"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629706595.8000836/events.out.tfevents.1629706595.74272264b15c.932.83 +3 -0
- model-bin/finetune/base/log/1629707241.594185/events.out.tfevents.1629707241.74272264b15c.932.85 +3 -0
- model-bin/finetune/base/log/1629707885.1581163/events.out.tfevents.1629707885.74272264b15c.932.87 +3 -0
- model-bin/finetune/base/log/1629708521.3637116/events.out.tfevents.1629708521.74272264b15c.932.89 +3 -0
- model-bin/finetune/base/log/1629709157.223267/events.out.tfevents.1629709157.74272264b15c.932.91 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629706595.74272264b15c.932.82 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629707241.74272264b15c.932.84 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629707885.74272264b15c.932.86 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629708521.74272264b15c.932.88 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629709157.74272264b15c.932.90 +3 -0
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/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 sha256:4275488caad469bf0ad60c54412cc7348a49c4d4f0e9cb465e70ad72d4f50d27
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/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 sha256:532e3531ae19fb5e90ba8c95182844c1515ccaf1c80b503e8876dadca556adad
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/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 sha256:87fc68a725bb3d06115349430ec68b24ca45a66f3bee09cf752293514ead9f51
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/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 sha256:19d417c9c70cf78d78adc95946a258e6bef56e4ac9753dbffb17338e7d7b5004
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/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 sha256:305ea61598bb48502cb647c2a8a675ffd20b6731385c7f50614f22a456e1db69
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -154464,11 +154464,800 @@
|
|
| 154464 |
"eval_steps_per_second": 0.685,
|
| 154465 |
"eval_wer": 0.20608132683494582,
|
| 154466 |
"step": 43306
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 154467 |
}
|
| 154468 |
],
|
| 154469 |
"max_steps": 625000,
|
| 154470 |
"num_train_epochs": 5000,
|
| 154471 |
-
"total_flos": 1.
|
| 154472 |
"trial_name": null,
|
| 154473 |
"trial_params": null
|
| 154474 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
+
"epoch": 351.0,
|
| 5 |
+
"global_step": 43928,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 154464 |
"eval_steps_per_second": 0.685,
|
| 154465 |
"eval_wer": 0.20608132683494582,
|
| 154466 |
"step": 43306
|
| 154467 |
+
},
|
| 154468 |
+
{
|
| 154469 |
+
"epoch": 349.03,
|
| 154470 |
+
"learning_rate": 9.32221153846154e-06,
|
| 154471 |
+
"loss": 0.3183,
|
| 154472 |
+
"step": 43310
|
| 154473 |
+
},
|
| 154474 |
+
{
|
| 154475 |
+
"epoch": 349.07,
|
| 154476 |
+
"learning_rate": 9.322131410256412e-06,
|
| 154477 |
+
"loss": 0.3538,
|
| 154478 |
+
"step": 43315
|
| 154479 |
+
},
|
| 154480 |
+
{
|
| 154481 |
+
"epoch": 349.11,
|
| 154482 |
+
"learning_rate": 9.322051282051283e-06,
|
| 154483 |
+
"loss": 0.3934,
|
| 154484 |
+
"step": 43320
|
| 154485 |
+
},
|
| 154486 |
+
{
|
| 154487 |
+
"epoch": 349.15,
|
| 154488 |
+
"learning_rate": 9.321971153846154e-06,
|
| 154489 |
+
"loss": 0.5114,
|
| 154490 |
+
"step": 43325
|
| 154491 |
+
},
|
| 154492 |
+
{
|
| 154493 |
+
"epoch": 349.19,
|
| 154494 |
+
"learning_rate": 9.321891025641027e-06,
|
| 154495 |
+
"loss": 1.0557,
|
| 154496 |
+
"step": 43330
|
| 154497 |
+
},
|
| 154498 |
+
{
|
| 154499 |
+
"epoch": 349.23,
|
| 154500 |
+
"learning_rate": 9.321810897435897e-06,
|
| 154501 |
+
"loss": 0.7281,
|
| 154502 |
+
"step": 43335
|
| 154503 |
+
},
|
| 154504 |
+
{
|
| 154505 |
+
"epoch": 349.27,
|
| 154506 |
+
"learning_rate": 9.32173076923077e-06,
|
| 154507 |
+
"loss": 0.3231,
|
| 154508 |
+
"step": 43340
|
| 154509 |
+
},
|
| 154510 |
+
{
|
| 154511 |
+
"epoch": 349.31,
|
| 154512 |
+
"learning_rate": 9.321650641025641e-06,
|
| 154513 |
+
"loss": 0.3539,
|
| 154514 |
+
"step": 43345
|
| 154515 |
+
},
|
| 154516 |
+
{
|
| 154517 |
+
"epoch": 349.35,
|
| 154518 |
+
"learning_rate": 9.321570512820513e-06,
|
| 154519 |
+
"loss": 0.5559,
|
| 154520 |
+
"step": 43350
|
| 154521 |
+
},
|
| 154522 |
+
{
|
| 154523 |
+
"epoch": 349.39,
|
| 154524 |
+
"learning_rate": 9.321490384615386e-06,
|
| 154525 |
+
"loss": 1.0684,
|
| 154526 |
+
"step": 43355
|
| 154527 |
+
},
|
| 154528 |
+
{
|
| 154529 |
+
"epoch": 349.43,
|
| 154530 |
+
"learning_rate": 9.321410256410257e-06,
|
| 154531 |
+
"loss": 0.5812,
|
| 154532 |
+
"step": 43360
|
| 154533 |
+
},
|
| 154534 |
+
{
|
| 154535 |
+
"epoch": 349.47,
|
| 154536 |
+
"learning_rate": 9.321330128205129e-06,
|
| 154537 |
+
"loss": 0.4631,
|
| 154538 |
+
"step": 43365
|
| 154539 |
+
},
|
| 154540 |
+
{
|
| 154541 |
+
"epoch": 349.51,
|
| 154542 |
+
"learning_rate": 9.32125e-06,
|
| 154543 |
+
"loss": 0.4378,
|
| 154544 |
+
"step": 43370
|
| 154545 |
+
},
|
| 154546 |
+
{
|
| 154547 |
+
"epoch": 349.55,
|
| 154548 |
+
"learning_rate": 9.321169871794873e-06,
|
| 154549 |
+
"loss": 0.4259,
|
| 154550 |
+
"step": 43375
|
| 154551 |
+
},
|
| 154552 |
+
{
|
| 154553 |
+
"epoch": 349.59,
|
| 154554 |
+
"learning_rate": 9.321089743589744e-06,
|
| 154555 |
+
"loss": 0.9408,
|
| 154556 |
+
"step": 43380
|
| 154557 |
+
},
|
| 154558 |
+
{
|
| 154559 |
+
"epoch": 349.63,
|
| 154560 |
+
"learning_rate": 9.321009615384616e-06,
|
| 154561 |
+
"loss": 0.6635,
|
| 154562 |
+
"step": 43385
|
| 154563 |
+
},
|
| 154564 |
+
{
|
| 154565 |
+
"epoch": 349.67,
|
| 154566 |
+
"learning_rate": 9.320929487179487e-06,
|
| 154567 |
+
"loss": 0.3909,
|
| 154568 |
+
"step": 43390
|
| 154569 |
+
},
|
| 154570 |
+
{
|
| 154571 |
+
"epoch": 349.71,
|
| 154572 |
+
"learning_rate": 9.32084935897436e-06,
|
| 154573 |
+
"loss": 0.3581,
|
| 154574 |
+
"step": 43395
|
| 154575 |
+
},
|
| 154576 |
+
{
|
| 154577 |
+
"epoch": 349.76,
|
| 154578 |
+
"learning_rate": 9.320769230769231e-06,
|
| 154579 |
+
"loss": 0.4577,
|
| 154580 |
+
"step": 43400
|
| 154581 |
+
},
|
| 154582 |
+
{
|
| 154583 |
+
"epoch": 349.8,
|
| 154584 |
+
"learning_rate": 9.320689102564103e-06,
|
| 154585 |
+
"loss": 0.9979,
|
| 154586 |
+
"step": 43405
|
| 154587 |
+
},
|
| 154588 |
+
{
|
| 154589 |
+
"epoch": 349.84,
|
| 154590 |
+
"learning_rate": 9.320608974358976e-06,
|
| 154591 |
+
"loss": 0.7277,
|
| 154592 |
+
"step": 43410
|
| 154593 |
+
},
|
| 154594 |
+
{
|
| 154595 |
+
"epoch": 349.88,
|
| 154596 |
+
"learning_rate": 9.320528846153847e-06,
|
| 154597 |
+
"loss": 0.37,
|
| 154598 |
+
"step": 43415
|
| 154599 |
+
},
|
| 154600 |
+
{
|
| 154601 |
+
"epoch": 349.92,
|
| 154602 |
+
"learning_rate": 9.320448717948719e-06,
|
| 154603 |
+
"loss": 0.383,
|
| 154604 |
+
"step": 43420
|
| 154605 |
+
},
|
| 154606 |
+
{
|
| 154607 |
+
"epoch": 349.96,
|
| 154608 |
+
"learning_rate": 9.32036858974359e-06,
|
| 154609 |
+
"loss": 0.516,
|
| 154610 |
+
"step": 43425
|
| 154611 |
+
},
|
| 154612 |
+
{
|
| 154613 |
+
"epoch": 350.0,
|
| 154614 |
+
"learning_rate": 9.320288461538463e-06,
|
| 154615 |
+
"loss": 1.2505,
|
| 154616 |
+
"step": 43430
|
| 154617 |
+
},
|
| 154618 |
+
{
|
| 154619 |
+
"epoch": 350.0,
|
| 154620 |
+
"eval_loss": 0.39978110790252686,
|
| 154621 |
+
"eval_runtime": 40.437,
|
| 154622 |
+
"eval_samples_per_second": 20.872,
|
| 154623 |
+
"eval_steps_per_second": 0.668,
|
| 154624 |
+
"eval_wer": 0.21337252020573108,
|
| 154625 |
+
"step": 43430
|
| 154626 |
+
},
|
| 154627 |
+
{
|
| 154628 |
+
"epoch": 350.04,
|
| 154629 |
+
"learning_rate": 9.320208333333334e-06,
|
| 154630 |
+
"loss": 0.3964,
|
| 154631 |
+
"step": 43435
|
| 154632 |
+
},
|
| 154633 |
+
{
|
| 154634 |
+
"epoch": 350.08,
|
| 154635 |
+
"learning_rate": 9.320128205128206e-06,
|
| 154636 |
+
"loss": 0.3537,
|
| 154637 |
+
"step": 43440
|
| 154638 |
+
},
|
| 154639 |
+
{
|
| 154640 |
+
"epoch": 350.12,
|
| 154641 |
+
"learning_rate": 9.320048076923077e-06,
|
| 154642 |
+
"loss": 0.3958,
|
| 154643 |
+
"step": 43445
|
| 154644 |
+
},
|
| 154645 |
+
{
|
| 154646 |
+
"epoch": 350.16,
|
| 154647 |
+
"learning_rate": 9.31996794871795e-06,
|
| 154648 |
+
"loss": 0.5112,
|
| 154649 |
+
"step": 43450
|
| 154650 |
+
},
|
| 154651 |
+
{
|
| 154652 |
+
"epoch": 350.2,
|
| 154653 |
+
"learning_rate": 9.319887820512821e-06,
|
| 154654 |
+
"loss": 1.3959,
|
| 154655 |
+
"step": 43455
|
| 154656 |
+
},
|
| 154657 |
+
{
|
| 154658 |
+
"epoch": 350.24,
|
| 154659 |
+
"learning_rate": 9.319807692307693e-06,
|
| 154660 |
+
"loss": 0.3648,
|
| 154661 |
+
"step": 43460
|
| 154662 |
+
},
|
| 154663 |
+
{
|
| 154664 |
+
"epoch": 350.28,
|
| 154665 |
+
"learning_rate": 9.319727564102566e-06,
|
| 154666 |
+
"loss": 0.2931,
|
| 154667 |
+
"step": 43465
|
| 154668 |
+
},
|
| 154669 |
+
{
|
| 154670 |
+
"epoch": 350.32,
|
| 154671 |
+
"learning_rate": 9.319647435897437e-06,
|
| 154672 |
+
"loss": 0.3845,
|
| 154673 |
+
"step": 43470
|
| 154674 |
+
},
|
| 154675 |
+
{
|
| 154676 |
+
"epoch": 350.36,
|
| 154677 |
+
"learning_rate": 9.319567307692309e-06,
|
| 154678 |
+
"loss": 0.4687,
|
| 154679 |
+
"step": 43475
|
| 154680 |
+
},
|
| 154681 |
+
{
|
| 154682 |
+
"epoch": 350.4,
|
| 154683 |
+
"learning_rate": 9.31948717948718e-06,
|
| 154684 |
+
"loss": 1.3989,
|
| 154685 |
+
"step": 43480
|
| 154686 |
+
},
|
| 154687 |
+
{
|
| 154688 |
+
"epoch": 350.44,
|
| 154689 |
+
"learning_rate": 9.319407051282053e-06,
|
| 154690 |
+
"loss": 0.395,
|
| 154691 |
+
"step": 43485
|
| 154692 |
+
},
|
| 154693 |
+
{
|
| 154694 |
+
"epoch": 350.48,
|
| 154695 |
+
"learning_rate": 9.319326923076923e-06,
|
| 154696 |
+
"loss": 0.3212,
|
| 154697 |
+
"step": 43490
|
| 154698 |
+
},
|
| 154699 |
+
{
|
| 154700 |
+
"epoch": 350.52,
|
| 154701 |
+
"learning_rate": 9.319246794871796e-06,
|
| 154702 |
+
"loss": 0.3875,
|
| 154703 |
+
"step": 43495
|
| 154704 |
+
},
|
| 154705 |
+
{
|
| 154706 |
+
"epoch": 350.56,
|
| 154707 |
+
"learning_rate": 9.319166666666667e-06,
|
| 154708 |
+
"loss": 0.6368,
|
| 154709 |
+
"step": 43500
|
| 154710 |
+
},
|
| 154711 |
+
{
|
| 154712 |
+
"epoch": 350.6,
|
| 154713 |
+
"learning_rate": 9.319086538461538e-06,
|
| 154714 |
+
"loss": 1.3578,
|
| 154715 |
+
"step": 43505
|
| 154716 |
+
},
|
| 154717 |
+
{
|
| 154718 |
+
"epoch": 350.64,
|
| 154719 |
+
"learning_rate": 9.319006410256411e-06,
|
| 154720 |
+
"loss": 0.3539,
|
| 154721 |
+
"step": 43510
|
| 154722 |
+
},
|
| 154723 |
+
{
|
| 154724 |
+
"epoch": 350.68,
|
| 154725 |
+
"learning_rate": 9.318926282051283e-06,
|
| 154726 |
+
"loss": 0.3661,
|
| 154727 |
+
"step": 43515
|
| 154728 |
+
},
|
| 154729 |
+
{
|
| 154730 |
+
"epoch": 350.72,
|
| 154731 |
+
"learning_rate": 9.318846153846154e-06,
|
| 154732 |
+
"loss": 0.4237,
|
| 154733 |
+
"step": 43520
|
| 154734 |
+
},
|
| 154735 |
+
{
|
| 154736 |
+
"epoch": 350.76,
|
| 154737 |
+
"learning_rate": 9.318766025641026e-06,
|
| 154738 |
+
"loss": 0.609,
|
| 154739 |
+
"step": 43525
|
| 154740 |
+
},
|
| 154741 |
+
{
|
| 154742 |
+
"epoch": 350.8,
|
| 154743 |
+
"learning_rate": 9.318685897435899e-06,
|
| 154744 |
+
"loss": 1.2773,
|
| 154745 |
+
"step": 43530
|
| 154746 |
+
},
|
| 154747 |
+
{
|
| 154748 |
+
"epoch": 350.84,
|
| 154749 |
+
"learning_rate": 9.31860576923077e-06,
|
| 154750 |
+
"loss": 0.3719,
|
| 154751 |
+
"step": 43535
|
| 154752 |
+
},
|
| 154753 |
+
{
|
| 154754 |
+
"epoch": 350.88,
|
| 154755 |
+
"learning_rate": 9.318525641025641e-06,
|
| 154756 |
+
"loss": 0.3371,
|
| 154757 |
+
"step": 43540
|
| 154758 |
+
},
|
| 154759 |
+
{
|
| 154760 |
+
"epoch": 350.92,
|
| 154761 |
+
"learning_rate": 9.318445512820513e-06,
|
| 154762 |
+
"loss": 0.329,
|
| 154763 |
+
"step": 43545
|
| 154764 |
+
},
|
| 154765 |
+
{
|
| 154766 |
+
"epoch": 350.96,
|
| 154767 |
+
"learning_rate": 9.318365384615386e-06,
|
| 154768 |
+
"loss": 0.4682,
|
| 154769 |
+
"step": 43550
|
| 154770 |
+
},
|
| 154771 |
+
{
|
| 154772 |
+
"epoch": 351.0,
|
| 154773 |
+
"eval_loss": 0.3992723822593689,
|
| 154774 |
+
"eval_runtime": 39.7734,
|
| 154775 |
+
"eval_samples_per_second": 21.22,
|
| 154776 |
+
"eval_steps_per_second": 0.679,
|
| 154777 |
+
"eval_wer": 0.20792792792792794,
|
| 154778 |
+
"step": 43554
|
| 154779 |
+
},
|
| 154780 |
+
{
|
| 154781 |
+
"epoch": 351.01,
|
| 154782 |
+
"learning_rate": 9.318285256410257e-06,
|
| 154783 |
+
"loss": 0.554,
|
| 154784 |
+
"step": 43555
|
| 154785 |
+
},
|
| 154786 |
+
{
|
| 154787 |
+
"epoch": 351.05,
|
| 154788 |
+
"learning_rate": 9.318205128205128e-06,
|
| 154789 |
+
"loss": 0.477,
|
| 154790 |
+
"step": 43560
|
| 154791 |
+
},
|
| 154792 |
+
{
|
| 154793 |
+
"epoch": 351.09,
|
| 154794 |
+
"learning_rate": 9.318125000000002e-06,
|
| 154795 |
+
"loss": 0.3928,
|
| 154796 |
+
"step": 43565
|
| 154797 |
+
},
|
| 154798 |
+
{
|
| 154799 |
+
"epoch": 351.13,
|
| 154800 |
+
"learning_rate": 9.318044871794873e-06,
|
| 154801 |
+
"loss": 0.3859,
|
| 154802 |
+
"step": 43570
|
| 154803 |
+
},
|
| 154804 |
+
{
|
| 154805 |
+
"epoch": 351.17,
|
| 154806 |
+
"learning_rate": 9.317964743589744e-06,
|
| 154807 |
+
"loss": 0.6592,
|
| 154808 |
+
"step": 43575
|
| 154809 |
+
},
|
| 154810 |
+
{
|
| 154811 |
+
"epoch": 351.21,
|
| 154812 |
+
"learning_rate": 9.317884615384616e-06,
|
| 154813 |
+
"loss": 1.3205,
|
| 154814 |
+
"step": 43580
|
| 154815 |
+
},
|
| 154816 |
+
{
|
| 154817 |
+
"epoch": 351.25,
|
| 154818 |
+
"learning_rate": 9.317804487179489e-06,
|
| 154819 |
+
"loss": 0.4071,
|
| 154820 |
+
"step": 43585
|
| 154821 |
+
},
|
| 154822 |
+
{
|
| 154823 |
+
"epoch": 351.29,
|
| 154824 |
+
"learning_rate": 9.31772435897436e-06,
|
| 154825 |
+
"loss": 0.4883,
|
| 154826 |
+
"step": 43590
|
| 154827 |
+
},
|
| 154828 |
+
{
|
| 154829 |
+
"epoch": 351.33,
|
| 154830 |
+
"learning_rate": 9.317644230769231e-06,
|
| 154831 |
+
"loss": 0.4414,
|
| 154832 |
+
"step": 43595
|
| 154833 |
+
},
|
| 154834 |
+
{
|
| 154835 |
+
"epoch": 351.37,
|
| 154836 |
+
"learning_rate": 9.317564102564103e-06,
|
| 154837 |
+
"loss": 0.5947,
|
| 154838 |
+
"step": 43600
|
| 154839 |
+
},
|
| 154840 |
+
{
|
| 154841 |
+
"epoch": 351.41,
|
| 154842 |
+
"learning_rate": 9.317483974358976e-06,
|
| 154843 |
+
"loss": 1.3214,
|
| 154844 |
+
"step": 43605
|
| 154845 |
+
},
|
| 154846 |
+
{
|
| 154847 |
+
"epoch": 351.45,
|
| 154848 |
+
"learning_rate": 9.317403846153847e-06,
|
| 154849 |
+
"loss": 0.3833,
|
| 154850 |
+
"step": 43610
|
| 154851 |
+
},
|
| 154852 |
+
{
|
| 154853 |
+
"epoch": 351.49,
|
| 154854 |
+
"learning_rate": 9.317323717948718e-06,
|
| 154855 |
+
"loss": 0.5028,
|
| 154856 |
+
"step": 43615
|
| 154857 |
+
},
|
| 154858 |
+
{
|
| 154859 |
+
"epoch": 351.53,
|
| 154860 |
+
"learning_rate": 9.317243589743592e-06,
|
| 154861 |
+
"loss": 0.4401,
|
| 154862 |
+
"step": 43620
|
| 154863 |
+
},
|
| 154864 |
+
{
|
| 154865 |
+
"epoch": 351.57,
|
| 154866 |
+
"learning_rate": 9.317163461538463e-06,
|
| 154867 |
+
"loss": 0.5422,
|
| 154868 |
+
"step": 43625
|
| 154869 |
+
},
|
| 154870 |
+
{
|
| 154871 |
+
"epoch": 351.61,
|
| 154872 |
+
"learning_rate": 9.317083333333334e-06,
|
| 154873 |
+
"loss": 1.2012,
|
| 154874 |
+
"step": 43630
|
| 154875 |
+
},
|
| 154876 |
+
{
|
| 154877 |
+
"epoch": 351.65,
|
| 154878 |
+
"learning_rate": 9.317003205128206e-06,
|
| 154879 |
+
"loss": 0.3402,
|
| 154880 |
+
"step": 43635
|
| 154881 |
+
},
|
| 154882 |
+
{
|
| 154883 |
+
"epoch": 351.69,
|
| 154884 |
+
"learning_rate": 9.316923076923079e-06,
|
| 154885 |
+
"loss": 0.4192,
|
| 154886 |
+
"step": 43640
|
| 154887 |
+
},
|
| 154888 |
+
{
|
| 154889 |
+
"epoch": 351.73,
|
| 154890 |
+
"learning_rate": 9.316842948717948e-06,
|
| 154891 |
+
"loss": 0.3876,
|
| 154892 |
+
"step": 43645
|
| 154893 |
+
},
|
| 154894 |
+
{
|
| 154895 |
+
"epoch": 351.77,
|
| 154896 |
+
"learning_rate": 9.316762820512821e-06,
|
| 154897 |
+
"loss": 0.5627,
|
| 154898 |
+
"step": 43650
|
| 154899 |
+
},
|
| 154900 |
+
{
|
| 154901 |
+
"epoch": 351.81,
|
| 154902 |
+
"learning_rate": 9.316682692307694e-06,
|
| 154903 |
+
"loss": 1.0915,
|
| 154904 |
+
"step": 43655
|
| 154905 |
+
},
|
| 154906 |
+
{
|
| 154907 |
+
"epoch": 351.85,
|
| 154908 |
+
"learning_rate": 9.316602564102564e-06,
|
| 154909 |
+
"loss": 0.4319,
|
| 154910 |
+
"step": 43660
|
| 154911 |
+
},
|
| 154912 |
+
{
|
| 154913 |
+
"epoch": 351.89,
|
| 154914 |
+
"learning_rate": 9.316522435897437e-06,
|
| 154915 |
+
"loss": 0.3954,
|
| 154916 |
+
"step": 43665
|
| 154917 |
+
},
|
| 154918 |
+
{
|
| 154919 |
+
"epoch": 351.93,
|
| 154920 |
+
"learning_rate": 9.316442307692309e-06,
|
| 154921 |
+
"loss": 0.3565,
|
| 154922 |
+
"step": 43670
|
| 154923 |
+
},
|
| 154924 |
+
{
|
| 154925 |
+
"epoch": 351.97,
|
| 154926 |
+
"learning_rate": 9.31636217948718e-06,
|
| 154927 |
+
"loss": 0.5501,
|
| 154928 |
+
"step": 43675
|
| 154929 |
+
},
|
| 154930 |
+
{
|
| 154931 |
+
"epoch": 352.0,
|
| 154932 |
+
"eval_loss": 0.41606399416923523,
|
| 154933 |
+
"eval_runtime": 38.4073,
|
| 154934 |
+
"eval_samples_per_second": 21.975,
|
| 154935 |
+
"eval_steps_per_second": 0.703,
|
| 154936 |
+
"eval_wer": 0.205798138869005,
|
| 154937 |
+
"step": 43678
|
| 154938 |
+
},
|
| 154939 |
+
{
|
| 154940 |
+
"epoch": 349.02,
|
| 154941 |
+
"learning_rate": 9.316282051282051e-06,
|
| 154942 |
+
"loss": 0.4183,
|
| 154943 |
+
"step": 43680
|
| 154944 |
+
},
|
| 154945 |
+
{
|
| 154946 |
+
"epoch": 349.06,
|
| 154947 |
+
"learning_rate": 9.316201923076924e-06,
|
| 154948 |
+
"loss": 0.3315,
|
| 154949 |
+
"step": 43685
|
| 154950 |
+
},
|
| 154951 |
+
{
|
| 154952 |
+
"epoch": 349.1,
|
| 154953 |
+
"learning_rate": 9.316121794871796e-06,
|
| 154954 |
+
"loss": 0.4277,
|
| 154955 |
+
"step": 43690
|
| 154956 |
+
},
|
| 154957 |
+
{
|
| 154958 |
+
"epoch": 349.14,
|
| 154959 |
+
"learning_rate": 9.316041666666667e-06,
|
| 154960 |
+
"loss": 0.3858,
|
| 154961 |
+
"step": 43695
|
| 154962 |
+
},
|
| 154963 |
+
{
|
| 154964 |
+
"epoch": 349.18,
|
| 154965 |
+
"learning_rate": 9.315961538461538e-06,
|
| 154966 |
+
"loss": 0.7705,
|
| 154967 |
+
"step": 43700
|
| 154968 |
+
},
|
| 154969 |
+
{
|
| 154970 |
+
"epoch": 349.22,
|
| 154971 |
+
"learning_rate": 9.315881410256411e-06,
|
| 154972 |
+
"loss": 1.1,
|
| 154973 |
+
"step": 43705
|
| 154974 |
+
},
|
| 154975 |
+
{
|
| 154976 |
+
"epoch": 349.26,
|
| 154977 |
+
"learning_rate": 9.315801282051283e-06,
|
| 154978 |
+
"loss": 0.452,
|
| 154979 |
+
"step": 43710
|
| 154980 |
+
},
|
| 154981 |
+
{
|
| 154982 |
+
"epoch": 349.3,
|
| 154983 |
+
"learning_rate": 9.315721153846154e-06,
|
| 154984 |
+
"loss": 0.4128,
|
| 154985 |
+
"step": 43715
|
| 154986 |
+
},
|
| 154987 |
+
{
|
| 154988 |
+
"epoch": 349.34,
|
| 154989 |
+
"learning_rate": 9.315641025641027e-06,
|
| 154990 |
+
"loss": 0.3649,
|
| 154991 |
+
"step": 43720
|
| 154992 |
+
},
|
| 154993 |
+
{
|
| 154994 |
+
"epoch": 349.38,
|
| 154995 |
+
"learning_rate": 9.315560897435899e-06,
|
| 154996 |
+
"loss": 0.5959,
|
| 154997 |
+
"step": 43725
|
| 154998 |
+
},
|
| 154999 |
+
{
|
| 155000 |
+
"epoch": 349.42,
|
| 155001 |
+
"learning_rate": 9.31548076923077e-06,
|
| 155002 |
+
"loss": 1.0803,
|
| 155003 |
+
"step": 43730
|
| 155004 |
+
},
|
| 155005 |
+
{
|
| 155006 |
+
"epoch": 349.46,
|
| 155007 |
+
"learning_rate": 9.315400641025641e-06,
|
| 155008 |
+
"loss": 0.4288,
|
| 155009 |
+
"step": 43735
|
| 155010 |
+
},
|
| 155011 |
+
{
|
| 155012 |
+
"epoch": 349.5,
|
| 155013 |
+
"learning_rate": 9.315320512820514e-06,
|
| 155014 |
+
"loss": 0.3629,
|
| 155015 |
+
"step": 43740
|
| 155016 |
+
},
|
| 155017 |
+
{
|
| 155018 |
+
"epoch": 349.54,
|
| 155019 |
+
"learning_rate": 9.315240384615386e-06,
|
| 155020 |
+
"loss": 0.4583,
|
| 155021 |
+
"step": 43745
|
| 155022 |
+
},
|
| 155023 |
+
{
|
| 155024 |
+
"epoch": 349.58,
|
| 155025 |
+
"learning_rate": 9.315160256410257e-06,
|
| 155026 |
+
"loss": 0.6611,
|
| 155027 |
+
"step": 43750
|
| 155028 |
+
},
|
| 155029 |
+
{
|
| 155030 |
+
"epoch": 349.62,
|
| 155031 |
+
"learning_rate": 9.315080128205128e-06,
|
| 155032 |
+
"loss": 1.168,
|
| 155033 |
+
"step": 43755
|
| 155034 |
+
},
|
| 155035 |
+
{
|
| 155036 |
+
"epoch": 349.66,
|
| 155037 |
+
"learning_rate": 9.315000000000001e-06,
|
| 155038 |
+
"loss": 0.411,
|
| 155039 |
+
"step": 43760
|
| 155040 |
+
},
|
| 155041 |
+
{
|
| 155042 |
+
"epoch": 349.7,
|
| 155043 |
+
"learning_rate": 9.314919871794873e-06,
|
| 155044 |
+
"loss": 0.3637,
|
| 155045 |
+
"step": 43765
|
| 155046 |
+
},
|
| 155047 |
+
{
|
| 155048 |
+
"epoch": 349.74,
|
| 155049 |
+
"learning_rate": 9.314839743589744e-06,
|
| 155050 |
+
"loss": 0.3932,
|
| 155051 |
+
"step": 43770
|
| 155052 |
+
},
|
| 155053 |
+
{
|
| 155054 |
+
"epoch": 349.78,
|
| 155055 |
+
"learning_rate": 9.314759615384617e-06,
|
| 155056 |
+
"loss": 0.7297,
|
| 155057 |
+
"step": 43775
|
| 155058 |
+
},
|
| 155059 |
+
{
|
| 155060 |
+
"epoch": 349.82,
|
| 155061 |
+
"learning_rate": 9.314695512820515e-06,
|
| 155062 |
+
"loss": 1.149,
|
| 155063 |
+
"step": 43780
|
| 155064 |
+
},
|
| 155065 |
+
{
|
| 155066 |
+
"epoch": 349.86,
|
| 155067 |
+
"learning_rate": 9.314615384615386e-06,
|
| 155068 |
+
"loss": 0.3523,
|
| 155069 |
+
"step": 43785
|
| 155070 |
+
},
|
| 155071 |
+
{
|
| 155072 |
+
"epoch": 349.9,
|
| 155073 |
+
"learning_rate": 9.314535256410257e-06,
|
| 155074 |
+
"loss": 0.2898,
|
| 155075 |
+
"step": 43790
|
| 155076 |
+
},
|
| 155077 |
+
{
|
| 155078 |
+
"epoch": 349.94,
|
| 155079 |
+
"learning_rate": 9.314455128205129e-06,
|
| 155080 |
+
"loss": 0.4869,
|
| 155081 |
+
"step": 43795
|
| 155082 |
+
},
|
| 155083 |
+
{
|
| 155084 |
+
"epoch": 349.98,
|
| 155085 |
+
"learning_rate": 9.314375000000002e-06,
|
| 155086 |
+
"loss": 0.6864,
|
| 155087 |
+
"step": 43800
|
| 155088 |
+
},
|
| 155089 |
+
{
|
| 155090 |
+
"epoch": 350.0,
|
| 155091 |
+
"eval_loss": 0.46590304374694824,
|
| 155092 |
+
"eval_runtime": 39.6318,
|
| 155093 |
+
"eval_samples_per_second": 21.296,
|
| 155094 |
+
"eval_steps_per_second": 0.681,
|
| 155095 |
+
"eval_wer": 0.21451432824711575,
|
| 155096 |
+
"step": 43803
|
| 155097 |
+
},
|
| 155098 |
+
{
|
| 155099 |
+
"epoch": 350.02,
|
| 155100 |
+
"learning_rate": 9.314294871794871e-06,
|
| 155101 |
+
"loss": 0.4529,
|
| 155102 |
+
"step": 43805
|
| 155103 |
+
},
|
| 155104 |
+
{
|
| 155105 |
+
"epoch": 350.06,
|
| 155106 |
+
"learning_rate": 9.314214743589744e-06,
|
| 155107 |
+
"loss": 0.3352,
|
| 155108 |
+
"step": 43810
|
| 155109 |
+
},
|
| 155110 |
+
{
|
| 155111 |
+
"epoch": 350.1,
|
| 155112 |
+
"learning_rate": 9.314134615384616e-06,
|
| 155113 |
+
"loss": 0.3689,
|
| 155114 |
+
"step": 43815
|
| 155115 |
+
},
|
| 155116 |
+
{
|
| 155117 |
+
"epoch": 350.14,
|
| 155118 |
+
"learning_rate": 9.314054487179487e-06,
|
| 155119 |
+
"loss": 0.3866,
|
| 155120 |
+
"step": 43820
|
| 155121 |
+
},
|
| 155122 |
+
{
|
| 155123 |
+
"epoch": 350.18,
|
| 155124 |
+
"learning_rate": 9.313974358974359e-06,
|
| 155125 |
+
"loss": 0.8155,
|
| 155126 |
+
"step": 43825
|
| 155127 |
+
},
|
| 155128 |
+
{
|
| 155129 |
+
"epoch": 350.22,
|
| 155130 |
+
"learning_rate": 9.313894230769232e-06,
|
| 155131 |
+
"loss": 1.1128,
|
| 155132 |
+
"step": 43830
|
| 155133 |
+
},
|
| 155134 |
+
{
|
| 155135 |
+
"epoch": 350.26,
|
| 155136 |
+
"learning_rate": 9.313814102564103e-06,
|
| 155137 |
+
"loss": 0.3854,
|
| 155138 |
+
"step": 43835
|
| 155139 |
+
},
|
| 155140 |
+
{
|
| 155141 |
+
"epoch": 350.3,
|
| 155142 |
+
"learning_rate": 9.313733974358974e-06,
|
| 155143 |
+
"loss": 0.3189,
|
| 155144 |
+
"step": 43840
|
| 155145 |
+
},
|
| 155146 |
+
{
|
| 155147 |
+
"epoch": 350.34,
|
| 155148 |
+
"learning_rate": 9.313653846153847e-06,
|
| 155149 |
+
"loss": 0.4679,
|
| 155150 |
+
"step": 43845
|
| 155151 |
+
},
|
| 155152 |
+
{
|
| 155153 |
+
"epoch": 350.38,
|
| 155154 |
+
"learning_rate": 9.313573717948719e-06,
|
| 155155 |
+
"loss": 0.7582,
|
| 155156 |
+
"step": 43850
|
| 155157 |
+
},
|
| 155158 |
+
{
|
| 155159 |
+
"epoch": 350.42,
|
| 155160 |
+
"learning_rate": 9.31349358974359e-06,
|
| 155161 |
+
"loss": 1.1655,
|
| 155162 |
+
"step": 43855
|
| 155163 |
+
},
|
| 155164 |
+
{
|
| 155165 |
+
"epoch": 350.46,
|
| 155166 |
+
"learning_rate": 9.313413461538461e-06,
|
| 155167 |
+
"loss": 0.4036,
|
| 155168 |
+
"step": 43860
|
| 155169 |
+
},
|
| 155170 |
+
{
|
| 155171 |
+
"epoch": 350.5,
|
| 155172 |
+
"learning_rate": 9.313333333333335e-06,
|
| 155173 |
+
"loss": 0.3391,
|
| 155174 |
+
"step": 43865
|
| 155175 |
+
},
|
| 155176 |
+
{
|
| 155177 |
+
"epoch": 350.54,
|
| 155178 |
+
"learning_rate": 9.313253205128206e-06,
|
| 155179 |
+
"loss": 0.3896,
|
| 155180 |
+
"step": 43870
|
| 155181 |
+
},
|
| 155182 |
+
{
|
| 155183 |
+
"epoch": 350.58,
|
| 155184 |
+
"learning_rate": 9.313173076923077e-06,
|
| 155185 |
+
"loss": 0.6911,
|
| 155186 |
+
"step": 43875
|
| 155187 |
+
},
|
| 155188 |
+
{
|
| 155189 |
+
"epoch": 350.62,
|
| 155190 |
+
"learning_rate": 9.31309294871795e-06,
|
| 155191 |
+
"loss": 1.0488,
|
| 155192 |
+
"step": 43880
|
| 155193 |
+
},
|
| 155194 |
+
{
|
| 155195 |
+
"epoch": 350.66,
|
| 155196 |
+
"learning_rate": 9.313012820512822e-06,
|
| 155197 |
+
"loss": 0.3502,
|
| 155198 |
+
"step": 43885
|
| 155199 |
+
},
|
| 155200 |
+
{
|
| 155201 |
+
"epoch": 350.7,
|
| 155202 |
+
"learning_rate": 9.312932692307693e-06,
|
| 155203 |
+
"loss": 0.323,
|
| 155204 |
+
"step": 43890
|
| 155205 |
+
},
|
| 155206 |
+
{
|
| 155207 |
+
"epoch": 350.74,
|
| 155208 |
+
"learning_rate": 9.312852564102564e-06,
|
| 155209 |
+
"loss": 0.3316,
|
| 155210 |
+
"step": 43895
|
| 155211 |
+
},
|
| 155212 |
+
{
|
| 155213 |
+
"epoch": 350.78,
|
| 155214 |
+
"learning_rate": 9.312772435897437e-06,
|
| 155215 |
+
"loss": 0.7101,
|
| 155216 |
+
"step": 43900
|
| 155217 |
+
},
|
| 155218 |
+
{
|
| 155219 |
+
"epoch": 350.82,
|
| 155220 |
+
"learning_rate": 9.312692307692309e-06,
|
| 155221 |
+
"loss": 1.0713,
|
| 155222 |
+
"step": 43905
|
| 155223 |
+
},
|
| 155224 |
+
{
|
| 155225 |
+
"epoch": 350.86,
|
| 155226 |
+
"learning_rate": 9.31261217948718e-06,
|
| 155227 |
+
"loss": 0.3449,
|
| 155228 |
+
"step": 43910
|
| 155229 |
+
},
|
| 155230 |
+
{
|
| 155231 |
+
"epoch": 350.9,
|
| 155232 |
+
"learning_rate": 9.312532051282051e-06,
|
| 155233 |
+
"loss": 0.3546,
|
| 155234 |
+
"step": 43915
|
| 155235 |
+
},
|
| 155236 |
+
{
|
| 155237 |
+
"epoch": 350.94,
|
| 155238 |
+
"learning_rate": 9.312451923076925e-06,
|
| 155239 |
+
"loss": 0.4361,
|
| 155240 |
+
"step": 43920
|
| 155241 |
+
},
|
| 155242 |
+
{
|
| 155243 |
+
"epoch": 350.98,
|
| 155244 |
+
"learning_rate": 9.312371794871796e-06,
|
| 155245 |
+
"loss": 0.8942,
|
| 155246 |
+
"step": 43925
|
| 155247 |
+
},
|
| 155248 |
+
{
|
| 155249 |
+
"epoch": 351.0,
|
| 155250 |
+
"eval_loss": 0.43212229013442993,
|
| 155251 |
+
"eval_runtime": 40.0877,
|
| 155252 |
+
"eval_samples_per_second": 21.054,
|
| 155253 |
+
"eval_steps_per_second": 0.674,
|
| 155254 |
+
"eval_wer": 0.2129204326566899,
|
| 155255 |
+
"step": 43928
|
| 155256 |
}
|
| 155257 |
],
|
| 155258 |
"max_steps": 625000,
|
| 155259 |
"num_train_epochs": 5000,
|
| 155260 |
+
"total_flos": 1.2359827217075426e+20,
|
| 155261 |
"trial_name": null,
|
| 155262 |
"trial_params": null
|
| 155263 |
}
|
model-bin/finetune/base/{checkpoint-43306 → checkpoint-43928}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629706595.8000836/events.out.tfevents.1629706595.74272264b15c.932.83
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38ddcb763717292849e08f9d520cc3ce35feb8a791d48c5620aecd4096dc6252
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629707241.594185/events.out.tfevents.1629707241.74272264b15c.932.85
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e417575244122a000b4c42451c7d22f3041fca3ca9448e6107c1fb038d9f7fc6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629707885.1581163/events.out.tfevents.1629707885.74272264b15c.932.87
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3705e4c4e78c4c26adc14fd5dd2a2a1119ea327537b0bee2b045339ea740b4b2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629708521.3637116/events.out.tfevents.1629708521.74272264b15c.932.89
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:09bb6eb82af2568df886a6fac12e24a057112780dda65183b34f4adb6045e71b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629709157.223267/events.out.tfevents.1629709157.74272264b15c.932.91
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6cfb6c25527fcfa984d708cd4ffd2752da96be98dc54bcd8eb4f30d5fdf851f4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629706595.74272264b15c.932.82
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3fb9a9de186434d5bbe81898392d49eb00692d01cff6add3d2775075b014f6ab
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629707241.74272264b15c.932.84
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a23562f8b06bde741c6353ed0a7f0111eb00c101516f5ad0f30fa3428689e24
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629707885.74272264b15c.932.86
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a8ea68953c41dbf3620d9254b63b6b4797c4480e5dab39d577cc856acad5aaec
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629708521.74272264b15c.932.88
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:25b2c1768f301602f909d0cbd09dae66518f18559f274eb5be68d6878a8dd5eb
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629709157.74272264b15c.932.90
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bb43c0ed902bb8d1971fc143f4fd6509fcedf3bec9e67be91d0f61a51197195d
|
| 3 |
+
size 8622
|