"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630132562.2337933/events.out.tfevents.1630132562.86bb0ddabf9b.4092.51 +3 -0
- model-bin/finetune/base/log/1630132940.836788/events.out.tfevents.1630132940.86bb0ddabf9b.4092.53 +3 -0
- model-bin/finetune/base/log/1630133330.209003/events.out.tfevents.1630133330.86bb0ddabf9b.4092.55 +3 -0
- model-bin/finetune/base/log/1630133714.3736/events.out.tfevents.1630133714.86bb0ddabf9b.4092.57 +3 -0
- model-bin/finetune/base/log/1630134097.5583656/events.out.tfevents.1630134097.86bb0ddabf9b.4092.59 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630132562.86bb0ddabf9b.4092.50 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630132940.86bb0ddabf9b.4092.52 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630133330.86bb0ddabf9b.4092.54 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630133714.86bb0ddabf9b.4092.56 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630134097.86bb0ddabf9b.4092.58 +3 -0
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/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 sha256:51fcbfe601df65783aad9ed90b143196bb9116032626928bb4e87c5585f1c83a
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/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:d3904af36b18b4d095fb0124e026e85f54816e0865cc1484799d44b454c39030
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/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:c1fa1ccd4d088fa375b9751da87ad6f087ce1ff3bd1852b863df97283f7f7a5c
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/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 sha256:2e6d50a58f036c734ac486a1b8b8de8f2a72641d68799526e48e85c9dc1db370
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/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:ecb97a7d6b086ed775dfd796d92ea4da8cacd42ec572a22c69cec1b8c8c780d8
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -253419,11 +253419,806 @@
|
|
| 253419 |
"eval_steps_per_second": 0.763,
|
| 253420 |
"eval_wer": 0.18734491315136476,
|
| 253421 |
"step": 121084
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 253422 |
}
|
| 253423 |
],
|
| 253424 |
"max_steps": 625000,
|
| 253425 |
"num_train_epochs": 5000,
|
| 253426 |
-
"total_flos": 3.
|
| 253427 |
"trial_name": null,
|
| 253428 |
"trial_params": null
|
| 253429 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 973.0,
|
| 5 |
+
"global_step": 121707,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 253419 |
"eval_steps_per_second": 0.763,
|
| 253420 |
"eval_wer": 0.18734491315136476,
|
| 253421 |
"step": 121084
|
| 253422 |
+
},
|
| 253423 |
+
{
|
| 253424 |
+
"epoch": 976.01,
|
| 253425 |
+
"learning_rate": 8.076314102564103e-06,
|
| 253426 |
+
"loss": 0.4885,
|
| 253427 |
+
"step": 121085
|
| 253428 |
+
},
|
| 253429 |
+
{
|
| 253430 |
+
"epoch": 976.05,
|
| 253431 |
+
"learning_rate": 8.076233974358975e-06,
|
| 253432 |
+
"loss": 0.2731,
|
| 253433 |
+
"step": 121090
|
| 253434 |
+
},
|
| 253435 |
+
{
|
| 253436 |
+
"epoch": 976.09,
|
| 253437 |
+
"learning_rate": 8.076153846153846e-06,
|
| 253438 |
+
"loss": 0.2863,
|
| 253439 |
+
"step": 121095
|
| 253440 |
+
},
|
| 253441 |
+
{
|
| 253442 |
+
"epoch": 976.13,
|
| 253443 |
+
"learning_rate": 8.076073717948717e-06,
|
| 253444 |
+
"loss": 0.3848,
|
| 253445 |
+
"step": 121100
|
| 253446 |
+
},
|
| 253447 |
+
{
|
| 253448 |
+
"epoch": 976.17,
|
| 253449 |
+
"learning_rate": 8.07599358974359e-06,
|
| 253450 |
+
"loss": 0.5399,
|
| 253451 |
+
"step": 121105
|
| 253452 |
+
},
|
| 253453 |
+
{
|
| 253454 |
+
"epoch": 976.21,
|
| 253455 |
+
"learning_rate": 8.075913461538462e-06,
|
| 253456 |
+
"loss": 1.2772,
|
| 253457 |
+
"step": 121110
|
| 253458 |
+
},
|
| 253459 |
+
{
|
| 253460 |
+
"epoch": 976.25,
|
| 253461 |
+
"learning_rate": 8.075833333333333e-06,
|
| 253462 |
+
"loss": 0.3456,
|
| 253463 |
+
"step": 121115
|
| 253464 |
+
},
|
| 253465 |
+
{
|
| 253466 |
+
"epoch": 976.29,
|
| 253467 |
+
"learning_rate": 8.075753205128206e-06,
|
| 253468 |
+
"loss": 0.2335,
|
| 253469 |
+
"step": 121120
|
| 253470 |
+
},
|
| 253471 |
+
{
|
| 253472 |
+
"epoch": 976.33,
|
| 253473 |
+
"learning_rate": 8.075673076923078e-06,
|
| 253474 |
+
"loss": 0.311,
|
| 253475 |
+
"step": 121125
|
| 253476 |
+
},
|
| 253477 |
+
{
|
| 253478 |
+
"epoch": 976.37,
|
| 253479 |
+
"learning_rate": 8.075592948717949e-06,
|
| 253480 |
+
"loss": 0.5154,
|
| 253481 |
+
"step": 121130
|
| 253482 |
+
},
|
| 253483 |
+
{
|
| 253484 |
+
"epoch": 976.41,
|
| 253485 |
+
"learning_rate": 8.07551282051282e-06,
|
| 253486 |
+
"loss": 1.08,
|
| 253487 |
+
"step": 121135
|
| 253488 |
+
},
|
| 253489 |
+
{
|
| 253490 |
+
"epoch": 976.45,
|
| 253491 |
+
"learning_rate": 8.075432692307693e-06,
|
| 253492 |
+
"loss": 0.3043,
|
| 253493 |
+
"step": 121140
|
| 253494 |
+
},
|
| 253495 |
+
{
|
| 253496 |
+
"epoch": 976.49,
|
| 253497 |
+
"learning_rate": 8.075352564102565e-06,
|
| 253498 |
+
"loss": 0.2545,
|
| 253499 |
+
"step": 121145
|
| 253500 |
+
},
|
| 253501 |
+
{
|
| 253502 |
+
"epoch": 976.53,
|
| 253503 |
+
"learning_rate": 8.075272435897436e-06,
|
| 253504 |
+
"loss": 0.3678,
|
| 253505 |
+
"step": 121150
|
| 253506 |
+
},
|
| 253507 |
+
{
|
| 253508 |
+
"epoch": 976.57,
|
| 253509 |
+
"learning_rate": 8.07519230769231e-06,
|
| 253510 |
+
"loss": 0.4868,
|
| 253511 |
+
"step": 121155
|
| 253512 |
+
},
|
| 253513 |
+
{
|
| 253514 |
+
"epoch": 976.61,
|
| 253515 |
+
"learning_rate": 8.07511217948718e-06,
|
| 253516 |
+
"loss": 1.1736,
|
| 253517 |
+
"step": 121160
|
| 253518 |
+
},
|
| 253519 |
+
{
|
| 253520 |
+
"epoch": 976.65,
|
| 253521 |
+
"learning_rate": 8.075032051282052e-06,
|
| 253522 |
+
"loss": 0.2969,
|
| 253523 |
+
"step": 121165
|
| 253524 |
+
},
|
| 253525 |
+
{
|
| 253526 |
+
"epoch": 976.69,
|
| 253527 |
+
"learning_rate": 8.074951923076923e-06,
|
| 253528 |
+
"loss": 0.3378,
|
| 253529 |
+
"step": 121170
|
| 253530 |
+
},
|
| 253531 |
+
{
|
| 253532 |
+
"epoch": 976.73,
|
| 253533 |
+
"learning_rate": 8.074871794871796e-06,
|
| 253534 |
+
"loss": 0.31,
|
| 253535 |
+
"step": 121175
|
| 253536 |
+
},
|
| 253537 |
+
{
|
| 253538 |
+
"epoch": 976.77,
|
| 253539 |
+
"learning_rate": 8.074791666666668e-06,
|
| 253540 |
+
"loss": 0.521,
|
| 253541 |
+
"step": 121180
|
| 253542 |
+
},
|
| 253543 |
+
{
|
| 253544 |
+
"epoch": 976.81,
|
| 253545 |
+
"learning_rate": 8.074711538461539e-06,
|
| 253546 |
+
"loss": 1.0544,
|
| 253547 |
+
"step": 121185
|
| 253548 |
+
},
|
| 253549 |
+
{
|
| 253550 |
+
"epoch": 976.85,
|
| 253551 |
+
"learning_rate": 8.07463141025641e-06,
|
| 253552 |
+
"loss": 0.291,
|
| 253553 |
+
"step": 121190
|
| 253554 |
+
},
|
| 253555 |
+
{
|
| 253556 |
+
"epoch": 976.9,
|
| 253557 |
+
"learning_rate": 8.074551282051283e-06,
|
| 253558 |
+
"loss": 0.2925,
|
| 253559 |
+
"step": 121195
|
| 253560 |
+
},
|
| 253561 |
+
{
|
| 253562 |
+
"epoch": 976.94,
|
| 253563 |
+
"learning_rate": 8.074471153846155e-06,
|
| 253564 |
+
"loss": 0.3824,
|
| 253565 |
+
"step": 121200
|
| 253566 |
+
},
|
| 253567 |
+
{
|
| 253568 |
+
"epoch": 976.98,
|
| 253569 |
+
"learning_rate": 8.074391025641026e-06,
|
| 253570 |
+
"loss": 0.5301,
|
| 253571 |
+
"step": 121205
|
| 253572 |
+
},
|
| 253573 |
+
{
|
| 253574 |
+
"epoch": 977.0,
|
| 253575 |
+
"eval_loss": 0.4441799521446228,
|
| 253576 |
+
"eval_runtime": 35.4497,
|
| 253577 |
+
"eval_samples_per_second": 23.752,
|
| 253578 |
+
"eval_steps_per_second": 0.762,
|
| 253579 |
+
"eval_wer": 0.17582178361512227,
|
| 253580 |
+
"step": 121208
|
| 253581 |
+
},
|
| 253582 |
+
{
|
| 253583 |
+
"epoch": 969.02,
|
| 253584 |
+
"learning_rate": 8.0743108974359e-06,
|
| 253585 |
+
"loss": 0.3701,
|
| 253586 |
+
"step": 121210
|
| 253587 |
+
},
|
| 253588 |
+
{
|
| 253589 |
+
"epoch": 969.06,
|
| 253590 |
+
"learning_rate": 8.07423076923077e-06,
|
| 253591 |
+
"loss": 0.3003,
|
| 253592 |
+
"step": 121215
|
| 253593 |
+
},
|
| 253594 |
+
{
|
| 253595 |
+
"epoch": 969.1,
|
| 253596 |
+
"learning_rate": 8.074150641025642e-06,
|
| 253597 |
+
"loss": 0.2455,
|
| 253598 |
+
"step": 121220
|
| 253599 |
+
},
|
| 253600 |
+
{
|
| 253601 |
+
"epoch": 969.14,
|
| 253602 |
+
"learning_rate": 8.074070512820513e-06,
|
| 253603 |
+
"loss": 0.311,
|
| 253604 |
+
"step": 121225
|
| 253605 |
+
},
|
| 253606 |
+
{
|
| 253607 |
+
"epoch": 969.18,
|
| 253608 |
+
"learning_rate": 8.073990384615386e-06,
|
| 253609 |
+
"loss": 0.6479,
|
| 253610 |
+
"step": 121230
|
| 253611 |
+
},
|
| 253612 |
+
{
|
| 253613 |
+
"epoch": 969.22,
|
| 253614 |
+
"learning_rate": 8.073910256410256e-06,
|
| 253615 |
+
"loss": 1.1652,
|
| 253616 |
+
"step": 121235
|
| 253617 |
+
},
|
| 253618 |
+
{
|
| 253619 |
+
"epoch": 969.26,
|
| 253620 |
+
"learning_rate": 8.073830128205129e-06,
|
| 253621 |
+
"loss": 0.3719,
|
| 253622 |
+
"step": 121240
|
| 253623 |
+
},
|
| 253624 |
+
{
|
| 253625 |
+
"epoch": 969.3,
|
| 253626 |
+
"learning_rate": 8.07375e-06,
|
| 253627 |
+
"loss": 0.2902,
|
| 253628 |
+
"step": 121245
|
| 253629 |
+
},
|
| 253630 |
+
{
|
| 253631 |
+
"epoch": 969.34,
|
| 253632 |
+
"learning_rate": 8.073669871794872e-06,
|
| 253633 |
+
"loss": 0.3676,
|
| 253634 |
+
"step": 121250
|
| 253635 |
+
},
|
| 253636 |
+
{
|
| 253637 |
+
"epoch": 969.38,
|
| 253638 |
+
"learning_rate": 8.073589743589745e-06,
|
| 253639 |
+
"loss": 0.6999,
|
| 253640 |
+
"step": 121255
|
| 253641 |
+
},
|
| 253642 |
+
{
|
| 253643 |
+
"epoch": 969.42,
|
| 253644 |
+
"learning_rate": 8.073509615384616e-06,
|
| 253645 |
+
"loss": 1.0091,
|
| 253646 |
+
"step": 121260
|
| 253647 |
+
},
|
| 253648 |
+
{
|
| 253649 |
+
"epoch": 969.46,
|
| 253650 |
+
"learning_rate": 8.073429487179488e-06,
|
| 253651 |
+
"loss": 0.3483,
|
| 253652 |
+
"step": 121265
|
| 253653 |
+
},
|
| 253654 |
+
{
|
| 253655 |
+
"epoch": 969.5,
|
| 253656 |
+
"learning_rate": 8.073349358974359e-06,
|
| 253657 |
+
"loss": 0.3236,
|
| 253658 |
+
"step": 121270
|
| 253659 |
+
},
|
| 253660 |
+
{
|
| 253661 |
+
"epoch": 969.54,
|
| 253662 |
+
"learning_rate": 8.073269230769232e-06,
|
| 253663 |
+
"loss": 0.3426,
|
| 253664 |
+
"step": 121275
|
| 253665 |
+
},
|
| 253666 |
+
{
|
| 253667 |
+
"epoch": 969.58,
|
| 253668 |
+
"learning_rate": 8.073189102564103e-06,
|
| 253669 |
+
"loss": 0.5984,
|
| 253670 |
+
"step": 121280
|
| 253671 |
+
},
|
| 253672 |
+
{
|
| 253673 |
+
"epoch": 969.62,
|
| 253674 |
+
"learning_rate": 8.073108974358975e-06,
|
| 253675 |
+
"loss": 0.9732,
|
| 253676 |
+
"step": 121285
|
| 253677 |
+
},
|
| 253678 |
+
{
|
| 253679 |
+
"epoch": 969.66,
|
| 253680 |
+
"learning_rate": 8.073028846153846e-06,
|
| 253681 |
+
"loss": 0.2848,
|
| 253682 |
+
"step": 121290
|
| 253683 |
+
},
|
| 253684 |
+
{
|
| 253685 |
+
"epoch": 969.7,
|
| 253686 |
+
"learning_rate": 8.072948717948719e-06,
|
| 253687 |
+
"loss": 0.2934,
|
| 253688 |
+
"step": 121295
|
| 253689 |
+
},
|
| 253690 |
+
{
|
| 253691 |
+
"epoch": 969.74,
|
| 253692 |
+
"learning_rate": 8.072884615384617e-06,
|
| 253693 |
+
"loss": 1.169,
|
| 253694 |
+
"step": 121300
|
| 253695 |
+
},
|
| 253696 |
+
{
|
| 253697 |
+
"epoch": 969.78,
|
| 253698 |
+
"learning_rate": 8.072804487179488e-06,
|
| 253699 |
+
"loss": 0.548,
|
| 253700 |
+
"step": 121305
|
| 253701 |
+
},
|
| 253702 |
+
{
|
| 253703 |
+
"epoch": 969.82,
|
| 253704 |
+
"learning_rate": 8.07272435897436e-06,
|
| 253705 |
+
"loss": 1.056,
|
| 253706 |
+
"step": 121310
|
| 253707 |
+
},
|
| 253708 |
+
{
|
| 253709 |
+
"epoch": 969.86,
|
| 253710 |
+
"learning_rate": 8.07264423076923e-06,
|
| 253711 |
+
"loss": 0.249,
|
| 253712 |
+
"step": 121315
|
| 253713 |
+
},
|
| 253714 |
+
{
|
| 253715 |
+
"epoch": 969.9,
|
| 253716 |
+
"learning_rate": 8.072564102564104e-06,
|
| 253717 |
+
"loss": 0.3025,
|
| 253718 |
+
"step": 121320
|
| 253719 |
+
},
|
| 253720 |
+
{
|
| 253721 |
+
"epoch": 969.94,
|
| 253722 |
+
"learning_rate": 8.072483974358975e-06,
|
| 253723 |
+
"loss": 0.3114,
|
| 253724 |
+
"step": 121325
|
| 253725 |
+
},
|
| 253726 |
+
{
|
| 253727 |
+
"epoch": 969.98,
|
| 253728 |
+
"learning_rate": 8.072403846153846e-06,
|
| 253729 |
+
"loss": 0.5554,
|
| 253730 |
+
"step": 121330
|
| 253731 |
+
},
|
| 253732 |
+
{
|
| 253733 |
+
"epoch": 970.0,
|
| 253734 |
+
"eval_loss": 0.3604688346385956,
|
| 253735 |
+
"eval_runtime": 35.6725,
|
| 253736 |
+
"eval_samples_per_second": 23.604,
|
| 253737 |
+
"eval_steps_per_second": 0.757,
|
| 253738 |
+
"eval_wer": 0.1775592828224407,
|
| 253739 |
+
"step": 121333
|
| 253740 |
+
},
|
| 253741 |
+
{
|
| 253742 |
+
"epoch": 978.02,
|
| 253743 |
+
"learning_rate": 8.07232371794872e-06,
|
| 253744 |
+
"loss": 0.363,
|
| 253745 |
+
"step": 121335
|
| 253746 |
+
},
|
| 253747 |
+
{
|
| 253748 |
+
"epoch": 978.06,
|
| 253749 |
+
"learning_rate": 8.07224358974359e-06,
|
| 253750 |
+
"loss": 0.3365,
|
| 253751 |
+
"step": 121340
|
| 253752 |
+
},
|
| 253753 |
+
{
|
| 253754 |
+
"epoch": 978.1,
|
| 253755 |
+
"learning_rate": 8.072163461538462e-06,
|
| 253756 |
+
"loss": 0.2485,
|
| 253757 |
+
"step": 121345
|
| 253758 |
+
},
|
| 253759 |
+
{
|
| 253760 |
+
"epoch": 978.14,
|
| 253761 |
+
"learning_rate": 8.072083333333333e-06,
|
| 253762 |
+
"loss": 0.4608,
|
| 253763 |
+
"step": 121350
|
| 253764 |
+
},
|
| 253765 |
+
{
|
| 253766 |
+
"epoch": 978.18,
|
| 253767 |
+
"learning_rate": 8.072003205128207e-06,
|
| 253768 |
+
"loss": 0.6248,
|
| 253769 |
+
"step": 121355
|
| 253770 |
+
},
|
| 253771 |
+
{
|
| 253772 |
+
"epoch": 978.22,
|
| 253773 |
+
"learning_rate": 8.071923076923078e-06,
|
| 253774 |
+
"loss": 1.1636,
|
| 253775 |
+
"step": 121360
|
| 253776 |
+
},
|
| 253777 |
+
{
|
| 253778 |
+
"epoch": 978.26,
|
| 253779 |
+
"learning_rate": 8.07184294871795e-06,
|
| 253780 |
+
"loss": 0.273,
|
| 253781 |
+
"step": 121365
|
| 253782 |
+
},
|
| 253783 |
+
{
|
| 253784 |
+
"epoch": 978.3,
|
| 253785 |
+
"learning_rate": 8.07176282051282e-06,
|
| 253786 |
+
"loss": 0.3089,
|
| 253787 |
+
"step": 121370
|
| 253788 |
+
},
|
| 253789 |
+
{
|
| 253790 |
+
"epoch": 978.34,
|
| 253791 |
+
"learning_rate": 8.071682692307694e-06,
|
| 253792 |
+
"loss": 0.323,
|
| 253793 |
+
"step": 121375
|
| 253794 |
+
},
|
| 253795 |
+
{
|
| 253796 |
+
"epoch": 978.38,
|
| 253797 |
+
"learning_rate": 8.071602564102565e-06,
|
| 253798 |
+
"loss": 0.599,
|
| 253799 |
+
"step": 121380
|
| 253800 |
+
},
|
| 253801 |
+
{
|
| 253802 |
+
"epoch": 978.42,
|
| 253803 |
+
"learning_rate": 8.071522435897436e-06,
|
| 253804 |
+
"loss": 1.039,
|
| 253805 |
+
"step": 121385
|
| 253806 |
+
},
|
| 253807 |
+
{
|
| 253808 |
+
"epoch": 978.46,
|
| 253809 |
+
"learning_rate": 8.07144230769231e-06,
|
| 253810 |
+
"loss": 0.2914,
|
| 253811 |
+
"step": 121390
|
| 253812 |
+
},
|
| 253813 |
+
{
|
| 253814 |
+
"epoch": 978.5,
|
| 253815 |
+
"learning_rate": 8.071362179487179e-06,
|
| 253816 |
+
"loss": 0.246,
|
| 253817 |
+
"step": 121395
|
| 253818 |
+
},
|
| 253819 |
+
{
|
| 253820 |
+
"epoch": 978.54,
|
| 253821 |
+
"learning_rate": 8.071282051282052e-06,
|
| 253822 |
+
"loss": 0.3878,
|
| 253823 |
+
"step": 121400
|
| 253824 |
+
},
|
| 253825 |
+
{
|
| 253826 |
+
"epoch": 978.58,
|
| 253827 |
+
"learning_rate": 8.071201923076924e-06,
|
| 253828 |
+
"loss": 0.6029,
|
| 253829 |
+
"step": 121405
|
| 253830 |
+
},
|
| 253831 |
+
{
|
| 253832 |
+
"epoch": 978.62,
|
| 253833 |
+
"learning_rate": 8.071121794871795e-06,
|
| 253834 |
+
"loss": 0.9838,
|
| 253835 |
+
"step": 121410
|
| 253836 |
+
},
|
| 253837 |
+
{
|
| 253838 |
+
"epoch": 978.66,
|
| 253839 |
+
"learning_rate": 8.071041666666666e-06,
|
| 253840 |
+
"loss": 0.301,
|
| 253841 |
+
"step": 121415
|
| 253842 |
+
},
|
| 253843 |
+
{
|
| 253844 |
+
"epoch": 978.7,
|
| 253845 |
+
"learning_rate": 8.07096153846154e-06,
|
| 253846 |
+
"loss": 0.254,
|
| 253847 |
+
"step": 121420
|
| 253848 |
+
},
|
| 253849 |
+
{
|
| 253850 |
+
"epoch": 978.74,
|
| 253851 |
+
"learning_rate": 8.07088141025641e-06,
|
| 253852 |
+
"loss": 0.3564,
|
| 253853 |
+
"step": 121425
|
| 253854 |
+
},
|
| 253855 |
+
{
|
| 253856 |
+
"epoch": 978.78,
|
| 253857 |
+
"learning_rate": 8.070801282051282e-06,
|
| 253858 |
+
"loss": 0.5654,
|
| 253859 |
+
"step": 121430
|
| 253860 |
+
},
|
| 253861 |
+
{
|
| 253862 |
+
"epoch": 978.82,
|
| 253863 |
+
"learning_rate": 8.070721153846155e-06,
|
| 253864 |
+
"loss": 0.9491,
|
| 253865 |
+
"step": 121435
|
| 253866 |
+
},
|
| 253867 |
+
{
|
| 253868 |
+
"epoch": 978.86,
|
| 253869 |
+
"learning_rate": 8.070641025641026e-06,
|
| 253870 |
+
"loss": 0.2507,
|
| 253871 |
+
"step": 121440
|
| 253872 |
+
},
|
| 253873 |
+
{
|
| 253874 |
+
"epoch": 978.9,
|
| 253875 |
+
"learning_rate": 8.070560897435898e-06,
|
| 253876 |
+
"loss": 0.5248,
|
| 253877 |
+
"step": 121445
|
| 253878 |
+
},
|
| 253879 |
+
{
|
| 253880 |
+
"epoch": 978.94,
|
| 253881 |
+
"learning_rate": 8.070480769230769e-06,
|
| 253882 |
+
"loss": 0.2928,
|
| 253883 |
+
"step": 121450
|
| 253884 |
+
},
|
| 253885 |
+
{
|
| 253886 |
+
"epoch": 978.98,
|
| 253887 |
+
"learning_rate": 8.070400641025642e-06,
|
| 253888 |
+
"loss": 0.6818,
|
| 253889 |
+
"step": 121455
|
| 253890 |
+
},
|
| 253891 |
+
{
|
| 253892 |
+
"epoch": 979.0,
|
| 253893 |
+
"eval_loss": 0.43977269530296326,
|
| 253894 |
+
"eval_runtime": 35.1739,
|
| 253895 |
+
"eval_samples_per_second": 23.938,
|
| 253896 |
+
"eval_steps_per_second": 0.768,
|
| 253897 |
+
"eval_wer": 0.18553898155632303,
|
| 253898 |
+
"step": 121457
|
| 253899 |
+
},
|
| 253900 |
+
{
|
| 253901 |
+
"epoch": 971.02,
|
| 253902 |
+
"learning_rate": 8.070320512820514e-06,
|
| 253903 |
+
"loss": 0.2944,
|
| 253904 |
+
"step": 121460
|
| 253905 |
+
},
|
| 253906 |
+
{
|
| 253907 |
+
"epoch": 971.06,
|
| 253908 |
+
"learning_rate": 8.070240384615385e-06,
|
| 253909 |
+
"loss": 0.2993,
|
| 253910 |
+
"step": 121465
|
| 253911 |
+
},
|
| 253912 |
+
{
|
| 253913 |
+
"epoch": 971.1,
|
| 253914 |
+
"learning_rate": 8.070160256410256e-06,
|
| 253915 |
+
"loss": 0.4033,
|
| 253916 |
+
"step": 121470
|
| 253917 |
+
},
|
| 253918 |
+
{
|
| 253919 |
+
"epoch": 971.14,
|
| 253920 |
+
"learning_rate": 8.07008012820513e-06,
|
| 253921 |
+
"loss": 0.3557,
|
| 253922 |
+
"step": 121475
|
| 253923 |
+
},
|
| 253924 |
+
{
|
| 253925 |
+
"epoch": 971.18,
|
| 253926 |
+
"learning_rate": 8.07e-06,
|
| 253927 |
+
"loss": 0.7841,
|
| 253928 |
+
"step": 121480
|
| 253929 |
+
},
|
| 253930 |
+
{
|
| 253931 |
+
"epoch": 971.22,
|
| 253932 |
+
"learning_rate": 8.069919871794872e-06,
|
| 253933 |
+
"loss": 0.8209,
|
| 253934 |
+
"step": 121485
|
| 253935 |
+
},
|
| 253936 |
+
{
|
| 253937 |
+
"epoch": 971.26,
|
| 253938 |
+
"learning_rate": 8.069839743589745e-06,
|
| 253939 |
+
"loss": 0.297,
|
| 253940 |
+
"step": 121490
|
| 253941 |
+
},
|
| 253942 |
+
{
|
| 253943 |
+
"epoch": 971.3,
|
| 253944 |
+
"learning_rate": 8.069759615384616e-06,
|
| 253945 |
+
"loss": 0.2987,
|
| 253946 |
+
"step": 121495
|
| 253947 |
+
},
|
| 253948 |
+
{
|
| 253949 |
+
"epoch": 971.34,
|
| 253950 |
+
"learning_rate": 8.069679487179488e-06,
|
| 253951 |
+
"loss": 0.3397,
|
| 253952 |
+
"step": 121500
|
| 253953 |
+
},
|
| 253954 |
+
{
|
| 253955 |
+
"epoch": 971.38,
|
| 253956 |
+
"learning_rate": 8.069599358974359e-06,
|
| 253957 |
+
"loss": 0.6693,
|
| 253958 |
+
"step": 121505
|
| 253959 |
+
},
|
| 253960 |
+
{
|
| 253961 |
+
"epoch": 971.42,
|
| 253962 |
+
"learning_rate": 8.069519230769232e-06,
|
| 253963 |
+
"loss": 0.7268,
|
| 253964 |
+
"step": 121510
|
| 253965 |
+
},
|
| 253966 |
+
{
|
| 253967 |
+
"epoch": 971.46,
|
| 253968 |
+
"learning_rate": 8.069439102564104e-06,
|
| 253969 |
+
"loss": 0.2656,
|
| 253970 |
+
"step": 121515
|
| 253971 |
+
},
|
| 253972 |
+
{
|
| 253973 |
+
"epoch": 971.5,
|
| 253974 |
+
"learning_rate": 8.069358974358975e-06,
|
| 253975 |
+
"loss": 0.3631,
|
| 253976 |
+
"step": 121520
|
| 253977 |
+
},
|
| 253978 |
+
{
|
| 253979 |
+
"epoch": 971.54,
|
| 253980 |
+
"learning_rate": 8.069278846153846e-06,
|
| 253981 |
+
"loss": 0.4355,
|
| 253982 |
+
"step": 121525
|
| 253983 |
+
},
|
| 253984 |
+
{
|
| 253985 |
+
"epoch": 971.58,
|
| 253986 |
+
"learning_rate": 8.06919871794872e-06,
|
| 253987 |
+
"loss": 0.6999,
|
| 253988 |
+
"step": 121530
|
| 253989 |
+
},
|
| 253990 |
+
{
|
| 253991 |
+
"epoch": 971.62,
|
| 253992 |
+
"learning_rate": 8.06911858974359e-06,
|
| 253993 |
+
"loss": 0.7744,
|
| 253994 |
+
"step": 121535
|
| 253995 |
+
},
|
| 253996 |
+
{
|
| 253997 |
+
"epoch": 971.66,
|
| 253998 |
+
"learning_rate": 8.069038461538462e-06,
|
| 253999 |
+
"loss": 0.3289,
|
| 254000 |
+
"step": 121540
|
| 254001 |
+
},
|
| 254002 |
+
{
|
| 254003 |
+
"epoch": 971.7,
|
| 254004 |
+
"learning_rate": 8.068958333333335e-06,
|
| 254005 |
+
"loss": 0.2904,
|
| 254006 |
+
"step": 121545
|
| 254007 |
+
},
|
| 254008 |
+
{
|
| 254009 |
+
"epoch": 971.74,
|
| 254010 |
+
"learning_rate": 8.068878205128205e-06,
|
| 254011 |
+
"loss": 0.2956,
|
| 254012 |
+
"step": 121550
|
| 254013 |
+
},
|
| 254014 |
+
{
|
| 254015 |
+
"epoch": 971.78,
|
| 254016 |
+
"learning_rate": 8.068798076923078e-06,
|
| 254017 |
+
"loss": 0.682,
|
| 254018 |
+
"step": 121555
|
| 254019 |
+
},
|
| 254020 |
+
{
|
| 254021 |
+
"epoch": 971.82,
|
| 254022 |
+
"learning_rate": 8.06871794871795e-06,
|
| 254023 |
+
"loss": 0.9232,
|
| 254024 |
+
"step": 121560
|
| 254025 |
+
},
|
| 254026 |
+
{
|
| 254027 |
+
"epoch": 971.86,
|
| 254028 |
+
"learning_rate": 8.06863782051282e-06,
|
| 254029 |
+
"loss": 0.346,
|
| 254030 |
+
"step": 121565
|
| 254031 |
+
},
|
| 254032 |
+
{
|
| 254033 |
+
"epoch": 971.9,
|
| 254034 |
+
"learning_rate": 8.068557692307692e-06,
|
| 254035 |
+
"loss": 0.2865,
|
| 254036 |
+
"step": 121570
|
| 254037 |
+
},
|
| 254038 |
+
{
|
| 254039 |
+
"epoch": 971.94,
|
| 254040 |
+
"learning_rate": 8.068477564102565e-06,
|
| 254041 |
+
"loss": 0.3959,
|
| 254042 |
+
"step": 121575
|
| 254043 |
+
},
|
| 254044 |
+
{
|
| 254045 |
+
"epoch": 971.98,
|
| 254046 |
+
"learning_rate": 8.068397435897436e-06,
|
| 254047 |
+
"loss": 0.7047,
|
| 254048 |
+
"step": 121580
|
| 254049 |
+
},
|
| 254050 |
+
{
|
| 254051 |
+
"epoch": 972.0,
|
| 254052 |
+
"eval_loss": 0.3686648905277252,
|
| 254053 |
+
"eval_runtime": 35.6826,
|
| 254054 |
+
"eval_samples_per_second": 23.597,
|
| 254055 |
+
"eval_steps_per_second": 0.757,
|
| 254056 |
+
"eval_wer": 0.18131747080854813,
|
| 254057 |
+
"step": 121582
|
| 254058 |
+
},
|
| 254059 |
+
{
|
| 254060 |
+
"epoch": 972.02,
|
| 254061 |
+
"learning_rate": 8.068317307692308e-06,
|
| 254062 |
+
"loss": 0.3209,
|
| 254063 |
+
"step": 121585
|
| 254064 |
+
},
|
| 254065 |
+
{
|
| 254066 |
+
"epoch": 972.06,
|
| 254067 |
+
"learning_rate": 8.06823717948718e-06,
|
| 254068 |
+
"loss": 0.2486,
|
| 254069 |
+
"step": 121590
|
| 254070 |
+
},
|
| 254071 |
+
{
|
| 254072 |
+
"epoch": 972.1,
|
| 254073 |
+
"learning_rate": 8.068157051282052e-06,
|
| 254074 |
+
"loss": 0.3214,
|
| 254075 |
+
"step": 121595
|
| 254076 |
+
},
|
| 254077 |
+
{
|
| 254078 |
+
"epoch": 972.14,
|
| 254079 |
+
"learning_rate": 8.068076923076923e-06,
|
| 254080 |
+
"loss": 0.2991,
|
| 254081 |
+
"step": 121600
|
| 254082 |
+
},
|
| 254083 |
+
{
|
| 254084 |
+
"epoch": 972.18,
|
| 254085 |
+
"learning_rate": 8.067996794871795e-06,
|
| 254086 |
+
"loss": 0.6965,
|
| 254087 |
+
"step": 121605
|
| 254088 |
+
},
|
| 254089 |
+
{
|
| 254090 |
+
"epoch": 972.22,
|
| 254091 |
+
"learning_rate": 8.067916666666668e-06,
|
| 254092 |
+
"loss": 0.784,
|
| 254093 |
+
"step": 121610
|
| 254094 |
+
},
|
| 254095 |
+
{
|
| 254096 |
+
"epoch": 972.26,
|
| 254097 |
+
"learning_rate": 8.06783653846154e-06,
|
| 254098 |
+
"loss": 0.3802,
|
| 254099 |
+
"step": 121615
|
| 254100 |
+
},
|
| 254101 |
+
{
|
| 254102 |
+
"epoch": 972.3,
|
| 254103 |
+
"learning_rate": 8.06775641025641e-06,
|
| 254104 |
+
"loss": 0.318,
|
| 254105 |
+
"step": 121620
|
| 254106 |
+
},
|
| 254107 |
+
{
|
| 254108 |
+
"epoch": 972.34,
|
| 254109 |
+
"learning_rate": 8.067676282051282e-06,
|
| 254110 |
+
"loss": 0.4142,
|
| 254111 |
+
"step": 121625
|
| 254112 |
+
},
|
| 254113 |
+
{
|
| 254114 |
+
"epoch": 972.38,
|
| 254115 |
+
"learning_rate": 8.067596153846155e-06,
|
| 254116 |
+
"loss": 0.6518,
|
| 254117 |
+
"step": 121630
|
| 254118 |
+
},
|
| 254119 |
+
{
|
| 254120 |
+
"epoch": 972.42,
|
| 254121 |
+
"learning_rate": 8.067516025641026e-06,
|
| 254122 |
+
"loss": 0.8874,
|
| 254123 |
+
"step": 121635
|
| 254124 |
+
},
|
| 254125 |
+
{
|
| 254126 |
+
"epoch": 972.46,
|
| 254127 |
+
"learning_rate": 8.067435897435898e-06,
|
| 254128 |
+
"loss": 0.3634,
|
| 254129 |
+
"step": 121640
|
| 254130 |
+
},
|
| 254131 |
+
{
|
| 254132 |
+
"epoch": 972.5,
|
| 254133 |
+
"learning_rate": 8.06735576923077e-06,
|
| 254134 |
+
"loss": 0.3147,
|
| 254135 |
+
"step": 121645
|
| 254136 |
+
},
|
| 254137 |
+
{
|
| 254138 |
+
"epoch": 972.54,
|
| 254139 |
+
"learning_rate": 8.067275641025642e-06,
|
| 254140 |
+
"loss": 0.3491,
|
| 254141 |
+
"step": 121650
|
| 254142 |
+
},
|
| 254143 |
+
{
|
| 254144 |
+
"epoch": 972.58,
|
| 254145 |
+
"learning_rate": 8.067195512820513e-06,
|
| 254146 |
+
"loss": 0.7093,
|
| 254147 |
+
"step": 121655
|
| 254148 |
+
},
|
| 254149 |
+
{
|
| 254150 |
+
"epoch": 972.62,
|
| 254151 |
+
"learning_rate": 8.067115384615385e-06,
|
| 254152 |
+
"loss": 0.9415,
|
| 254153 |
+
"step": 121660
|
| 254154 |
+
},
|
| 254155 |
+
{
|
| 254156 |
+
"epoch": 972.66,
|
| 254157 |
+
"learning_rate": 8.067035256410258e-06,
|
| 254158 |
+
"loss": 0.3118,
|
| 254159 |
+
"step": 121665
|
| 254160 |
+
},
|
| 254161 |
+
{
|
| 254162 |
+
"epoch": 972.7,
|
| 254163 |
+
"learning_rate": 8.06695512820513e-06,
|
| 254164 |
+
"loss": 0.2936,
|
| 254165 |
+
"step": 121670
|
| 254166 |
+
},
|
| 254167 |
+
{
|
| 254168 |
+
"epoch": 972.74,
|
| 254169 |
+
"learning_rate": 8.066875e-06,
|
| 254170 |
+
"loss": 0.338,
|
| 254171 |
+
"step": 121675
|
| 254172 |
+
},
|
| 254173 |
+
{
|
| 254174 |
+
"epoch": 972.78,
|
| 254175 |
+
"learning_rate": 8.066794871794874e-06,
|
| 254176 |
+
"loss": 0.6909,
|
| 254177 |
+
"step": 121680
|
| 254178 |
+
},
|
| 254179 |
+
{
|
| 254180 |
+
"epoch": 972.82,
|
| 254181 |
+
"learning_rate": 8.066714743589745e-06,
|
| 254182 |
+
"loss": 0.7566,
|
| 254183 |
+
"step": 121685
|
| 254184 |
+
},
|
| 254185 |
+
{
|
| 254186 |
+
"epoch": 972.86,
|
| 254187 |
+
"learning_rate": 8.066634615384616e-06,
|
| 254188 |
+
"loss": 0.2324,
|
| 254189 |
+
"step": 121690
|
| 254190 |
+
},
|
| 254191 |
+
{
|
| 254192 |
+
"epoch": 972.9,
|
| 254193 |
+
"learning_rate": 8.066554487179488e-06,
|
| 254194 |
+
"loss": 0.309,
|
| 254195 |
+
"step": 121695
|
| 254196 |
+
},
|
| 254197 |
+
{
|
| 254198 |
+
"epoch": 972.94,
|
| 254199 |
+
"learning_rate": 8.06647435897436e-06,
|
| 254200 |
+
"loss": 0.367,
|
| 254201 |
+
"step": 121700
|
| 254202 |
+
},
|
| 254203 |
+
{
|
| 254204 |
+
"epoch": 972.98,
|
| 254205 |
+
"learning_rate": 8.06639423076923e-06,
|
| 254206 |
+
"loss": 0.9295,
|
| 254207 |
+
"step": 121705
|
| 254208 |
+
},
|
| 254209 |
+
{
|
| 254210 |
+
"epoch": 973.0,
|
| 254211 |
+
"eval_loss": 0.3698074519634247,
|
| 254212 |
+
"eval_runtime": 37.0962,
|
| 254213 |
+
"eval_samples_per_second": 22.698,
|
| 254214 |
+
"eval_steps_per_second": 0.728,
|
| 254215 |
+
"eval_wer": 0.18579426172148356,
|
| 254216 |
+
"step": 121707
|
| 254217 |
}
|
| 254218 |
],
|
| 254219 |
"max_steps": 625000,
|
| 254220 |
"num_train_epochs": 5000,
|
| 254221 |
+
"total_flos": 3.425064251533899e+20,
|
| 254222 |
"trial_name": null,
|
| 254223 |
"trial_params": null
|
| 254224 |
}
|
model-bin/finetune/base/{checkpoint-121084 → checkpoint-121707}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630132562.2337933/events.out.tfevents.1630132562.86bb0ddabf9b.4092.51
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:baddbba9e71714999abac43eec83bf67c7533fa1604d8205e55509dda3ec6247
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630132940.836788/events.out.tfevents.1630132940.86bb0ddabf9b.4092.53
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8fa409f40a52697f7113a591f335ed4829d67463862bb2cb043c4eb843e87e91
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630133330.209003/events.out.tfevents.1630133330.86bb0ddabf9b.4092.55
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:70d936ec4eca46cf5a46ce6ea0444453fcb73a523136686f76dba8b6202dbc21
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630133714.3736/events.out.tfevents.1630133714.86bb0ddabf9b.4092.57
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1476550766e11c54c6b5a6eaa11fa671f5a33d9a5c6feba04057397204dfdf0f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630134097.5583656/events.out.tfevents.1630134097.86bb0ddabf9b.4092.59
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:babd24ecf25f909721bcf44a4929f50942d73cbb5c12f82e1d49ad88d6fd3685
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630132562.86bb0ddabf9b.4092.50
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:45e83d1153142d32dac1279afb71f50560619d3196cde2e0d55e9daec9cef2da
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630132940.86bb0ddabf9b.4092.52
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1d70449ee1a35ac9e05b353210136e33592e5b2e2db9a3037aee364c6f9f7f00
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630133330.86bb0ddabf9b.4092.54
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:16cc3ddada4eec5c151b06e9ad33a8320e77e41bceb204f77fc798dc2aa4e2cd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630133714.86bb0ddabf9b.4092.56
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bebae397e40701784ff5201669ce666cce1c985cecdfb3adb1657096f97ae639
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630134097.86bb0ddabf9b.4092.58
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:84f7af9e0a043b7745405261785cfadf7bc068c1fda0fd55408003dfe277d65a
|
| 3 |
+
size 8622
|