"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630163053.748851/events.out.tfevents.1630163053.86bb0ddabf9b.4092.201 +3 -0
- model-bin/finetune/base/log/1630163453.0593839/events.out.tfevents.1630163453.86bb0ddabf9b.4092.203 +3 -0
- model-bin/finetune/base/log/1630163842.4436672/events.out.tfevents.1630163842.86bb0ddabf9b.4092.205 +3 -0
- model-bin/finetune/base/log/1630164227.907955/events.out.tfevents.1630164227.86bb0ddabf9b.4092.207 +3 -0
- model-bin/finetune/base/log/1630164623.2772548/events.out.tfevents.1630164623.86bb0ddabf9b.4092.209 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630163053.86bb0ddabf9b.4092.200 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630163453.86bb0ddabf9b.4092.202 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630163842.86bb0ddabf9b.4092.204 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630164227.86bb0ddabf9b.4092.206 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630164623.86bb0ddabf9b.4092.208 +3 -0
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/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 sha256:653b939f79425281e6ecdeb60ca3148b9ac988addb5763424283925616a7cbd0
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/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 sha256:26987437825a04d245ab97e24b0ef772ee77f1ed2f78f71fd9b417f9245608da
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/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 sha256:6bb292ef71a8a3cd8967ac863441c3ada437a8e0e54ae949d89887fb3fabdeac
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/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 sha256:9e2501abdc077c7a803c491659edc2d02989ba8148008e328737de3065f37a00
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/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 sha256:54593ffd46b39a7b550f5b28ab713225dbbb87010aa83b606fc65a9fe845d5ac
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17105075053609722,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-129674",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -265302,11 +265302,800 @@
|
|
| 265302 |
"eval_steps_per_second": 0.742,
|
| 265303 |
"eval_wer": 0.18075693360101122,
|
| 265304 |
"step": 130420
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 265305 |
}
|
| 265306 |
],
|
| 265307 |
-
"max_steps":
|
| 265308 |
"num_train_epochs": 5000,
|
| 265309 |
-
"total_flos": 3.
|
| 265310 |
"trial_name": null,
|
| 265311 |
"trial_params": null
|
| 265312 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17105075053609722,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-129674",
|
| 4 |
+
"epoch": 1047.996015936255,
|
| 5 |
+
"global_step": 131042,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 265302 |
"eval_steps_per_second": 0.742,
|
| 265303 |
"eval_wer": 0.18075693360101122,
|
| 265304 |
"step": 130420
|
| 265305 |
+
},
|
| 265306 |
+
{
|
| 265307 |
+
"epoch": 1043.04,
|
| 265308 |
+
"learning_rate": 7.926698717948719e-06,
|
| 265309 |
+
"loss": 0.2924,
|
| 265310 |
+
"step": 130425
|
| 265311 |
+
},
|
| 265312 |
+
{
|
| 265313 |
+
"epoch": 1043.08,
|
| 265314 |
+
"learning_rate": 7.92661858974359e-06,
|
| 265315 |
+
"loss": 0.2821,
|
| 265316 |
+
"step": 130430
|
| 265317 |
+
},
|
| 265318 |
+
{
|
| 265319 |
+
"epoch": 1043.12,
|
| 265320 |
+
"learning_rate": 7.926538461538463e-06,
|
| 265321 |
+
"loss": 0.2655,
|
| 265322 |
+
"step": 130435
|
| 265323 |
+
},
|
| 265324 |
+
{
|
| 265325 |
+
"epoch": 1043.16,
|
| 265326 |
+
"learning_rate": 7.926458333333334e-06,
|
| 265327 |
+
"loss": 0.4237,
|
| 265328 |
+
"step": 130440
|
| 265329 |
+
},
|
| 265330 |
+
{
|
| 265331 |
+
"epoch": 1043.2,
|
| 265332 |
+
"learning_rate": 7.926378205128206e-06,
|
| 265333 |
+
"loss": 1.2634,
|
| 265334 |
+
"step": 130445
|
| 265335 |
+
},
|
| 265336 |
+
{
|
| 265337 |
+
"epoch": 1043.24,
|
| 265338 |
+
"learning_rate": 7.926298076923079e-06,
|
| 265339 |
+
"loss": 0.286,
|
| 265340 |
+
"step": 130450
|
| 265341 |
+
},
|
| 265342 |
+
{
|
| 265343 |
+
"epoch": 1043.28,
|
| 265344 |
+
"learning_rate": 7.926217948717948e-06,
|
| 265345 |
+
"loss": 0.3288,
|
| 265346 |
+
"step": 130455
|
| 265347 |
+
},
|
| 265348 |
+
{
|
| 265349 |
+
"epoch": 1043.32,
|
| 265350 |
+
"learning_rate": 7.926137820512821e-06,
|
| 265351 |
+
"loss": 0.2873,
|
| 265352 |
+
"step": 130460
|
| 265353 |
+
},
|
| 265354 |
+
{
|
| 265355 |
+
"epoch": 1043.36,
|
| 265356 |
+
"learning_rate": 7.926057692307693e-06,
|
| 265357 |
+
"loss": 0.4448,
|
| 265358 |
+
"step": 130465
|
| 265359 |
+
},
|
| 265360 |
+
{
|
| 265361 |
+
"epoch": 1043.4,
|
| 265362 |
+
"learning_rate": 7.925977564102564e-06,
|
| 265363 |
+
"loss": 1.2558,
|
| 265364 |
+
"step": 130470
|
| 265365 |
+
},
|
| 265366 |
+
{
|
| 265367 |
+
"epoch": 1043.44,
|
| 265368 |
+
"learning_rate": 7.925897435897436e-06,
|
| 265369 |
+
"loss": 0.3083,
|
| 265370 |
+
"step": 130475
|
| 265371 |
+
},
|
| 265372 |
+
{
|
| 265373 |
+
"epoch": 1043.48,
|
| 265374 |
+
"learning_rate": 7.925817307692309e-06,
|
| 265375 |
+
"loss": 0.3517,
|
| 265376 |
+
"step": 130480
|
| 265377 |
+
},
|
| 265378 |
+
{
|
| 265379 |
+
"epoch": 1043.52,
|
| 265380 |
+
"learning_rate": 7.92573717948718e-06,
|
| 265381 |
+
"loss": 0.3243,
|
| 265382 |
+
"step": 130485
|
| 265383 |
+
},
|
| 265384 |
+
{
|
| 265385 |
+
"epoch": 1043.56,
|
| 265386 |
+
"learning_rate": 7.925657051282051e-06,
|
| 265387 |
+
"loss": 0.4209,
|
| 265388 |
+
"step": 130490
|
| 265389 |
+
},
|
| 265390 |
+
{
|
| 265391 |
+
"epoch": 1043.6,
|
| 265392 |
+
"learning_rate": 7.925576923076924e-06,
|
| 265393 |
+
"loss": 1.0973,
|
| 265394 |
+
"step": 130495
|
| 265395 |
+
},
|
| 265396 |
+
{
|
| 265397 |
+
"epoch": 1043.64,
|
| 265398 |
+
"learning_rate": 7.925496794871796e-06,
|
| 265399 |
+
"loss": 0.311,
|
| 265400 |
+
"step": 130500
|
| 265401 |
+
},
|
| 265402 |
+
{
|
| 265403 |
+
"epoch": 1043.68,
|
| 265404 |
+
"learning_rate": 7.925416666666667e-06,
|
| 265405 |
+
"loss": 0.3225,
|
| 265406 |
+
"step": 130505
|
| 265407 |
+
},
|
| 265408 |
+
{
|
| 265409 |
+
"epoch": 1043.72,
|
| 265410 |
+
"learning_rate": 7.925336538461538e-06,
|
| 265411 |
+
"loss": 0.3059,
|
| 265412 |
+
"step": 130510
|
| 265413 |
+
},
|
| 265414 |
+
{
|
| 265415 |
+
"epoch": 1043.76,
|
| 265416 |
+
"learning_rate": 7.925256410256412e-06,
|
| 265417 |
+
"loss": 0.4018,
|
| 265418 |
+
"step": 130515
|
| 265419 |
+
},
|
| 265420 |
+
{
|
| 265421 |
+
"epoch": 1043.8,
|
| 265422 |
+
"learning_rate": 7.925176282051283e-06,
|
| 265423 |
+
"loss": 1.1179,
|
| 265424 |
+
"step": 130520
|
| 265425 |
+
},
|
| 265426 |
+
{
|
| 265427 |
+
"epoch": 1043.84,
|
| 265428 |
+
"learning_rate": 7.925096153846154e-06,
|
| 265429 |
+
"loss": 0.2868,
|
| 265430 |
+
"step": 130525
|
| 265431 |
+
},
|
| 265432 |
+
{
|
| 265433 |
+
"epoch": 1043.88,
|
| 265434 |
+
"learning_rate": 7.925016025641026e-06,
|
| 265435 |
+
"loss": 0.2716,
|
| 265436 |
+
"step": 130530
|
| 265437 |
+
},
|
| 265438 |
+
{
|
| 265439 |
+
"epoch": 1043.92,
|
| 265440 |
+
"learning_rate": 7.924935897435899e-06,
|
| 265441 |
+
"loss": 0.3303,
|
| 265442 |
+
"step": 130535
|
| 265443 |
+
},
|
| 265444 |
+
{
|
| 265445 |
+
"epoch": 1043.96,
|
| 265446 |
+
"learning_rate": 7.92485576923077e-06,
|
| 265447 |
+
"loss": 0.4517,
|
| 265448 |
+
"step": 130540
|
| 265449 |
+
},
|
| 265450 |
+
{
|
| 265451 |
+
"epoch": 1044.0,
|
| 265452 |
+
"learning_rate": 7.924775641025641e-06,
|
| 265453 |
+
"loss": 1.1773,
|
| 265454 |
+
"step": 130545
|
| 265455 |
+
},
|
| 265456 |
+
{
|
| 265457 |
+
"epoch": 1044.0,
|
| 265458 |
+
"eval_loss": 0.3330843150615692,
|
| 265459 |
+
"eval_runtime": 36.8417,
|
| 265460 |
+
"eval_samples_per_second": 22.692,
|
| 265461 |
+
"eval_steps_per_second": 0.733,
|
| 265462 |
+
"eval_wer": 0.1764533606677405,
|
| 265463 |
+
"step": 130545
|
| 265464 |
+
},
|
| 265465 |
+
{
|
| 265466 |
+
"epoch": 1052.04,
|
| 265467 |
+
"learning_rate": 7.924695512820514e-06,
|
| 265468 |
+
"loss": 0.3384,
|
| 265469 |
+
"step": 130550
|
| 265470 |
+
},
|
| 265471 |
+
{
|
| 265472 |
+
"epoch": 1052.08,
|
| 265473 |
+
"learning_rate": 7.924615384615386e-06,
|
| 265474 |
+
"loss": 0.2406,
|
| 265475 |
+
"step": 130555
|
| 265476 |
+
},
|
| 265477 |
+
{
|
| 265478 |
+
"epoch": 1052.12,
|
| 265479 |
+
"learning_rate": 7.924535256410257e-06,
|
| 265480 |
+
"loss": 0.3477,
|
| 265481 |
+
"step": 130560
|
| 265482 |
+
},
|
| 265483 |
+
{
|
| 265484 |
+
"epoch": 1052.16,
|
| 265485 |
+
"learning_rate": 7.924455128205128e-06,
|
| 265486 |
+
"loss": 0.4641,
|
| 265487 |
+
"step": 130565
|
| 265488 |
+
},
|
| 265489 |
+
{
|
| 265490 |
+
"epoch": 1052.2,
|
| 265491 |
+
"learning_rate": 7.924375000000002e-06,
|
| 265492 |
+
"loss": 1.1962,
|
| 265493 |
+
"step": 130570
|
| 265494 |
+
},
|
| 265495 |
+
{
|
| 265496 |
+
"epoch": 1052.24,
|
| 265497 |
+
"learning_rate": 7.924294871794871e-06,
|
| 265498 |
+
"loss": 0.3233,
|
| 265499 |
+
"step": 130575
|
| 265500 |
+
},
|
| 265501 |
+
{
|
| 265502 |
+
"epoch": 1052.28,
|
| 265503 |
+
"learning_rate": 7.924214743589744e-06,
|
| 265504 |
+
"loss": 0.2914,
|
| 265505 |
+
"step": 130580
|
| 265506 |
+
},
|
| 265507 |
+
{
|
| 265508 |
+
"epoch": 1052.32,
|
| 265509 |
+
"learning_rate": 7.924134615384616e-06,
|
| 265510 |
+
"loss": 0.2446,
|
| 265511 |
+
"step": 130585
|
| 265512 |
+
},
|
| 265513 |
+
{
|
| 265514 |
+
"epoch": 1052.36,
|
| 265515 |
+
"learning_rate": 7.924054487179487e-06,
|
| 265516 |
+
"loss": 0.3771,
|
| 265517 |
+
"step": 130590
|
| 265518 |
+
},
|
| 265519 |
+
{
|
| 265520 |
+
"epoch": 1052.4,
|
| 265521 |
+
"learning_rate": 7.92397435897436e-06,
|
| 265522 |
+
"loss": 1.2242,
|
| 265523 |
+
"step": 130595
|
| 265524 |
+
},
|
| 265525 |
+
{
|
| 265526 |
+
"epoch": 1052.44,
|
| 265527 |
+
"learning_rate": 7.923894230769231e-06,
|
| 265528 |
+
"loss": 0.2807,
|
| 265529 |
+
"step": 130600
|
| 265530 |
+
},
|
| 265531 |
+
{
|
| 265532 |
+
"epoch": 1052.48,
|
| 265533 |
+
"learning_rate": 7.923814102564103e-06,
|
| 265534 |
+
"loss": 0.3239,
|
| 265535 |
+
"step": 130605
|
| 265536 |
+
},
|
| 265537 |
+
{
|
| 265538 |
+
"epoch": 1052.52,
|
| 265539 |
+
"learning_rate": 7.923733974358974e-06,
|
| 265540 |
+
"loss": 0.3105,
|
| 265541 |
+
"step": 130610
|
| 265542 |
+
},
|
| 265543 |
+
{
|
| 265544 |
+
"epoch": 1052.56,
|
| 265545 |
+
"learning_rate": 7.923653846153847e-06,
|
| 265546 |
+
"loss": 0.5188,
|
| 265547 |
+
"step": 130615
|
| 265548 |
+
},
|
| 265549 |
+
{
|
| 265550 |
+
"epoch": 1052.6,
|
| 265551 |
+
"learning_rate": 7.923573717948719e-06,
|
| 265552 |
+
"loss": 1.2514,
|
| 265553 |
+
"step": 130620
|
| 265554 |
+
},
|
| 265555 |
+
{
|
| 265556 |
+
"epoch": 1052.64,
|
| 265557 |
+
"learning_rate": 7.92349358974359e-06,
|
| 265558 |
+
"loss": 0.3024,
|
| 265559 |
+
"step": 130625
|
| 265560 |
+
},
|
| 265561 |
+
{
|
| 265562 |
+
"epoch": 1052.68,
|
| 265563 |
+
"learning_rate": 7.923413461538461e-06,
|
| 265564 |
+
"loss": 0.2717,
|
| 265565 |
+
"step": 130630
|
| 265566 |
+
},
|
| 265567 |
+
{
|
| 265568 |
+
"epoch": 1052.72,
|
| 265569 |
+
"learning_rate": 7.923333333333334e-06,
|
| 265570 |
+
"loss": 0.3497,
|
| 265571 |
+
"step": 130635
|
| 265572 |
+
},
|
| 265573 |
+
{
|
| 265574 |
+
"epoch": 1052.76,
|
| 265575 |
+
"learning_rate": 7.923253205128206e-06,
|
| 265576 |
+
"loss": 0.4023,
|
| 265577 |
+
"step": 130640
|
| 265578 |
+
},
|
| 265579 |
+
{
|
| 265580 |
+
"epoch": 1052.8,
|
| 265581 |
+
"learning_rate": 7.923173076923077e-06,
|
| 265582 |
+
"loss": 1.178,
|
| 265583 |
+
"step": 130645
|
| 265584 |
+
},
|
| 265585 |
+
{
|
| 265586 |
+
"epoch": 1052.84,
|
| 265587 |
+
"learning_rate": 7.92309294871795e-06,
|
| 265588 |
+
"loss": 0.3058,
|
| 265589 |
+
"step": 130650
|
| 265590 |
+
},
|
| 265591 |
+
{
|
| 265592 |
+
"epoch": 1052.88,
|
| 265593 |
+
"learning_rate": 7.923012820512821e-06,
|
| 265594 |
+
"loss": 0.3344,
|
| 265595 |
+
"step": 130655
|
| 265596 |
+
},
|
| 265597 |
+
{
|
| 265598 |
+
"epoch": 1052.92,
|
| 265599 |
+
"learning_rate": 7.922932692307693e-06,
|
| 265600 |
+
"loss": 0.3237,
|
| 265601 |
+
"step": 130660
|
| 265602 |
+
},
|
| 265603 |
+
{
|
| 265604 |
+
"epoch": 1052.96,
|
| 265605 |
+
"learning_rate": 7.922852564102564e-06,
|
| 265606 |
+
"loss": 0.5455,
|
| 265607 |
+
"step": 130665
|
| 265608 |
+
},
|
| 265609 |
+
{
|
| 265610 |
+
"epoch": 1053.0,
|
| 265611 |
+
"eval_loss": 0.37590697407722473,
|
| 265612 |
+
"eval_runtime": 37.2613,
|
| 265613 |
+
"eval_samples_per_second": 22.436,
|
| 265614 |
+
"eval_steps_per_second": 0.725,
|
| 265615 |
+
"eval_wer": 0.17155110793423875,
|
| 265616 |
+
"step": 130669
|
| 265617 |
+
},
|
| 265618 |
+
{
|
| 265619 |
+
"epoch": 1053.01,
|
| 265620 |
+
"learning_rate": 7.922772435897437e-06,
|
| 265621 |
+
"loss": 0.5027,
|
| 265622 |
+
"step": 130670
|
| 265623 |
+
},
|
| 265624 |
+
{
|
| 265625 |
+
"epoch": 1053.05,
|
| 265626 |
+
"learning_rate": 7.922692307692309e-06,
|
| 265627 |
+
"loss": 0.2415,
|
| 265628 |
+
"step": 130675
|
| 265629 |
+
},
|
| 265630 |
+
{
|
| 265631 |
+
"epoch": 1053.09,
|
| 265632 |
+
"learning_rate": 7.92261217948718e-06,
|
| 265633 |
+
"loss": 0.2721,
|
| 265634 |
+
"step": 130680
|
| 265635 |
+
},
|
| 265636 |
+
{
|
| 265637 |
+
"epoch": 1053.13,
|
| 265638 |
+
"learning_rate": 7.922532051282051e-06,
|
| 265639 |
+
"loss": 0.3606,
|
| 265640 |
+
"step": 130685
|
| 265641 |
+
},
|
| 265642 |
+
{
|
| 265643 |
+
"epoch": 1053.17,
|
| 265644 |
+
"learning_rate": 7.922451923076924e-06,
|
| 265645 |
+
"loss": 0.5634,
|
| 265646 |
+
"step": 130690
|
| 265647 |
+
},
|
| 265648 |
+
{
|
| 265649 |
+
"epoch": 1053.21,
|
| 265650 |
+
"learning_rate": 7.922371794871796e-06,
|
| 265651 |
+
"loss": 1.0906,
|
| 265652 |
+
"step": 130695
|
| 265653 |
+
},
|
| 265654 |
+
{
|
| 265655 |
+
"epoch": 1053.25,
|
| 265656 |
+
"learning_rate": 7.922291666666667e-06,
|
| 265657 |
+
"loss": 0.3278,
|
| 265658 |
+
"step": 130700
|
| 265659 |
+
},
|
| 265660 |
+
{
|
| 265661 |
+
"epoch": 1053.29,
|
| 265662 |
+
"learning_rate": 7.92221153846154e-06,
|
| 265663 |
+
"loss": 0.2634,
|
| 265664 |
+
"step": 130705
|
| 265665 |
+
},
|
| 265666 |
+
{
|
| 265667 |
+
"epoch": 1053.33,
|
| 265668 |
+
"learning_rate": 7.922131410256411e-06,
|
| 265669 |
+
"loss": 0.3169,
|
| 265670 |
+
"step": 130710
|
| 265671 |
+
},
|
| 265672 |
+
{
|
| 265673 |
+
"epoch": 1053.37,
|
| 265674 |
+
"learning_rate": 7.922051282051283e-06,
|
| 265675 |
+
"loss": 0.6403,
|
| 265676 |
+
"step": 130715
|
| 265677 |
+
},
|
| 265678 |
+
{
|
| 265679 |
+
"epoch": 1053.41,
|
| 265680 |
+
"learning_rate": 7.921971153846154e-06,
|
| 265681 |
+
"loss": 1.2072,
|
| 265682 |
+
"step": 130720
|
| 265683 |
+
},
|
| 265684 |
+
{
|
| 265685 |
+
"epoch": 1053.45,
|
| 265686 |
+
"learning_rate": 7.921891025641027e-06,
|
| 265687 |
+
"loss": 0.3458,
|
| 265688 |
+
"step": 130725
|
| 265689 |
+
},
|
| 265690 |
+
{
|
| 265691 |
+
"epoch": 1053.49,
|
| 265692 |
+
"learning_rate": 7.921810897435897e-06,
|
| 265693 |
+
"loss": 0.2612,
|
| 265694 |
+
"step": 130730
|
| 265695 |
+
},
|
| 265696 |
+
{
|
| 265697 |
+
"epoch": 1053.53,
|
| 265698 |
+
"learning_rate": 7.92173076923077e-06,
|
| 265699 |
+
"loss": 0.3228,
|
| 265700 |
+
"step": 130735
|
| 265701 |
+
},
|
| 265702 |
+
{
|
| 265703 |
+
"epoch": 1053.57,
|
| 265704 |
+
"learning_rate": 7.921650641025643e-06,
|
| 265705 |
+
"loss": 0.5138,
|
| 265706 |
+
"step": 130740
|
| 265707 |
+
},
|
| 265708 |
+
{
|
| 265709 |
+
"epoch": 1053.61,
|
| 265710 |
+
"learning_rate": 7.921570512820513e-06,
|
| 265711 |
+
"loss": 1.0249,
|
| 265712 |
+
"step": 130745
|
| 265713 |
+
},
|
| 265714 |
+
{
|
| 265715 |
+
"epoch": 1053.65,
|
| 265716 |
+
"learning_rate": 7.921490384615386e-06,
|
| 265717 |
+
"loss": 0.2914,
|
| 265718 |
+
"step": 130750
|
| 265719 |
+
},
|
| 265720 |
+
{
|
| 265721 |
+
"epoch": 1053.69,
|
| 265722 |
+
"learning_rate": 7.921410256410257e-06,
|
| 265723 |
+
"loss": 0.2793,
|
| 265724 |
+
"step": 130755
|
| 265725 |
+
},
|
| 265726 |
+
{
|
| 265727 |
+
"epoch": 1053.73,
|
| 265728 |
+
"learning_rate": 7.921330128205128e-06,
|
| 265729 |
+
"loss": 0.3124,
|
| 265730 |
+
"step": 130760
|
| 265731 |
+
},
|
| 265732 |
+
{
|
| 265733 |
+
"epoch": 1053.77,
|
| 265734 |
+
"learning_rate": 7.92125e-06,
|
| 265735 |
+
"loss": 0.4488,
|
| 265736 |
+
"step": 130765
|
| 265737 |
+
},
|
| 265738 |
+
{
|
| 265739 |
+
"epoch": 1053.81,
|
| 265740 |
+
"learning_rate": 7.921169871794873e-06,
|
| 265741 |
+
"loss": 0.9875,
|
| 265742 |
+
"step": 130770
|
| 265743 |
+
},
|
| 265744 |
+
{
|
| 265745 |
+
"epoch": 1053.85,
|
| 265746 |
+
"learning_rate": 7.921089743589744e-06,
|
| 265747 |
+
"loss": 0.294,
|
| 265748 |
+
"step": 130775
|
| 265749 |
+
},
|
| 265750 |
+
{
|
| 265751 |
+
"epoch": 1053.89,
|
| 265752 |
+
"learning_rate": 7.921009615384616e-06,
|
| 265753 |
+
"loss": 0.2695,
|
| 265754 |
+
"step": 130780
|
| 265755 |
+
},
|
| 265756 |
+
{
|
| 265757 |
+
"epoch": 1053.93,
|
| 265758 |
+
"learning_rate": 7.920929487179487e-06,
|
| 265759 |
+
"loss": 0.3278,
|
| 265760 |
+
"step": 130785
|
| 265761 |
+
},
|
| 265762 |
+
{
|
| 265763 |
+
"epoch": 1053.97,
|
| 265764 |
+
"learning_rate": 7.92084935897436e-06,
|
| 265765 |
+
"loss": 0.642,
|
| 265766 |
+
"step": 130790
|
| 265767 |
+
},
|
| 265768 |
+
{
|
| 265769 |
+
"epoch": 1054.0,
|
| 265770 |
+
"eval_loss": 0.3659282624721527,
|
| 265771 |
+
"eval_runtime": 34.6363,
|
| 265772 |
+
"eval_samples_per_second": 24.137,
|
| 265773 |
+
"eval_steps_per_second": 0.78,
|
| 265774 |
+
"eval_wer": 0.18181818181818182,
|
| 265775 |
+
"step": 130793
|
| 265776 |
+
},
|
| 265777 |
+
{
|
| 265778 |
+
"epoch": 1054.02,
|
| 265779 |
+
"learning_rate": 7.920769230769231e-06,
|
| 265780 |
+
"loss": 0.3544,
|
| 265781 |
+
"step": 130795
|
| 265782 |
+
},
|
| 265783 |
+
{
|
| 265784 |
+
"epoch": 1054.06,
|
| 265785 |
+
"learning_rate": 7.920689102564103e-06,
|
| 265786 |
+
"loss": 0.2746,
|
| 265787 |
+
"step": 130800
|
| 265788 |
+
},
|
| 265789 |
+
{
|
| 265790 |
+
"epoch": 1054.1,
|
| 265791 |
+
"learning_rate": 7.920608974358976e-06,
|
| 265792 |
+
"loss": 0.3017,
|
| 265793 |
+
"step": 130805
|
| 265794 |
+
},
|
| 265795 |
+
{
|
| 265796 |
+
"epoch": 1054.14,
|
| 265797 |
+
"learning_rate": 7.920528846153847e-06,
|
| 265798 |
+
"loss": 0.3507,
|
| 265799 |
+
"step": 130810
|
| 265800 |
+
},
|
| 265801 |
+
{
|
| 265802 |
+
"epoch": 1054.18,
|
| 265803 |
+
"learning_rate": 7.920448717948718e-06,
|
| 265804 |
+
"loss": 0.6455,
|
| 265805 |
+
"step": 130815
|
| 265806 |
+
},
|
| 265807 |
+
{
|
| 265808 |
+
"epoch": 1054.22,
|
| 265809 |
+
"learning_rate": 7.92036858974359e-06,
|
| 265810 |
+
"loss": 0.9787,
|
| 265811 |
+
"step": 130820
|
| 265812 |
+
},
|
| 265813 |
+
{
|
| 265814 |
+
"epoch": 1054.26,
|
| 265815 |
+
"learning_rate": 7.920288461538463e-06,
|
| 265816 |
+
"loss": 0.2413,
|
| 265817 |
+
"step": 130825
|
| 265818 |
+
},
|
| 265819 |
+
{
|
| 265820 |
+
"epoch": 1054.3,
|
| 265821 |
+
"learning_rate": 7.920208333333334e-06,
|
| 265822 |
+
"loss": 0.2911,
|
| 265823 |
+
"step": 130830
|
| 265824 |
+
},
|
| 265825 |
+
{
|
| 265826 |
+
"epoch": 1054.34,
|
| 265827 |
+
"learning_rate": 7.920128205128206e-06,
|
| 265828 |
+
"loss": 0.3419,
|
| 265829 |
+
"step": 130835
|
| 265830 |
+
},
|
| 265831 |
+
{
|
| 265832 |
+
"epoch": 1054.38,
|
| 265833 |
+
"learning_rate": 7.920048076923079e-06,
|
| 265834 |
+
"loss": 0.635,
|
| 265835 |
+
"step": 130840
|
| 265836 |
+
},
|
| 265837 |
+
{
|
| 265838 |
+
"epoch": 1054.42,
|
| 265839 |
+
"learning_rate": 7.91996794871795e-06,
|
| 265840 |
+
"loss": 1.065,
|
| 265841 |
+
"step": 130845
|
| 265842 |
+
},
|
| 265843 |
+
{
|
| 265844 |
+
"epoch": 1054.46,
|
| 265845 |
+
"learning_rate": 7.919887820512821e-06,
|
| 265846 |
+
"loss": 0.3015,
|
| 265847 |
+
"step": 130850
|
| 265848 |
+
},
|
| 265849 |
+
{
|
| 265850 |
+
"epoch": 1054.5,
|
| 265851 |
+
"learning_rate": 7.919807692307693e-06,
|
| 265852 |
+
"loss": 0.3013,
|
| 265853 |
+
"step": 130855
|
| 265854 |
+
},
|
| 265855 |
+
{
|
| 265856 |
+
"epoch": 1054.54,
|
| 265857 |
+
"learning_rate": 7.919727564102566e-06,
|
| 265858 |
+
"loss": 0.3144,
|
| 265859 |
+
"step": 130860
|
| 265860 |
+
},
|
| 265861 |
+
{
|
| 265862 |
+
"epoch": 1054.58,
|
| 265863 |
+
"learning_rate": 7.919647435897435e-06,
|
| 265864 |
+
"loss": 0.6121,
|
| 265865 |
+
"step": 130865
|
| 265866 |
+
},
|
| 265867 |
+
{
|
| 265868 |
+
"epoch": 1054.62,
|
| 265869 |
+
"learning_rate": 7.919567307692308e-06,
|
| 265870 |
+
"loss": 0.9506,
|
| 265871 |
+
"step": 130870
|
| 265872 |
+
},
|
| 265873 |
+
{
|
| 265874 |
+
"epoch": 1054.66,
|
| 265875 |
+
"learning_rate": 7.91948717948718e-06,
|
| 265876 |
+
"loss": 0.2685,
|
| 265877 |
+
"step": 130875
|
| 265878 |
+
},
|
| 265879 |
+
{
|
| 265880 |
+
"epoch": 1054.7,
|
| 265881 |
+
"learning_rate": 7.919407051282051e-06,
|
| 265882 |
+
"loss": 0.3364,
|
| 265883 |
+
"step": 130880
|
| 265884 |
+
},
|
| 265885 |
+
{
|
| 265886 |
+
"epoch": 1054.74,
|
| 265887 |
+
"learning_rate": 7.919326923076923e-06,
|
| 265888 |
+
"loss": 0.3002,
|
| 265889 |
+
"step": 130885
|
| 265890 |
+
},
|
| 265891 |
+
{
|
| 265892 |
+
"epoch": 1054.78,
|
| 265893 |
+
"learning_rate": 7.919246794871796e-06,
|
| 265894 |
+
"loss": 0.5299,
|
| 265895 |
+
"step": 130890
|
| 265896 |
+
},
|
| 265897 |
+
{
|
| 265898 |
+
"epoch": 1054.82,
|
| 265899 |
+
"learning_rate": 7.919166666666667e-06,
|
| 265900 |
+
"loss": 0.9988,
|
| 265901 |
+
"step": 130895
|
| 265902 |
+
},
|
| 265903 |
+
{
|
| 265904 |
+
"epoch": 1054.86,
|
| 265905 |
+
"learning_rate": 7.919086538461538e-06,
|
| 265906 |
+
"loss": 0.2798,
|
| 265907 |
+
"step": 130900
|
| 265908 |
+
},
|
| 265909 |
+
{
|
| 265910 |
+
"epoch": 1054.9,
|
| 265911 |
+
"learning_rate": 7.919006410256411e-06,
|
| 265912 |
+
"loss": 0.2805,
|
| 265913 |
+
"step": 130905
|
| 265914 |
+
},
|
| 265915 |
+
{
|
| 265916 |
+
"epoch": 1054.94,
|
| 265917 |
+
"learning_rate": 7.918926282051283e-06,
|
| 265918 |
+
"loss": 0.2999,
|
| 265919 |
+
"step": 130910
|
| 265920 |
+
},
|
| 265921 |
+
{
|
| 265922 |
+
"epoch": 1054.98,
|
| 265923 |
+
"learning_rate": 7.918846153846154e-06,
|
| 265924 |
+
"loss": 0.6856,
|
| 265925 |
+
"step": 130915
|
| 265926 |
+
},
|
| 265927 |
+
{
|
| 265928 |
+
"epoch": 1055.0,
|
| 265929 |
+
"eval_loss": 0.4071400761604309,
|
| 265930 |
+
"eval_runtime": 36.0409,
|
| 265931 |
+
"eval_samples_per_second": 23.168,
|
| 265932 |
+
"eval_steps_per_second": 0.749,
|
| 265933 |
+
"eval_wer": 0.17985185185185185,
|
| 265934 |
+
"step": 130917
|
| 265935 |
+
},
|
| 265936 |
+
{
|
| 265937 |
+
"epoch": 1047.02,
|
| 265938 |
+
"learning_rate": 7.918766025641025e-06,
|
| 265939 |
+
"loss": 0.3123,
|
| 265940 |
+
"step": 130920
|
| 265941 |
+
},
|
| 265942 |
+
{
|
| 265943 |
+
"epoch": 1047.06,
|
| 265944 |
+
"learning_rate": 7.918685897435898e-06,
|
| 265945 |
+
"loss": 0.2641,
|
| 265946 |
+
"step": 130925
|
| 265947 |
+
},
|
| 265948 |
+
{
|
| 265949 |
+
"epoch": 1047.1,
|
| 265950 |
+
"learning_rate": 7.91860576923077e-06,
|
| 265951 |
+
"loss": 0.3179,
|
| 265952 |
+
"step": 130930
|
| 265953 |
+
},
|
| 265954 |
+
{
|
| 265955 |
+
"epoch": 1047.14,
|
| 265956 |
+
"learning_rate": 7.918525641025641e-06,
|
| 265957 |
+
"loss": 0.3773,
|
| 265958 |
+
"step": 130935
|
| 265959 |
+
},
|
| 265960 |
+
{
|
| 265961 |
+
"epoch": 1047.18,
|
| 265962 |
+
"learning_rate": 7.918445512820514e-06,
|
| 265963 |
+
"loss": 0.6746,
|
| 265964 |
+
"step": 130940
|
| 265965 |
+
},
|
| 265966 |
+
{
|
| 265967 |
+
"epoch": 1047.22,
|
| 265968 |
+
"learning_rate": 7.918365384615386e-06,
|
| 265969 |
+
"loss": 0.8553,
|
| 265970 |
+
"step": 130945
|
| 265971 |
+
},
|
| 265972 |
+
{
|
| 265973 |
+
"epoch": 1047.26,
|
| 265974 |
+
"learning_rate": 7.918285256410257e-06,
|
| 265975 |
+
"loss": 0.2629,
|
| 265976 |
+
"step": 130950
|
| 265977 |
+
},
|
| 265978 |
+
{
|
| 265979 |
+
"epoch": 1047.3,
|
| 265980 |
+
"learning_rate": 7.918205128205128e-06,
|
| 265981 |
+
"loss": 0.2765,
|
| 265982 |
+
"step": 130955
|
| 265983 |
+
},
|
| 265984 |
+
{
|
| 265985 |
+
"epoch": 1047.34,
|
| 265986 |
+
"learning_rate": 7.918125000000001e-06,
|
| 265987 |
+
"loss": 0.344,
|
| 265988 |
+
"step": 130960
|
| 265989 |
+
},
|
| 265990 |
+
{
|
| 265991 |
+
"epoch": 1047.38,
|
| 265992 |
+
"learning_rate": 7.918044871794873e-06,
|
| 265993 |
+
"loss": 0.6662,
|
| 265994 |
+
"step": 130965
|
| 265995 |
+
},
|
| 265996 |
+
{
|
| 265997 |
+
"epoch": 1047.42,
|
| 265998 |
+
"learning_rate": 7.917964743589744e-06,
|
| 265999 |
+
"loss": 0.8144,
|
| 266000 |
+
"step": 130970
|
| 266001 |
+
},
|
| 266002 |
+
{
|
| 266003 |
+
"epoch": 1047.46,
|
| 266004 |
+
"learning_rate": 7.917884615384615e-06,
|
| 266005 |
+
"loss": 0.2599,
|
| 266006 |
+
"step": 130975
|
| 266007 |
+
},
|
| 266008 |
+
{
|
| 266009 |
+
"epoch": 1047.5,
|
| 266010 |
+
"learning_rate": 7.917804487179489e-06,
|
| 266011 |
+
"loss": 0.3034,
|
| 266012 |
+
"step": 130980
|
| 266013 |
+
},
|
| 266014 |
+
{
|
| 266015 |
+
"epoch": 1047.54,
|
| 266016 |
+
"learning_rate": 7.91772435897436e-06,
|
| 266017 |
+
"loss": 0.3557,
|
| 266018 |
+
"step": 130985
|
| 266019 |
+
},
|
| 266020 |
+
{
|
| 266021 |
+
"epoch": 1047.58,
|
| 266022 |
+
"learning_rate": 7.917644230769231e-06,
|
| 266023 |
+
"loss": 0.6239,
|
| 266024 |
+
"step": 130990
|
| 266025 |
+
},
|
| 266026 |
+
{
|
| 266027 |
+
"epoch": 1047.62,
|
| 266028 |
+
"learning_rate": 7.917564102564104e-06,
|
| 266029 |
+
"loss": 0.7923,
|
| 266030 |
+
"step": 130995
|
| 266031 |
+
},
|
| 266032 |
+
{
|
| 266033 |
+
"epoch": 1047.66,
|
| 266034 |
+
"learning_rate": 7.917483974358976e-06,
|
| 266035 |
+
"loss": 0.2728,
|
| 266036 |
+
"step": 131000
|
| 266037 |
+
},
|
| 266038 |
+
{
|
| 266039 |
+
"epoch": 1047.7,
|
| 266040 |
+
"learning_rate": 7.917403846153847e-06,
|
| 266041 |
+
"loss": 0.2946,
|
| 266042 |
+
"step": 131005
|
| 266043 |
+
},
|
| 266044 |
+
{
|
| 266045 |
+
"epoch": 1047.74,
|
| 266046 |
+
"learning_rate": 7.917323717948718e-06,
|
| 266047 |
+
"loss": 0.3642,
|
| 266048 |
+
"step": 131010
|
| 266049 |
+
},
|
| 266050 |
+
{
|
| 266051 |
+
"epoch": 1047.78,
|
| 266052 |
+
"learning_rate": 7.917243589743591e-06,
|
| 266053 |
+
"loss": 0.7808,
|
| 266054 |
+
"step": 131015
|
| 266055 |
+
},
|
| 266056 |
+
{
|
| 266057 |
+
"epoch": 1047.82,
|
| 266058 |
+
"learning_rate": 7.917163461538461e-06,
|
| 266059 |
+
"loss": 0.8567,
|
| 266060 |
+
"step": 131020
|
| 266061 |
+
},
|
| 266062 |
+
{
|
| 266063 |
+
"epoch": 1047.86,
|
| 266064 |
+
"learning_rate": 7.917083333333334e-06,
|
| 266065 |
+
"loss": 0.3216,
|
| 266066 |
+
"step": 131025
|
| 266067 |
+
},
|
| 266068 |
+
{
|
| 266069 |
+
"epoch": 1047.9,
|
| 266070 |
+
"learning_rate": 7.917003205128205e-06,
|
| 266071 |
+
"loss": 0.3074,
|
| 266072 |
+
"step": 131030
|
| 266073 |
+
},
|
| 266074 |
+
{
|
| 266075 |
+
"epoch": 1047.94,
|
| 266076 |
+
"learning_rate": 7.916923076923077e-06,
|
| 266077 |
+
"loss": 0.4341,
|
| 266078 |
+
"step": 131035
|
| 266079 |
+
},
|
| 266080 |
+
{
|
| 266081 |
+
"epoch": 1047.98,
|
| 266082 |
+
"learning_rate": 7.91684294871795e-06,
|
| 266083 |
+
"loss": 0.8587,
|
| 266084 |
+
"step": 131040
|
| 266085 |
+
},
|
| 266086 |
+
{
|
| 266087 |
+
"epoch": 1048.0,
|
| 266088 |
+
"eval_loss": 0.35665926337242126,
|
| 266089 |
+
"eval_runtime": 34.609,
|
| 266090 |
+
"eval_samples_per_second": 24.127,
|
| 266091 |
+
"eval_steps_per_second": 0.78,
|
| 266092 |
+
"eval_wer": 0.1811192429954905,
|
| 266093 |
+
"step": 131042
|
| 266094 |
}
|
| 266095 |
],
|
| 266096 |
+
"max_steps": 625000,
|
| 266097 |
"num_train_epochs": 5000,
|
| 266098 |
+
"total_flos": 3.68759774497066e+20,
|
| 266099 |
"trial_name": null,
|
| 266100 |
"trial_params": null
|
| 266101 |
}
|
model-bin/finetune/base/{checkpoint-130420 → checkpoint-131042}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630163053.748851/events.out.tfevents.1630163053.86bb0ddabf9b.4092.201
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d147df7b211e85e8866e8b3eb7431558c10019dc44837c948a910119ef896008
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630163453.0593839/events.out.tfevents.1630163453.86bb0ddabf9b.4092.203
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0139a90b6ce9ec610432846b6f197b7981f141e00653c07de282e5e94a580f61
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630163842.4436672/events.out.tfevents.1630163842.86bb0ddabf9b.4092.205
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e46c124a69b925953c6cf715f18aa64040b3ff5635dd580a755bb72ef60ed167
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630164227.907955/events.out.tfevents.1630164227.86bb0ddabf9b.4092.207
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:66b00d09ba049da096f21e57a9a1ae7b52c2defed919ac01203532e445dec403
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630164623.2772548/events.out.tfevents.1630164623.86bb0ddabf9b.4092.209
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:559bb880cec80281977293d1a6bf946bfb263eacd5906987333146ecf170b4ff
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630163053.86bb0ddabf9b.4092.200
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fc2db19a5c2d7210f35f7b0f2c36361b147fd01da5b9a0f5b9442102a200a44b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630163453.86bb0ddabf9b.4092.202
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3aedc94451aa2ac63426955c440cb170d4f9c403a6aa4c1263f5af35306dc72b
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630163842.86bb0ddabf9b.4092.204
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:50b53cd1ee694c219abccca9a8f4466ff476d7283f2abfefe0e68d1eb6b3c1d6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630164227.86bb0ddabf9b.4092.206
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bb1c52504c395f6d00e0eeb971b275e159e1e92e0b3fe394bc322018c1c2bfd1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630164623.86bb0ddabf9b.4092.208
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4eb9b1f365440b00151912f0320922c8e491b426639dabe5ce8eaaf5f0754813
|
| 3 |
+
size 8622
|