 165425 |
+
"step": 51920
|
| 165426 |
+
},
|
| 165427 |
+
{
|
| 165428 |
+
"epoch": 418.27,
|
| 165429 |
+
"learning_rate": 9.18419871794872e-06,
|
| 165430 |
+
"loss": 0.6036,
|
| 165431 |
+
"step": 51925
|
| 165432 |
+
},
|
| 165433 |
+
{
|
| 165434 |
+
"epoch": 418.31,
|
| 165435 |
+
"learning_rate": 9.184118589743591e-06,
|
| 165436 |
+
"loss": 0.3595,
|
| 165437 |
+
"step": 51930
|
| 165438 |
+
},
|
| 165439 |
+
{
|
| 165440 |
+
"epoch": 418.35,
|
| 165441 |
+
"learning_rate": 9.184038461538462e-06,
|
| 165442 |
+
"loss": 0.4328,
|
| 165443 |
+
"step": 51935
|
| 165444 |
+
},
|
| 165445 |
+
{
|
| 165446 |
+
"epoch": 418.39,
|
| 165447 |
+
"learning_rate": 9.183958333333334e-06,
|
| 165448 |
+
"loss": 1.0082,
|
| 165449 |
+
"step": 51940
|
| 165450 |
+
},
|
| 165451 |
+
{
|
| 165452 |
+
"epoch": 418.43,
|
| 165453 |
+
"learning_rate": 9.183878205128207e-06,
|
| 165454 |
+
"loss": 0.7708,
|
| 165455 |
+
"step": 51945
|
| 165456 |
+
},
|
| 165457 |
+
{
|
| 165458 |
+
"epoch": 418.47,
|
| 165459 |
+
"learning_rate": 9.183798076923076e-06,
|
| 165460 |
+
"loss": 0.3597,
|
| 165461 |
+
"step": 51950
|
| 165462 |
+
},
|
| 165463 |
+
{
|
| 165464 |
+
"epoch": 418.51,
|
| 165465 |
+
"learning_rate": 9.18371794871795e-06,
|
| 165466 |
+
"loss": 0.6259,
|
| 165467 |
+
"step": 51955
|
| 165468 |
+
},
|
| 165469 |
+
{
|
| 165470 |
+
"epoch": 418.55,
|
| 165471 |
+
"learning_rate": 9.183637820512821e-06,
|
| 165472 |
+
"loss": 0.4406,
|
| 165473 |
+
"step": 51960
|
| 165474 |
+
},
|
| 165475 |
+
{
|
| 165476 |
+
"epoch": 418.59,
|
| 165477 |
+
"learning_rate": 9.183557692307692e-06,
|
| 165478 |
+
"loss": 0.8934,
|
| 165479 |
+
"step": 51965
|
| 165480 |
+
},
|
| 165481 |
+
{
|
| 165482 |
+
"epoch": 418.63,
|
| 165483 |
+
"learning_rate": 9.183477564102565e-06,
|
| 165484 |
+
"loss": 0.6403,
|
| 165485 |
+
"step": 51970
|
| 165486 |
+
},
|
| 165487 |
+
{
|
| 165488 |
+
"epoch": 418.67,
|
| 165489 |
+
"learning_rate": 9.183397435897437e-06,
|
| 165490 |
+
"loss": 0.3313,
|
| 165491 |
+
"step": 51975
|
| 165492 |
+
},
|
| 165493 |
+
{
|
| 165494 |
+
"epoch": 418.71,
|
| 165495 |
+
"learning_rate": 9.183317307692308e-06,
|
| 165496 |
+
"loss": 0.346,
|
| 165497 |
+
"step": 51980
|
| 165498 |
+
},
|
| 165499 |
+
{
|
| 165500 |
+
"epoch": 418.76,
|
| 165501 |
+
"learning_rate": 9.18323717948718e-06,
|
| 165502 |
+
"loss": 0.3728,
|
| 165503 |
+
"step": 51985
|
| 165504 |
+
},
|
| 165505 |
+
{
|
| 165506 |
+
"epoch": 418.8,
|
| 165507 |
+
"learning_rate": 9.183157051282052e-06,
|
| 165508 |
+
"loss": 0.9497,
|
| 165509 |
+
"step": 51990
|
| 165510 |
+
},
|
| 165511 |
+
{
|
| 165512 |
+
"epoch": 418.84,
|
| 165513 |
+
"learning_rate": 9.183076923076924e-06,
|
| 165514 |
+
"loss": 0.7863,
|
| 165515 |
+
"step": 51995
|
| 165516 |
+
},
|
| 165517 |
+
{
|
| 165518 |
+
"epoch": 418.88,
|
| 165519 |
+
"learning_rate": 9.182996794871795e-06,
|
| 165520 |
+
"loss": 0.3685,
|
| 165521 |
+
"step": 52000
|
| 165522 |
+
},
|
| 165523 |
+
{
|
| 165524 |
+
"epoch": 418.92,
|
| 165525 |
+
"learning_rate": 9.182916666666666e-06,
|
| 165526 |
+
"loss": 0.3457,
|
| 165527 |
+
"step": 52005
|
| 165528 |
+
},
|
| 165529 |
+
{
|
| 165530 |
+
"epoch": 418.96,
|
| 165531 |
+
"learning_rate": 9.18283653846154e-06,
|
| 165532 |
+
"loss": 0.5205,
|
| 165533 |
+
"step": 52010
|
| 165534 |
+
},
|
| 165535 |
+
{
|
| 165536 |
+
"epoch": 419.0,
|
| 165537 |
+
"learning_rate": 9.182756410256411e-06,
|
| 165538 |
+
"loss": 1.2548,
|
| 165539 |
+
"step": 52015
|
| 165540 |
+
},
|
| 165541 |
+
{
|
| 165542 |
+
"epoch": 419.0,
|
| 165543 |
+
"eval_loss": 0.38904717564582825,
|
| 165544 |
+
"eval_runtime": 40.5611,
|
| 165545 |
+
"eval_samples_per_second": 20.685,
|
| 165546 |
+
"eval_steps_per_second": 0.666,
|
| 165547 |
+
"eval_wer": 0.20132061628760087,
|
| 165548 |
+
"step": 52015
|
| 165549 |
}
|
| 165550 |
],
|
| 165551 |
"max_steps": 620000,
|
| 165552 |
"num_train_epochs": 5000,
|
| 165553 |
+
"total_flos": 1.463606633594667e+20,
|
| 165554 |
"trial_name": null,
|
| 165555 |
"trial_params": null
|
| 165556 |
}
|
model-bin/finetune/base/{checkpoint-51395 → checkpoint-52015}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629749097.2738104/events.out.tfevents.1629749097.74272264b15c.932.213
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:05635f0a577f4f327a56bf14328091448cee7f962e90bbb4ae74254e4b0ceb39
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629749718.7735815/events.out.tfevents.1629749718.74272264b15c.932.215
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cdc3759d78a74a1af2330d53d13c368e32b3277b47112a91686bf97d51e54842
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629750364.5758934/events.out.tfevents.1629750364.74272264b15c.932.217
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9868e2fcb7d38a903e85089889db4b98cceff2bdeecea54cfd2e799d624f84f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629751114.6670787/events.out.tfevents.1629751114.74272264b15c.932.219
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0fa165fe62d903b3b0b11f5b6906763634a0d2d1b330b6ad9c1b5376190be81c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629751765.2580316/events.out.tfevents.1629751765.74272264b15c.932.221
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b2c79c19aaf1b8ea72f9b5d1a2889fb5326ab2d45c67e0b6b0f6157df4718202
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629749097.74272264b15c.932.212
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fc2dcbb7311c486ff1d8fa8a6160ff82b8096386b97c023c4edbe672954106b3
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629749718.74272264b15c.932.214
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68c18f28264b2796a237845fa0b99b910c6af91ccf4e90dee6c0d9337c9001ac
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629750364.74272264b15c.932.216
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:59f157d9e385766f6bbe85a0045c9910b1d60f3ddd0fdeb8bf9352455183ca3e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629751114.74272264b15c.932.218
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14b219a635ba0d2b02e27b94fe8d62a67c47b7701b0d10327a322228f5885a0b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629751765.74272264b15c.932.220
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ea09d15a61fd36d54177898ba59369fcf0f4c85f94b2aac9e1ff33b3df0b889f
|
| 3 |
+
size 8622
|