"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/trainer_state.json +509 -5
- model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629638257.4324517/events.out.tfevents.1629638257.91f414e9573a.903.141 +3 -0
- model-bin/finetune/base/log/1629638929.1884677/events.out.tfevents.1629638929.91f414e9573a.903.143 +3 -0
- model-bin/finetune/base/log/1629639583.460282/events.out.tfevents.1629639583.91f414e9573a.903.145 +3 -0
- model-bin/finetune/base/log/1629640240.3225508/events.out.tfevents.1629640240.91f414e9573a.903.147 +3 -0
- model-bin/finetune/base/log/1629640914.9915497/events.out.tfevents.1629640914.91f414e9573a.903.149 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629638257.91f414e9573a.903.140 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629638929.91f414e9573a.903.142 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629639583.91f414e9573a.903.144 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629640240.91f414e9573a.903.146 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629640914.91f414e9573a.903.148 +3 -0
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/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 sha256:03bd02a954efb88f4b15d7bdfa118f386332dd00f932508d0252cb23181c3dc3
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/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 sha256:2c1a8dd78f17e85de2ecf6f48be9edf89da0385ceafb7b95742de12dc47dca8b
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/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 sha256:514d919f1942d981214ed16138f885cc0c36b82d97dc4608422d1d18dc5e0541
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/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:cb527b7fc4c649bd8e37525e28368483ae8b83da3bf7dba64ef24b1c7439c24e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/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:bf9145632cfbd511ed0eab8ce21c09dca20533c227aba62bd0237190bc45b7e8
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/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,
|
|
@@ -143277,11 +143277,515 @@
|
|
| 143277 |
"eval_steps_per_second": 0.655,
|
| 143278 |
"eval_wer": 0.22114444695089858,
|
| 143279 |
"step": 31228
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 143280 |
}
|
| 143281 |
],
|
| 143282 |
"max_steps": 625000,
|
| 143283 |
"num_train_epochs": 5000,
|
| 143284 |
-
"total_flos": 8.
|
| 143285 |
"trial_name": null,
|
| 143286 |
"trial_params": null
|
| 143287 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.20553532410779315,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-31352",
|
| 4 |
+
"epoch": 255.0,
|
| 5 |
+
"global_step": 31975,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 143277 |
"eval_steps_per_second": 0.655,
|
| 143278 |
"eval_wer": 0.22114444695089858,
|
| 143279 |
"step": 31228
|
| 143280 |
+
},
|
| 143281 |
+
{
|
| 143282 |
+
"epoch": 251.02,
|
| 143283 |
+
"learning_rate": 9.511809369951535e-06,
|
| 143284 |
+
"loss": 0.3917,
|
| 143285 |
+
"step": 31230
|
| 143286 |
+
},
|
| 143287 |
+
{
|
| 143288 |
+
"epoch": 251.1,
|
| 143289 |
+
"learning_rate": 9.511647819063005e-06,
|
| 143290 |
+
"loss": 0.4211,
|
| 143291 |
+
"step": 31240
|
| 143292 |
+
},
|
| 143293 |
+
{
|
| 143294 |
+
"epoch": 251.18,
|
| 143295 |
+
"learning_rate": 9.511486268174475e-06,
|
| 143296 |
+
"loss": 0.5998,
|
| 143297 |
+
"step": 31250
|
| 143298 |
+
},
|
| 143299 |
+
{
|
| 143300 |
+
"epoch": 251.26,
|
| 143301 |
+
"learning_rate": 9.511324717285945e-06,
|
| 143302 |
+
"loss": 0.7724,
|
| 143303 |
+
"step": 31260
|
| 143304 |
+
},
|
| 143305 |
+
{
|
| 143306 |
+
"epoch": 251.34,
|
| 143307 |
+
"learning_rate": 9.511163166397415e-06,
|
| 143308 |
+
"loss": 0.4741,
|
| 143309 |
+
"step": 31270
|
| 143310 |
+
},
|
| 143311 |
+
{
|
| 143312 |
+
"epoch": 251.42,
|
| 143313 |
+
"learning_rate": 9.511001615508885e-06,
|
| 143314 |
+
"loss": 0.911,
|
| 143315 |
+
"step": 31280
|
| 143316 |
+
},
|
| 143317 |
+
{
|
| 143318 |
+
"epoch": 251.5,
|
| 143319 |
+
"learning_rate": 9.510840064620355e-06,
|
| 143320 |
+
"loss": 0.367,
|
| 143321 |
+
"step": 31290
|
| 143322 |
+
},
|
| 143323 |
+
{
|
| 143324 |
+
"epoch": 251.58,
|
| 143325 |
+
"learning_rate": 9.510678513731825e-06,
|
| 143326 |
+
"loss": 0.6541,
|
| 143327 |
+
"step": 31300
|
| 143328 |
+
},
|
| 143329 |
+
{
|
| 143330 |
+
"epoch": 251.66,
|
| 143331 |
+
"learning_rate": 9.510516962843297e-06,
|
| 143332 |
+
"loss": 0.7644,
|
| 143333 |
+
"step": 31310
|
| 143334 |
+
},
|
| 143335 |
+
{
|
| 143336 |
+
"epoch": 251.74,
|
| 143337 |
+
"learning_rate": 9.510355411954767e-06,
|
| 143338 |
+
"loss": 0.4301,
|
| 143339 |
+
"step": 31320
|
| 143340 |
+
},
|
| 143341 |
+
{
|
| 143342 |
+
"epoch": 251.82,
|
| 143343 |
+
"learning_rate": 9.510193861066237e-06,
|
| 143344 |
+
"loss": 1.1003,
|
| 143345 |
+
"step": 31330
|
| 143346 |
+
},
|
| 143347 |
+
{
|
| 143348 |
+
"epoch": 251.9,
|
| 143349 |
+
"learning_rate": 9.510032310177707e-06,
|
| 143350 |
+
"loss": 0.4245,
|
| 143351 |
+
"step": 31340
|
| 143352 |
+
},
|
| 143353 |
+
{
|
| 143354 |
+
"epoch": 251.98,
|
| 143355 |
+
"learning_rate": 9.509870759289177e-06,
|
| 143356 |
+
"loss": 0.743,
|
| 143357 |
+
"step": 31350
|
| 143358 |
+
},
|
| 143359 |
+
{
|
| 143360 |
+
"epoch": 252.0,
|
| 143361 |
+
"eval_loss": 0.4088016152381897,
|
| 143362 |
+
"eval_runtime": 40.379,
|
| 143363 |
+
"eval_samples_per_second": 20.605,
|
| 143364 |
+
"eval_steps_per_second": 0.644,
|
| 143365 |
+
"eval_wer": 0.20553532410779315,
|
| 143366 |
+
"step": 31352
|
| 143367 |
+
},
|
| 143368 |
+
{
|
| 143369 |
+
"epoch": 250.06,
|
| 143370 |
+
"learning_rate": 9.509709208400647e-06,
|
| 143371 |
+
"loss": 0.3956,
|
| 143372 |
+
"step": 31360
|
| 143373 |
+
},
|
| 143374 |
+
{
|
| 143375 |
+
"epoch": 250.14,
|
| 143376 |
+
"learning_rate": 9.509547657512117e-06,
|
| 143377 |
+
"loss": 0.4407,
|
| 143378 |
+
"step": 31370
|
| 143379 |
+
},
|
| 143380 |
+
{
|
| 143381 |
+
"epoch": 250.22,
|
| 143382 |
+
"learning_rate": 9.509386106623587e-06,
|
| 143383 |
+
"loss": 0.9137,
|
| 143384 |
+
"step": 31380
|
| 143385 |
+
},
|
| 143386 |
+
{
|
| 143387 |
+
"epoch": 250.3,
|
| 143388 |
+
"learning_rate": 9.509224555735057e-06,
|
| 143389 |
+
"loss": 0.4242,
|
| 143390 |
+
"step": 31390
|
| 143391 |
+
},
|
| 143392 |
+
{
|
| 143393 |
+
"epoch": 250.38,
|
| 143394 |
+
"learning_rate": 9.509063004846527e-06,
|
| 143395 |
+
"loss": 0.7312,
|
| 143396 |
+
"step": 31400
|
| 143397 |
+
},
|
| 143398 |
+
{
|
| 143399 |
+
"epoch": 250.46,
|
| 143400 |
+
"learning_rate": 9.508901453957997e-06,
|
| 143401 |
+
"loss": 0.7856,
|
| 143402 |
+
"step": 31410
|
| 143403 |
+
},
|
| 143404 |
+
{
|
| 143405 |
+
"epoch": 250.54,
|
| 143406 |
+
"learning_rate": 9.508739903069467e-06,
|
| 143407 |
+
"loss": 0.4817,
|
| 143408 |
+
"step": 31420
|
| 143409 |
+
},
|
| 143410 |
+
{
|
| 143411 |
+
"epoch": 250.62,
|
| 143412 |
+
"learning_rate": 9.508578352180938e-06,
|
| 143413 |
+
"loss": 1.0326,
|
| 143414 |
+
"step": 31430
|
| 143415 |
+
},
|
| 143416 |
+
{
|
| 143417 |
+
"epoch": 250.7,
|
| 143418 |
+
"learning_rate": 9.508416801292408e-06,
|
| 143419 |
+
"loss": 0.404,
|
| 143420 |
+
"step": 31440
|
| 143421 |
+
},
|
| 143422 |
+
{
|
| 143423 |
+
"epoch": 250.78,
|
| 143424 |
+
"learning_rate": 9.508255250403878e-06,
|
| 143425 |
+
"loss": 0.6966,
|
| 143426 |
+
"step": 31450
|
| 143427 |
+
},
|
| 143428 |
+
{
|
| 143429 |
+
"epoch": 250.86,
|
| 143430 |
+
"learning_rate": 9.508093699515348e-06,
|
| 143431 |
+
"loss": 0.7209,
|
| 143432 |
+
"step": 31460
|
| 143433 |
+
},
|
| 143434 |
+
{
|
| 143435 |
+
"epoch": 250.94,
|
| 143436 |
+
"learning_rate": 9.507932148626818e-06,
|
| 143437 |
+
"loss": 0.4792,
|
| 143438 |
+
"step": 31470
|
| 143439 |
+
},
|
| 143440 |
+
{
|
| 143441 |
+
"epoch": 251.0,
|
| 143442 |
+
"eval_loss": 0.5309865474700928,
|
| 143443 |
+
"eval_runtime": 39.4727,
|
| 143444 |
+
"eval_samples_per_second": 21.078,
|
| 143445 |
+
"eval_steps_per_second": 0.659,
|
| 143446 |
+
"eval_wer": 0.21732954545454544,
|
| 143447 |
+
"step": 31477
|
| 143448 |
+
},
|
| 143449 |
+
{
|
| 143450 |
+
"epoch": 251.02,
|
| 143451 |
+
"learning_rate": 9.507786752827141e-06,
|
| 143452 |
+
"loss": 1.0989,
|
| 143453 |
+
"step": 31480
|
| 143454 |
+
},
|
| 143455 |
+
{
|
| 143456 |
+
"epoch": 251.1,
|
| 143457 |
+
"learning_rate": 9.507625201938611e-06,
|
| 143458 |
+
"loss": 0.4682,
|
| 143459 |
+
"step": 31490
|
| 143460 |
+
},
|
| 143461 |
+
{
|
| 143462 |
+
"epoch": 251.18,
|
| 143463 |
+
"learning_rate": 9.507463651050081e-06,
|
| 143464 |
+
"loss": 0.7449,
|
| 143465 |
+
"step": 31500
|
| 143466 |
+
},
|
| 143467 |
+
{
|
| 143468 |
+
"epoch": 251.26,
|
| 143469 |
+
"learning_rate": 9.507302100161551e-06,
|
| 143470 |
+
"loss": 0.742,
|
| 143471 |
+
"step": 31510
|
| 143472 |
+
},
|
| 143473 |
+
{
|
| 143474 |
+
"epoch": 251.34,
|
| 143475 |
+
"learning_rate": 9.507140549273021e-06,
|
| 143476 |
+
"loss": 0.4308,
|
| 143477 |
+
"step": 31520
|
| 143478 |
+
},
|
| 143479 |
+
{
|
| 143480 |
+
"epoch": 251.42,
|
| 143481 |
+
"learning_rate": 9.506978998384491e-06,
|
| 143482 |
+
"loss": 0.8553,
|
| 143483 |
+
"step": 31530
|
| 143484 |
+
},
|
| 143485 |
+
{
|
| 143486 |
+
"epoch": 251.5,
|
| 143487 |
+
"learning_rate": 9.506817447495961e-06,
|
| 143488 |
+
"loss": 0.4713,
|
| 143489 |
+
"step": 31540
|
| 143490 |
+
},
|
| 143491 |
+
{
|
| 143492 |
+
"epoch": 251.58,
|
| 143493 |
+
"learning_rate": 9.506655896607431e-06,
|
| 143494 |
+
"loss": 0.6502,
|
| 143495 |
+
"step": 31550
|
| 143496 |
+
},
|
| 143497 |
+
{
|
| 143498 |
+
"epoch": 251.66,
|
| 143499 |
+
"learning_rate": 9.506494345718901e-06,
|
| 143500 |
+
"loss": 0.6602,
|
| 143501 |
+
"step": 31560
|
| 143502 |
+
},
|
| 143503 |
+
{
|
| 143504 |
+
"epoch": 251.74,
|
| 143505 |
+
"learning_rate": 9.506332794830371e-06,
|
| 143506 |
+
"loss": 0.4611,
|
| 143507 |
+
"step": 31570
|
| 143508 |
+
},
|
| 143509 |
+
{
|
| 143510 |
+
"epoch": 251.82,
|
| 143511 |
+
"learning_rate": 9.506171243941843e-06,
|
| 143512 |
+
"loss": 1.0343,
|
| 143513 |
+
"step": 31580
|
| 143514 |
+
},
|
| 143515 |
+
{
|
| 143516 |
+
"epoch": 251.9,
|
| 143517 |
+
"learning_rate": 9.506025848142166e-06,
|
| 143518 |
+
"loss": 1.2558,
|
| 143519 |
+
"step": 31590
|
| 143520 |
+
},
|
| 143521 |
+
{
|
| 143522 |
+
"epoch": 251.98,
|
| 143523 |
+
"learning_rate": 9.505864297253636e-06,
|
| 143524 |
+
"loss": 0.6931,
|
| 143525 |
+
"step": 31600
|
| 143526 |
+
},
|
| 143527 |
+
{
|
| 143528 |
+
"epoch": 252.0,
|
| 143529 |
+
"eval_loss": 0.522945761680603,
|
| 143530 |
+
"eval_runtime": 40.2186,
|
| 143531 |
+
"eval_samples_per_second": 20.687,
|
| 143532 |
+
"eval_steps_per_second": 0.646,
|
| 143533 |
+
"eval_wer": 0.2220187518312335,
|
| 143534 |
+
"step": 31602
|
| 143535 |
+
},
|
| 143536 |
+
{
|
| 143537 |
+
"epoch": 254.06,
|
| 143538 |
+
"learning_rate": 9.505702746365106e-06,
|
| 143539 |
+
"loss": 0.3789,
|
| 143540 |
+
"step": 31610
|
| 143541 |
+
},
|
| 143542 |
+
{
|
| 143543 |
+
"epoch": 254.14,
|
| 143544 |
+
"learning_rate": 9.505541195476576e-06,
|
| 143545 |
+
"loss": 0.4063,
|
| 143546 |
+
"step": 31620
|
| 143547 |
+
},
|
| 143548 |
+
{
|
| 143549 |
+
"epoch": 254.22,
|
| 143550 |
+
"learning_rate": 9.505379644588046e-06,
|
| 143551 |
+
"loss": 0.9326,
|
| 143552 |
+
"step": 31630
|
| 143553 |
+
},
|
| 143554 |
+
{
|
| 143555 |
+
"epoch": 254.31,
|
| 143556 |
+
"learning_rate": 9.505218093699516e-06,
|
| 143557 |
+
"loss": 0.4117,
|
| 143558 |
+
"step": 31640
|
| 143559 |
+
},
|
| 143560 |
+
{
|
| 143561 |
+
"epoch": 254.39,
|
| 143562 |
+
"learning_rate": 9.505056542810986e-06,
|
| 143563 |
+
"loss": 0.69,
|
| 143564 |
+
"step": 31650
|
| 143565 |
+
},
|
| 143566 |
+
{
|
| 143567 |
+
"epoch": 254.47,
|
| 143568 |
+
"learning_rate": 9.504894991922456e-06,
|
| 143569 |
+
"loss": 0.8041,
|
| 143570 |
+
"step": 31660
|
| 143571 |
+
},
|
| 143572 |
+
{
|
| 143573 |
+
"epoch": 254.55,
|
| 143574 |
+
"learning_rate": 9.504733441033926e-06,
|
| 143575 |
+
"loss": 0.4179,
|
| 143576 |
+
"step": 31670
|
| 143577 |
+
},
|
| 143578 |
+
{
|
| 143579 |
+
"epoch": 254.63,
|
| 143580 |
+
"learning_rate": 9.504571890145396e-06,
|
| 143581 |
+
"loss": 0.9579,
|
| 143582 |
+
"step": 31680
|
| 143583 |
+
},
|
| 143584 |
+
{
|
| 143585 |
+
"epoch": 254.71,
|
| 143586 |
+
"learning_rate": 9.504410339256867e-06,
|
| 143587 |
+
"loss": 0.3987,
|
| 143588 |
+
"step": 31690
|
| 143589 |
+
},
|
| 143590 |
+
{
|
| 143591 |
+
"epoch": 254.79,
|
| 143592 |
+
"learning_rate": 9.504248788368337e-06,
|
| 143593 |
+
"loss": 0.6974,
|
| 143594 |
+
"step": 31700
|
| 143595 |
+
},
|
| 143596 |
+
{
|
| 143597 |
+
"epoch": 254.87,
|
| 143598 |
+
"learning_rate": 9.504087237479807e-06,
|
| 143599 |
+
"loss": 0.7365,
|
| 143600 |
+
"step": 31710
|
| 143601 |
+
},
|
| 143602 |
+
{
|
| 143603 |
+
"epoch": 254.95,
|
| 143604 |
+
"learning_rate": 9.503925686591277e-06,
|
| 143605 |
+
"loss": 0.416,
|
| 143606 |
+
"step": 31720
|
| 143607 |
+
},
|
| 143608 |
+
{
|
| 143609 |
+
"epoch": 255.0,
|
| 143610 |
+
"eval_loss": 0.4134102463722229,
|
| 143611 |
+
"eval_runtime": 40.4669,
|
| 143612 |
+
"eval_samples_per_second": 20.56,
|
| 143613 |
+
"eval_steps_per_second": 0.643,
|
| 143614 |
+
"eval_wer": 0.2129027544385183,
|
| 143615 |
+
"step": 31726
|
| 143616 |
+
},
|
| 143617 |
+
{
|
| 143618 |
+
"epoch": 255.03,
|
| 143619 |
+
"learning_rate": 9.503764135702747e-06,
|
| 143620 |
+
"loss": 0.5083,
|
| 143621 |
+
"step": 31730
|
| 143622 |
+
},
|
| 143623 |
+
{
|
| 143624 |
+
"epoch": 255.11,
|
| 143625 |
+
"learning_rate": 9.503602584814217e-06,
|
| 143626 |
+
"loss": 0.3727,
|
| 143627 |
+
"step": 31740
|
| 143628 |
+
},
|
| 143629 |
+
{
|
| 143630 |
+
"epoch": 255.19,
|
| 143631 |
+
"learning_rate": 9.503441033925687e-06,
|
| 143632 |
+
"loss": 0.8556,
|
| 143633 |
+
"step": 31750
|
| 143634 |
+
},
|
| 143635 |
+
{
|
| 143636 |
+
"epoch": 255.27,
|
| 143637 |
+
"learning_rate": 9.503279483037157e-06,
|
| 143638 |
+
"loss": 0.571,
|
| 143639 |
+
"step": 31760
|
| 143640 |
+
},
|
| 143641 |
+
{
|
| 143642 |
+
"epoch": 255.35,
|
| 143643 |
+
"learning_rate": 9.503117932148627e-06,
|
| 143644 |
+
"loss": 0.425,
|
| 143645 |
+
"step": 31770
|
| 143646 |
+
},
|
| 143647 |
+
{
|
| 143648 |
+
"epoch": 255.43,
|
| 143649 |
+
"learning_rate": 9.502956381260097e-06,
|
| 143650 |
+
"loss": 0.9308,
|
| 143651 |
+
"step": 31780
|
| 143652 |
+
},
|
| 143653 |
+
{
|
| 143654 |
+
"epoch": 255.51,
|
| 143655 |
+
"learning_rate": 9.502794830371567e-06,
|
| 143656 |
+
"loss": 0.4105,
|
| 143657 |
+
"step": 31790
|
| 143658 |
+
},
|
| 143659 |
+
{
|
| 143660 |
+
"epoch": 255.59,
|
| 143661 |
+
"learning_rate": 9.502633279483037e-06,
|
| 143662 |
+
"loss": 0.8542,
|
| 143663 |
+
"step": 31800
|
| 143664 |
+
},
|
| 143665 |
+
{
|
| 143666 |
+
"epoch": 255.67,
|
| 143667 |
+
"learning_rate": 9.502471728594507e-06,
|
| 143668 |
+
"loss": 0.5851,
|
| 143669 |
+
"step": 31810
|
| 143670 |
+
},
|
| 143671 |
+
{
|
| 143672 |
+
"epoch": 255.76,
|
| 143673 |
+
"learning_rate": 9.502310177705979e-06,
|
| 143674 |
+
"loss": 0.4563,
|
| 143675 |
+
"step": 31820
|
| 143676 |
+
},
|
| 143677 |
+
{
|
| 143678 |
+
"epoch": 255.84,
|
| 143679 |
+
"learning_rate": 9.502148626817449e-06,
|
| 143680 |
+
"loss": 0.8857,
|
| 143681 |
+
"step": 31830
|
| 143682 |
+
},
|
| 143683 |
+
{
|
| 143684 |
+
"epoch": 255.92,
|
| 143685 |
+
"learning_rate": 9.501987075928919e-06,
|
| 143686 |
+
"loss": 0.3693,
|
| 143687 |
+
"step": 31840
|
| 143688 |
+
},
|
| 143689 |
+
{
|
| 143690 |
+
"epoch": 256.0,
|
| 143691 |
+
"learning_rate": 9.501825525040389e-06,
|
| 143692 |
+
"loss": 0.9257,
|
| 143693 |
+
"step": 31850
|
| 143694 |
+
},
|
| 143695 |
+
{
|
| 143696 |
+
"epoch": 256.0,
|
| 143697 |
+
"eval_loss": 0.46124252676963806,
|
| 143698 |
+
"eval_runtime": 40.6948,
|
| 143699 |
+
"eval_samples_per_second": 20.42,
|
| 143700 |
+
"eval_steps_per_second": 0.639,
|
| 143701 |
+
"eval_wer": 0.21019724599925568,
|
| 143702 |
+
"step": 31850
|
| 143703 |
+
},
|
| 143704 |
+
{
|
| 143705 |
+
"epoch": 254.08,
|
| 143706 |
+
"learning_rate": 9.501663974151859e-06,
|
| 143707 |
+
"loss": 0.411,
|
| 143708 |
+
"step": 31860
|
| 143709 |
+
},
|
| 143710 |
+
{
|
| 143711 |
+
"epoch": 254.16,
|
| 143712 |
+
"learning_rate": 9.501502423263329e-06,
|
| 143713 |
+
"loss": 0.5186,
|
| 143714 |
+
"step": 31870
|
| 143715 |
+
},
|
| 143716 |
+
{
|
| 143717 |
+
"epoch": 254.24,
|
| 143718 |
+
"learning_rate": 9.501340872374799e-06,
|
| 143719 |
+
"loss": 0.9338,
|
| 143720 |
+
"step": 31880
|
| 143721 |
+
},
|
| 143722 |
+
{
|
| 143723 |
+
"epoch": 254.32,
|
| 143724 |
+
"learning_rate": 9.501179321486269e-06,
|
| 143725 |
+
"loss": 0.4827,
|
| 143726 |
+
"step": 31890
|
| 143727 |
+
},
|
| 143728 |
+
{
|
| 143729 |
+
"epoch": 254.4,
|
| 143730 |
+
"learning_rate": 9.501017770597738e-06,
|
| 143731 |
+
"loss": 0.9752,
|
| 143732 |
+
"step": 31900
|
| 143733 |
+
},
|
| 143734 |
+
{
|
| 143735 |
+
"epoch": 254.48,
|
| 143736 |
+
"learning_rate": 9.500856219709208e-06,
|
| 143737 |
+
"loss": 0.483,
|
| 143738 |
+
"step": 31910
|
| 143739 |
+
},
|
| 143740 |
+
{
|
| 143741 |
+
"epoch": 254.56,
|
| 143742 |
+
"learning_rate": 9.500694668820678e-06,
|
| 143743 |
+
"loss": 0.5003,
|
| 143744 |
+
"step": 31920
|
| 143745 |
+
},
|
| 143746 |
+
{
|
| 143747 |
+
"epoch": 254.64,
|
| 143748 |
+
"learning_rate": 9.500533117932148e-06,
|
| 143749 |
+
"loss": 0.9161,
|
| 143750 |
+
"step": 31930
|
| 143751 |
+
},
|
| 143752 |
+
{
|
| 143753 |
+
"epoch": 254.72,
|
| 143754 |
+
"learning_rate": 9.500371567043618e-06,
|
| 143755 |
+
"loss": 0.3675,
|
| 143756 |
+
"step": 31940
|
| 143757 |
+
},
|
| 143758 |
+
{
|
| 143759 |
+
"epoch": 254.8,
|
| 143760 |
+
"learning_rate": 9.500210016155088e-06,
|
| 143761 |
+
"loss": 0.997,
|
| 143762 |
+
"step": 31950
|
| 143763 |
+
},
|
| 143764 |
+
{
|
| 143765 |
+
"epoch": 254.88,
|
| 143766 |
+
"learning_rate": 9.50004846526656e-06,
|
| 143767 |
+
"loss": 0.443,
|
| 143768 |
+
"step": 31960
|
| 143769 |
+
},
|
| 143770 |
+
{
|
| 143771 |
+
"epoch": 254.96,
|
| 143772 |
+
"learning_rate": 9.49988691437803e-06,
|
| 143773 |
+
"loss": 0.4837,
|
| 143774 |
+
"step": 31970
|
| 143775 |
+
},
|
| 143776 |
+
{
|
| 143777 |
+
"epoch": 255.0,
|
| 143778 |
+
"eval_loss": 0.44418424367904663,
|
| 143779 |
+
"eval_runtime": 39.231,
|
| 143780 |
+
"eval_samples_per_second": 21.182,
|
| 143781 |
+
"eval_steps_per_second": 0.663,
|
| 143782 |
+
"eval_wer": 0.2137329286798179,
|
| 143783 |
+
"step": 31975
|
| 143784 |
}
|
| 143785 |
],
|
| 143786 |
"max_steps": 625000,
|
| 143787 |
"num_train_epochs": 5000,
|
| 143788 |
+
"total_flos": 8.997442302766028e+19,
|
| 143789 |
"trial_name": null,
|
| 143790 |
"trial_params": null
|
| 143791 |
}
|
model-bin/finetune/base/{checkpoint-31228 → checkpoint-31975}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629638257.4324517/events.out.tfevents.1629638257.91f414e9573a.903.141
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:00c31b87bcc0f756f28a7955cf961d2ed30c52bb87ac60c6102c68ede896d392
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629638929.1884677/events.out.tfevents.1629638929.91f414e9573a.903.143
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e9aa4f67141254db3cbc26bbdcbd3b4a88e7248f63e5419c2d4108e929bc237b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629639583.460282/events.out.tfevents.1629639583.91f414e9573a.903.145
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2e68aef71fdc77061cdccd97beaae4e03076f03246673a4a66bbbefa9857b79d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629640240.3225508/events.out.tfevents.1629640240.91f414e9573a.903.147
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f75373af3d7aed4ef0ea26a06c40c8d62b24311f05fe8f2baaa4d43a9cf281c9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629640914.9915497/events.out.tfevents.1629640914.91f414e9573a.903.149
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6426a52166dd4d537950fb428f6007eb3b6936730762a72ea9117469b6cc7042
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629638257.91f414e9573a.903.140
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f8fc3a368bb205b7674df6455426b0ee7169f8a1d172cd175300b3f7f01c9a1
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629638929.91f414e9573a.903.142
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a32f0ce6d9a143ca03ffe899a5dc4ea81132013bf7021f652634f633ccc255ad
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629639583.91f414e9573a.903.144
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d8897279124968ddd873ab3075f397b2509742b16ab1d22b1e8d28f29c9a8c7b
|
| 3 |
+
size 6543
|
model-bin/finetune/base/log/events.out.tfevents.1629640240.91f414e9573a.903.146
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f2aeaad4e9f12b0354565e205536e2c85cb312a15520c9d7444f26e2a038475
|
| 3 |
+
size 6703
|
model-bin/finetune/base/log/events.out.tfevents.1629640914.91f414e9573a.903.148
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08b6c7009141885b22f4f6c23cf686f56f51843259157cc7b46dc88b134711dd
|
| 3 |
+
size 6543
|