"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629928346.7422721/events.out.tfevents.1629928346.7e498afd5545.7645.165 +3 -0
- model-bin/finetune/base/log/1629928816.2265396/events.out.tfevents.1629928816.7e498afd5545.7645.167 +3 -0
- model-bin/finetune/base/log/1629929279.5633097/events.out.tfevents.1629929279.7e498afd5545.7645.169 +3 -0
- model-bin/finetune/base/log/1629929743.3866482/events.out.tfevents.1629929743.7e498afd5545.7645.171 +3 -0
- model-bin/finetune/base/log/1629930214.058052/events.out.tfevents.1629930214.7e498afd5545.7645.173 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629928346.7e498afd5545.7645.164 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629928816.7e498afd5545.7645.166 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629929279.7e498afd5545.7645.168 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629929743.7e498afd5545.7645.170 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629930214.7e498afd5545.7645.172 +3 -0
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/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 sha256:4cf8b784263d4a72e2214afab913db466a83f2d0cf77a41a39cf01dd543f169a
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/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:b2be6ca945e3dd0e43577e0637b9cf6d87a2f309fc3cdf8288463ca55c399177
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/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:b343c273f252ce92e634a1a086aeffa058f7b3307f9211ab0f20a39a15b6f527
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/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:c35d331c08033f7866283c2091a8ec97b1c59d0ee2ad0d4d551d4f1157dffdb8
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/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:e8b9fd83e3261dc1f13417e46e117ca33e9815431fca14cb0942f3be59f93d76
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch": 696.
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -209403,11 +209403,806 @@
|
|
| 209403 |
"eval_steps_per_second": 0.644,
|
| 209404 |
"eval_wer": 0.18736416098328712,
|
| 209405 |
"step": 86488
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 209406 |
}
|
| 209407 |
],
|
| 209408 |
-
"max_steps":
|
| 209409 |
"num_train_epochs": 5000,
|
| 209410 |
-
"total_flos": 2.
|
| 209411 |
"trial_name": null,
|
| 209412 |
"trial_params": null
|
| 209413 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 696.0,
|
| 5 |
+
"global_step": 87113,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 209403 |
"eval_steps_per_second": 0.644,
|
| 209404 |
"eval_wer": 0.18736416098328712,
|
| 209405 |
"step": 86488
|
| 209406 |
+
},
|
| 209407 |
+
{
|
| 209408 |
+
"epoch": 691.02,
|
| 209409 |
+
"learning_rate": 8.619450726979e-06,
|
| 209410 |
+
"loss": 0.3771,
|
| 209411 |
+
"step": 86490
|
| 209412 |
+
},
|
| 209413 |
+
{
|
| 209414 |
+
"epoch": 691.06,
|
| 209415 |
+
"learning_rate": 8.619369951534734e-06,
|
| 209416 |
+
"loss": 0.316,
|
| 209417 |
+
"step": 86495
|
| 209418 |
+
},
|
| 209419 |
+
{
|
| 209420 |
+
"epoch": 691.1,
|
| 209421 |
+
"learning_rate": 8.61928917609047e-06,
|
| 209422 |
+
"loss": 0.2947,
|
| 209423 |
+
"step": 86500
|
| 209424 |
+
},
|
| 209425 |
+
{
|
| 209426 |
+
"epoch": 691.14,
|
| 209427 |
+
"learning_rate": 8.619208400646204e-06,
|
| 209428 |
+
"loss": 0.3675,
|
| 209429 |
+
"step": 86505
|
| 209430 |
+
},
|
| 209431 |
+
{
|
| 209432 |
+
"epoch": 691.18,
|
| 209433 |
+
"learning_rate": 8.61912762520194e-06,
|
| 209434 |
+
"loss": 0.6921,
|
| 209435 |
+
"step": 86510
|
| 209436 |
+
},
|
| 209437 |
+
{
|
| 209438 |
+
"epoch": 691.22,
|
| 209439 |
+
"learning_rate": 8.619046849757674e-06,
|
| 209440 |
+
"loss": 1.0663,
|
| 209441 |
+
"step": 86515
|
| 209442 |
+
},
|
| 209443 |
+
{
|
| 209444 |
+
"epoch": 691.26,
|
| 209445 |
+
"learning_rate": 8.61896607431341e-06,
|
| 209446 |
+
"loss": 0.2239,
|
| 209447 |
+
"step": 86520
|
| 209448 |
+
},
|
| 209449 |
+
{
|
| 209450 |
+
"epoch": 691.3,
|
| 209451 |
+
"learning_rate": 8.618885298869143e-06,
|
| 209452 |
+
"loss": 0.2704,
|
| 209453 |
+
"step": 86525
|
| 209454 |
+
},
|
| 209455 |
+
{
|
| 209456 |
+
"epoch": 691.34,
|
| 209457 |
+
"learning_rate": 8.61880452342488e-06,
|
| 209458 |
+
"loss": 0.3656,
|
| 209459 |
+
"step": 86530
|
| 209460 |
+
},
|
| 209461 |
+
{
|
| 209462 |
+
"epoch": 691.38,
|
| 209463 |
+
"learning_rate": 8.618723747980613e-06,
|
| 209464 |
+
"loss": 0.6328,
|
| 209465 |
+
"step": 86535
|
| 209466 |
+
},
|
| 209467 |
+
{
|
| 209468 |
+
"epoch": 691.42,
|
| 209469 |
+
"learning_rate": 8.61864297253635e-06,
|
| 209470 |
+
"loss": 0.8947,
|
| 209471 |
+
"step": 86540
|
| 209472 |
+
},
|
| 209473 |
+
{
|
| 209474 |
+
"epoch": 691.46,
|
| 209475 |
+
"learning_rate": 8.618562197092085e-06,
|
| 209476 |
+
"loss": 0.2775,
|
| 209477 |
+
"step": 86545
|
| 209478 |
+
},
|
| 209479 |
+
{
|
| 209480 |
+
"epoch": 691.5,
|
| 209481 |
+
"learning_rate": 8.61848142164782e-06,
|
| 209482 |
+
"loss": 0.3381,
|
| 209483 |
+
"step": 86550
|
| 209484 |
+
},
|
| 209485 |
+
{
|
| 209486 |
+
"epoch": 691.54,
|
| 209487 |
+
"learning_rate": 8.618400646203555e-06,
|
| 209488 |
+
"loss": 0.3351,
|
| 209489 |
+
"step": 86555
|
| 209490 |
+
},
|
| 209491 |
+
{
|
| 209492 |
+
"epoch": 691.58,
|
| 209493 |
+
"learning_rate": 8.61831987075929e-06,
|
| 209494 |
+
"loss": 0.6931,
|
| 209495 |
+
"step": 86560
|
| 209496 |
+
},
|
| 209497 |
+
{
|
| 209498 |
+
"epoch": 691.62,
|
| 209499 |
+
"learning_rate": 8.618239095315025e-06,
|
| 209500 |
+
"loss": 1.0264,
|
| 209501 |
+
"step": 86565
|
| 209502 |
+
},
|
| 209503 |
+
{
|
| 209504 |
+
"epoch": 691.66,
|
| 209505 |
+
"learning_rate": 8.61815831987076e-06,
|
| 209506 |
+
"loss": 0.3469,
|
| 209507 |
+
"step": 86570
|
| 209508 |
+
},
|
| 209509 |
+
{
|
| 209510 |
+
"epoch": 691.7,
|
| 209511 |
+
"learning_rate": 8.618077544426495e-06,
|
| 209512 |
+
"loss": 0.2688,
|
| 209513 |
+
"step": 86575
|
| 209514 |
+
},
|
| 209515 |
+
{
|
| 209516 |
+
"epoch": 691.74,
|
| 209517 |
+
"learning_rate": 8.61799676898223e-06,
|
| 209518 |
+
"loss": 0.3317,
|
| 209519 |
+
"step": 86580
|
| 209520 |
+
},
|
| 209521 |
+
{
|
| 209522 |
+
"epoch": 691.78,
|
| 209523 |
+
"learning_rate": 8.617915993537965e-06,
|
| 209524 |
+
"loss": 0.5662,
|
| 209525 |
+
"step": 86585
|
| 209526 |
+
},
|
| 209527 |
+
{
|
| 209528 |
+
"epoch": 691.82,
|
| 209529 |
+
"learning_rate": 8.6178352180937e-06,
|
| 209530 |
+
"loss": 0.9705,
|
| 209531 |
+
"step": 86590
|
| 209532 |
+
},
|
| 209533 |
+
{
|
| 209534 |
+
"epoch": 691.86,
|
| 209535 |
+
"learning_rate": 8.617754442649435e-06,
|
| 209536 |
+
"loss": 0.2945,
|
| 209537 |
+
"step": 86595
|
| 209538 |
+
},
|
| 209539 |
+
{
|
| 209540 |
+
"epoch": 691.9,
|
| 209541 |
+
"learning_rate": 8.61767366720517e-06,
|
| 209542 |
+
"loss": 0.284,
|
| 209543 |
+
"step": 86600
|
| 209544 |
+
},
|
| 209545 |
+
{
|
| 209546 |
+
"epoch": 691.94,
|
| 209547 |
+
"learning_rate": 8.617592891760905e-06,
|
| 209548 |
+
"loss": 0.5205,
|
| 209549 |
+
"step": 86605
|
| 209550 |
+
},
|
| 209551 |
+
{
|
| 209552 |
+
"epoch": 691.98,
|
| 209553 |
+
"learning_rate": 8.61751211631664e-06,
|
| 209554 |
+
"loss": 0.6463,
|
| 209555 |
+
"step": 86610
|
| 209556 |
+
},
|
| 209557 |
+
{
|
| 209558 |
+
"epoch": 692.0,
|
| 209559 |
+
"eval_loss": 0.39119741320610046,
|
| 209560 |
+
"eval_runtime": 42.4865,
|
| 209561 |
+
"eval_samples_per_second": 19.771,
|
| 209562 |
+
"eval_steps_per_second": 0.635,
|
| 209563 |
+
"eval_wer": 0.1851824656729569,
|
| 209564 |
+
"step": 86613
|
| 209565 |
+
},
|
| 209566 |
+
{
|
| 209567 |
+
"epoch": 692.02,
|
| 209568 |
+
"learning_rate": 8.617431340872375e-06,
|
| 209569 |
+
"loss": 0.3333,
|
| 209570 |
+
"step": 86615
|
| 209571 |
+
},
|
| 209572 |
+
{
|
| 209573 |
+
"epoch": 692.06,
|
| 209574 |
+
"learning_rate": 8.61735056542811e-06,
|
| 209575 |
+
"loss": 0.3179,
|
| 209576 |
+
"step": 86620
|
| 209577 |
+
},
|
| 209578 |
+
{
|
| 209579 |
+
"epoch": 692.1,
|
| 209580 |
+
"learning_rate": 8.617269789983845e-06,
|
| 209581 |
+
"loss": 0.2953,
|
| 209582 |
+
"step": 86625
|
| 209583 |
+
},
|
| 209584 |
+
{
|
| 209585 |
+
"epoch": 692.14,
|
| 209586 |
+
"learning_rate": 8.61718901453958e-06,
|
| 209587 |
+
"loss": 0.3652,
|
| 209588 |
+
"step": 86630
|
| 209589 |
+
},
|
| 209590 |
+
{
|
| 209591 |
+
"epoch": 692.18,
|
| 209592 |
+
"learning_rate": 8.617108239095315e-06,
|
| 209593 |
+
"loss": 0.6494,
|
| 209594 |
+
"step": 86635
|
| 209595 |
+
},
|
| 209596 |
+
{
|
| 209597 |
+
"epoch": 692.22,
|
| 209598 |
+
"learning_rate": 8.61702746365105e-06,
|
| 209599 |
+
"loss": 0.9831,
|
| 209600 |
+
"step": 86640
|
| 209601 |
+
},
|
| 209602 |
+
{
|
| 209603 |
+
"epoch": 692.26,
|
| 209604 |
+
"learning_rate": 8.616946688206785e-06,
|
| 209605 |
+
"loss": 0.3179,
|
| 209606 |
+
"step": 86645
|
| 209607 |
+
},
|
| 209608 |
+
{
|
| 209609 |
+
"epoch": 692.3,
|
| 209610 |
+
"learning_rate": 8.61686591276252e-06,
|
| 209611 |
+
"loss": 0.3476,
|
| 209612 |
+
"step": 86650
|
| 209613 |
+
},
|
| 209614 |
+
{
|
| 209615 |
+
"epoch": 692.34,
|
| 209616 |
+
"learning_rate": 8.616785137318257e-06,
|
| 209617 |
+
"loss": 0.3987,
|
| 209618 |
+
"step": 86655
|
| 209619 |
+
},
|
| 209620 |
+
{
|
| 209621 |
+
"epoch": 692.38,
|
| 209622 |
+
"learning_rate": 8.61670436187399e-06,
|
| 209623 |
+
"loss": 0.7323,
|
| 209624 |
+
"step": 86660
|
| 209625 |
+
},
|
| 209626 |
+
{
|
| 209627 |
+
"epoch": 692.42,
|
| 209628 |
+
"learning_rate": 8.616623586429727e-06,
|
| 209629 |
+
"loss": 1.2547,
|
| 209630 |
+
"step": 86665
|
| 209631 |
+
},
|
| 209632 |
+
{
|
| 209633 |
+
"epoch": 692.46,
|
| 209634 |
+
"learning_rate": 8.61654281098546e-06,
|
| 209635 |
+
"loss": 0.2914,
|
| 209636 |
+
"step": 86670
|
| 209637 |
+
},
|
| 209638 |
+
{
|
| 209639 |
+
"epoch": 692.5,
|
| 209640 |
+
"learning_rate": 8.616462035541196e-06,
|
| 209641 |
+
"loss": 0.2999,
|
| 209642 |
+
"step": 86675
|
| 209643 |
+
},
|
| 209644 |
+
{
|
| 209645 |
+
"epoch": 692.54,
|
| 209646 |
+
"learning_rate": 8.61638126009693e-06,
|
| 209647 |
+
"loss": 0.3679,
|
| 209648 |
+
"step": 86680
|
| 209649 |
+
},
|
| 209650 |
+
{
|
| 209651 |
+
"epoch": 692.58,
|
| 209652 |
+
"learning_rate": 8.616300484652666e-06,
|
| 209653 |
+
"loss": 0.6552,
|
| 209654 |
+
"step": 86685
|
| 209655 |
+
},
|
| 209656 |
+
{
|
| 209657 |
+
"epoch": 692.62,
|
| 209658 |
+
"learning_rate": 8.6162197092084e-06,
|
| 209659 |
+
"loss": 1.1263,
|
| 209660 |
+
"step": 86690
|
| 209661 |
+
},
|
| 209662 |
+
{
|
| 209663 |
+
"epoch": 692.66,
|
| 209664 |
+
"learning_rate": 8.616138933764136e-06,
|
| 209665 |
+
"loss": 0.2866,
|
| 209666 |
+
"step": 86695
|
| 209667 |
+
},
|
| 209668 |
+
{
|
| 209669 |
+
"epoch": 692.7,
|
| 209670 |
+
"learning_rate": 8.61605815831987e-06,
|
| 209671 |
+
"loss": 0.2683,
|
| 209672 |
+
"step": 86700
|
| 209673 |
+
},
|
| 209674 |
+
{
|
| 209675 |
+
"epoch": 692.74,
|
| 209676 |
+
"learning_rate": 8.615977382875606e-06,
|
| 209677 |
+
"loss": 0.3497,
|
| 209678 |
+
"step": 86705
|
| 209679 |
+
},
|
| 209680 |
+
{
|
| 209681 |
+
"epoch": 692.78,
|
| 209682 |
+
"learning_rate": 8.61589660743134e-06,
|
| 209683 |
+
"loss": 0.6652,
|
| 209684 |
+
"step": 86710
|
| 209685 |
+
},
|
| 209686 |
+
{
|
| 209687 |
+
"epoch": 692.82,
|
| 209688 |
+
"learning_rate": 8.615815831987076e-06,
|
| 209689 |
+
"loss": 1.145,
|
| 209690 |
+
"step": 86715
|
| 209691 |
+
},
|
| 209692 |
+
{
|
| 209693 |
+
"epoch": 692.86,
|
| 209694 |
+
"learning_rate": 8.615735056542812e-06,
|
| 209695 |
+
"loss": 0.2773,
|
| 209696 |
+
"step": 86720
|
| 209697 |
+
},
|
| 209698 |
+
{
|
| 209699 |
+
"epoch": 692.9,
|
| 209700 |
+
"learning_rate": 8.615654281098546e-06,
|
| 209701 |
+
"loss": 0.321,
|
| 209702 |
+
"step": 86725
|
| 209703 |
+
},
|
| 209704 |
+
{
|
| 209705 |
+
"epoch": 692.94,
|
| 209706 |
+
"learning_rate": 8.615573505654282e-06,
|
| 209707 |
+
"loss": 0.3368,
|
| 209708 |
+
"step": 86730
|
| 209709 |
+
},
|
| 209710 |
+
{
|
| 209711 |
+
"epoch": 692.98,
|
| 209712 |
+
"learning_rate": 8.615492730210016e-06,
|
| 209713 |
+
"loss": 0.6519,
|
| 209714 |
+
"step": 86735
|
| 209715 |
+
},
|
| 209716 |
+
{
|
| 209717 |
+
"epoch": 693.0,
|
| 209718 |
+
"eval_loss": 0.3798106014728546,
|
| 209719 |
+
"eval_runtime": 40.1988,
|
| 209720 |
+
"eval_samples_per_second": 20.896,
|
| 209721 |
+
"eval_steps_per_second": 0.672,
|
| 209722 |
+
"eval_wer": 0.19495548961424333,
|
| 209723 |
+
"step": 86738
|
| 209724 |
+
},
|
| 209725 |
+
{
|
| 209726 |
+
"epoch": 693.02,
|
| 209727 |
+
"learning_rate": 8.615411954765752e-06,
|
| 209728 |
+
"loss": 0.4544,
|
| 209729 |
+
"step": 86740
|
| 209730 |
+
},
|
| 209731 |
+
{
|
| 209732 |
+
"epoch": 693.06,
|
| 209733 |
+
"learning_rate": 8.615331179321486e-06,
|
| 209734 |
+
"loss": 0.2771,
|
| 209735 |
+
"step": 86745
|
| 209736 |
+
},
|
| 209737 |
+
{
|
| 209738 |
+
"epoch": 693.1,
|
| 209739 |
+
"learning_rate": 8.615250403877222e-06,
|
| 209740 |
+
"loss": 0.3675,
|
| 209741 |
+
"step": 86750
|
| 209742 |
+
},
|
| 209743 |
+
{
|
| 209744 |
+
"epoch": 693.14,
|
| 209745 |
+
"learning_rate": 8.615169628432956e-06,
|
| 209746 |
+
"loss": 0.3438,
|
| 209747 |
+
"step": 86755
|
| 209748 |
+
},
|
| 209749 |
+
{
|
| 209750 |
+
"epoch": 693.18,
|
| 209751 |
+
"learning_rate": 8.615088852988692e-06,
|
| 209752 |
+
"loss": 0.6171,
|
| 209753 |
+
"step": 86760
|
| 209754 |
+
},
|
| 209755 |
+
{
|
| 209756 |
+
"epoch": 693.22,
|
| 209757 |
+
"learning_rate": 8.615008077544426e-06,
|
| 209758 |
+
"loss": 1.0492,
|
| 209759 |
+
"step": 86765
|
| 209760 |
+
},
|
| 209761 |
+
{
|
| 209762 |
+
"epoch": 693.26,
|
| 209763 |
+
"learning_rate": 8.614927302100162e-06,
|
| 209764 |
+
"loss": 0.3273,
|
| 209765 |
+
"step": 86770
|
| 209766 |
+
},
|
| 209767 |
+
{
|
| 209768 |
+
"epoch": 693.3,
|
| 209769 |
+
"learning_rate": 8.614846526655898e-06,
|
| 209770 |
+
"loss": 0.3558,
|
| 209771 |
+
"step": 86775
|
| 209772 |
+
},
|
| 209773 |
+
{
|
| 209774 |
+
"epoch": 693.34,
|
| 209775 |
+
"learning_rate": 8.614765751211632e-06,
|
| 209776 |
+
"loss": 0.3772,
|
| 209777 |
+
"step": 86780
|
| 209778 |
+
},
|
| 209779 |
+
{
|
| 209780 |
+
"epoch": 693.38,
|
| 209781 |
+
"learning_rate": 8.614684975767368e-06,
|
| 209782 |
+
"loss": 0.6212,
|
| 209783 |
+
"step": 86785
|
| 209784 |
+
},
|
| 209785 |
+
{
|
| 209786 |
+
"epoch": 693.42,
|
| 209787 |
+
"learning_rate": 8.614604200323102e-06,
|
| 209788 |
+
"loss": 1.1256,
|
| 209789 |
+
"step": 86790
|
| 209790 |
+
},
|
| 209791 |
+
{
|
| 209792 |
+
"epoch": 693.46,
|
| 209793 |
+
"learning_rate": 8.614523424878838e-06,
|
| 209794 |
+
"loss": 0.3408,
|
| 209795 |
+
"step": 86795
|
| 209796 |
+
},
|
| 209797 |
+
{
|
| 209798 |
+
"epoch": 693.5,
|
| 209799 |
+
"learning_rate": 8.614442649434572e-06,
|
| 209800 |
+
"loss": 0.3713,
|
| 209801 |
+
"step": 86800
|
| 209802 |
+
},
|
| 209803 |
+
{
|
| 209804 |
+
"epoch": 693.54,
|
| 209805 |
+
"learning_rate": 8.614361873990308e-06,
|
| 209806 |
+
"loss": 0.3121,
|
| 209807 |
+
"step": 86805
|
| 209808 |
+
},
|
| 209809 |
+
{
|
| 209810 |
+
"epoch": 693.58,
|
| 209811 |
+
"learning_rate": 8.614281098546042e-06,
|
| 209812 |
+
"loss": 0.6541,
|
| 209813 |
+
"step": 86810
|
| 209814 |
+
},
|
| 209815 |
+
{
|
| 209816 |
+
"epoch": 693.62,
|
| 209817 |
+
"learning_rate": 8.614200323101778e-06,
|
| 209818 |
+
"loss": 1.1208,
|
| 209819 |
+
"step": 86815
|
| 209820 |
+
},
|
| 209821 |
+
{
|
| 209822 |
+
"epoch": 693.66,
|
| 209823 |
+
"learning_rate": 8.614119547657512e-06,
|
| 209824 |
+
"loss": 0.3341,
|
| 209825 |
+
"step": 86820
|
| 209826 |
+
},
|
| 209827 |
+
{
|
| 209828 |
+
"epoch": 693.7,
|
| 209829 |
+
"learning_rate": 8.614038772213248e-06,
|
| 209830 |
+
"loss": 0.3526,
|
| 209831 |
+
"step": 86825
|
| 209832 |
+
},
|
| 209833 |
+
{
|
| 209834 |
+
"epoch": 693.74,
|
| 209835 |
+
"learning_rate": 8.613957996768984e-06,
|
| 209836 |
+
"loss": 0.3218,
|
| 209837 |
+
"step": 86830
|
| 209838 |
+
},
|
| 209839 |
+
{
|
| 209840 |
+
"epoch": 693.78,
|
| 209841 |
+
"learning_rate": 8.613877221324718e-06,
|
| 209842 |
+
"loss": 0.635,
|
| 209843 |
+
"step": 86835
|
| 209844 |
+
},
|
| 209845 |
+
{
|
| 209846 |
+
"epoch": 693.82,
|
| 209847 |
+
"learning_rate": 8.613796445880454e-06,
|
| 209848 |
+
"loss": 0.9507,
|
| 209849 |
+
"step": 86840
|
| 209850 |
+
},
|
| 209851 |
+
{
|
| 209852 |
+
"epoch": 693.86,
|
| 209853 |
+
"learning_rate": 8.613715670436188e-06,
|
| 209854 |
+
"loss": 0.2826,
|
| 209855 |
+
"step": 86845
|
| 209856 |
+
},
|
| 209857 |
+
{
|
| 209858 |
+
"epoch": 693.9,
|
| 209859 |
+
"learning_rate": 8.613634894991924e-06,
|
| 209860 |
+
"loss": 0.28,
|
| 209861 |
+
"step": 86850
|
| 209862 |
+
},
|
| 209863 |
+
{
|
| 209864 |
+
"epoch": 693.94,
|
| 209865 |
+
"learning_rate": 8.613554119547658e-06,
|
| 209866 |
+
"loss": 0.4393,
|
| 209867 |
+
"step": 86855
|
| 209868 |
+
},
|
| 209869 |
+
{
|
| 209870 |
+
"epoch": 693.98,
|
| 209871 |
+
"learning_rate": 8.613473344103394e-06,
|
| 209872 |
+
"loss": 0.6824,
|
| 209873 |
+
"step": 86860
|
| 209874 |
+
},
|
| 209875 |
+
{
|
| 209876 |
+
"epoch": 694.0,
|
| 209877 |
+
"eval_loss": 0.4369931221008301,
|
| 209878 |
+
"eval_runtime": 43.2065,
|
| 209879 |
+
"eval_samples_per_second": 19.442,
|
| 209880 |
+
"eval_steps_per_second": 0.625,
|
| 209881 |
+
"eval_wer": 0.18327764056370768,
|
| 209882 |
+
"step": 86863
|
| 209883 |
+
},
|
| 209884 |
+
{
|
| 209885 |
+
"epoch": 694.02,
|
| 209886 |
+
"learning_rate": 8.613392568659128e-06,
|
| 209887 |
+
"loss": 0.3783,
|
| 209888 |
+
"step": 86865
|
| 209889 |
+
},
|
| 209890 |
+
{
|
| 209891 |
+
"epoch": 694.06,
|
| 209892 |
+
"learning_rate": 8.613311793214864e-06,
|
| 209893 |
+
"loss": 0.3584,
|
| 209894 |
+
"step": 86870
|
| 209895 |
+
},
|
| 209896 |
+
{
|
| 209897 |
+
"epoch": 694.1,
|
| 209898 |
+
"learning_rate": 8.613231017770598e-06,
|
| 209899 |
+
"loss": 0.3092,
|
| 209900 |
+
"step": 86875
|
| 209901 |
+
},
|
| 209902 |
+
{
|
| 209903 |
+
"epoch": 694.14,
|
| 209904 |
+
"learning_rate": 8.613150242326334e-06,
|
| 209905 |
+
"loss": 0.3691,
|
| 209906 |
+
"step": 86880
|
| 209907 |
+
},
|
| 209908 |
+
{
|
| 209909 |
+
"epoch": 694.18,
|
| 209910 |
+
"learning_rate": 8.613069466882068e-06,
|
| 209911 |
+
"loss": 0.6162,
|
| 209912 |
+
"step": 86885
|
| 209913 |
+
},
|
| 209914 |
+
{
|
| 209915 |
+
"epoch": 694.22,
|
| 209916 |
+
"learning_rate": 8.612988691437803e-06,
|
| 209917 |
+
"loss": 0.9997,
|
| 209918 |
+
"step": 86890
|
| 209919 |
+
},
|
| 209920 |
+
{
|
| 209921 |
+
"epoch": 694.26,
|
| 209922 |
+
"learning_rate": 8.61290791599354e-06,
|
| 209923 |
+
"loss": 0.2906,
|
| 209924 |
+
"step": 86895
|
| 209925 |
+
},
|
| 209926 |
+
{
|
| 209927 |
+
"epoch": 694.3,
|
| 209928 |
+
"learning_rate": 8.612827140549273e-06,
|
| 209929 |
+
"loss": 0.3346,
|
| 209930 |
+
"step": 86900
|
| 209931 |
+
},
|
| 209932 |
+
{
|
| 209933 |
+
"epoch": 694.34,
|
| 209934 |
+
"learning_rate": 8.61274636510501e-06,
|
| 209935 |
+
"loss": 0.3532,
|
| 209936 |
+
"step": 86905
|
| 209937 |
+
},
|
| 209938 |
+
{
|
| 209939 |
+
"epoch": 694.38,
|
| 209940 |
+
"learning_rate": 8.612665589660743e-06,
|
| 209941 |
+
"loss": 0.7297,
|
| 209942 |
+
"step": 86910
|
| 209943 |
+
},
|
| 209944 |
+
{
|
| 209945 |
+
"epoch": 694.42,
|
| 209946 |
+
"learning_rate": 8.61258481421648e-06,
|
| 209947 |
+
"loss": 0.9558,
|
| 209948 |
+
"step": 86915
|
| 209949 |
+
},
|
| 209950 |
+
{
|
| 209951 |
+
"epoch": 694.46,
|
| 209952 |
+
"learning_rate": 8.612504038772213e-06,
|
| 209953 |
+
"loss": 0.2938,
|
| 209954 |
+
"step": 86920
|
| 209955 |
+
},
|
| 209956 |
+
{
|
| 209957 |
+
"epoch": 694.5,
|
| 209958 |
+
"learning_rate": 8.61242326332795e-06,
|
| 209959 |
+
"loss": 0.274,
|
| 209960 |
+
"step": 86925
|
| 209961 |
+
},
|
| 209962 |
+
{
|
| 209963 |
+
"epoch": 694.54,
|
| 209964 |
+
"learning_rate": 8.612342487883683e-06,
|
| 209965 |
+
"loss": 0.3146,
|
| 209966 |
+
"step": 86930
|
| 209967 |
+
},
|
| 209968 |
+
{
|
| 209969 |
+
"epoch": 694.58,
|
| 209970 |
+
"learning_rate": 8.61226171243942e-06,
|
| 209971 |
+
"loss": 0.6603,
|
| 209972 |
+
"step": 86935
|
| 209973 |
+
},
|
| 209974 |
+
{
|
| 209975 |
+
"epoch": 694.62,
|
| 209976 |
+
"learning_rate": 8.612180936995153e-06,
|
| 209977 |
+
"loss": 1.0825,
|
| 209978 |
+
"step": 86940
|
| 209979 |
+
},
|
| 209980 |
+
{
|
| 209981 |
+
"epoch": 694.66,
|
| 209982 |
+
"learning_rate": 8.61210016155089e-06,
|
| 209983 |
+
"loss": 0.2996,
|
| 209984 |
+
"step": 86945
|
| 209985 |
+
},
|
| 209986 |
+
{
|
| 209987 |
+
"epoch": 694.7,
|
| 209988 |
+
"learning_rate": 8.612019386106625e-06,
|
| 209989 |
+
"loss": 0.3106,
|
| 209990 |
+
"step": 86950
|
| 209991 |
+
},
|
| 209992 |
+
{
|
| 209993 |
+
"epoch": 694.74,
|
| 209994 |
+
"learning_rate": 8.61193861066236e-06,
|
| 209995 |
+
"loss": 0.3602,
|
| 209996 |
+
"step": 86955
|
| 209997 |
+
},
|
| 209998 |
+
{
|
| 209999 |
+
"epoch": 694.78,
|
| 210000 |
+
"learning_rate": 8.611857835218095e-06,
|
| 210001 |
+
"loss": 0.634,
|
| 210002 |
+
"step": 86960
|
| 210003 |
+
},
|
| 210004 |
+
{
|
| 210005 |
+
"epoch": 694.82,
|
| 210006 |
+
"learning_rate": 8.61177705977383e-06,
|
| 210007 |
+
"loss": 1.0922,
|
| 210008 |
+
"step": 86965
|
| 210009 |
+
},
|
| 210010 |
+
{
|
| 210011 |
+
"epoch": 694.86,
|
| 210012 |
+
"learning_rate": 8.611696284329565e-06,
|
| 210013 |
+
"loss": 0.3165,
|
| 210014 |
+
"step": 86970
|
| 210015 |
+
},
|
| 210016 |
+
{
|
| 210017 |
+
"epoch": 694.9,
|
| 210018 |
+
"learning_rate": 8.611615508885299e-06,
|
| 210019 |
+
"loss": 0.3214,
|
| 210020 |
+
"step": 86975
|
| 210021 |
+
},
|
| 210022 |
+
{
|
| 210023 |
+
"epoch": 694.94,
|
| 210024 |
+
"learning_rate": 8.611534733441035e-06,
|
| 210025 |
+
"loss": 0.4198,
|
| 210026 |
+
"step": 86980
|
| 210027 |
+
},
|
| 210028 |
+
{
|
| 210029 |
+
"epoch": 694.98,
|
| 210030 |
+
"learning_rate": 8.611453957996769e-06,
|
| 210031 |
+
"loss": 0.6903,
|
| 210032 |
+
"step": 86985
|
| 210033 |
+
},
|
| 210034 |
+
{
|
| 210035 |
+
"epoch": 695.0,
|
| 210036 |
+
"eval_loss": 0.33145707845687866,
|
| 210037 |
+
"eval_runtime": 42.8641,
|
| 210038 |
+
"eval_samples_per_second": 19.597,
|
| 210039 |
+
"eval_steps_per_second": 0.63,
|
| 210040 |
+
"eval_wer": 0.1826943671021632,
|
| 210041 |
+
"step": 86988
|
| 210042 |
+
},
|
| 210043 |
+
{
|
| 210044 |
+
"epoch": 695.02,
|
| 210045 |
+
"learning_rate": 8.611373182552505e-06,
|
| 210046 |
+
"loss": 0.3287,
|
| 210047 |
+
"step": 86990
|
| 210048 |
+
},
|
| 210049 |
+
{
|
| 210050 |
+
"epoch": 695.06,
|
| 210051 |
+
"learning_rate": 8.611292407108239e-06,
|
| 210052 |
+
"loss": 0.3216,
|
| 210053 |
+
"step": 86995
|
| 210054 |
+
},
|
| 210055 |
+
{
|
| 210056 |
+
"epoch": 695.1,
|
| 210057 |
+
"learning_rate": 8.611211631663975e-06,
|
| 210058 |
+
"loss": 0.2636,
|
| 210059 |
+
"step": 87000
|
| 210060 |
+
},
|
| 210061 |
+
{
|
| 210062 |
+
"epoch": 695.14,
|
| 210063 |
+
"learning_rate": 8.61113085621971e-06,
|
| 210064 |
+
"loss": 0.4143,
|
| 210065 |
+
"step": 87005
|
| 210066 |
+
},
|
| 210067 |
+
{
|
| 210068 |
+
"epoch": 695.18,
|
| 210069 |
+
"learning_rate": 8.611050080775445e-06,
|
| 210070 |
+
"loss": 0.7097,
|
| 210071 |
+
"step": 87010
|
| 210072 |
+
},
|
| 210073 |
+
{
|
| 210074 |
+
"epoch": 695.22,
|
| 210075 |
+
"learning_rate": 8.61096930533118e-06,
|
| 210076 |
+
"loss": 0.9721,
|
| 210077 |
+
"step": 87015
|
| 210078 |
+
},
|
| 210079 |
+
{
|
| 210080 |
+
"epoch": 695.26,
|
| 210081 |
+
"learning_rate": 8.610888529886915e-06,
|
| 210082 |
+
"loss": 0.2953,
|
| 210083 |
+
"step": 87020
|
| 210084 |
+
},
|
| 210085 |
+
{
|
| 210086 |
+
"epoch": 695.3,
|
| 210087 |
+
"learning_rate": 8.61080775444265e-06,
|
| 210088 |
+
"loss": 0.3665,
|
| 210089 |
+
"step": 87025
|
| 210090 |
+
},
|
| 210091 |
+
{
|
| 210092 |
+
"epoch": 695.34,
|
| 210093 |
+
"learning_rate": 8.610726978998385e-06,
|
| 210094 |
+
"loss": 0.3368,
|
| 210095 |
+
"step": 87030
|
| 210096 |
+
},
|
| 210097 |
+
{
|
| 210098 |
+
"epoch": 695.38,
|
| 210099 |
+
"learning_rate": 8.61064620355412e-06,
|
| 210100 |
+
"loss": 0.7114,
|
| 210101 |
+
"step": 87035
|
| 210102 |
+
},
|
| 210103 |
+
{
|
| 210104 |
+
"epoch": 695.42,
|
| 210105 |
+
"learning_rate": 8.610565428109855e-06,
|
| 210106 |
+
"loss": 1.138,
|
| 210107 |
+
"step": 87040
|
| 210108 |
+
},
|
| 210109 |
+
{
|
| 210110 |
+
"epoch": 695.46,
|
| 210111 |
+
"learning_rate": 8.61048465266559e-06,
|
| 210112 |
+
"loss": 0.2662,
|
| 210113 |
+
"step": 87045
|
| 210114 |
+
},
|
| 210115 |
+
{
|
| 210116 |
+
"epoch": 695.5,
|
| 210117 |
+
"learning_rate": 8.610403877221325e-06,
|
| 210118 |
+
"loss": 0.2978,
|
| 210119 |
+
"step": 87050
|
| 210120 |
+
},
|
| 210121 |
+
{
|
| 210122 |
+
"epoch": 695.54,
|
| 210123 |
+
"learning_rate": 8.61032310177706e-06,
|
| 210124 |
+
"loss": 0.4894,
|
| 210125 |
+
"step": 87055
|
| 210126 |
+
},
|
| 210127 |
+
{
|
| 210128 |
+
"epoch": 695.58,
|
| 210129 |
+
"learning_rate": 8.610242326332795e-06,
|
| 210130 |
+
"loss": 0.7232,
|
| 210131 |
+
"step": 87060
|
| 210132 |
+
},
|
| 210133 |
+
{
|
| 210134 |
+
"epoch": 695.62,
|
| 210135 |
+
"learning_rate": 8.61016155088853e-06,
|
| 210136 |
+
"loss": 1.1415,
|
| 210137 |
+
"step": 87065
|
| 210138 |
+
},
|
| 210139 |
+
{
|
| 210140 |
+
"epoch": 695.66,
|
| 210141 |
+
"learning_rate": 8.610080775444266e-06,
|
| 210142 |
+
"loss": 0.2633,
|
| 210143 |
+
"step": 87070
|
| 210144 |
+
},
|
| 210145 |
+
{
|
| 210146 |
+
"epoch": 695.7,
|
| 210147 |
+
"learning_rate": 8.61e-06,
|
| 210148 |
+
"loss": 0.3416,
|
| 210149 |
+
"step": 87075
|
| 210150 |
+
},
|
| 210151 |
+
{
|
| 210152 |
+
"epoch": 695.74,
|
| 210153 |
+
"learning_rate": 8.609919224555736e-06,
|
| 210154 |
+
"loss": 0.3643,
|
| 210155 |
+
"step": 87080
|
| 210156 |
+
},
|
| 210157 |
+
{
|
| 210158 |
+
"epoch": 695.78,
|
| 210159 |
+
"learning_rate": 8.60983844911147e-06,
|
| 210160 |
+
"loss": 0.7117,
|
| 210161 |
+
"step": 87085
|
| 210162 |
+
},
|
| 210163 |
+
{
|
| 210164 |
+
"epoch": 695.82,
|
| 210165 |
+
"learning_rate": 8.609757673667206e-06,
|
| 210166 |
+
"loss": 1.1024,
|
| 210167 |
+
"step": 87090
|
| 210168 |
+
},
|
| 210169 |
+
{
|
| 210170 |
+
"epoch": 695.86,
|
| 210171 |
+
"learning_rate": 8.60967689822294e-06,
|
| 210172 |
+
"loss": 0.2822,
|
| 210173 |
+
"step": 87095
|
| 210174 |
+
},
|
| 210175 |
+
{
|
| 210176 |
+
"epoch": 695.9,
|
| 210177 |
+
"learning_rate": 8.609596122778676e-06,
|
| 210178 |
+
"loss": 0.2877,
|
| 210179 |
+
"step": 87100
|
| 210180 |
+
},
|
| 210181 |
+
{
|
| 210182 |
+
"epoch": 695.94,
|
| 210183 |
+
"learning_rate": 8.60951534733441e-06,
|
| 210184 |
+
"loss": 0.3892,
|
| 210185 |
+
"step": 87105
|
| 210186 |
+
},
|
| 210187 |
+
{
|
| 210188 |
+
"epoch": 695.98,
|
| 210189 |
+
"learning_rate": 8.609434571890146e-06,
|
| 210190 |
+
"loss": 0.611,
|
| 210191 |
+
"step": 87110
|
| 210192 |
+
},
|
| 210193 |
+
{
|
| 210194 |
+
"epoch": 696.0,
|
| 210195 |
+
"eval_loss": 0.3427501618862152,
|
| 210196 |
+
"eval_runtime": 41.411,
|
| 210197 |
+
"eval_samples_per_second": 20.26,
|
| 210198 |
+
"eval_steps_per_second": 0.652,
|
| 210199 |
+
"eval_wer": 0.18761075014766687,
|
| 210200 |
+
"step": 87113
|
| 210201 |
}
|
| 210202 |
],
|
| 210203 |
+
"max_steps": 625000,
|
| 210204 |
"num_train_epochs": 5000,
|
| 210205 |
+
"total_flos": 2.4515887698594654e+20,
|
| 210206 |
"trial_name": null,
|
| 210207 |
"trial_params": null
|
| 210208 |
}
|
model-bin/finetune/base/{checkpoint-86488 → checkpoint-87113}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629928346.7422721/events.out.tfevents.1629928346.7e498afd5545.7645.165
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:41423bd5fa5c9d1c3211ce8cdbd4631edc46706e3a4a5636825256a3a50f9af6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629928816.2265396/events.out.tfevents.1629928816.7e498afd5545.7645.167
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:85b0d6fbaf1dd6e1f9f9c389e89afba0713a8c4ba899d3b3e34cb5aaabe8418e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629929279.5633097/events.out.tfevents.1629929279.7e498afd5545.7645.169
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:260cd6814b90e142e504e6d80cd12a6797ea206ad5caa09a01b409dfd1cf3408
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629929743.3866482/events.out.tfevents.1629929743.7e498afd5545.7645.171
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:efb8fcc91c0c841616d1bf169dce61c536cbc5015928a5956f7db126a9ee6faa
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629930214.058052/events.out.tfevents.1629930214.7e498afd5545.7645.173
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f8ddbf6dee37a465f51abfa8a82d58096ecde3f6fd4134a0ae0a56674f091fd5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629928346.7e498afd5545.7645.164
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:56e1a028e1a71bd02f2bcd85552205bd95d83069190177272c717441ab6e8e20
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629928816.7e498afd5545.7645.166
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e2fa522d7dd04a1287281059e9f8ca0ee361cdebb92d082f85411b3878c2f06a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629929279.7e498afd5545.7645.168
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:15222be30a5eed8ad5a388deda9a431defcf5f4a13b4a00820455aadd3108e14
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629929743.7e498afd5545.7645.170
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a0caa33dee4b59411bf007844acfe49f4bac168198227974a1772fb904018f99
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629930214.7e498afd5545.7645.172
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:85918693ef6a677aae434f385abf79e52dca789333c92d93f417bca7161d84c1
|
| 3 |
+
size 8622
|