"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/trainer_state.json +510 -6
- model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629597755.4176855/events.out.tfevents.1629597755.91f414e9573a.903.21 +3 -0
- model-bin/finetune/base/log/1629598414.5427365/events.out.tfevents.1629598414.91f414e9573a.903.23 +3 -0
- model-bin/finetune/base/log/1629599084.569025/events.out.tfevents.1629599084.91f414e9573a.903.25 +3 -0
- model-bin/finetune/base/log/1629599824.9825046/events.out.tfevents.1629599824.91f414e9573a.903.27 +3 -0
- model-bin/finetune/base/log/1629600485.658627/events.out.tfevents.1629600485.91f414e9573a.903.29 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629597755.91f414e9573a.903.20 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629598414.91f414e9573a.903.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629599084.91f414e9573a.903.24 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629599824.91f414e9573a.903.26 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629600485.91f414e9573a.903.28 +3 -0
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/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 sha256:32569797a14066213cf5e900b24e06f379a27b0ba4090c69593219763860c6c4
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/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 sha256:62b6ffeabe1971b6aa303964a297ae94f1a184097d837c0462e29b0976a5f3e8
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/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:bdf498fbaa2ec1e014b2b4fe686f1f5101b506c87aa83efe481e246bcc3f4714
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/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 sha256:95a869d830030200dd624e63a3f6089c404550e17c9119b6c96710429dbfccdf
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/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 sha256:5ff7b22b4c07c99c42cb1ee9ebacd90ef4c0304144be6de1f23ba542b0a8d423
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -138261,11 +138261,515 @@
|
|
| 138261 |
"eval_steps_per_second": 0.677,
|
| 138262 |
"eval_wer": 0.24058212688546957,
|
| 138263 |
"step": 23765
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 138264 |
}
|
| 138265 |
],
|
| 138266 |
-
"max_steps":
|
| 138267 |
"num_train_epochs": 5000,
|
| 138268 |
-
"total_flos": 6.
|
| 138269 |
"trial_name": null,
|
| 138270 |
"trial_params": null
|
| 138271 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.2172567645365573,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-23890",
|
| 4 |
+
"epoch": 196.99598393574297,
|
| 5 |
+
"global_step": 24512,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 138261 |
"eval_steps_per_second": 0.677,
|
| 138262 |
"eval_wer": 0.24058212688546957,
|
| 138263 |
"step": 23765
|
| 138264 |
+
},
|
| 138265 |
+
{
|
| 138266 |
+
"epoch": 190.04,
|
| 138267 |
+
"learning_rate": 9.632294022617124e-06,
|
| 138268 |
+
"loss": 0.5074,
|
| 138269 |
+
"step": 23770
|
| 138270 |
+
},
|
| 138271 |
+
{
|
| 138272 |
+
"epoch": 190.12,
|
| 138273 |
+
"learning_rate": 9.632132471728596e-06,
|
| 138274 |
+
"loss": 0.4812,
|
| 138275 |
+
"step": 23780
|
| 138276 |
+
},
|
| 138277 |
+
{
|
| 138278 |
+
"epoch": 190.2,
|
| 138279 |
+
"learning_rate": 9.631970920840066e-06,
|
| 138280 |
+
"loss": 1.0191,
|
| 138281 |
+
"step": 23790
|
| 138282 |
+
},
|
| 138283 |
+
{
|
| 138284 |
+
"epoch": 190.28,
|
| 138285 |
+
"learning_rate": 9.631809369951536e-06,
|
| 138286 |
+
"loss": 0.5023,
|
| 138287 |
+
"step": 23800
|
| 138288 |
+
},
|
| 138289 |
+
{
|
| 138290 |
+
"epoch": 190.36,
|
| 138291 |
+
"learning_rate": 9.631647819063006e-06,
|
| 138292 |
+
"loss": 0.6277,
|
| 138293 |
+
"step": 23810
|
| 138294 |
+
},
|
| 138295 |
+
{
|
| 138296 |
+
"epoch": 190.44,
|
| 138297 |
+
"learning_rate": 9.631486268174476e-06,
|
| 138298 |
+
"loss": 0.9313,
|
| 138299 |
+
"step": 23820
|
| 138300 |
+
},
|
| 138301 |
+
{
|
| 138302 |
+
"epoch": 190.52,
|
| 138303 |
+
"learning_rate": 9.631324717285946e-06,
|
| 138304 |
+
"loss": 0.5032,
|
| 138305 |
+
"step": 23830
|
| 138306 |
+
},
|
| 138307 |
+
{
|
| 138308 |
+
"epoch": 190.6,
|
| 138309 |
+
"learning_rate": 9.631163166397416e-06,
|
| 138310 |
+
"loss": 1.0566,
|
| 138311 |
+
"step": 23840
|
| 138312 |
+
},
|
| 138313 |
+
{
|
| 138314 |
+
"epoch": 190.68,
|
| 138315 |
+
"learning_rate": 9.631001615508886e-06,
|
| 138316 |
+
"loss": 0.4492,
|
| 138317 |
+
"step": 23850
|
| 138318 |
+
},
|
| 138319 |
+
{
|
| 138320 |
+
"epoch": 190.76,
|
| 138321 |
+
"learning_rate": 9.630840064620356e-06,
|
| 138322 |
+
"loss": 0.5329,
|
| 138323 |
+
"step": 23860
|
| 138324 |
+
},
|
| 138325 |
+
{
|
| 138326 |
+
"epoch": 190.84,
|
| 138327 |
+
"learning_rate": 9.630678513731826e-06,
|
| 138328 |
+
"loss": 0.9454,
|
| 138329 |
+
"step": 23870
|
| 138330 |
+
},
|
| 138331 |
+
{
|
| 138332 |
+
"epoch": 190.92,
|
| 138333 |
+
"learning_rate": 9.630516962843296e-06,
|
| 138334 |
+
"loss": 0.4499,
|
| 138335 |
+
"step": 23880
|
| 138336 |
+
},
|
| 138337 |
+
{
|
| 138338 |
+
"epoch": 191.0,
|
| 138339 |
+
"learning_rate": 9.630355411954767e-06,
|
| 138340 |
+
"loss": 0.9996,
|
| 138341 |
+
"step": 23890
|
| 138342 |
+
},
|
| 138343 |
+
{
|
| 138344 |
+
"epoch": 191.0,
|
| 138345 |
+
"eval_loss": 0.5291967988014221,
|
| 138346 |
+
"eval_runtime": 41.96,
|
| 138347 |
+
"eval_samples_per_second": 19.924,
|
| 138348 |
+
"eval_steps_per_second": 0.643,
|
| 138349 |
+
"eval_wer": 0.2172567645365573,
|
| 138350 |
+
"step": 23890
|
| 138351 |
+
},
|
| 138352 |
+
{
|
| 138353 |
+
"epoch": 191.08,
|
| 138354 |
+
"learning_rate": 9.630193861066237e-06,
|
| 138355 |
+
"loss": 0.4907,
|
| 138356 |
+
"step": 23900
|
| 138357 |
+
},
|
| 138358 |
+
{
|
| 138359 |
+
"epoch": 191.16,
|
| 138360 |
+
"learning_rate": 9.630032310177707e-06,
|
| 138361 |
+
"loss": 0.4867,
|
| 138362 |
+
"step": 23910
|
| 138363 |
+
},
|
| 138364 |
+
{
|
| 138365 |
+
"epoch": 191.24,
|
| 138366 |
+
"learning_rate": 9.629870759289177e-06,
|
| 138367 |
+
"loss": 1.1891,
|
| 138368 |
+
"step": 23920
|
| 138369 |
+
},
|
| 138370 |
+
{
|
| 138371 |
+
"epoch": 191.32,
|
| 138372 |
+
"learning_rate": 9.629709208400647e-06,
|
| 138373 |
+
"loss": 0.3992,
|
| 138374 |
+
"step": 23930
|
| 138375 |
+
},
|
| 138376 |
+
{
|
| 138377 |
+
"epoch": 191.4,
|
| 138378 |
+
"learning_rate": 9.629547657512117e-06,
|
| 138379 |
+
"loss": 1.0267,
|
| 138380 |
+
"step": 23940
|
| 138381 |
+
},
|
| 138382 |
+
{
|
| 138383 |
+
"epoch": 191.48,
|
| 138384 |
+
"learning_rate": 9.629386106623587e-06,
|
| 138385 |
+
"loss": 0.4272,
|
| 138386 |
+
"step": 23950
|
| 138387 |
+
},
|
| 138388 |
+
{
|
| 138389 |
+
"epoch": 191.56,
|
| 138390 |
+
"learning_rate": 9.629224555735057e-06,
|
| 138391 |
+
"loss": 0.5949,
|
| 138392 |
+
"step": 23960
|
| 138393 |
+
},
|
| 138394 |
+
{
|
| 138395 |
+
"epoch": 191.64,
|
| 138396 |
+
"learning_rate": 9.629063004846527e-06,
|
| 138397 |
+
"loss": 0.9722,
|
| 138398 |
+
"step": 23970
|
| 138399 |
+
},
|
| 138400 |
+
{
|
| 138401 |
+
"epoch": 191.72,
|
| 138402 |
+
"learning_rate": 9.628901453957997e-06,
|
| 138403 |
+
"loss": 0.5117,
|
| 138404 |
+
"step": 23980
|
| 138405 |
+
},
|
| 138406 |
+
{
|
| 138407 |
+
"epoch": 191.8,
|
| 138408 |
+
"learning_rate": 9.628739903069467e-06,
|
| 138409 |
+
"loss": 1.0096,
|
| 138410 |
+
"step": 23990
|
| 138411 |
+
},
|
| 138412 |
+
{
|
| 138413 |
+
"epoch": 191.88,
|
| 138414 |
+
"learning_rate": 9.628578352180937e-06,
|
| 138415 |
+
"loss": 0.4353,
|
| 138416 |
+
"step": 24000
|
| 138417 |
+
},
|
| 138418 |
+
{
|
| 138419 |
+
"epoch": 191.96,
|
| 138420 |
+
"learning_rate": 9.628416801292409e-06,
|
| 138421 |
+
"loss": 0.5322,
|
| 138422 |
+
"step": 24010
|
| 138423 |
+
},
|
| 138424 |
+
{
|
| 138425 |
+
"epoch": 192.0,
|
| 138426 |
+
"eval_loss": 0.47713422775268555,
|
| 138427 |
+
"eval_runtime": 40.8067,
|
| 138428 |
+
"eval_samples_per_second": 20.487,
|
| 138429 |
+
"eval_steps_per_second": 0.662,
|
| 138430 |
+
"eval_wer": 0.23508771929824562,
|
| 138431 |
+
"step": 24015
|
| 138432 |
+
},
|
| 138433 |
+
{
|
| 138434 |
+
"epoch": 193.04,
|
| 138435 |
+
"learning_rate": 9.62827140549273e-06,
|
| 138436 |
+
"loss": 0.4565,
|
| 138437 |
+
"step": 24020
|
| 138438 |
+
},
|
| 138439 |
+
{
|
| 138440 |
+
"epoch": 193.12,
|
| 138441 |
+
"learning_rate": 9.6281098546042e-06,
|
| 138442 |
+
"loss": 0.4753,
|
| 138443 |
+
"step": 24030
|
| 138444 |
+
},
|
| 138445 |
+
{
|
| 138446 |
+
"epoch": 193.2,
|
| 138447 |
+
"learning_rate": 9.62794830371567e-06,
|
| 138448 |
+
"loss": 1.1457,
|
| 138449 |
+
"step": 24040
|
| 138450 |
+
},
|
| 138451 |
+
{
|
| 138452 |
+
"epoch": 193.28,
|
| 138453 |
+
"learning_rate": 9.627786752827142e-06,
|
| 138454 |
+
"loss": 0.4221,
|
| 138455 |
+
"step": 24050
|
| 138456 |
+
},
|
| 138457 |
+
{
|
| 138458 |
+
"epoch": 193.36,
|
| 138459 |
+
"learning_rate": 9.627625201938612e-06,
|
| 138460 |
+
"loss": 0.5413,
|
| 138461 |
+
"step": 24060
|
| 138462 |
+
},
|
| 138463 |
+
{
|
| 138464 |
+
"epoch": 193.44,
|
| 138465 |
+
"learning_rate": 9.627463651050082e-06,
|
| 138466 |
+
"loss": 0.9784,
|
| 138467 |
+
"step": 24070
|
| 138468 |
+
},
|
| 138469 |
+
{
|
| 138470 |
+
"epoch": 193.52,
|
| 138471 |
+
"learning_rate": 9.627302100161552e-06,
|
| 138472 |
+
"loss": 0.4663,
|
| 138473 |
+
"step": 24080
|
| 138474 |
+
},
|
| 138475 |
+
{
|
| 138476 |
+
"epoch": 193.6,
|
| 138477 |
+
"learning_rate": 9.627140549273022e-06,
|
| 138478 |
+
"loss": 1.0874,
|
| 138479 |
+
"step": 24090
|
| 138480 |
+
},
|
| 138481 |
+
{
|
| 138482 |
+
"epoch": 193.68,
|
| 138483 |
+
"learning_rate": 9.626978998384492e-06,
|
| 138484 |
+
"loss": 0.4995,
|
| 138485 |
+
"step": 24100
|
| 138486 |
+
},
|
| 138487 |
+
{
|
| 138488 |
+
"epoch": 193.76,
|
| 138489 |
+
"learning_rate": 9.626817447495962e-06,
|
| 138490 |
+
"loss": 0.4804,
|
| 138491 |
+
"step": 24110
|
| 138492 |
+
},
|
| 138493 |
+
{
|
| 138494 |
+
"epoch": 193.84,
|
| 138495 |
+
"learning_rate": 9.626655896607432e-06,
|
| 138496 |
+
"loss": 0.874,
|
| 138497 |
+
"step": 24120
|
| 138498 |
+
},
|
| 138499 |
+
{
|
| 138500 |
+
"epoch": 193.92,
|
| 138501 |
+
"learning_rate": 9.626494345718902e-06,
|
| 138502 |
+
"loss": 0.5074,
|
| 138503 |
+
"step": 24130
|
| 138504 |
+
},
|
| 138505 |
+
{
|
| 138506 |
+
"epoch": 194.0,
|
| 138507 |
+
"eval_loss": 0.5371707677841187,
|
| 138508 |
+
"eval_runtime": 40.5936,
|
| 138509 |
+
"eval_samples_per_second": 20.619,
|
| 138510 |
+
"eval_steps_per_second": 0.665,
|
| 138511 |
+
"eval_wer": 0.22862439747868002,
|
| 138512 |
+
"step": 24139
|
| 138513 |
+
},
|
| 138514 |
+
{
|
| 138515 |
+
"epoch": 194.01,
|
| 138516 |
+
"learning_rate": 9.626332794830372e-06,
|
| 138517 |
+
"loss": 0.5268,
|
| 138518 |
+
"step": 24140
|
| 138519 |
+
},
|
| 138520 |
+
{
|
| 138521 |
+
"epoch": 194.09,
|
| 138522 |
+
"learning_rate": 9.626171243941842e-06,
|
| 138523 |
+
"loss": 0.4241,
|
| 138524 |
+
"step": 24150
|
| 138525 |
+
},
|
| 138526 |
+
{
|
| 138527 |
+
"epoch": 194.17,
|
| 138528 |
+
"learning_rate": 9.626009693053313e-06,
|
| 138529 |
+
"loss": 0.6227,
|
| 138530 |
+
"step": 24160
|
| 138531 |
+
},
|
| 138532 |
+
{
|
| 138533 |
+
"epoch": 194.25,
|
| 138534 |
+
"learning_rate": 9.625848142164783e-06,
|
| 138535 |
+
"loss": 0.9652,
|
| 138536 |
+
"step": 24170
|
| 138537 |
+
},
|
| 138538 |
+
{
|
| 138539 |
+
"epoch": 194.33,
|
| 138540 |
+
"learning_rate": 9.625686591276253e-06,
|
| 138541 |
+
"loss": 0.4683,
|
| 138542 |
+
"step": 24180
|
| 138543 |
+
},
|
| 138544 |
+
{
|
| 138545 |
+
"epoch": 194.41,
|
| 138546 |
+
"learning_rate": 9.625525040387723e-06,
|
| 138547 |
+
"loss": 1.1767,
|
| 138548 |
+
"step": 24190
|
| 138549 |
+
},
|
| 138550 |
+
{
|
| 138551 |
+
"epoch": 194.49,
|
| 138552 |
+
"learning_rate": 9.625363489499193e-06,
|
| 138553 |
+
"loss": 0.4377,
|
| 138554 |
+
"step": 24200
|
| 138555 |
+
},
|
| 138556 |
+
{
|
| 138557 |
+
"epoch": 194.57,
|
| 138558 |
+
"learning_rate": 9.625201938610663e-06,
|
| 138559 |
+
"loss": 0.6596,
|
| 138560 |
+
"step": 24210
|
| 138561 |
+
},
|
| 138562 |
+
{
|
| 138563 |
+
"epoch": 194.65,
|
| 138564 |
+
"learning_rate": 9.625040387722133e-06,
|
| 138565 |
+
"loss": 0.9457,
|
| 138566 |
+
"step": 24220
|
| 138567 |
+
},
|
| 138568 |
+
{
|
| 138569 |
+
"epoch": 194.73,
|
| 138570 |
+
"learning_rate": 9.624878836833603e-06,
|
| 138571 |
+
"loss": 0.4555,
|
| 138572 |
+
"step": 24230
|
| 138573 |
+
},
|
| 138574 |
+
{
|
| 138575 |
+
"epoch": 194.81,
|
| 138576 |
+
"learning_rate": 9.624717285945073e-06,
|
| 138577 |
+
"loss": 0.9888,
|
| 138578 |
+
"step": 24240
|
| 138579 |
+
},
|
| 138580 |
+
{
|
| 138581 |
+
"epoch": 194.89,
|
| 138582 |
+
"learning_rate": 9.624555735056543e-06,
|
| 138583 |
+
"loss": 0.4755,
|
| 138584 |
+
"step": 24250
|
| 138585 |
+
},
|
| 138586 |
+
{
|
| 138587 |
+
"epoch": 194.97,
|
| 138588 |
+
"learning_rate": 9.624394184168013e-06,
|
| 138589 |
+
"loss": 0.5908,
|
| 138590 |
+
"step": 24260
|
| 138591 |
+
},
|
| 138592 |
+
{
|
| 138593 |
+
"epoch": 195.0,
|
| 138594 |
+
"eval_loss": 0.5001394748687744,
|
| 138595 |
+
"eval_runtime": 40.3014,
|
| 138596 |
+
"eval_samples_per_second": 20.769,
|
| 138597 |
+
"eval_steps_per_second": 0.67,
|
| 138598 |
+
"eval_wer": 0.22985141828005404,
|
| 138599 |
+
"step": 24263
|
| 138600 |
+
},
|
| 138601 |
+
{
|
| 138602 |
+
"epoch": 194.06,
|
| 138603 |
+
"learning_rate": 9.624232633279483e-06,
|
| 138604 |
+
"loss": 0.4893,
|
| 138605 |
+
"step": 24270
|
| 138606 |
+
},
|
| 138607 |
+
{
|
| 138608 |
+
"epoch": 194.14,
|
| 138609 |
+
"learning_rate": 9.624071082390955e-06,
|
| 138610 |
+
"loss": 0.4607,
|
| 138611 |
+
"step": 24280
|
| 138612 |
+
},
|
| 138613 |
+
{
|
| 138614 |
+
"epoch": 194.22,
|
| 138615 |
+
"learning_rate": 9.623909531502425e-06,
|
| 138616 |
+
"loss": 1.0318,
|
| 138617 |
+
"step": 24290
|
| 138618 |
+
},
|
| 138619 |
+
{
|
| 138620 |
+
"epoch": 194.3,
|
| 138621 |
+
"learning_rate": 9.623747980613895e-06,
|
| 138622 |
+
"loss": 0.4608,
|
| 138623 |
+
"step": 24300
|
| 138624 |
+
},
|
| 138625 |
+
{
|
| 138626 |
+
"epoch": 194.38,
|
| 138627 |
+
"learning_rate": 9.623586429725365e-06,
|
| 138628 |
+
"loss": 0.6567,
|
| 138629 |
+
"step": 24310
|
| 138630 |
+
},
|
| 138631 |
+
{
|
| 138632 |
+
"epoch": 194.46,
|
| 138633 |
+
"learning_rate": 9.623424878836835e-06,
|
| 138634 |
+
"loss": 0.8709,
|
| 138635 |
+
"step": 24320
|
| 138636 |
+
},
|
| 138637 |
+
{
|
| 138638 |
+
"epoch": 194.54,
|
| 138639 |
+
"learning_rate": 9.623263327948305e-06,
|
| 138640 |
+
"loss": 0.4206,
|
| 138641 |
+
"step": 24330
|
| 138642 |
+
},
|
| 138643 |
+
{
|
| 138644 |
+
"epoch": 194.62,
|
| 138645 |
+
"learning_rate": 9.623101777059774e-06,
|
| 138646 |
+
"loss": 0.9308,
|
| 138647 |
+
"step": 24340
|
| 138648 |
+
},
|
| 138649 |
+
{
|
| 138650 |
+
"epoch": 194.7,
|
| 138651 |
+
"learning_rate": 9.622940226171244e-06,
|
| 138652 |
+
"loss": 0.3828,
|
| 138653 |
+
"step": 24350
|
| 138654 |
+
},
|
| 138655 |
+
{
|
| 138656 |
+
"epoch": 194.78,
|
| 138657 |
+
"learning_rate": 9.622778675282714e-06,
|
| 138658 |
+
"loss": 0.7304,
|
| 138659 |
+
"step": 24360
|
| 138660 |
+
},
|
| 138661 |
+
{
|
| 138662 |
+
"epoch": 194.86,
|
| 138663 |
+
"learning_rate": 9.622617124394184e-06,
|
| 138664 |
+
"loss": 0.8561,
|
| 138665 |
+
"step": 24370
|
| 138666 |
+
},
|
| 138667 |
+
{
|
| 138668 |
+
"epoch": 194.94,
|
| 138669 |
+
"learning_rate": 9.622455573505654e-06,
|
| 138670 |
+
"loss": 0.5146,
|
| 138671 |
+
"step": 24380
|
| 138672 |
+
},
|
| 138673 |
+
{
|
| 138674 |
+
"epoch": 195.0,
|
| 138675 |
+
"eval_loss": 0.6156412959098816,
|
| 138676 |
+
"eval_runtime": 39.0293,
|
| 138677 |
+
"eval_samples_per_second": 21.65,
|
| 138678 |
+
"eval_steps_per_second": 0.692,
|
| 138679 |
+
"eval_wer": 0.23226561357320463,
|
| 138680 |
+
"step": 24388
|
| 138681 |
+
},
|
| 138682 |
+
{
|
| 138683 |
+
"epoch": 196.02,
|
| 138684 |
+
"learning_rate": 9.622294022617124e-06,
|
| 138685 |
+
"loss": 0.4826,
|
| 138686 |
+
"step": 24390
|
| 138687 |
+
},
|
| 138688 |
+
{
|
| 138689 |
+
"epoch": 196.1,
|
| 138690 |
+
"learning_rate": 9.622132471728596e-06,
|
| 138691 |
+
"loss": 0.4913,
|
| 138692 |
+
"step": 24400
|
| 138693 |
+
},
|
| 138694 |
+
{
|
| 138695 |
+
"epoch": 196.18,
|
| 138696 |
+
"learning_rate": 9.621970920840066e-06,
|
| 138697 |
+
"loss": 0.6569,
|
| 138698 |
+
"step": 24410
|
| 138699 |
+
},
|
| 138700 |
+
{
|
| 138701 |
+
"epoch": 196.26,
|
| 138702 |
+
"learning_rate": 9.621809369951536e-06,
|
| 138703 |
+
"loss": 0.9062,
|
| 138704 |
+
"step": 24420
|
| 138705 |
+
},
|
| 138706 |
+
{
|
| 138707 |
+
"epoch": 196.34,
|
| 138708 |
+
"learning_rate": 9.621647819063006e-06,
|
| 138709 |
+
"loss": 0.4574,
|
| 138710 |
+
"step": 24430
|
| 138711 |
+
},
|
| 138712 |
+
{
|
| 138713 |
+
"epoch": 196.42,
|
| 138714 |
+
"learning_rate": 9.621486268174476e-06,
|
| 138715 |
+
"loss": 0.9443,
|
| 138716 |
+
"step": 24440
|
| 138717 |
+
},
|
| 138718 |
+
{
|
| 138719 |
+
"epoch": 196.5,
|
| 138720 |
+
"learning_rate": 9.621324717285946e-06,
|
| 138721 |
+
"loss": 0.4282,
|
| 138722 |
+
"step": 24450
|
| 138723 |
+
},
|
| 138724 |
+
{
|
| 138725 |
+
"epoch": 196.58,
|
| 138726 |
+
"learning_rate": 9.621163166397416e-06,
|
| 138727 |
+
"loss": 0.692,
|
| 138728 |
+
"step": 24460
|
| 138729 |
+
},
|
| 138730 |
+
{
|
| 138731 |
+
"epoch": 196.66,
|
| 138732 |
+
"learning_rate": 9.621001615508886e-06,
|
| 138733 |
+
"loss": 0.9446,
|
| 138734 |
+
"step": 24470
|
| 138735 |
+
},
|
| 138736 |
+
{
|
| 138737 |
+
"epoch": 196.74,
|
| 138738 |
+
"learning_rate": 9.620840064620356e-06,
|
| 138739 |
+
"loss": 0.4262,
|
| 138740 |
+
"step": 24480
|
| 138741 |
+
},
|
| 138742 |
+
{
|
| 138743 |
+
"epoch": 196.82,
|
| 138744 |
+
"learning_rate": 9.620678513731826e-06,
|
| 138745 |
+
"loss": 1.078,
|
| 138746 |
+
"step": 24490
|
| 138747 |
+
},
|
| 138748 |
+
{
|
| 138749 |
+
"epoch": 196.9,
|
| 138750 |
+
"learning_rate": 9.620516962843296e-06,
|
| 138751 |
+
"loss": 0.4244,
|
| 138752 |
+
"step": 24500
|
| 138753 |
+
},
|
| 138754 |
+
{
|
| 138755 |
+
"epoch": 196.98,
|
| 138756 |
+
"learning_rate": 9.620355411954767e-06,
|
| 138757 |
+
"loss": 0.725,
|
| 138758 |
+
"step": 24510
|
| 138759 |
+
},
|
| 138760 |
+
{
|
| 138761 |
+
"epoch": 197.0,
|
| 138762 |
+
"eval_loss": 0.4438294768333435,
|
| 138763 |
+
"eval_runtime": 40.621,
|
| 138764 |
+
"eval_samples_per_second": 20.802,
|
| 138765 |
+
"eval_steps_per_second": 0.665,
|
| 138766 |
+
"eval_wer": 0.22446705786228924,
|
| 138767 |
+
"step": 24512
|
| 138768 |
}
|
| 138769 |
],
|
| 138770 |
+
"max_steps": 620000,
|
| 138771 |
"num_train_epochs": 5000,
|
| 138772 |
+
"total_flos": 6.897191142499402e+19,
|
| 138773 |
"trial_name": null,
|
| 138774 |
"trial_params": null
|
| 138775 |
}
|
model-bin/finetune/base/{checkpoint-23765 → checkpoint-24512}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629597755.4176855/events.out.tfevents.1629597755.91f414e9573a.903.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:204679b101a0c5f892bc649c90369737b44e3ecfde9448cbc8ca17a8a3904142
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629598414.5427365/events.out.tfevents.1629598414.91f414e9573a.903.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3da9d374136319f66312676ca3c1c11083fce653b63cd66772fff2d1e57168e7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629599084.569025/events.out.tfevents.1629599084.91f414e9573a.903.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e80b1d242140ee9faa433b6cd30ef7d8d1db4b159f799064aa809abf555177ba
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629599824.9825046/events.out.tfevents.1629599824.91f414e9573a.903.27
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9e089502762fb12dd9f995aa5c94d3aaec3d8c6a78e7c816f1d47d79f1bec70
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629600485.658627/events.out.tfevents.1629600485.91f414e9573a.903.29
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e24626fb6b3c7c685122d0a81b0d168e510c2c64b261c89a8b6be9a785f076bc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629597755.91f414e9573a.903.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d8ed9880a1c5eb2f61d1638b2e92912fc253d6f3f86bd4b349f755d9a6644a1
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629598414.91f414e9573a.903.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8fd03485e28661d5f85eaea3bb5a58d02afe2cc4323f6849953973cbe16a720a
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629599084.91f414e9573a.903.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e471c712a001b0df7c75e547e9681d7d554eedd547f590362208124c32c49207
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629599824.91f414e9573a.903.26
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cd8e6301e56ae9d430fde360ef795147dd3884936d2ee69cd87442466c36b983
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629600485.91f414e9573a.903.28
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff6ec515590bb4c04d62b6a6071b98376eab5814badb8ddbf64e7ce8eef9f9d4
|
| 3 |
+
size 6703
|