"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630179532.3843727/events.out.tfevents.1630179532.86bb0ddabf9b.4092.281 +3 -0
- model-bin/finetune/base/log/1630179923.6632285/events.out.tfevents.1630179923.86bb0ddabf9b.4092.283 +3 -0
- model-bin/finetune/base/log/1630180319.8798604/events.out.tfevents.1630180319.86bb0ddabf9b.4092.285 +3 -0
- model-bin/finetune/base/log/1630180707.9549763/events.out.tfevents.1630180707.86bb0ddabf9b.4092.287 +3 -0
- model-bin/finetune/base/log/1630181095.4206262/events.out.tfevents.1630181095.86bb0ddabf9b.4092.289 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630179532.86bb0ddabf9b.4092.280 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630179923.86bb0ddabf9b.4092.282 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630180319.86bb0ddabf9b.4092.284 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630180707.86bb0ddabf9b.4092.286 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630181095.86bb0ddabf9b.4092.288 +3 -0
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/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:a3f6fe274f491d2f102f8f164964b4422ae4f7d02a0b0d62c53fce16fd85196e
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/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 sha256:432426b4107e7d521ac9e9114b6a8d2b1c84b3949d741b39c0ff852a76d398db
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/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 sha256:3d67134db135497c84a3a978dfc47cb114c83fb298257788d379d9d15f5c9200
|
| 3 |
+
size 14439
|
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/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 sha256:3908491cb65cf55e8133021f081a21eb083f24a621260a82ac94ef5a912a15f0
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/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:a6f9648e0ebdeaf33cc36f06a036f7cd1a47eaf8d8b8086212b1b2803614924d
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -271632,11 +271632,800 @@
|
|
| 271632 |
"eval_steps_per_second": 0.741,
|
| 271633 |
"eval_wer": 0.1816446402349486,
|
| 271634 |
"step": 135397
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 271635 |
}
|
| 271636 |
],
|
| 271637 |
"max_steps": 620000,
|
| 271638 |
"num_train_epochs": 5000,
|
| 271639 |
-
"total_flos": 3.
|
| 271640 |
"trial_name": null,
|
| 271641 |
"trial_params": null
|
| 271642 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1689111747851003,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
+
"epoch": 1096.0,
|
| 5 |
+
"global_step": 136018,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 271632 |
"eval_steps_per_second": 0.741,
|
| 271633 |
"eval_wer": 0.1816446402349486,
|
| 271634 |
"step": 135397
|
| 271635 |
+
},
|
| 271636 |
+
{
|
| 271637 |
+
"epoch": 1083.02,
|
| 271638 |
+
"learning_rate": 7.847003205128204e-06,
|
| 271639 |
+
"loss": 0.3065,
|
| 271640 |
+
"step": 135400
|
| 271641 |
+
},
|
| 271642 |
+
{
|
| 271643 |
+
"epoch": 1083.06,
|
| 271644 |
+
"learning_rate": 7.846923076923078e-06,
|
| 271645 |
+
"loss": 0.294,
|
| 271646 |
+
"step": 135405
|
| 271647 |
+
},
|
| 271648 |
+
{
|
| 271649 |
+
"epoch": 1083.1,
|
| 271650 |
+
"learning_rate": 7.84684294871795e-06,
|
| 271651 |
+
"loss": 0.3027,
|
| 271652 |
+
"step": 135410
|
| 271653 |
+
},
|
| 271654 |
+
{
|
| 271655 |
+
"epoch": 1083.14,
|
| 271656 |
+
"learning_rate": 7.84676282051282e-06,
|
| 271657 |
+
"loss": 0.3488,
|
| 271658 |
+
"step": 135415
|
| 271659 |
+
},
|
| 271660 |
+
{
|
| 271661 |
+
"epoch": 1083.18,
|
| 271662 |
+
"learning_rate": 7.846682692307693e-06,
|
| 271663 |
+
"loss": 0.7898,
|
| 271664 |
+
"step": 135420
|
| 271665 |
+
},
|
| 271666 |
+
{
|
| 271667 |
+
"epoch": 1083.22,
|
| 271668 |
+
"learning_rate": 7.846602564102565e-06,
|
| 271669 |
+
"loss": 0.8657,
|
| 271670 |
+
"step": 135425
|
| 271671 |
+
},
|
| 271672 |
+
{
|
| 271673 |
+
"epoch": 1083.26,
|
| 271674 |
+
"learning_rate": 7.846522435897436e-06,
|
| 271675 |
+
"loss": 0.2766,
|
| 271676 |
+
"step": 135430
|
| 271677 |
+
},
|
| 271678 |
+
{
|
| 271679 |
+
"epoch": 1083.3,
|
| 271680 |
+
"learning_rate": 7.846442307692307e-06,
|
| 271681 |
+
"loss": 0.2873,
|
| 271682 |
+
"step": 135435
|
| 271683 |
+
},
|
| 271684 |
+
{
|
| 271685 |
+
"epoch": 1083.34,
|
| 271686 |
+
"learning_rate": 7.84636217948718e-06,
|
| 271687 |
+
"loss": 0.3926,
|
| 271688 |
+
"step": 135440
|
| 271689 |
+
},
|
| 271690 |
+
{
|
| 271691 |
+
"epoch": 1083.38,
|
| 271692 |
+
"learning_rate": 7.846282051282052e-06,
|
| 271693 |
+
"loss": 0.6891,
|
| 271694 |
+
"step": 135445
|
| 271695 |
+
},
|
| 271696 |
+
{
|
| 271697 |
+
"epoch": 1083.42,
|
| 271698 |
+
"learning_rate": 7.846201923076923e-06,
|
| 271699 |
+
"loss": 0.754,
|
| 271700 |
+
"step": 135450
|
| 271701 |
+
},
|
| 271702 |
+
{
|
| 271703 |
+
"epoch": 1083.46,
|
| 271704 |
+
"learning_rate": 7.846121794871796e-06,
|
| 271705 |
+
"loss": 0.2743,
|
| 271706 |
+
"step": 135455
|
| 271707 |
+
},
|
| 271708 |
+
{
|
| 271709 |
+
"epoch": 1083.5,
|
| 271710 |
+
"learning_rate": 7.846041666666668e-06,
|
| 271711 |
+
"loss": 0.2947,
|
| 271712 |
+
"step": 135460
|
| 271713 |
+
},
|
| 271714 |
+
{
|
| 271715 |
+
"epoch": 1083.54,
|
| 271716 |
+
"learning_rate": 7.845961538461539e-06,
|
| 271717 |
+
"loss": 0.3073,
|
| 271718 |
+
"step": 135465
|
| 271719 |
+
},
|
| 271720 |
+
{
|
| 271721 |
+
"epoch": 1083.58,
|
| 271722 |
+
"learning_rate": 7.84588141025641e-06,
|
| 271723 |
+
"loss": 0.6519,
|
| 271724 |
+
"step": 135470
|
| 271725 |
+
},
|
| 271726 |
+
{
|
| 271727 |
+
"epoch": 1083.62,
|
| 271728 |
+
"learning_rate": 7.845801282051283e-06,
|
| 271729 |
+
"loss": 0.851,
|
| 271730 |
+
"step": 135475
|
| 271731 |
+
},
|
| 271732 |
+
{
|
| 271733 |
+
"epoch": 1083.66,
|
| 271734 |
+
"learning_rate": 7.845721153846155e-06,
|
| 271735 |
+
"loss": 0.29,
|
| 271736 |
+
"step": 135480
|
| 271737 |
+
},
|
| 271738 |
+
{
|
| 271739 |
+
"epoch": 1083.7,
|
| 271740 |
+
"learning_rate": 7.845641025641026e-06,
|
| 271741 |
+
"loss": 0.2615,
|
| 271742 |
+
"step": 135485
|
| 271743 |
+
},
|
| 271744 |
+
{
|
| 271745 |
+
"epoch": 1083.74,
|
| 271746 |
+
"learning_rate": 7.845560897435897e-06,
|
| 271747 |
+
"loss": 0.3377,
|
| 271748 |
+
"step": 135490
|
| 271749 |
+
},
|
| 271750 |
+
{
|
| 271751 |
+
"epoch": 1083.78,
|
| 271752 |
+
"learning_rate": 7.84548076923077e-06,
|
| 271753 |
+
"loss": 0.7867,
|
| 271754 |
+
"step": 135495
|
| 271755 |
+
},
|
| 271756 |
+
{
|
| 271757 |
+
"epoch": 1083.82,
|
| 271758 |
+
"learning_rate": 7.845400641025642e-06,
|
| 271759 |
+
"loss": 0.9239,
|
| 271760 |
+
"step": 135500
|
| 271761 |
+
},
|
| 271762 |
+
{
|
| 271763 |
+
"epoch": 1083.86,
|
| 271764 |
+
"learning_rate": 7.845320512820513e-06,
|
| 271765 |
+
"loss": 0.2768,
|
| 271766 |
+
"step": 135505
|
| 271767 |
+
},
|
| 271768 |
+
{
|
| 271769 |
+
"epoch": 1083.9,
|
| 271770 |
+
"learning_rate": 7.845240384615386e-06,
|
| 271771 |
+
"loss": 0.3035,
|
| 271772 |
+
"step": 135510
|
| 271773 |
+
},
|
| 271774 |
+
{
|
| 271775 |
+
"epoch": 1083.94,
|
| 271776 |
+
"learning_rate": 7.845160256410258e-06,
|
| 271777 |
+
"loss": 0.3213,
|
| 271778 |
+
"step": 135515
|
| 271779 |
+
},
|
| 271780 |
+
{
|
| 271781 |
+
"epoch": 1083.98,
|
| 271782 |
+
"learning_rate": 7.845080128205129e-06,
|
| 271783 |
+
"loss": 0.8615,
|
| 271784 |
+
"step": 135520
|
| 271785 |
+
},
|
| 271786 |
+
{
|
| 271787 |
+
"epoch": 1084.0,
|
| 271788 |
+
"eval_loss": 0.3909757137298584,
|
| 271789 |
+
"eval_runtime": 37.5538,
|
| 271790 |
+
"eval_samples_per_second": 22.448,
|
| 271791 |
+
"eval_steps_per_second": 0.719,
|
| 271792 |
+
"eval_wer": 0.17052023121387283,
|
| 271793 |
+
"step": 135522
|
| 271794 |
+
},
|
| 271795 |
+
{
|
| 271796 |
+
"epoch": 1092.02,
|
| 271797 |
+
"learning_rate": 7.845e-06,
|
| 271798 |
+
"loss": 0.28,
|
| 271799 |
+
"step": 135525
|
| 271800 |
+
},
|
| 271801 |
+
{
|
| 271802 |
+
"epoch": 1092.06,
|
| 271803 |
+
"learning_rate": 7.844919871794873e-06,
|
| 271804 |
+
"loss": 0.2947,
|
| 271805 |
+
"step": 135530
|
| 271806 |
+
},
|
| 271807 |
+
{
|
| 271808 |
+
"epoch": 1092.1,
|
| 271809 |
+
"learning_rate": 7.844839743589745e-06,
|
| 271810 |
+
"loss": 0.3206,
|
| 271811 |
+
"step": 135535
|
| 271812 |
+
},
|
| 271813 |
+
{
|
| 271814 |
+
"epoch": 1092.14,
|
| 271815 |
+
"learning_rate": 7.844759615384616e-06,
|
| 271816 |
+
"loss": 0.3456,
|
| 271817 |
+
"step": 135540
|
| 271818 |
+
},
|
| 271819 |
+
{
|
| 271820 |
+
"epoch": 1092.18,
|
| 271821 |
+
"learning_rate": 7.844679487179487e-06,
|
| 271822 |
+
"loss": 0.795,
|
| 271823 |
+
"step": 135545
|
| 271824 |
+
},
|
| 271825 |
+
{
|
| 271826 |
+
"epoch": 1092.22,
|
| 271827 |
+
"learning_rate": 7.84459935897436e-06,
|
| 271828 |
+
"loss": 0.8684,
|
| 271829 |
+
"step": 135550
|
| 271830 |
+
},
|
| 271831 |
+
{
|
| 271832 |
+
"epoch": 1092.27,
|
| 271833 |
+
"learning_rate": 7.844519230769232e-06,
|
| 271834 |
+
"loss": 0.2677,
|
| 271835 |
+
"step": 135555
|
| 271836 |
+
},
|
| 271837 |
+
{
|
| 271838 |
+
"epoch": 1092.31,
|
| 271839 |
+
"learning_rate": 7.844439102564103e-06,
|
| 271840 |
+
"loss": 0.2882,
|
| 271841 |
+
"step": 135560
|
| 271842 |
+
},
|
| 271843 |
+
{
|
| 271844 |
+
"epoch": 1092.35,
|
| 271845 |
+
"learning_rate": 7.844358974358976e-06,
|
| 271846 |
+
"loss": 0.3792,
|
| 271847 |
+
"step": 135565
|
| 271848 |
+
},
|
| 271849 |
+
{
|
| 271850 |
+
"epoch": 1092.39,
|
| 271851 |
+
"learning_rate": 7.844278846153846e-06,
|
| 271852 |
+
"loss": 0.7045,
|
| 271853 |
+
"step": 135570
|
| 271854 |
+
},
|
| 271855 |
+
{
|
| 271856 |
+
"epoch": 1092.43,
|
| 271857 |
+
"learning_rate": 7.844198717948719e-06,
|
| 271858 |
+
"loss": 0.8184,
|
| 271859 |
+
"step": 135575
|
| 271860 |
+
},
|
| 271861 |
+
{
|
| 271862 |
+
"epoch": 1092.47,
|
| 271863 |
+
"learning_rate": 7.84411858974359e-06,
|
| 271864 |
+
"loss": 0.2846,
|
| 271865 |
+
"step": 135580
|
| 271866 |
+
},
|
| 271867 |
+
{
|
| 271868 |
+
"epoch": 1092.51,
|
| 271869 |
+
"learning_rate": 7.844038461538462e-06,
|
| 271870 |
+
"loss": 0.3111,
|
| 271871 |
+
"step": 135585
|
| 271872 |
+
},
|
| 271873 |
+
{
|
| 271874 |
+
"epoch": 1092.55,
|
| 271875 |
+
"learning_rate": 7.843958333333333e-06,
|
| 271876 |
+
"loss": 0.3708,
|
| 271877 |
+
"step": 135590
|
| 271878 |
+
},
|
| 271879 |
+
{
|
| 271880 |
+
"epoch": 1092.59,
|
| 271881 |
+
"learning_rate": 7.843878205128206e-06,
|
| 271882 |
+
"loss": 0.8047,
|
| 271883 |
+
"step": 135595
|
| 271884 |
+
},
|
| 271885 |
+
{
|
| 271886 |
+
"epoch": 1092.63,
|
| 271887 |
+
"learning_rate": 7.843798076923077e-06,
|
| 271888 |
+
"loss": 0.8642,
|
| 271889 |
+
"step": 135600
|
| 271890 |
+
},
|
| 271891 |
+
{
|
| 271892 |
+
"epoch": 1092.67,
|
| 271893 |
+
"learning_rate": 7.843717948717949e-06,
|
| 271894 |
+
"loss": 0.2369,
|
| 271895 |
+
"step": 135605
|
| 271896 |
+
},
|
| 271897 |
+
{
|
| 271898 |
+
"epoch": 1092.71,
|
| 271899 |
+
"learning_rate": 7.843637820512822e-06,
|
| 271900 |
+
"loss": 0.2866,
|
| 271901 |
+
"step": 135610
|
| 271902 |
+
},
|
| 271903 |
+
{
|
| 271904 |
+
"epoch": 1092.75,
|
| 271905 |
+
"learning_rate": 7.843557692307693e-06,
|
| 271906 |
+
"loss": 0.3743,
|
| 271907 |
+
"step": 135615
|
| 271908 |
+
},
|
| 271909 |
+
{
|
| 271910 |
+
"epoch": 1092.79,
|
| 271911 |
+
"learning_rate": 7.843477564102565e-06,
|
| 271912 |
+
"loss": 0.6897,
|
| 271913 |
+
"step": 135620
|
| 271914 |
+
},
|
| 271915 |
+
{
|
| 271916 |
+
"epoch": 1092.83,
|
| 271917 |
+
"learning_rate": 7.843397435897436e-06,
|
| 271918 |
+
"loss": 0.8442,
|
| 271919 |
+
"step": 135625
|
| 271920 |
+
},
|
| 271921 |
+
{
|
| 271922 |
+
"epoch": 1092.87,
|
| 271923 |
+
"learning_rate": 7.843317307692309e-06,
|
| 271924 |
+
"loss": 0.2325,
|
| 271925 |
+
"step": 135630
|
| 271926 |
+
},
|
| 271927 |
+
{
|
| 271928 |
+
"epoch": 1092.91,
|
| 271929 |
+
"learning_rate": 7.84323717948718e-06,
|
| 271930 |
+
"loss": 0.3242,
|
| 271931 |
+
"step": 135635
|
| 271932 |
+
},
|
| 271933 |
+
{
|
| 271934 |
+
"epoch": 1092.95,
|
| 271935 |
+
"learning_rate": 7.843157051282052e-06,
|
| 271936 |
+
"loss": 0.3931,
|
| 271937 |
+
"step": 135640
|
| 271938 |
+
},
|
| 271939 |
+
{
|
| 271940 |
+
"epoch": 1092.99,
|
| 271941 |
+
"learning_rate": 7.843076923076923e-06,
|
| 271942 |
+
"loss": 0.8294,
|
| 271943 |
+
"step": 135645
|
| 271944 |
+
},
|
| 271945 |
+
{
|
| 271946 |
+
"epoch": 1093.0,
|
| 271947 |
+
"eval_loss": 0.36988407373428345,
|
| 271948 |
+
"eval_runtime": 35.9696,
|
| 271949 |
+
"eval_samples_per_second": 23.436,
|
| 271950 |
+
"eval_steps_per_second": 0.751,
|
| 271951 |
+
"eval_wer": 0.17609686822632484,
|
| 271952 |
+
"step": 135646
|
| 271953 |
+
},
|
| 271954 |
+
{
|
| 271955 |
+
"epoch": 1093.03,
|
| 271956 |
+
"learning_rate": 7.842996794871796e-06,
|
| 271957 |
+
"loss": 0.2782,
|
| 271958 |
+
"step": 135650
|
| 271959 |
+
},
|
| 271960 |
+
{
|
| 271961 |
+
"epoch": 1093.07,
|
| 271962 |
+
"learning_rate": 7.842916666666667e-06,
|
| 271963 |
+
"loss": 0.2226,
|
| 271964 |
+
"step": 135655
|
| 271965 |
+
},
|
| 271966 |
+
{
|
| 271967 |
+
"epoch": 1093.11,
|
| 271968 |
+
"learning_rate": 7.842836538461539e-06,
|
| 271969 |
+
"loss": 0.2889,
|
| 271970 |
+
"step": 135660
|
| 271971 |
+
},
|
| 271972 |
+
{
|
| 271973 |
+
"epoch": 1093.15,
|
| 271974 |
+
"learning_rate": 7.842756410256412e-06,
|
| 271975 |
+
"loss": 0.4144,
|
| 271976 |
+
"step": 135665
|
| 271977 |
+
},
|
| 271978 |
+
{
|
| 271979 |
+
"epoch": 1093.19,
|
| 271980 |
+
"learning_rate": 7.842676282051283e-06,
|
| 271981 |
+
"loss": 0.9336,
|
| 271982 |
+
"step": 135670
|
| 271983 |
+
},
|
| 271984 |
+
{
|
| 271985 |
+
"epoch": 1093.23,
|
| 271986 |
+
"learning_rate": 7.842596153846155e-06,
|
| 271987 |
+
"loss": 0.6257,
|
| 271988 |
+
"step": 135675
|
| 271989 |
+
},
|
| 271990 |
+
{
|
| 271991 |
+
"epoch": 1093.27,
|
| 271992 |
+
"learning_rate": 7.842516025641026e-06,
|
| 271993 |
+
"loss": 0.2613,
|
| 271994 |
+
"step": 135680
|
| 271995 |
+
},
|
| 271996 |
+
{
|
| 271997 |
+
"epoch": 1093.31,
|
| 271998 |
+
"learning_rate": 7.842435897435899e-06,
|
| 271999 |
+
"loss": 0.2648,
|
| 272000 |
+
"step": 135685
|
| 272001 |
+
},
|
| 272002 |
+
{
|
| 272003 |
+
"epoch": 1093.35,
|
| 272004 |
+
"learning_rate": 7.842355769230769e-06,
|
| 272005 |
+
"loss": 0.3987,
|
| 272006 |
+
"step": 135690
|
| 272007 |
+
},
|
| 272008 |
+
{
|
| 272009 |
+
"epoch": 1093.4,
|
| 272010 |
+
"learning_rate": 7.842275641025642e-06,
|
| 272011 |
+
"loss": 0.8109,
|
| 272012 |
+
"step": 135695
|
| 272013 |
+
},
|
| 272014 |
+
{
|
| 272015 |
+
"epoch": 1093.44,
|
| 272016 |
+
"learning_rate": 7.842195512820513e-06,
|
| 272017 |
+
"loss": 0.6588,
|
| 272018 |
+
"step": 135700
|
| 272019 |
+
},
|
| 272020 |
+
{
|
| 272021 |
+
"epoch": 1093.48,
|
| 272022 |
+
"learning_rate": 7.842115384615384e-06,
|
| 272023 |
+
"loss": 0.3614,
|
| 272024 |
+
"step": 135705
|
| 272025 |
+
},
|
| 272026 |
+
{
|
| 272027 |
+
"epoch": 1093.52,
|
| 272028 |
+
"learning_rate": 7.842035256410257e-06,
|
| 272029 |
+
"loss": 0.2922,
|
| 272030 |
+
"step": 135710
|
| 272031 |
+
},
|
| 272032 |
+
{
|
| 272033 |
+
"epoch": 1093.56,
|
| 272034 |
+
"learning_rate": 7.841955128205129e-06,
|
| 272035 |
+
"loss": 0.3946,
|
| 272036 |
+
"step": 135715
|
| 272037 |
+
},
|
| 272038 |
+
{
|
| 272039 |
+
"epoch": 1093.6,
|
| 272040 |
+
"learning_rate": 7.841875e-06,
|
| 272041 |
+
"loss": 0.8577,
|
| 272042 |
+
"step": 135720
|
| 272043 |
+
},
|
| 272044 |
+
{
|
| 272045 |
+
"epoch": 1093.64,
|
| 272046 |
+
"learning_rate": 7.841794871794872e-06,
|
| 272047 |
+
"loss": 0.5763,
|
| 272048 |
+
"step": 135725
|
| 272049 |
+
},
|
| 272050 |
+
{
|
| 272051 |
+
"epoch": 1093.68,
|
| 272052 |
+
"learning_rate": 7.841714743589745e-06,
|
| 272053 |
+
"loss": 0.2408,
|
| 272054 |
+
"step": 135730
|
| 272055 |
+
},
|
| 272056 |
+
{
|
| 272057 |
+
"epoch": 1093.72,
|
| 272058 |
+
"learning_rate": 7.841634615384616e-06,
|
| 272059 |
+
"loss": 0.2981,
|
| 272060 |
+
"step": 135735
|
| 272061 |
+
},
|
| 272062 |
+
{
|
| 272063 |
+
"epoch": 1093.76,
|
| 272064 |
+
"learning_rate": 7.841554487179487e-06,
|
| 272065 |
+
"loss": 0.3766,
|
| 272066 |
+
"step": 135740
|
| 272067 |
+
},
|
| 272068 |
+
{
|
| 272069 |
+
"epoch": 1093.8,
|
| 272070 |
+
"learning_rate": 7.841474358974359e-06,
|
| 272071 |
+
"loss": 1.0882,
|
| 272072 |
+
"step": 135745
|
| 272073 |
+
},
|
| 272074 |
+
{
|
| 272075 |
+
"epoch": 1093.84,
|
| 272076 |
+
"learning_rate": 7.841394230769232e-06,
|
| 272077 |
+
"loss": 0.6584,
|
| 272078 |
+
"step": 135750
|
| 272079 |
+
},
|
| 272080 |
+
{
|
| 272081 |
+
"epoch": 1093.88,
|
| 272082 |
+
"learning_rate": 7.841314102564103e-06,
|
| 272083 |
+
"loss": 0.3232,
|
| 272084 |
+
"step": 135755
|
| 272085 |
+
},
|
| 272086 |
+
{
|
| 272087 |
+
"epoch": 1093.92,
|
| 272088 |
+
"learning_rate": 7.841233974358974e-06,
|
| 272089 |
+
"loss": 0.2835,
|
| 272090 |
+
"step": 135760
|
| 272091 |
+
},
|
| 272092 |
+
{
|
| 272093 |
+
"epoch": 1093.96,
|
| 272094 |
+
"learning_rate": 7.841153846153848e-06,
|
| 272095 |
+
"loss": 0.4187,
|
| 272096 |
+
"step": 135765
|
| 272097 |
+
},
|
| 272098 |
+
{
|
| 272099 |
+
"epoch": 1094.0,
|
| 272100 |
+
"learning_rate": 7.841073717948719e-06,
|
| 272101 |
+
"loss": 1.1544,
|
| 272102 |
+
"step": 135770
|
| 272103 |
+
},
|
| 272104 |
+
{
|
| 272105 |
+
"epoch": 1094.0,
|
| 272106 |
+
"eval_loss": 0.5360937714576721,
|
| 272107 |
+
"eval_runtime": 37.4324,
|
| 272108 |
+
"eval_samples_per_second": 22.521,
|
| 272109 |
+
"eval_steps_per_second": 0.721,
|
| 272110 |
+
"eval_wer": 0.17736414206247325,
|
| 272111 |
+
"step": 135770
|
| 272112 |
+
},
|
| 272113 |
+
{
|
| 272114 |
+
"epoch": 1094.04,
|
| 272115 |
+
"learning_rate": 7.84099358974359e-06,
|
| 272116 |
+
"loss": 0.284,
|
| 272117 |
+
"step": 135775
|
| 272118 |
+
},
|
| 272119 |
+
{
|
| 272120 |
+
"epoch": 1094.08,
|
| 272121 |
+
"learning_rate": 7.840913461538462e-06,
|
| 272122 |
+
"loss": 0.3002,
|
| 272123 |
+
"step": 135780
|
| 272124 |
+
},
|
| 272125 |
+
{
|
| 272126 |
+
"epoch": 1094.12,
|
| 272127 |
+
"learning_rate": 7.840833333333335e-06,
|
| 272128 |
+
"loss": 0.3255,
|
| 272129 |
+
"step": 135785
|
| 272130 |
+
},
|
| 272131 |
+
{
|
| 272132 |
+
"epoch": 1094.16,
|
| 272133 |
+
"learning_rate": 7.840753205128206e-06,
|
| 272134 |
+
"loss": 0.4187,
|
| 272135 |
+
"step": 135790
|
| 272136 |
+
},
|
| 272137 |
+
{
|
| 272138 |
+
"epoch": 1094.2,
|
| 272139 |
+
"learning_rate": 7.840673076923077e-06,
|
| 272140 |
+
"loss": 1.1222,
|
| 272141 |
+
"step": 135795
|
| 272142 |
+
},
|
| 272143 |
+
{
|
| 272144 |
+
"epoch": 1094.24,
|
| 272145 |
+
"learning_rate": 7.840592948717949e-06,
|
| 272146 |
+
"loss": 0.3768,
|
| 272147 |
+
"step": 135800
|
| 272148 |
+
},
|
| 272149 |
+
{
|
| 272150 |
+
"epoch": 1094.28,
|
| 272151 |
+
"learning_rate": 7.840512820512822e-06,
|
| 272152 |
+
"loss": 0.274,
|
| 272153 |
+
"step": 135805
|
| 272154 |
+
},
|
| 272155 |
+
{
|
| 272156 |
+
"epoch": 1094.32,
|
| 272157 |
+
"learning_rate": 7.840432692307693e-06,
|
| 272158 |
+
"loss": 0.2465,
|
| 272159 |
+
"step": 135810
|
| 272160 |
+
},
|
| 272161 |
+
{
|
| 272162 |
+
"epoch": 1094.36,
|
| 272163 |
+
"learning_rate": 7.840352564102564e-06,
|
| 272164 |
+
"loss": 0.3905,
|
| 272165 |
+
"step": 135815
|
| 272166 |
+
},
|
| 272167 |
+
{
|
| 272168 |
+
"epoch": 1094.4,
|
| 272169 |
+
"learning_rate": 7.840272435897438e-06,
|
| 272170 |
+
"loss": 1.1815,
|
| 272171 |
+
"step": 135820
|
| 272172 |
+
},
|
| 272173 |
+
{
|
| 272174 |
+
"epoch": 1094.44,
|
| 272175 |
+
"learning_rate": 7.840192307692309e-06,
|
| 272176 |
+
"loss": 0.2768,
|
| 272177 |
+
"step": 135825
|
| 272178 |
+
},
|
| 272179 |
+
{
|
| 272180 |
+
"epoch": 1094.48,
|
| 272181 |
+
"learning_rate": 7.84011217948718e-06,
|
| 272182 |
+
"loss": 0.2598,
|
| 272183 |
+
"step": 135830
|
| 272184 |
+
},
|
| 272185 |
+
{
|
| 272186 |
+
"epoch": 1094.52,
|
| 272187 |
+
"learning_rate": 7.840032051282052e-06,
|
| 272188 |
+
"loss": 0.2602,
|
| 272189 |
+
"step": 135835
|
| 272190 |
+
},
|
| 272191 |
+
{
|
| 272192 |
+
"epoch": 1094.56,
|
| 272193 |
+
"learning_rate": 7.839951923076925e-06,
|
| 272194 |
+
"loss": 0.4186,
|
| 272195 |
+
"step": 135840
|
| 272196 |
+
},
|
| 272197 |
+
{
|
| 272198 |
+
"epoch": 1094.6,
|
| 272199 |
+
"learning_rate": 7.839871794871794e-06,
|
| 272200 |
+
"loss": 1.0874,
|
| 272201 |
+
"step": 135845
|
| 272202 |
+
},
|
| 272203 |
+
{
|
| 272204 |
+
"epoch": 1094.64,
|
| 272205 |
+
"learning_rate": 7.839791666666667e-06,
|
| 272206 |
+
"loss": 0.2706,
|
| 272207 |
+
"step": 135850
|
| 272208 |
+
},
|
| 272209 |
+
{
|
| 272210 |
+
"epoch": 1094.68,
|
| 272211 |
+
"learning_rate": 7.83971153846154e-06,
|
| 272212 |
+
"loss": 0.2828,
|
| 272213 |
+
"step": 135855
|
| 272214 |
+
},
|
| 272215 |
+
{
|
| 272216 |
+
"epoch": 1094.72,
|
| 272217 |
+
"learning_rate": 7.83963141025641e-06,
|
| 272218 |
+
"loss": 0.3435,
|
| 272219 |
+
"step": 135860
|
| 272220 |
+
},
|
| 272221 |
+
{
|
| 272222 |
+
"epoch": 1094.76,
|
| 272223 |
+
"learning_rate": 7.839551282051283e-06,
|
| 272224 |
+
"loss": 0.365,
|
| 272225 |
+
"step": 135865
|
| 272226 |
+
},
|
| 272227 |
+
{
|
| 272228 |
+
"epoch": 1094.8,
|
| 272229 |
+
"learning_rate": 7.839471153846155e-06,
|
| 272230 |
+
"loss": 1.0529,
|
| 272231 |
+
"step": 135870
|
| 272232 |
+
},
|
| 272233 |
+
{
|
| 272234 |
+
"epoch": 1094.84,
|
| 272235 |
+
"learning_rate": 7.839391025641026e-06,
|
| 272236 |
+
"loss": 0.3155,
|
| 272237 |
+
"step": 135875
|
| 272238 |
+
},
|
| 272239 |
+
{
|
| 272240 |
+
"epoch": 1094.88,
|
| 272241 |
+
"learning_rate": 7.839310897435897e-06,
|
| 272242 |
+
"loss": 0.2624,
|
| 272243 |
+
"step": 135880
|
| 272244 |
+
},
|
| 272245 |
+
{
|
| 272246 |
+
"epoch": 1094.92,
|
| 272247 |
+
"learning_rate": 7.83923076923077e-06,
|
| 272248 |
+
"loss": 0.3118,
|
| 272249 |
+
"step": 135885
|
| 272250 |
+
},
|
| 272251 |
+
{
|
| 272252 |
+
"epoch": 1094.96,
|
| 272253 |
+
"learning_rate": 7.839150641025642e-06,
|
| 272254 |
+
"loss": 0.4909,
|
| 272255 |
+
"step": 135890
|
| 272256 |
+
},
|
| 272257 |
+
{
|
| 272258 |
+
"epoch": 1095.0,
|
| 272259 |
+
"eval_loss": 0.3657137453556061,
|
| 272260 |
+
"eval_runtime": 35.8538,
|
| 272261 |
+
"eval_samples_per_second": 23.512,
|
| 272262 |
+
"eval_steps_per_second": 0.753,
|
| 272263 |
+
"eval_wer": 0.1897031191281473,
|
| 272264 |
+
"step": 135894
|
| 272265 |
+
},
|
| 272266 |
+
{
|
| 272267 |
+
"epoch": 1095.01,
|
| 272268 |
+
"learning_rate": 7.839070512820513e-06,
|
| 272269 |
+
"loss": 0.3456,
|
| 272270 |
+
"step": 135895
|
| 272271 |
+
},
|
| 272272 |
+
{
|
| 272273 |
+
"epoch": 1095.05,
|
| 272274 |
+
"learning_rate": 7.838990384615384e-06,
|
| 272275 |
+
"loss": 0.2858,
|
| 272276 |
+
"step": 135900
|
| 272277 |
+
},
|
| 272278 |
+
{
|
| 272279 |
+
"epoch": 1095.09,
|
| 272280 |
+
"learning_rate": 7.838910256410257e-06,
|
| 272281 |
+
"loss": 0.2599,
|
| 272282 |
+
"step": 135905
|
| 272283 |
+
},
|
| 272284 |
+
{
|
| 272285 |
+
"epoch": 1095.13,
|
| 272286 |
+
"learning_rate": 7.838830128205129e-06,
|
| 272287 |
+
"loss": 0.2929,
|
| 272288 |
+
"step": 135910
|
| 272289 |
+
},
|
| 272290 |
+
{
|
| 272291 |
+
"epoch": 1095.17,
|
| 272292 |
+
"learning_rate": 7.83875e-06,
|
| 272293 |
+
"loss": 0.4546,
|
| 272294 |
+
"step": 135915
|
| 272295 |
+
},
|
| 272296 |
+
{
|
| 272297 |
+
"epoch": 1095.21,
|
| 272298 |
+
"learning_rate": 7.838669871794873e-06,
|
| 272299 |
+
"loss": 1.1072,
|
| 272300 |
+
"step": 135920
|
| 272301 |
+
},
|
| 272302 |
+
{
|
| 272303 |
+
"epoch": 1095.25,
|
| 272304 |
+
"learning_rate": 7.838589743589745e-06,
|
| 272305 |
+
"loss": 0.2675,
|
| 272306 |
+
"step": 135925
|
| 272307 |
+
},
|
| 272308 |
+
{
|
| 272309 |
+
"epoch": 1095.29,
|
| 272310 |
+
"learning_rate": 7.838509615384616e-06,
|
| 272311 |
+
"loss": 0.249,
|
| 272312 |
+
"step": 135930
|
| 272313 |
+
},
|
| 272314 |
+
{
|
| 272315 |
+
"epoch": 1095.33,
|
| 272316 |
+
"learning_rate": 7.838429487179487e-06,
|
| 272317 |
+
"loss": 0.3207,
|
| 272318 |
+
"step": 135935
|
| 272319 |
+
},
|
| 272320 |
+
{
|
| 272321 |
+
"epoch": 1095.37,
|
| 272322 |
+
"learning_rate": 7.83834935897436e-06,
|
| 272323 |
+
"loss": 0.4878,
|
| 272324 |
+
"step": 135940
|
| 272325 |
+
},
|
| 272326 |
+
{
|
| 272327 |
+
"epoch": 1095.41,
|
| 272328 |
+
"learning_rate": 7.838269230769232e-06,
|
| 272329 |
+
"loss": 1.059,
|
| 272330 |
+
"step": 135945
|
| 272331 |
+
},
|
| 272332 |
+
{
|
| 272333 |
+
"epoch": 1095.45,
|
| 272334 |
+
"learning_rate": 7.838189102564103e-06,
|
| 272335 |
+
"loss": 0.2822,
|
| 272336 |
+
"step": 135950
|
| 272337 |
+
},
|
| 272338 |
+
{
|
| 272339 |
+
"epoch": 1095.49,
|
| 272340 |
+
"learning_rate": 7.838108974358976e-06,
|
| 272341 |
+
"loss": 0.2435,
|
| 272342 |
+
"step": 135955
|
| 272343 |
+
},
|
| 272344 |
+
{
|
| 272345 |
+
"epoch": 1095.53,
|
| 272346 |
+
"learning_rate": 7.838028846153847e-06,
|
| 272347 |
+
"loss": 0.3643,
|
| 272348 |
+
"step": 135960
|
| 272349 |
+
},
|
| 272350 |
+
{
|
| 272351 |
+
"epoch": 1095.57,
|
| 272352 |
+
"learning_rate": 7.837948717948719e-06,
|
| 272353 |
+
"loss": 0.4741,
|
| 272354 |
+
"step": 135965
|
| 272355 |
+
},
|
| 272356 |
+
{
|
| 272357 |
+
"epoch": 1095.61,
|
| 272358 |
+
"learning_rate": 7.83786858974359e-06,
|
| 272359 |
+
"loss": 1.0892,
|
| 272360 |
+
"step": 135970
|
| 272361 |
+
},
|
| 272362 |
+
{
|
| 272363 |
+
"epoch": 1095.65,
|
| 272364 |
+
"learning_rate": 7.837788461538463e-06,
|
| 272365 |
+
"loss": 0.4263,
|
| 272366 |
+
"step": 135975
|
| 272367 |
+
},
|
| 272368 |
+
{
|
| 272369 |
+
"epoch": 1095.69,
|
| 272370 |
+
"learning_rate": 7.837708333333333e-06,
|
| 272371 |
+
"loss": 0.3049,
|
| 272372 |
+
"step": 135980
|
| 272373 |
+
},
|
| 272374 |
+
{
|
| 272375 |
+
"epoch": 1095.73,
|
| 272376 |
+
"learning_rate": 7.837628205128206e-06,
|
| 272377 |
+
"loss": 0.3588,
|
| 272378 |
+
"step": 135985
|
| 272379 |
+
},
|
| 272380 |
+
{
|
| 272381 |
+
"epoch": 1095.77,
|
| 272382 |
+
"learning_rate": 7.837548076923077e-06,
|
| 272383 |
+
"loss": 0.5209,
|
| 272384 |
+
"step": 135990
|
| 272385 |
+
},
|
| 272386 |
+
{
|
| 272387 |
+
"epoch": 1095.81,
|
| 272388 |
+
"learning_rate": 7.837467948717949e-06,
|
| 272389 |
+
"loss": 1.0809,
|
| 272390 |
+
"step": 135995
|
| 272391 |
+
},
|
| 272392 |
+
{
|
| 272393 |
+
"epoch": 1095.85,
|
| 272394 |
+
"learning_rate": 7.83738782051282e-06,
|
| 272395 |
+
"loss": 0.2727,
|
| 272396 |
+
"step": 136000
|
| 272397 |
+
},
|
| 272398 |
+
{
|
| 272399 |
+
"epoch": 1095.9,
|
| 272400 |
+
"learning_rate": 7.837307692307693e-06,
|
| 272401 |
+
"loss": 0.3043,
|
| 272402 |
+
"step": 136005
|
| 272403 |
+
},
|
| 272404 |
+
{
|
| 272405 |
+
"epoch": 1095.94,
|
| 272406 |
+
"learning_rate": 7.837227564102564e-06,
|
| 272407 |
+
"loss": 0.4021,
|
| 272408 |
+
"step": 136010
|
| 272409 |
+
},
|
| 272410 |
+
{
|
| 272411 |
+
"epoch": 1095.98,
|
| 272412 |
+
"learning_rate": 7.837147435897436e-06,
|
| 272413 |
+
"loss": 0.6525,
|
| 272414 |
+
"step": 136015
|
| 272415 |
+
},
|
| 272416 |
+
{
|
| 272417 |
+
"epoch": 1096.0,
|
| 272418 |
+
"eval_loss": 0.36106330156326294,
|
| 272419 |
+
"eval_runtime": 37.0373,
|
| 272420 |
+
"eval_samples_per_second": 22.761,
|
| 272421 |
+
"eval_steps_per_second": 0.729,
|
| 272422 |
+
"eval_wer": 0.17515510027413073,
|
| 272423 |
+
"step": 136018
|
| 272424 |
}
|
| 272425 |
],
|
| 272426 |
"max_steps": 620000,
|
| 272427 |
"num_train_epochs": 5000,
|
| 272428 |
+
"total_flos": 3.827546148750689e+20,
|
| 272429 |
"trial_name": null,
|
| 272430 |
"trial_params": null
|
| 272431 |
}
|
model-bin/finetune/base/{checkpoint-135397 → checkpoint-136018}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630179532.3843727/events.out.tfevents.1630179532.86bb0ddabf9b.4092.281
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ef49fd1be1d4fd262a2ef36e5d97778d72d62e9fc83f581f834fd55a0fc6c28e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630179923.6632285/events.out.tfevents.1630179923.86bb0ddabf9b.4092.283
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da56798ff7ed9c2f7ce059305af4f3227f421a438433716ab5dee940ce9d9313
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630180319.8798604/events.out.tfevents.1630180319.86bb0ddabf9b.4092.285
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da2add2ab4033faa10baea4f60d70df3631278111fa8e98e959a79e35490440e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630180707.9549763/events.out.tfevents.1630180707.86bb0ddabf9b.4092.287
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2c9b50fb7bf089fd20f3acbcb946cf69a8905f56979b44c360f596a13023b8e1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630181095.4206262/events.out.tfevents.1630181095.86bb0ddabf9b.4092.289
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:be18e7e7f28d59ae329d9e39c9b598346b404405d5fff33e7dc1ac123c6b6d91
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630179532.86bb0ddabf9b.4092.280
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:826c9ab2b10680c4b1db33e5e936e2c88f52f0153b3b24f052dbceef43c83ae7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630179923.86bb0ddabf9b.4092.282
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:99420d52aa8f74aaa85c3da6f8493b0901c3817b635625c45082424f87c1bbce
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630180319.86bb0ddabf9b.4092.284
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bb1ba94fac627f7430e91e8c70b996a734fde462563f516c84d06fc40362abe7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630180707.86bb0ddabf9b.4092.286
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:787b930089e3f8930079fa07480925a96554ea9077575a54ebe57531f594b585
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630181095.86bb0ddabf9b.4092.288
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:220143d9f35515a0e4c581030267d68f315299207b6787a0518b43aaa60d297d
|
| 3 |
+
size 8622
|