"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630107172.831585/events.out.tfevents.1630107172.86bb0ddabf9b.1042.1 +3 -0
- model-bin/finetune/base/log/1630107597.3949003/events.out.tfevents.1630107597.86bb0ddabf9b.1042.3 +3 -0
- model-bin/finetune/base/log/1630108127.983279/events.out.tfevents.1630108127.86bb0ddabf9b.1042.5 +3 -0
- model-bin/finetune/base/log/1630108542.9879258/events.out.tfevents.1630108542.86bb0ddabf9b.1042.7 +3 -0
- model-bin/finetune/base/log/1630108960.8081708/events.out.tfevents.1630108960.86bb0ddabf9b.1042.9 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630107172.86bb0ddabf9b.1042.0 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630107597.86bb0ddabf9b.1042.2 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630108127.86bb0ddabf9b.1042.4 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630108542.86bb0ddabf9b.1042.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630108960.86bb0ddabf9b.1042.8 +3 -0
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/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 sha256:3dc554dec301ccc8dca1e759108d10f280284ba56da831c7924e4f12d8241e81
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/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 sha256:5144d589a8f3be429110fd3438d71c50cbe2b5d96f43ebff77a52854f4e4c466
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/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 sha256:06429aa76c5c31c9f61e7d1a7c32603b81cc59ed4d9203338065436cd62cc39b
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/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:d93d43bbeecc2e2dd56d18ef4fdffea967e794dca54efdfd4a801ca408b71336
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/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:ffe03c0c156b224ba155a03dcb28f915736509e98b9e680245bcc2241ec2d0d4
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/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,
|
|
@@ -244398,11 +244398,800 @@
|
|
| 244398 |
"eval_steps_per_second": 0.661,
|
| 244399 |
"eval_wer": 0.18297995933778682,
|
| 244400 |
"step": 113992
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 244401 |
}
|
| 244402 |
],
|
| 244403 |
-
"max_steps":
|
| 244404 |
"num_train_epochs": 5000,
|
| 244405 |
-
"total_flos": 3.
|
| 244406 |
"trial_name": null,
|
| 244407 |
"trial_params": null
|
| 244408 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 923.995983935743,
|
| 5 |
+
"global_step": 114613,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 244398 |
"eval_steps_per_second": 0.661,
|
| 244399 |
"eval_wer": 0.18297995933778682,
|
| 244400 |
"step": 113992
|
| 244401 |
+
},
|
| 244402 |
+
{
|
| 244403 |
+
"epoch": 911.02,
|
| 244404 |
+
"learning_rate": 8.189903846153846e-06,
|
| 244405 |
+
"loss": 0.3429,
|
| 244406 |
+
"step": 113995
|
| 244407 |
+
},
|
| 244408 |
+
{
|
| 244409 |
+
"epoch": 911.06,
|
| 244410 |
+
"learning_rate": 8.189823717948718e-06,
|
| 244411 |
+
"loss": 0.3061,
|
| 244412 |
+
"step": 114000
|
| 244413 |
+
},
|
| 244414 |
+
{
|
| 244415 |
+
"epoch": 911.1,
|
| 244416 |
+
"learning_rate": 8.189743589743591e-06,
|
| 244417 |
+
"loss": 0.2955,
|
| 244418 |
+
"step": 114005
|
| 244419 |
+
},
|
| 244420 |
+
{
|
| 244421 |
+
"epoch": 911.14,
|
| 244422 |
+
"learning_rate": 8.189663461538462e-06,
|
| 244423 |
+
"loss": 0.3999,
|
| 244424 |
+
"step": 114010
|
| 244425 |
+
},
|
| 244426 |
+
{
|
| 244427 |
+
"epoch": 911.18,
|
| 244428 |
+
"learning_rate": 8.189583333333334e-06,
|
| 244429 |
+
"loss": 0.7552,
|
| 244430 |
+
"step": 114015
|
| 244431 |
+
},
|
| 244432 |
+
{
|
| 244433 |
+
"epoch": 911.22,
|
| 244434 |
+
"learning_rate": 8.189503205128205e-06,
|
| 244435 |
+
"loss": 0.7735,
|
| 244436 |
+
"step": 114020
|
| 244437 |
+
},
|
| 244438 |
+
{
|
| 244439 |
+
"epoch": 911.26,
|
| 244440 |
+
"learning_rate": 8.189423076923078e-06,
|
| 244441 |
+
"loss": 0.2605,
|
| 244442 |
+
"step": 114025
|
| 244443 |
+
},
|
| 244444 |
+
{
|
| 244445 |
+
"epoch": 911.3,
|
| 244446 |
+
"learning_rate": 8.18934294871795e-06,
|
| 244447 |
+
"loss": 0.3436,
|
| 244448 |
+
"step": 114030
|
| 244449 |
+
},
|
| 244450 |
+
{
|
| 244451 |
+
"epoch": 911.34,
|
| 244452 |
+
"learning_rate": 8.18926282051282e-06,
|
| 244453 |
+
"loss": 0.3272,
|
| 244454 |
+
"step": 114035
|
| 244455 |
+
},
|
| 244456 |
+
{
|
| 244457 |
+
"epoch": 911.38,
|
| 244458 |
+
"learning_rate": 8.189182692307694e-06,
|
| 244459 |
+
"loss": 0.7182,
|
| 244460 |
+
"step": 114040
|
| 244461 |
+
},
|
| 244462 |
+
{
|
| 244463 |
+
"epoch": 911.42,
|
| 244464 |
+
"learning_rate": 8.189102564102565e-06,
|
| 244465 |
+
"loss": 0.8876,
|
| 244466 |
+
"step": 114045
|
| 244467 |
+
},
|
| 244468 |
+
{
|
| 244469 |
+
"epoch": 911.46,
|
| 244470 |
+
"learning_rate": 8.189022435897437e-06,
|
| 244471 |
+
"loss": 0.2841,
|
| 244472 |
+
"step": 114050
|
| 244473 |
+
},
|
| 244474 |
+
{
|
| 244475 |
+
"epoch": 911.5,
|
| 244476 |
+
"learning_rate": 8.188942307692308e-06,
|
| 244477 |
+
"loss": 0.2787,
|
| 244478 |
+
"step": 114055
|
| 244479 |
+
},
|
| 244480 |
+
{
|
| 244481 |
+
"epoch": 911.54,
|
| 244482 |
+
"learning_rate": 8.188862179487181e-06,
|
| 244483 |
+
"loss": 0.3579,
|
| 244484 |
+
"step": 114060
|
| 244485 |
+
},
|
| 244486 |
+
{
|
| 244487 |
+
"epoch": 911.58,
|
| 244488 |
+
"learning_rate": 8.188782051282052e-06,
|
| 244489 |
+
"loss": 0.7733,
|
| 244490 |
+
"step": 114065
|
| 244491 |
+
},
|
| 244492 |
+
{
|
| 244493 |
+
"epoch": 911.62,
|
| 244494 |
+
"learning_rate": 8.188701923076924e-06,
|
| 244495 |
+
"loss": 0.9861,
|
| 244496 |
+
"step": 114070
|
| 244497 |
+
},
|
| 244498 |
+
{
|
| 244499 |
+
"epoch": 911.66,
|
| 244500 |
+
"learning_rate": 8.188621794871797e-06,
|
| 244501 |
+
"loss": 0.2945,
|
| 244502 |
+
"step": 114075
|
| 244503 |
+
},
|
| 244504 |
+
{
|
| 244505 |
+
"epoch": 911.7,
|
| 244506 |
+
"learning_rate": 8.188541666666668e-06,
|
| 244507 |
+
"loss": 0.3552,
|
| 244508 |
+
"step": 114080
|
| 244509 |
+
},
|
| 244510 |
+
{
|
| 244511 |
+
"epoch": 911.74,
|
| 244512 |
+
"learning_rate": 8.18846153846154e-06,
|
| 244513 |
+
"loss": 0.3424,
|
| 244514 |
+
"step": 114085
|
| 244515 |
+
},
|
| 244516 |
+
{
|
| 244517 |
+
"epoch": 911.78,
|
| 244518 |
+
"learning_rate": 8.18838141025641e-06,
|
| 244519 |
+
"loss": 0.6975,
|
| 244520 |
+
"step": 114090
|
| 244521 |
+
},
|
| 244522 |
+
{
|
| 244523 |
+
"epoch": 911.82,
|
| 244524 |
+
"learning_rate": 8.188301282051284e-06,
|
| 244525 |
+
"loss": 0.8733,
|
| 244526 |
+
"step": 114095
|
| 244527 |
+
},
|
| 244528 |
+
{
|
| 244529 |
+
"epoch": 911.86,
|
| 244530 |
+
"learning_rate": 8.188221153846153e-06,
|
| 244531 |
+
"loss": 0.298,
|
| 244532 |
+
"step": 114100
|
| 244533 |
+
},
|
| 244534 |
+
{
|
| 244535 |
+
"epoch": 911.9,
|
| 244536 |
+
"learning_rate": 8.188141025641027e-06,
|
| 244537 |
+
"loss": 0.298,
|
| 244538 |
+
"step": 114105
|
| 244539 |
+
},
|
| 244540 |
+
{
|
| 244541 |
+
"epoch": 911.94,
|
| 244542 |
+
"learning_rate": 8.188060897435898e-06,
|
| 244543 |
+
"loss": 0.4223,
|
| 244544 |
+
"step": 114110
|
| 244545 |
+
},
|
| 244546 |
+
{
|
| 244547 |
+
"epoch": 911.98,
|
| 244548 |
+
"learning_rate": 8.18798076923077e-06,
|
| 244549 |
+
"loss": 0.7582,
|
| 244550 |
+
"step": 114115
|
| 244551 |
+
},
|
| 244552 |
+
{
|
| 244553 |
+
"epoch": 912.0,
|
| 244554 |
+
"eval_loss": 0.4367474317550659,
|
| 244555 |
+
"eval_runtime": 39.6856,
|
| 244556 |
+
"eval_samples_per_second": 21.066,
|
| 244557 |
+
"eval_steps_per_second": 0.68,
|
| 244558 |
+
"eval_wer": 0.1899942163100058,
|
| 244559 |
+
"step": 114117
|
| 244560 |
+
},
|
| 244561 |
+
{
|
| 244562 |
+
"epoch": 920.02,
|
| 244563 |
+
"learning_rate": 8.18790064102564e-06,
|
| 244564 |
+
"loss": 0.2965,
|
| 244565 |
+
"step": 114120
|
| 244566 |
+
},
|
| 244567 |
+
{
|
| 244568 |
+
"epoch": 920.06,
|
| 244569 |
+
"learning_rate": 8.187820512820514e-06,
|
| 244570 |
+
"loss": 0.3006,
|
| 244571 |
+
"step": 114125
|
| 244572 |
+
},
|
| 244573 |
+
{
|
| 244574 |
+
"epoch": 920.1,
|
| 244575 |
+
"learning_rate": 8.187740384615385e-06,
|
| 244576 |
+
"loss": 0.2453,
|
| 244577 |
+
"step": 114130
|
| 244578 |
+
},
|
| 244579 |
+
{
|
| 244580 |
+
"epoch": 920.14,
|
| 244581 |
+
"learning_rate": 8.187660256410256e-06,
|
| 244582 |
+
"loss": 0.3644,
|
| 244583 |
+
"step": 114135
|
| 244584 |
+
},
|
| 244585 |
+
{
|
| 244586 |
+
"epoch": 920.18,
|
| 244587 |
+
"learning_rate": 8.18758012820513e-06,
|
| 244588 |
+
"loss": 0.8205,
|
| 244589 |
+
"step": 114140
|
| 244590 |
+
},
|
| 244591 |
+
{
|
| 244592 |
+
"epoch": 920.22,
|
| 244593 |
+
"learning_rate": 8.1875e-06,
|
| 244594 |
+
"loss": 0.8451,
|
| 244595 |
+
"step": 114145
|
| 244596 |
+
},
|
| 244597 |
+
{
|
| 244598 |
+
"epoch": 920.27,
|
| 244599 |
+
"learning_rate": 8.187419871794872e-06,
|
| 244600 |
+
"loss": 0.2713,
|
| 244601 |
+
"step": 114150
|
| 244602 |
+
},
|
| 244603 |
+
{
|
| 244604 |
+
"epoch": 920.31,
|
| 244605 |
+
"learning_rate": 8.187339743589744e-06,
|
| 244606 |
+
"loss": 0.2795,
|
| 244607 |
+
"step": 114155
|
| 244608 |
+
},
|
| 244609 |
+
{
|
| 244610 |
+
"epoch": 920.35,
|
| 244611 |
+
"learning_rate": 8.187259615384617e-06,
|
| 244612 |
+
"loss": 0.3686,
|
| 244613 |
+
"step": 114160
|
| 244614 |
+
},
|
| 244615 |
+
{
|
| 244616 |
+
"epoch": 920.39,
|
| 244617 |
+
"learning_rate": 8.187179487179488e-06,
|
| 244618 |
+
"loss": 0.8049,
|
| 244619 |
+
"step": 114165
|
| 244620 |
+
},
|
| 244621 |
+
{
|
| 244622 |
+
"epoch": 920.43,
|
| 244623 |
+
"learning_rate": 8.18709935897436e-06,
|
| 244624 |
+
"loss": 0.9125,
|
| 244625 |
+
"step": 114170
|
| 244626 |
+
},
|
| 244627 |
+
{
|
| 244628 |
+
"epoch": 920.47,
|
| 244629 |
+
"learning_rate": 8.18701923076923e-06,
|
| 244630 |
+
"loss": 0.237,
|
| 244631 |
+
"step": 114175
|
| 244632 |
+
},
|
| 244633 |
+
{
|
| 244634 |
+
"epoch": 920.51,
|
| 244635 |
+
"learning_rate": 8.186939102564104e-06,
|
| 244636 |
+
"loss": 0.3213,
|
| 244637 |
+
"step": 114180
|
| 244638 |
+
},
|
| 244639 |
+
{
|
| 244640 |
+
"epoch": 920.55,
|
| 244641 |
+
"learning_rate": 8.186858974358975e-06,
|
| 244642 |
+
"loss": 0.3379,
|
| 244643 |
+
"step": 114185
|
| 244644 |
+
},
|
| 244645 |
+
{
|
| 244646 |
+
"epoch": 920.59,
|
| 244647 |
+
"learning_rate": 8.186778846153846e-06,
|
| 244648 |
+
"loss": 0.6896,
|
| 244649 |
+
"step": 114190
|
| 244650 |
+
},
|
| 244651 |
+
{
|
| 244652 |
+
"epoch": 920.63,
|
| 244653 |
+
"learning_rate": 8.18669871794872e-06,
|
| 244654 |
+
"loss": 0.8142,
|
| 244655 |
+
"step": 114195
|
| 244656 |
+
},
|
| 244657 |
+
{
|
| 244658 |
+
"epoch": 920.67,
|
| 244659 |
+
"learning_rate": 8.18661858974359e-06,
|
| 244660 |
+
"loss": 0.327,
|
| 244661 |
+
"step": 114200
|
| 244662 |
+
},
|
| 244663 |
+
{
|
| 244664 |
+
"epoch": 920.71,
|
| 244665 |
+
"learning_rate": 8.186538461538462e-06,
|
| 244666 |
+
"loss": 0.3287,
|
| 244667 |
+
"step": 114205
|
| 244668 |
+
},
|
| 244669 |
+
{
|
| 244670 |
+
"epoch": 920.75,
|
| 244671 |
+
"learning_rate": 8.186458333333334e-06,
|
| 244672 |
+
"loss": 0.35,
|
| 244673 |
+
"step": 114210
|
| 244674 |
+
},
|
| 244675 |
+
{
|
| 244676 |
+
"epoch": 920.79,
|
| 244677 |
+
"learning_rate": 8.186378205128207e-06,
|
| 244678 |
+
"loss": 0.8775,
|
| 244679 |
+
"step": 114215
|
| 244680 |
+
},
|
| 244681 |
+
{
|
| 244682 |
+
"epoch": 920.83,
|
| 244683 |
+
"learning_rate": 8.186298076923076e-06,
|
| 244684 |
+
"loss": 0.8496,
|
| 244685 |
+
"step": 114220
|
| 244686 |
+
},
|
| 244687 |
+
{
|
| 244688 |
+
"epoch": 920.87,
|
| 244689 |
+
"learning_rate": 8.18621794871795e-06,
|
| 244690 |
+
"loss": 0.2979,
|
| 244691 |
+
"step": 114225
|
| 244692 |
+
},
|
| 244693 |
+
{
|
| 244694 |
+
"epoch": 920.91,
|
| 244695 |
+
"learning_rate": 8.186137820512822e-06,
|
| 244696 |
+
"loss": 0.2862,
|
| 244697 |
+
"step": 114230
|
| 244698 |
+
},
|
| 244699 |
+
{
|
| 244700 |
+
"epoch": 920.95,
|
| 244701 |
+
"learning_rate": 8.186057692307692e-06,
|
| 244702 |
+
"loss": 0.3481,
|
| 244703 |
+
"step": 114235
|
| 244704 |
+
},
|
| 244705 |
+
{
|
| 244706 |
+
"epoch": 920.99,
|
| 244707 |
+
"learning_rate": 8.185977564102565e-06,
|
| 244708 |
+
"loss": 0.7926,
|
| 244709 |
+
"step": 114240
|
| 244710 |
+
},
|
| 244711 |
+
{
|
| 244712 |
+
"epoch": 921.0,
|
| 244713 |
+
"eval_loss": 0.47359028458595276,
|
| 244714 |
+
"eval_runtime": 48.9621,
|
| 244715 |
+
"eval_samples_per_second": 17.095,
|
| 244716 |
+
"eval_steps_per_second": 0.551,
|
| 244717 |
+
"eval_wer": 0.17835125448028674,
|
| 244718 |
+
"step": 114241
|
| 244719 |
+
},
|
| 244720 |
+
{
|
| 244721 |
+
"epoch": 921.03,
|
| 244722 |
+
"learning_rate": 8.185897435897436e-06,
|
| 244723 |
+
"loss": 0.3033,
|
| 244724 |
+
"step": 114245
|
| 244725 |
+
},
|
| 244726 |
+
{
|
| 244727 |
+
"epoch": 921.07,
|
| 244728 |
+
"learning_rate": 8.185817307692308e-06,
|
| 244729 |
+
"loss": 0.2835,
|
| 244730 |
+
"step": 114250
|
| 244731 |
+
},
|
| 244732 |
+
{
|
| 244733 |
+
"epoch": 921.11,
|
| 244734 |
+
"learning_rate": 8.18573717948718e-06,
|
| 244735 |
+
"loss": 0.2973,
|
| 244736 |
+
"step": 114255
|
| 244737 |
+
},
|
| 244738 |
+
{
|
| 244739 |
+
"epoch": 921.15,
|
| 244740 |
+
"learning_rate": 8.185657051282052e-06,
|
| 244741 |
+
"loss": 0.3924,
|
| 244742 |
+
"step": 114260
|
| 244743 |
+
},
|
| 244744 |
+
{
|
| 244745 |
+
"epoch": 921.19,
|
| 244746 |
+
"learning_rate": 8.185576923076924e-06,
|
| 244747 |
+
"loss": 0.8991,
|
| 244748 |
+
"step": 114265
|
| 244749 |
+
},
|
| 244750 |
+
{
|
| 244751 |
+
"epoch": 921.23,
|
| 244752 |
+
"learning_rate": 8.185496794871795e-06,
|
| 244753 |
+
"loss": 0.7195,
|
| 244754 |
+
"step": 114270
|
| 244755 |
+
},
|
| 244756 |
+
{
|
| 244757 |
+
"epoch": 921.27,
|
| 244758 |
+
"learning_rate": 8.185416666666666e-06,
|
| 244759 |
+
"loss": 0.2874,
|
| 244760 |
+
"step": 114275
|
| 244761 |
+
},
|
| 244762 |
+
{
|
| 244763 |
+
"epoch": 921.31,
|
| 244764 |
+
"learning_rate": 8.18533653846154e-06,
|
| 244765 |
+
"loss": 0.3402,
|
| 244766 |
+
"step": 114280
|
| 244767 |
+
},
|
| 244768 |
+
{
|
| 244769 |
+
"epoch": 921.35,
|
| 244770 |
+
"learning_rate": 8.18525641025641e-06,
|
| 244771 |
+
"loss": 0.4055,
|
| 244772 |
+
"step": 114285
|
| 244773 |
+
},
|
| 244774 |
+
{
|
| 244775 |
+
"epoch": 921.39,
|
| 244776 |
+
"learning_rate": 8.185176282051282e-06,
|
| 244777 |
+
"loss": 1.2911,
|
| 244778 |
+
"step": 114290
|
| 244779 |
+
},
|
| 244780 |
+
{
|
| 244781 |
+
"epoch": 921.43,
|
| 244782 |
+
"learning_rate": 8.185096153846155e-06,
|
| 244783 |
+
"loss": 0.6954,
|
| 244784 |
+
"step": 114295
|
| 244785 |
+
},
|
| 244786 |
+
{
|
| 244787 |
+
"epoch": 921.47,
|
| 244788 |
+
"learning_rate": 8.185016025641026e-06,
|
| 244789 |
+
"loss": 0.298,
|
| 244790 |
+
"step": 114300
|
| 244791 |
+
},
|
| 244792 |
+
{
|
| 244793 |
+
"epoch": 921.51,
|
| 244794 |
+
"learning_rate": 8.184935897435898e-06,
|
| 244795 |
+
"loss": 0.2934,
|
| 244796 |
+
"step": 114305
|
| 244797 |
+
},
|
| 244798 |
+
{
|
| 244799 |
+
"epoch": 921.55,
|
| 244800 |
+
"learning_rate": 8.18485576923077e-06,
|
| 244801 |
+
"loss": 0.4285,
|
| 244802 |
+
"step": 114310
|
| 244803 |
+
},
|
| 244804 |
+
{
|
| 244805 |
+
"epoch": 921.59,
|
| 244806 |
+
"learning_rate": 8.184775641025642e-06,
|
| 244807 |
+
"loss": 0.8933,
|
| 244808 |
+
"step": 114315
|
| 244809 |
+
},
|
| 244810 |
+
{
|
| 244811 |
+
"epoch": 921.63,
|
| 244812 |
+
"learning_rate": 8.184695512820514e-06,
|
| 244813 |
+
"loss": 0.9413,
|
| 244814 |
+
"step": 114320
|
| 244815 |
+
},
|
| 244816 |
+
{
|
| 244817 |
+
"epoch": 921.67,
|
| 244818 |
+
"learning_rate": 8.184615384615385e-06,
|
| 244819 |
+
"loss": 0.2615,
|
| 244820 |
+
"step": 114325
|
| 244821 |
+
},
|
| 244822 |
+
{
|
| 244823 |
+
"epoch": 921.71,
|
| 244824 |
+
"learning_rate": 8.184535256410258e-06,
|
| 244825 |
+
"loss": 0.3053,
|
| 244826 |
+
"step": 114330
|
| 244827 |
+
},
|
| 244828 |
+
{
|
| 244829 |
+
"epoch": 921.76,
|
| 244830 |
+
"learning_rate": 8.18445512820513e-06,
|
| 244831 |
+
"loss": 0.4418,
|
| 244832 |
+
"step": 114335
|
| 244833 |
+
},
|
| 244834 |
+
{
|
| 244835 |
+
"epoch": 921.8,
|
| 244836 |
+
"learning_rate": 8.184375e-06,
|
| 244837 |
+
"loss": 0.7972,
|
| 244838 |
+
"step": 114340
|
| 244839 |
+
},
|
| 244840 |
+
{
|
| 244841 |
+
"epoch": 921.84,
|
| 244842 |
+
"learning_rate": 8.184294871794872e-06,
|
| 244843 |
+
"loss": 0.697,
|
| 244844 |
+
"step": 114345
|
| 244845 |
+
},
|
| 244846 |
+
{
|
| 244847 |
+
"epoch": 921.88,
|
| 244848 |
+
"learning_rate": 8.184214743589745e-06,
|
| 244849 |
+
"loss": 0.2833,
|
| 244850 |
+
"step": 114350
|
| 244851 |
+
},
|
| 244852 |
+
{
|
| 244853 |
+
"epoch": 921.92,
|
| 244854 |
+
"learning_rate": 8.184134615384617e-06,
|
| 244855 |
+
"loss": 0.3443,
|
| 244856 |
+
"step": 114355
|
| 244857 |
+
},
|
| 244858 |
+
{
|
| 244859 |
+
"epoch": 921.96,
|
| 244860 |
+
"learning_rate": 8.184054487179488e-06,
|
| 244861 |
+
"loss": 0.4554,
|
| 244862 |
+
"step": 114360
|
| 244863 |
+
},
|
| 244864 |
+
{
|
| 244865 |
+
"epoch": 922.0,
|
| 244866 |
+
"learning_rate": 8.18397435897436e-06,
|
| 244867 |
+
"loss": 1.1524,
|
| 244868 |
+
"step": 114365
|
| 244869 |
+
},
|
| 244870 |
+
{
|
| 244871 |
+
"epoch": 922.0,
|
| 244872 |
+
"eval_loss": 0.4351113736629486,
|
| 244873 |
+
"eval_runtime": 39.0293,
|
| 244874 |
+
"eval_samples_per_second": 21.599,
|
| 244875 |
+
"eval_steps_per_second": 0.692,
|
| 244876 |
+
"eval_wer": 0.18565555718690355,
|
| 244877 |
+
"step": 114365
|
| 244878 |
+
},
|
| 244879 |
+
{
|
| 244880 |
+
"epoch": 922.04,
|
| 244881 |
+
"learning_rate": 8.183894230769232e-06,
|
| 244882 |
+
"loss": 0.3181,
|
| 244883 |
+
"step": 114370
|
| 244884 |
+
},
|
| 244885 |
+
{
|
| 244886 |
+
"epoch": 922.08,
|
| 244887 |
+
"learning_rate": 8.183814102564102e-06,
|
| 244888 |
+
"loss": 0.287,
|
| 244889 |
+
"step": 114375
|
| 244890 |
+
},
|
| 244891 |
+
{
|
| 244892 |
+
"epoch": 922.12,
|
| 244893 |
+
"learning_rate": 8.183733974358975e-06,
|
| 244894 |
+
"loss": 0.3079,
|
| 244895 |
+
"step": 114380
|
| 244896 |
+
},
|
| 244897 |
+
{
|
| 244898 |
+
"epoch": 922.16,
|
| 244899 |
+
"learning_rate": 8.183653846153848e-06,
|
| 244900 |
+
"loss": 0.3877,
|
| 244901 |
+
"step": 114385
|
| 244902 |
+
},
|
| 244903 |
+
{
|
| 244904 |
+
"epoch": 922.2,
|
| 244905 |
+
"learning_rate": 8.183573717948718e-06,
|
| 244906 |
+
"loss": 1.0876,
|
| 244907 |
+
"step": 114390
|
| 244908 |
+
},
|
| 244909 |
+
{
|
| 244910 |
+
"epoch": 922.24,
|
| 244911 |
+
"learning_rate": 8.18349358974359e-06,
|
| 244912 |
+
"loss": 0.315,
|
| 244913 |
+
"step": 114395
|
| 244914 |
+
},
|
| 244915 |
+
{
|
| 244916 |
+
"epoch": 922.28,
|
| 244917 |
+
"learning_rate": 8.183413461538462e-06,
|
| 244918 |
+
"loss": 0.2827,
|
| 244919 |
+
"step": 114400
|
| 244920 |
+
},
|
| 244921 |
+
{
|
| 244922 |
+
"epoch": 922.32,
|
| 244923 |
+
"learning_rate": 8.183333333333333e-06,
|
| 244924 |
+
"loss": 0.3118,
|
| 244925 |
+
"step": 114405
|
| 244926 |
+
},
|
| 244927 |
+
{
|
| 244928 |
+
"epoch": 922.36,
|
| 244929 |
+
"learning_rate": 8.183253205128205e-06,
|
| 244930 |
+
"loss": 0.4451,
|
| 244931 |
+
"step": 114410
|
| 244932 |
+
},
|
| 244933 |
+
{
|
| 244934 |
+
"epoch": 922.4,
|
| 244935 |
+
"learning_rate": 8.183189102564102e-06,
|
| 244936 |
+
"loss": 1.3298,
|
| 244937 |
+
"step": 114415
|
| 244938 |
+
},
|
| 244939 |
+
{
|
| 244940 |
+
"epoch": 922.44,
|
| 244941 |
+
"learning_rate": 8.183108974358975e-06,
|
| 244942 |
+
"loss": 0.2886,
|
| 244943 |
+
"step": 114420
|
| 244944 |
+
},
|
| 244945 |
+
{
|
| 244946 |
+
"epoch": 922.48,
|
| 244947 |
+
"learning_rate": 8.183028846153847e-06,
|
| 244948 |
+
"loss": 0.2573,
|
| 244949 |
+
"step": 114425
|
| 244950 |
+
},
|
| 244951 |
+
{
|
| 244952 |
+
"epoch": 922.52,
|
| 244953 |
+
"learning_rate": 8.182948717948718e-06,
|
| 244954 |
+
"loss": 0.3227,
|
| 244955 |
+
"step": 114430
|
| 244956 |
+
},
|
| 244957 |
+
{
|
| 244958 |
+
"epoch": 922.56,
|
| 244959 |
+
"learning_rate": 8.18286858974359e-06,
|
| 244960 |
+
"loss": 0.5028,
|
| 244961 |
+
"step": 114435
|
| 244962 |
+
},
|
| 244963 |
+
{
|
| 244964 |
+
"epoch": 922.6,
|
| 244965 |
+
"learning_rate": 8.182788461538462e-06,
|
| 244966 |
+
"loss": 1.246,
|
| 244967 |
+
"step": 114440
|
| 244968 |
+
},
|
| 244969 |
+
{
|
| 244970 |
+
"epoch": 922.64,
|
| 244971 |
+
"learning_rate": 8.182708333333334e-06,
|
| 244972 |
+
"loss": 0.4493,
|
| 244973 |
+
"step": 114445
|
| 244974 |
+
},
|
| 244975 |
+
{
|
| 244976 |
+
"epoch": 922.68,
|
| 244977 |
+
"learning_rate": 8.182628205128205e-06,
|
| 244978 |
+
"loss": 0.3027,
|
| 244979 |
+
"step": 114450
|
| 244980 |
+
},
|
| 244981 |
+
{
|
| 244982 |
+
"epoch": 922.72,
|
| 244983 |
+
"learning_rate": 8.182548076923078e-06,
|
| 244984 |
+
"loss": 0.3715,
|
| 244985 |
+
"step": 114455
|
| 244986 |
+
},
|
| 244987 |
+
{
|
| 244988 |
+
"epoch": 922.76,
|
| 244989 |
+
"learning_rate": 8.18246794871795e-06,
|
| 244990 |
+
"loss": 0.4394,
|
| 244991 |
+
"step": 114460
|
| 244992 |
+
},
|
| 244993 |
+
{
|
| 244994 |
+
"epoch": 922.8,
|
| 244995 |
+
"learning_rate": 8.182387820512821e-06,
|
| 244996 |
+
"loss": 1.129,
|
| 244997 |
+
"step": 114465
|
| 244998 |
+
},
|
| 244999 |
+
{
|
| 245000 |
+
"epoch": 922.84,
|
| 245001 |
+
"learning_rate": 8.182307692307692e-06,
|
| 245002 |
+
"loss": 0.3216,
|
| 245003 |
+
"step": 114470
|
| 245004 |
+
},
|
| 245005 |
+
{
|
| 245006 |
+
"epoch": 922.88,
|
| 245007 |
+
"learning_rate": 8.182227564102565e-06,
|
| 245008 |
+
"loss": 0.3146,
|
| 245009 |
+
"step": 114475
|
| 245010 |
+
},
|
| 245011 |
+
{
|
| 245012 |
+
"epoch": 922.92,
|
| 245013 |
+
"learning_rate": 8.182147435897437e-06,
|
| 245014 |
+
"loss": 0.3341,
|
| 245015 |
+
"step": 114480
|
| 245016 |
+
},
|
| 245017 |
+
{
|
| 245018 |
+
"epoch": 922.96,
|
| 245019 |
+
"learning_rate": 8.182067307692308e-06,
|
| 245020 |
+
"loss": 0.4186,
|
| 245021 |
+
"step": 114485
|
| 245022 |
+
},
|
| 245023 |
+
{
|
| 245024 |
+
"epoch": 923.0,
|
| 245025 |
+
"eval_loss": 0.4149532616138458,
|
| 245026 |
+
"eval_runtime": 39.8191,
|
| 245027 |
+
"eval_samples_per_second": 21.171,
|
| 245028 |
+
"eval_steps_per_second": 0.678,
|
| 245029 |
+
"eval_wer": 0.18946368551804765,
|
| 245030 |
+
"step": 114489
|
| 245031 |
+
},
|
| 245032 |
+
{
|
| 245033 |
+
"epoch": 923.01,
|
| 245034 |
+
"learning_rate": 8.18198717948718e-06,
|
| 245035 |
+
"loss": 0.3366,
|
| 245036 |
+
"step": 114490
|
| 245037 |
+
},
|
| 245038 |
+
{
|
| 245039 |
+
"epoch": 923.05,
|
| 245040 |
+
"learning_rate": 8.181907051282052e-06,
|
| 245041 |
+
"loss": 0.2907,
|
| 245042 |
+
"step": 114495
|
| 245043 |
+
},
|
| 245044 |
+
{
|
| 245045 |
+
"epoch": 923.09,
|
| 245046 |
+
"learning_rate": 8.181826923076924e-06,
|
| 245047 |
+
"loss": 0.3085,
|
| 245048 |
+
"step": 114500
|
| 245049 |
+
},
|
| 245050 |
+
{
|
| 245051 |
+
"epoch": 923.13,
|
| 245052 |
+
"learning_rate": 8.181746794871795e-06,
|
| 245053 |
+
"loss": 0.3446,
|
| 245054 |
+
"step": 114505
|
| 245055 |
+
},
|
| 245056 |
+
{
|
| 245057 |
+
"epoch": 923.17,
|
| 245058 |
+
"learning_rate": 8.181666666666668e-06,
|
| 245059 |
+
"loss": 0.4964,
|
| 245060 |
+
"step": 114510
|
| 245061 |
+
},
|
| 245062 |
+
{
|
| 245063 |
+
"epoch": 923.21,
|
| 245064 |
+
"learning_rate": 8.18158653846154e-06,
|
| 245065 |
+
"loss": 1.0242,
|
| 245066 |
+
"step": 114515
|
| 245067 |
+
},
|
| 245068 |
+
{
|
| 245069 |
+
"epoch": 923.25,
|
| 245070 |
+
"learning_rate": 8.181506410256411e-06,
|
| 245071 |
+
"loss": 0.3272,
|
| 245072 |
+
"step": 114520
|
| 245073 |
+
},
|
| 245074 |
+
{
|
| 245075 |
+
"epoch": 923.29,
|
| 245076 |
+
"learning_rate": 8.181426282051282e-06,
|
| 245077 |
+
"loss": 0.325,
|
| 245078 |
+
"step": 114525
|
| 245079 |
+
},
|
| 245080 |
+
{
|
| 245081 |
+
"epoch": 923.33,
|
| 245082 |
+
"learning_rate": 8.181346153846155e-06,
|
| 245083 |
+
"loss": 0.3969,
|
| 245084 |
+
"step": 114530
|
| 245085 |
+
},
|
| 245086 |
+
{
|
| 245087 |
+
"epoch": 923.37,
|
| 245088 |
+
"learning_rate": 8.181266025641027e-06,
|
| 245089 |
+
"loss": 0.4844,
|
| 245090 |
+
"step": 114535
|
| 245091 |
+
},
|
| 245092 |
+
{
|
| 245093 |
+
"epoch": 923.41,
|
| 245094 |
+
"learning_rate": 8.181185897435898e-06,
|
| 245095 |
+
"loss": 1.1032,
|
| 245096 |
+
"step": 114540
|
| 245097 |
+
},
|
| 245098 |
+
{
|
| 245099 |
+
"epoch": 923.45,
|
| 245100 |
+
"learning_rate": 8.18110576923077e-06,
|
| 245101 |
+
"loss": 0.2643,
|
| 245102 |
+
"step": 114545
|
| 245103 |
+
},
|
| 245104 |
+
{
|
| 245105 |
+
"epoch": 923.49,
|
| 245106 |
+
"learning_rate": 8.181025641025642e-06,
|
| 245107 |
+
"loss": 0.2698,
|
| 245108 |
+
"step": 114550
|
| 245109 |
+
},
|
| 245110 |
+
{
|
| 245111 |
+
"epoch": 923.53,
|
| 245112 |
+
"learning_rate": 8.180945512820514e-06,
|
| 245113 |
+
"loss": 0.2943,
|
| 245114 |
+
"step": 114555
|
| 245115 |
+
},
|
| 245116 |
+
{
|
| 245117 |
+
"epoch": 923.57,
|
| 245118 |
+
"learning_rate": 8.180865384615385e-06,
|
| 245119 |
+
"loss": 0.4995,
|
| 245120 |
+
"step": 114560
|
| 245121 |
+
},
|
| 245122 |
+
{
|
| 245123 |
+
"epoch": 923.61,
|
| 245124 |
+
"learning_rate": 8.180785256410258e-06,
|
| 245125 |
+
"loss": 1.1026,
|
| 245126 |
+
"step": 114565
|
| 245127 |
+
},
|
| 245128 |
+
{
|
| 245129 |
+
"epoch": 923.65,
|
| 245130 |
+
"learning_rate": 8.180705128205128e-06,
|
| 245131 |
+
"loss": 0.2717,
|
| 245132 |
+
"step": 114570
|
| 245133 |
+
},
|
| 245134 |
+
{
|
| 245135 |
+
"epoch": 923.69,
|
| 245136 |
+
"learning_rate": 8.180625000000001e-06,
|
| 245137 |
+
"loss": 0.2867,
|
| 245138 |
+
"step": 114575
|
| 245139 |
+
},
|
| 245140 |
+
{
|
| 245141 |
+
"epoch": 923.73,
|
| 245142 |
+
"learning_rate": 8.180544871794872e-06,
|
| 245143 |
+
"loss": 0.3375,
|
| 245144 |
+
"step": 114580
|
| 245145 |
+
},
|
| 245146 |
+
{
|
| 245147 |
+
"epoch": 923.77,
|
| 245148 |
+
"learning_rate": 8.180464743589744e-06,
|
| 245149 |
+
"loss": 0.5839,
|
| 245150 |
+
"step": 114585
|
| 245151 |
+
},
|
| 245152 |
+
{
|
| 245153 |
+
"epoch": 923.81,
|
| 245154 |
+
"learning_rate": 8.180384615384615e-06,
|
| 245155 |
+
"loss": 1.0434,
|
| 245156 |
+
"step": 114590
|
| 245157 |
+
},
|
| 245158 |
+
{
|
| 245159 |
+
"epoch": 923.85,
|
| 245160 |
+
"learning_rate": 8.180304487179488e-06,
|
| 245161 |
+
"loss": 0.3545,
|
| 245162 |
+
"step": 114595
|
| 245163 |
+
},
|
| 245164 |
+
{
|
| 245165 |
+
"epoch": 923.89,
|
| 245166 |
+
"learning_rate": 8.18022435897436e-06,
|
| 245167 |
+
"loss": 0.3341,
|
| 245168 |
+
"step": 114600
|
| 245169 |
+
},
|
| 245170 |
+
{
|
| 245171 |
+
"epoch": 923.93,
|
| 245172 |
+
"learning_rate": 8.18014423076923e-06,
|
| 245173 |
+
"loss": 0.3615,
|
| 245174 |
+
"step": 114605
|
| 245175 |
+
},
|
| 245176 |
+
{
|
| 245177 |
+
"epoch": 923.97,
|
| 245178 |
+
"learning_rate": 8.180064102564104e-06,
|
| 245179 |
+
"loss": 0.6272,
|
| 245180 |
+
"step": 114610
|
| 245181 |
+
},
|
| 245182 |
+
{
|
| 245183 |
+
"epoch": 924.0,
|
| 245184 |
+
"eval_loss": 0.3545716106891632,
|
| 245185 |
+
"eval_runtime": 39.8654,
|
| 245186 |
+
"eval_samples_per_second": 21.146,
|
| 245187 |
+
"eval_steps_per_second": 0.677,
|
| 245188 |
+
"eval_wer": 0.18582472364713531,
|
| 245189 |
+
"step": 114613
|
| 245190 |
}
|
| 245191 |
],
|
| 245192 |
+
"max_steps": 620000,
|
| 245193 |
"num_train_epochs": 5000,
|
| 245194 |
+
"total_flos": 3.225517960789425e+20,
|
| 245195 |
"trial_name": null,
|
| 245196 |
"trial_params": null
|
| 245197 |
}
|
model-bin/finetune/base/{checkpoint-113992 → checkpoint-114613}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630107172.831585/events.out.tfevents.1630107172.86bb0ddabf9b.1042.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6aeda773d1e5f19ef742dd91a92e470ccc30e14faaca77578dfa690f8ba7e209
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630107597.3949003/events.out.tfevents.1630107597.86bb0ddabf9b.1042.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b6f8c9b033347f38e86107122356cffdfc05824f5b13d69389580da12c3af0f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630108127.983279/events.out.tfevents.1630108127.86bb0ddabf9b.1042.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0d1387b80bfffc1109dc85dc636d6e602a32e5cdb20e80fd35c0154a6fc7bfaa
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630108542.9879258/events.out.tfevents.1630108542.86bb0ddabf9b.1042.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b13b91412aeecb26c21fb11c54caaeef8910b368662e63848bdab048df4974e0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630108960.8081708/events.out.tfevents.1630108960.86bb0ddabf9b.1042.9
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9de072a4f71eac89e5082ec377cee3ad05553b5e574f5235aebd7b9ffe492130
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630107172.86bb0ddabf9b.1042.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6dafb199912054a3d32e56492b11806f1f40499977fb9baca81bc4b2db159a66
|
| 3 |
+
size 8630
|
model-bin/finetune/base/log/events.out.tfevents.1630107597.86bb0ddabf9b.1042.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:85a44d609ba96a704a9993014a97d1360f7ebfd4e37098729a5ebcf531ad8613
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630108127.86bb0ddabf9b.1042.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4249aa82cc943be75382adc2fa3444c76cd80d9275bb29f3810df2c65a5c7d41
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630108542.86bb0ddabf9b.1042.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:98e40bee306491bc7675a24ae3340bdc28af0293266cc25fe946014ef80e9200
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630108960.86bb0ddabf9b.1042.8
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2cded565932eb070edcb2a75c1ab2265dd5eb139681ba17383a44fb5241dfad1
|
| 3 |
+
size 8622
|