"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/trainer_state.json +420 -3
- model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629661703.623691/events.out.tfevents.1629661703.91f414e9573a.903.211 +3 -0
- model-bin/finetune/base/log/1629662363.6404548/events.out.tfevents.1629662363.91f414e9573a.903.213 +3 -0
- model-bin/finetune/base/log/1629663014.4377518/events.out.tfevents.1629663014.91f414e9573a.903.215 +3 -0
- model-bin/finetune/base/log/1629663672.0698838/events.out.tfevents.1629663672.91f414e9573a.903.217 +3 -0
- model-bin/finetune/base/log/1629664322.8475788/events.out.tfevents.1629664322.91f414e9573a.903.219 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629661703.91f414e9573a.903.210 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629662363.91f414e9573a.903.212 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629663014.91f414e9573a.903.214 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629663672.91f414e9573a.903.216 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629664322.91f414e9573a.903.218 +3 -0
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3eb4738a7b7717d3ff9e7071fd5cb9b6d52672d86c742d7f1dec0c9302503be1
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/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:b9fb229bb33c15a9986ebda12e6e5e7ff864541e022a83bb7bfc9e05b14f915c
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4e96eb9b698f8fb31c54762aa19f531af254240b50ae1b740f7c323b97453f3d
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/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 sha256:9d9cd3fedc5e7eb14c9de984a70f18c3445b012f7a0208035cf4a96e7423bac5
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/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 sha256:486cec8b4c21c5cec6ad713608d048a82c6deb89ae83e7705fbfeec621fbe8ce
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -146295,11 +146295,428 @@
|
|
| 146295 |
"eval_steps_per_second": 0.671,
|
| 146296 |
"eval_wer": 0.20595987699516766,
|
| 146297 |
"step": 35710
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 146298 |
}
|
| 146299 |
],
|
| 146300 |
"max_steps": 620000,
|
| 146301 |
"num_train_epochs": 5000,
|
| 146302 |
-
"total_flos": 1.
|
| 146303 |
"trial_name": null,
|
| 146304 |
"trial_params": null
|
| 146305 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.20553532410779315,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
+
"epoch": 292.99598393574297,
|
| 5 |
+
"global_step": 36333,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 146295 |
"eval_steps_per_second": 0.671,
|
| 146296 |
"eval_wer": 0.20595987699516766,
|
| 146297 |
"step": 35710
|
| 146298 |
+
},
|
| 146299 |
+
{
|
| 146300 |
+
"epoch": 285.08,
|
| 146301 |
+
"learning_rate": 9.439305331179323e-06,
|
| 146302 |
+
"loss": 0.4005,
|
| 146303 |
+
"step": 35720
|
| 146304 |
+
},
|
| 146305 |
+
{
|
| 146306 |
+
"epoch": 285.16,
|
| 146307 |
+
"learning_rate": 9.439143780290793e-06,
|
| 146308 |
+
"loss": 0.4237,
|
| 146309 |
+
"step": 35730
|
| 146310 |
+
},
|
| 146311 |
+
{
|
| 146312 |
+
"epoch": 285.24,
|
| 146313 |
+
"learning_rate": 9.438982229402263e-06,
|
| 146314 |
+
"loss": 0.8435,
|
| 146315 |
+
"step": 35740
|
| 146316 |
+
},
|
| 146317 |
+
{
|
| 146318 |
+
"epoch": 285.32,
|
| 146319 |
+
"learning_rate": 9.438820678513733e-06,
|
| 146320 |
+
"loss": 0.4195,
|
| 146321 |
+
"step": 35750
|
| 146322 |
+
},
|
| 146323 |
+
{
|
| 146324 |
+
"epoch": 285.4,
|
| 146325 |
+
"learning_rate": 9.438659127625203e-06,
|
| 146326 |
+
"loss": 1.0739,
|
| 146327 |
+
"step": 35760
|
| 146328 |
+
},
|
| 146329 |
+
{
|
| 146330 |
+
"epoch": 285.48,
|
| 146331 |
+
"learning_rate": 9.438497576736673e-06,
|
| 146332 |
+
"loss": 0.4082,
|
| 146333 |
+
"step": 35770
|
| 146334 |
+
},
|
| 146335 |
+
{
|
| 146336 |
+
"epoch": 285.56,
|
| 146337 |
+
"learning_rate": 9.438336025848143e-06,
|
| 146338 |
+
"loss": 0.5064,
|
| 146339 |
+
"step": 35780
|
| 146340 |
+
},
|
| 146341 |
+
{
|
| 146342 |
+
"epoch": 285.64,
|
| 146343 |
+
"learning_rate": 9.438174474959613e-06,
|
| 146344 |
+
"loss": 0.8912,
|
| 146345 |
+
"step": 35790
|
| 146346 |
+
},
|
| 146347 |
+
{
|
| 146348 |
+
"epoch": 285.72,
|
| 146349 |
+
"learning_rate": 9.438012924071083e-06,
|
| 146350 |
+
"loss": 0.4114,
|
| 146351 |
+
"step": 35800
|
| 146352 |
+
},
|
| 146353 |
+
{
|
| 146354 |
+
"epoch": 285.8,
|
| 146355 |
+
"learning_rate": 9.437851373182553e-06,
|
| 146356 |
+
"loss": 1.0904,
|
| 146357 |
+
"step": 35810
|
| 146358 |
+
},
|
| 146359 |
+
{
|
| 146360 |
+
"epoch": 285.88,
|
| 146361 |
+
"learning_rate": 9.437689822294025e-06,
|
| 146362 |
+
"loss": 0.4086,
|
| 146363 |
+
"step": 35820
|
| 146364 |
+
},
|
| 146365 |
+
{
|
| 146366 |
+
"epoch": 285.96,
|
| 146367 |
+
"learning_rate": 9.437528271405495e-06,
|
| 146368 |
+
"loss": 0.5455,
|
| 146369 |
+
"step": 35830
|
| 146370 |
+
},
|
| 146371 |
+
{
|
| 146372 |
+
"epoch": 286.0,
|
| 146373 |
+
"eval_loss": 0.46360307931900024,
|
| 146374 |
+
"eval_runtime": 42.165,
|
| 146375 |
+
"eval_samples_per_second": 19.922,
|
| 146376 |
+
"eval_steps_per_second": 0.64,
|
| 146377 |
+
"eval_wer": 0.21165557286503328,
|
| 146378 |
+
"step": 35835
|
| 146379 |
+
},
|
| 146380 |
+
{
|
| 146381 |
+
"epoch": 288.04,
|
| 146382 |
+
"learning_rate": 9.437366720516965e-06,
|
| 146383 |
+
"loss": 0.4096,
|
| 146384 |
+
"step": 35840
|
| 146385 |
+
},
|
| 146386 |
+
{
|
| 146387 |
+
"epoch": 288.12,
|
| 146388 |
+
"learning_rate": 9.437205169628435e-06,
|
| 146389 |
+
"loss": 0.3787,
|
| 146390 |
+
"step": 35850
|
| 146391 |
+
},
|
| 146392 |
+
{
|
| 146393 |
+
"epoch": 288.2,
|
| 146394 |
+
"learning_rate": 9.437043618739905e-06,
|
| 146395 |
+
"loss": 0.8851,
|
| 146396 |
+
"step": 35860
|
| 146397 |
+
},
|
| 146398 |
+
{
|
| 146399 |
+
"epoch": 288.28,
|
| 146400 |
+
"learning_rate": 9.436882067851374e-06,
|
| 146401 |
+
"loss": 0.3719,
|
| 146402 |
+
"step": 35870
|
| 146403 |
+
},
|
| 146404 |
+
{
|
| 146405 |
+
"epoch": 288.36,
|
| 146406 |
+
"learning_rate": 9.436720516962844e-06,
|
| 146407 |
+
"loss": 0.4929,
|
| 146408 |
+
"step": 35880
|
| 146409 |
+
},
|
| 146410 |
+
{
|
| 146411 |
+
"epoch": 288.44,
|
| 146412 |
+
"learning_rate": 9.436558966074314e-06,
|
| 146413 |
+
"loss": 0.8879,
|
| 146414 |
+
"step": 35890
|
| 146415 |
+
},
|
| 146416 |
+
{
|
| 146417 |
+
"epoch": 288.52,
|
| 146418 |
+
"learning_rate": 9.436397415185784e-06,
|
| 146419 |
+
"loss": 0.3472,
|
| 146420 |
+
"step": 35900
|
| 146421 |
+
},
|
| 146422 |
+
{
|
| 146423 |
+
"epoch": 288.6,
|
| 146424 |
+
"learning_rate": 9.436235864297254e-06,
|
| 146425 |
+
"loss": 1.0493,
|
| 146426 |
+
"step": 35910
|
| 146427 |
+
},
|
| 146428 |
+
{
|
| 146429 |
+
"epoch": 288.68,
|
| 146430 |
+
"learning_rate": 9.436074313408724e-06,
|
| 146431 |
+
"loss": 0.3832,
|
| 146432 |
+
"step": 35920
|
| 146433 |
+
},
|
| 146434 |
+
{
|
| 146435 |
+
"epoch": 288.76,
|
| 146436 |
+
"learning_rate": 9.435912762520194e-06,
|
| 146437 |
+
"loss": 0.4808,
|
| 146438 |
+
"step": 35930
|
| 146439 |
+
},
|
| 146440 |
+
{
|
| 146441 |
+
"epoch": 288.84,
|
| 146442 |
+
"learning_rate": 9.435751211631666e-06,
|
| 146443 |
+
"loss": 0.8737,
|
| 146444 |
+
"step": 35940
|
| 146445 |
+
},
|
| 146446 |
+
{
|
| 146447 |
+
"epoch": 288.92,
|
| 146448 |
+
"learning_rate": 9.435589660743136e-06,
|
| 146449 |
+
"loss": 0.3942,
|
| 146450 |
+
"step": 35950
|
| 146451 |
+
},
|
| 146452 |
+
{
|
| 146453 |
+
"epoch": 289.0,
|
| 146454 |
+
"eval_loss": 0.5137014985084534,
|
| 146455 |
+
"eval_runtime": 41.691,
|
| 146456 |
+
"eval_samples_per_second": 20.148,
|
| 146457 |
+
"eval_steps_per_second": 0.648,
|
| 146458 |
+
"eval_wer": 0.2143329172479624,
|
| 146459 |
+
"step": 35959
|
| 146460 |
+
},
|
| 146461 |
+
{
|
| 146462 |
+
"epoch": 287.01,
|
| 146463 |
+
"learning_rate": 9.435428109854606e-06,
|
| 146464 |
+
"loss": 0.5178,
|
| 146465 |
+
"step": 35960
|
| 146466 |
+
},
|
| 146467 |
+
{
|
| 146468 |
+
"epoch": 287.09,
|
| 146469 |
+
"learning_rate": 9.435266558966076e-06,
|
| 146470 |
+
"loss": 0.3734,
|
| 146471 |
+
"step": 35970
|
| 146472 |
+
},
|
| 146473 |
+
{
|
| 146474 |
+
"epoch": 287.17,
|
| 146475 |
+
"learning_rate": 9.435105008077546e-06,
|
| 146476 |
+
"loss": 0.5814,
|
| 146477 |
+
"step": 35980
|
| 146478 |
+
},
|
| 146479 |
+
{
|
| 146480 |
+
"epoch": 287.25,
|
| 146481 |
+
"learning_rate": 9.434943457189016e-06,
|
| 146482 |
+
"loss": 0.9011,
|
| 146483 |
+
"step": 35990
|
| 146484 |
+
},
|
| 146485 |
+
{
|
| 146486 |
+
"epoch": 287.33,
|
| 146487 |
+
"learning_rate": 9.434781906300486e-06,
|
| 146488 |
+
"loss": 0.4301,
|
| 146489 |
+
"step": 36000
|
| 146490 |
+
},
|
| 146491 |
+
{
|
| 146492 |
+
"epoch": 287.41,
|
| 146493 |
+
"learning_rate": 9.434620355411956e-06,
|
| 146494 |
+
"loss": 0.9715,
|
| 146495 |
+
"step": 36010
|
| 146496 |
+
},
|
| 146497 |
+
{
|
| 146498 |
+
"epoch": 287.49,
|
| 146499 |
+
"learning_rate": 9.434458804523426e-06,
|
| 146500 |
+
"loss": 0.3657,
|
| 146501 |
+
"step": 36020
|
| 146502 |
+
},
|
| 146503 |
+
{
|
| 146504 |
+
"epoch": 287.57,
|
| 146505 |
+
"learning_rate": 9.434297253634896e-06,
|
| 146506 |
+
"loss": 0.4817,
|
| 146507 |
+
"step": 36030
|
| 146508 |
+
},
|
| 146509 |
+
{
|
| 146510 |
+
"epoch": 287.65,
|
| 146511 |
+
"learning_rate": 9.434135702746366e-06,
|
| 146512 |
+
"loss": 0.905,
|
| 146513 |
+
"step": 36040
|
| 146514 |
+
},
|
| 146515 |
+
{
|
| 146516 |
+
"epoch": 287.73,
|
| 146517 |
+
"learning_rate": 9.433974151857836e-06,
|
| 146518 |
+
"loss": 0.3928,
|
| 146519 |
+
"step": 36050
|
| 146520 |
+
},
|
| 146521 |
+
{
|
| 146522 |
+
"epoch": 287.81,
|
| 146523 |
+
"learning_rate": 9.433812600969306e-06,
|
| 146524 |
+
"loss": 0.9375,
|
| 146525 |
+
"step": 36060
|
| 146526 |
+
},
|
| 146527 |
+
{
|
| 146528 |
+
"epoch": 287.89,
|
| 146529 |
+
"learning_rate": 9.433651050080776e-06,
|
| 146530 |
+
"loss": 0.3678,
|
| 146531 |
+
"step": 36070
|
| 146532 |
+
},
|
| 146533 |
+
{
|
| 146534 |
+
"epoch": 287.97,
|
| 146535 |
+
"learning_rate": 9.433489499192247e-06,
|
| 146536 |
+
"loss": 0.527,
|
| 146537 |
+
"step": 36080
|
| 146538 |
+
},
|
| 146539 |
+
{
|
| 146540 |
+
"epoch": 288.0,
|
| 146541 |
+
"eval_loss": 0.4413570463657379,
|
| 146542 |
+
"eval_runtime": 40.6098,
|
| 146543 |
+
"eval_samples_per_second": 20.709,
|
| 146544 |
+
"eval_steps_per_second": 0.665,
|
| 146545 |
+
"eval_wer": 0.22419077521118339,
|
| 146546 |
+
"step": 36084
|
| 146547 |
+
},
|
| 146548 |
+
{
|
| 146549 |
+
"epoch": 288.05,
|
| 146550 |
+
"learning_rate": 9.433327948303717e-06,
|
| 146551 |
+
"loss": 0.3747,
|
| 146552 |
+
"step": 36090
|
| 146553 |
+
},
|
| 146554 |
+
{
|
| 146555 |
+
"epoch": 288.13,
|
| 146556 |
+
"learning_rate": 9.433166397415187e-06,
|
| 146557 |
+
"loss": 0.3968,
|
| 146558 |
+
"step": 36100
|
| 146559 |
+
},
|
| 146560 |
+
{
|
| 146561 |
+
"epoch": 288.21,
|
| 146562 |
+
"learning_rate": 9.433004846526657e-06,
|
| 146563 |
+
"loss": 0.9139,
|
| 146564 |
+
"step": 36110
|
| 146565 |
+
},
|
| 146566 |
+
{
|
| 146567 |
+
"epoch": 288.29,
|
| 146568 |
+
"learning_rate": 9.432843295638127e-06,
|
| 146569 |
+
"loss": 0.4236,
|
| 146570 |
+
"step": 36120
|
| 146571 |
+
},
|
| 146572 |
+
{
|
| 146573 |
+
"epoch": 288.37,
|
| 146574 |
+
"learning_rate": 9.432681744749597e-06,
|
| 146575 |
+
"loss": 0.518,
|
| 146576 |
+
"step": 36130
|
| 146577 |
+
},
|
| 146578 |
+
{
|
| 146579 |
+
"epoch": 288.45,
|
| 146580 |
+
"learning_rate": 9.432520193861067e-06,
|
| 146581 |
+
"loss": 0.9399,
|
| 146582 |
+
"step": 36140
|
| 146583 |
+
},
|
| 146584 |
+
{
|
| 146585 |
+
"epoch": 288.53,
|
| 146586 |
+
"learning_rate": 9.432358642972537e-06,
|
| 146587 |
+
"loss": 0.37,
|
| 146588 |
+
"step": 36150
|
| 146589 |
+
},
|
| 146590 |
+
{
|
| 146591 |
+
"epoch": 288.61,
|
| 146592 |
+
"learning_rate": 9.432197092084007e-06,
|
| 146593 |
+
"loss": 0.993,
|
| 146594 |
+
"step": 36160
|
| 146595 |
+
},
|
| 146596 |
+
{
|
| 146597 |
+
"epoch": 288.69,
|
| 146598 |
+
"learning_rate": 9.432035541195477e-06,
|
| 146599 |
+
"loss": 0.3595,
|
| 146600 |
+
"step": 36170
|
| 146601 |
+
},
|
| 146602 |
+
{
|
| 146603 |
+
"epoch": 288.77,
|
| 146604 |
+
"learning_rate": 9.431873990306947e-06,
|
| 146605 |
+
"loss": 0.5414,
|
| 146606 |
+
"step": 36180
|
| 146607 |
+
},
|
| 146608 |
+
{
|
| 146609 |
+
"epoch": 288.85,
|
| 146610 |
+
"learning_rate": 9.431712439418417e-06,
|
| 146611 |
+
"loss": 0.7993,
|
| 146612 |
+
"step": 36190
|
| 146613 |
+
},
|
| 146614 |
+
{
|
| 146615 |
+
"epoch": 288.93,
|
| 146616 |
+
"learning_rate": 9.431550888529887e-06,
|
| 146617 |
+
"loss": 0.3905,
|
| 146618 |
+
"step": 36200
|
| 146619 |
+
},
|
| 146620 |
+
{
|
| 146621 |
+
"epoch": 289.0,
|
| 146622 |
+
"eval_loss": 0.48133236169815063,
|
| 146623 |
+
"eval_runtime": 41.4591,
|
| 146624 |
+
"eval_samples_per_second": 20.285,
|
| 146625 |
+
"eval_steps_per_second": 0.651,
|
| 146626 |
+
"eval_wer": 0.21214545454545455,
|
| 146627 |
+
"step": 36209
|
| 146628 |
+
},
|
| 146629 |
+
{
|
| 146630 |
+
"epoch": 292.01,
|
| 146631 |
+
"learning_rate": 9.431389337641359e-06,
|
| 146632 |
+
"loss": 0.4171,
|
| 146633 |
+
"step": 36210
|
| 146634 |
+
},
|
| 146635 |
+
{
|
| 146636 |
+
"epoch": 292.09,
|
| 146637 |
+
"learning_rate": 9.431227786752829e-06,
|
| 146638 |
+
"loss": 0.3464,
|
| 146639 |
+
"step": 36220
|
| 146640 |
+
},
|
| 146641 |
+
{
|
| 146642 |
+
"epoch": 292.17,
|
| 146643 |
+
"learning_rate": 9.431066235864299e-06,
|
| 146644 |
+
"loss": 0.5731,
|
| 146645 |
+
"step": 36230
|
| 146646 |
+
},
|
| 146647 |
+
{
|
| 146648 |
+
"epoch": 292.25,
|
| 146649 |
+
"learning_rate": 9.430904684975769e-06,
|
| 146650 |
+
"loss": 0.8759,
|
| 146651 |
+
"step": 36240
|
| 146652 |
+
},
|
| 146653 |
+
{
|
| 146654 |
+
"epoch": 292.33,
|
| 146655 |
+
"learning_rate": 9.430743134087239e-06,
|
| 146656 |
+
"loss": 0.372,
|
| 146657 |
+
"step": 36250
|
| 146658 |
+
},
|
| 146659 |
+
{
|
| 146660 |
+
"epoch": 292.41,
|
| 146661 |
+
"learning_rate": 9.430581583198709e-06,
|
| 146662 |
+
"loss": 0.9577,
|
| 146663 |
+
"step": 36260
|
| 146664 |
+
},
|
| 146665 |
+
{
|
| 146666 |
+
"epoch": 292.49,
|
| 146667 |
+
"learning_rate": 9.430420032310179e-06,
|
| 146668 |
+
"loss": 0.362,
|
| 146669 |
+
"step": 36270
|
| 146670 |
+
},
|
| 146671 |
+
{
|
| 146672 |
+
"epoch": 292.57,
|
| 146673 |
+
"learning_rate": 9.430258481421649e-06,
|
| 146674 |
+
"loss": 0.5399,
|
| 146675 |
+
"step": 36280
|
| 146676 |
+
},
|
| 146677 |
+
{
|
| 146678 |
+
"epoch": 292.65,
|
| 146679 |
+
"learning_rate": 9.430096930533119e-06,
|
| 146680 |
+
"loss": 0.8912,
|
| 146681 |
+
"step": 36290
|
| 146682 |
+
},
|
| 146683 |
+
{
|
| 146684 |
+
"epoch": 292.73,
|
| 146685 |
+
"learning_rate": 9.429935379644589e-06,
|
| 146686 |
+
"loss": 0.4154,
|
| 146687 |
+
"step": 36300
|
| 146688 |
+
},
|
| 146689 |
+
{
|
| 146690 |
+
"epoch": 292.81,
|
| 146691 |
+
"learning_rate": 9.429773828756058e-06,
|
| 146692 |
+
"loss": 0.9742,
|
| 146693 |
+
"step": 36310
|
| 146694 |
+
},
|
| 146695 |
+
{
|
| 146696 |
+
"epoch": 292.89,
|
| 146697 |
+
"learning_rate": 9.429612277867528e-06,
|
| 146698 |
+
"loss": 0.4224,
|
| 146699 |
+
"step": 36320
|
| 146700 |
+
},
|
| 146701 |
+
{
|
| 146702 |
+
"epoch": 292.97,
|
| 146703 |
+
"learning_rate": 9.429450726978998e-06,
|
| 146704 |
+
"loss": 0.5349,
|
| 146705 |
+
"step": 36330
|
| 146706 |
+
},
|
| 146707 |
+
{
|
| 146708 |
+
"epoch": 293.0,
|
| 146709 |
+
"eval_loss": 0.4554632008075714,
|
| 146710 |
+
"eval_runtime": 38.9753,
|
| 146711 |
+
"eval_samples_per_second": 21.552,
|
| 146712 |
+
"eval_steps_per_second": 0.693,
|
| 146713 |
+
"eval_wer": 0.21878310882397456,
|
| 146714 |
+
"step": 36333
|
| 146715 |
}
|
| 146716 |
],
|
| 146717 |
"max_steps": 620000,
|
| 146718 |
"num_train_epochs": 5000,
|
| 146719 |
+
"total_flos": 1.0224169251081573e+20,
|
| 146720 |
"trial_name": null,
|
| 146721 |
"trial_params": null
|
| 146722 |
}
|
model-bin/finetune/base/{checkpoint-35710 → checkpoint-36333}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629661703.623691/events.out.tfevents.1629661703.91f414e9573a.903.211
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a87bfd60193ac9452578c6cd1c22a81bd7368e0aa2fa770176da66c764ce1b80
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629662363.6404548/events.out.tfevents.1629662363.91f414e9573a.903.213
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:71e42ff294c5ee1dfd13f3712dc343c1c7b4c7455353f5c4b05b91eabe10ec3e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629663014.4377518/events.out.tfevents.1629663014.91f414e9573a.903.215
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f9e3136cb13436ccb9e9e5b7d1ede06eb28f862c4a47f7b6573a6d22653b2ebf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629663672.0698838/events.out.tfevents.1629663672.91f414e9573a.903.217
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a205e40a66db69789ebc92c720abb25c8657cf9cc57915d65041dc385c986226
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629664322.8475788/events.out.tfevents.1629664322.91f414e9573a.903.219
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3346b988d1f6e7058ec12d80221a306663fcfde86e7c41f6986d6539ca16eafc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629661703.91f414e9573a.903.210
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:99e6d62fb1dc3c8a85d1f4e3321aadb826d8602d15de0c311da412c8e99c79c5
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629662363.91f414e9573a.903.212
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:26273b08f571c98ed15dbe0e31ae63b57c7f24626650ce03dcb9309da9ad2e51
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629663014.91f414e9573a.903.214
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6eb62519436cc680fed70ecda686c66c69684dde43fcaa910e5e83abb6d3d27b
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629663672.91f414e9573a.903.216
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb9a22ff9e335d2003a791bc13a3c9fdc19b4d6935295450c34e018003badeb2
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629664322.91f414e9573a.903.218
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8079f0970fa26648a8540bfc4b8f6dc6ee557a4d6f9f1e2e7c5a404a63c5acb2
|
| 3 |
+
size 6703
|