"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629837488.925602/events.out.tfevents.1629837488.c435e1c5ee04.920.221 +3 -0
- model-bin/finetune/base/log/1629838113.24123/events.out.tfevents.1629838113.c435e1c5ee04.920.223 +3 -0
- model-bin/finetune/base/log/1629838737.3102496/events.out.tfevents.1629838737.c435e1c5ee04.920.225 +3 -0
- model-bin/finetune/base/log/1629839378.965553/events.out.tfevents.1629839378.c435e1c5ee04.920.227 +3 -0
- model-bin/finetune/base/log/1629840015.2190616/events.out.tfevents.1629840015.c435e1c5ee04.920.229 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629837488.c435e1c5ee04.920.220 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629838113.c435e1c5ee04.920.222 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629838737.c435e1c5ee04.920.224 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629839378.c435e1c5ee04.920.226 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629840015.c435e1c5ee04.920.228 +3 -0
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/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 sha256:6bf43d968dbe02ae7a22a40fea3f3b35c92eea6da15a59b6a0bf2b3a511f4915
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/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:fb02808d8842791dc4116b2aee9127299d59ea802c1a61ba6ba3092f36c849da
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/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:e38e660e121c61cfbe53b56b9b19b73618319ff2d4b4f0c08beadc29f3062f73
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/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 sha256:6497b3560f4dce675de0441236126359cb27ad7919d73715dda52b9f6fbfb5a6
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/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 sha256:4ca10994f54dd3af978a50aed2d65798b9b5d3bd7145e83c529f6a03bd9f42bf
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1855363713557883,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-67447",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -185337,11 +185337,806 @@
|
|
| 185337 |
"eval_steps_per_second": 0.705,
|
| 185338 |
"eval_wer": 0.192304851931172,
|
| 185339 |
"step": 67572
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 185340 |
}
|
| 185341 |
],
|
| 185342 |
-
"max_steps":
|
| 185343 |
"num_train_epochs": 5000,
|
| 185344 |
-
"total_flos": 1.
|
| 185345 |
"trial_name": null,
|
| 185346 |
"trial_params": null
|
| 185347 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1855363713557883,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-67447",
|
| 4 |
+
"epoch": 548.995983935743,
|
| 5 |
+
"global_step": 68195,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 185337 |
"eval_steps_per_second": 0.705,
|
| 185338 |
"eval_wer": 0.192304851931172,
|
| 185339 |
"step": 67572
|
| 185340 |
+
},
|
| 185341 |
+
{
|
| 185342 |
+
"epoch": 540.02,
|
| 185343 |
+
"learning_rate": 8.93349358974359e-06,
|
| 185344 |
+
"loss": 0.3757,
|
| 185345 |
+
"step": 67575
|
| 185346 |
+
},
|
| 185347 |
+
{
|
| 185348 |
+
"epoch": 540.06,
|
| 185349 |
+
"learning_rate": 8.933413461538463e-06,
|
| 185350 |
+
"loss": 0.3056,
|
| 185351 |
+
"step": 67580
|
| 185352 |
+
},
|
| 185353 |
+
{
|
| 185354 |
+
"epoch": 540.1,
|
| 185355 |
+
"learning_rate": 8.933333333333333e-06,
|
| 185356 |
+
"loss": 0.3568,
|
| 185357 |
+
"step": 67585
|
| 185358 |
+
},
|
| 185359 |
+
{
|
| 185360 |
+
"epoch": 540.14,
|
| 185361 |
+
"learning_rate": 8.933253205128206e-06,
|
| 185362 |
+
"loss": 0.4253,
|
| 185363 |
+
"step": 67590
|
| 185364 |
+
},
|
| 185365 |
+
{
|
| 185366 |
+
"epoch": 540.18,
|
| 185367 |
+
"learning_rate": 8.933173076923077e-06,
|
| 185368 |
+
"loss": 0.9211,
|
| 185369 |
+
"step": 67595
|
| 185370 |
+
},
|
| 185371 |
+
{
|
| 185372 |
+
"epoch": 540.22,
|
| 185373 |
+
"learning_rate": 8.933092948717949e-06,
|
| 185374 |
+
"loss": 1.0014,
|
| 185375 |
+
"step": 67600
|
| 185376 |
+
},
|
| 185377 |
+
{
|
| 185378 |
+
"epoch": 540.26,
|
| 185379 |
+
"learning_rate": 8.93301282051282e-06,
|
| 185380 |
+
"loss": 0.5676,
|
| 185381 |
+
"step": 67605
|
| 185382 |
+
},
|
| 185383 |
+
{
|
| 185384 |
+
"epoch": 540.3,
|
| 185385 |
+
"learning_rate": 8.932932692307693e-06,
|
| 185386 |
+
"loss": 0.3648,
|
| 185387 |
+
"step": 67610
|
| 185388 |
+
},
|
| 185389 |
+
{
|
| 185390 |
+
"epoch": 540.34,
|
| 185391 |
+
"learning_rate": 8.932852564102565e-06,
|
| 185392 |
+
"loss": 0.4003,
|
| 185393 |
+
"step": 67615
|
| 185394 |
+
},
|
| 185395 |
+
{
|
| 185396 |
+
"epoch": 540.38,
|
| 185397 |
+
"learning_rate": 8.932772435897436e-06,
|
| 185398 |
+
"loss": 0.8113,
|
| 185399 |
+
"step": 67620
|
| 185400 |
+
},
|
| 185401 |
+
{
|
| 185402 |
+
"epoch": 540.42,
|
| 185403 |
+
"learning_rate": 8.932692307692309e-06,
|
| 185404 |
+
"loss": 1.0352,
|
| 185405 |
+
"step": 67625
|
| 185406 |
+
},
|
| 185407 |
+
{
|
| 185408 |
+
"epoch": 540.46,
|
| 185409 |
+
"learning_rate": 8.93261217948718e-06,
|
| 185410 |
+
"loss": 0.3178,
|
| 185411 |
+
"step": 67630
|
| 185412 |
+
},
|
| 185413 |
+
{
|
| 185414 |
+
"epoch": 540.5,
|
| 185415 |
+
"learning_rate": 8.932532051282052e-06,
|
| 185416 |
+
"loss": 0.5683,
|
| 185417 |
+
"step": 67635
|
| 185418 |
+
},
|
| 185419 |
+
{
|
| 185420 |
+
"epoch": 540.54,
|
| 185421 |
+
"learning_rate": 8.932451923076923e-06,
|
| 185422 |
+
"loss": 0.3428,
|
| 185423 |
+
"step": 67640
|
| 185424 |
+
},
|
| 185425 |
+
{
|
| 185426 |
+
"epoch": 540.58,
|
| 185427 |
+
"learning_rate": 8.932371794871796e-06,
|
| 185428 |
+
"loss": 0.7413,
|
| 185429 |
+
"step": 67645
|
| 185430 |
+
},
|
| 185431 |
+
{
|
| 185432 |
+
"epoch": 540.62,
|
| 185433 |
+
"learning_rate": 8.932291666666668e-06,
|
| 185434 |
+
"loss": 0.8567,
|
| 185435 |
+
"step": 67650
|
| 185436 |
+
},
|
| 185437 |
+
{
|
| 185438 |
+
"epoch": 540.66,
|
| 185439 |
+
"learning_rate": 8.932211538461539e-06,
|
| 185440 |
+
"loss": 0.314,
|
| 185441 |
+
"step": 67655
|
| 185442 |
+
},
|
| 185443 |
+
{
|
| 185444 |
+
"epoch": 540.7,
|
| 185445 |
+
"learning_rate": 8.93213141025641e-06,
|
| 185446 |
+
"loss": 0.2992,
|
| 185447 |
+
"step": 67660
|
| 185448 |
+
},
|
| 185449 |
+
{
|
| 185450 |
+
"epoch": 540.74,
|
| 185451 |
+
"learning_rate": 8.932051282051283e-06,
|
| 185452 |
+
"loss": 0.3365,
|
| 185453 |
+
"step": 67665
|
| 185454 |
+
},
|
| 185455 |
+
{
|
| 185456 |
+
"epoch": 540.78,
|
| 185457 |
+
"learning_rate": 8.931971153846155e-06,
|
| 185458 |
+
"loss": 0.7477,
|
| 185459 |
+
"step": 67670
|
| 185460 |
+
},
|
| 185461 |
+
{
|
| 185462 |
+
"epoch": 540.82,
|
| 185463 |
+
"learning_rate": 8.931891025641026e-06,
|
| 185464 |
+
"loss": 0.9346,
|
| 185465 |
+
"step": 67675
|
| 185466 |
+
},
|
| 185467 |
+
{
|
| 185468 |
+
"epoch": 540.86,
|
| 185469 |
+
"learning_rate": 8.931810897435899e-06,
|
| 185470 |
+
"loss": 0.3394,
|
| 185471 |
+
"step": 67680
|
| 185472 |
+
},
|
| 185473 |
+
{
|
| 185474 |
+
"epoch": 540.9,
|
| 185475 |
+
"learning_rate": 8.93173076923077e-06,
|
| 185476 |
+
"loss": 0.2948,
|
| 185477 |
+
"step": 67685
|
| 185478 |
+
},
|
| 185479 |
+
{
|
| 185480 |
+
"epoch": 540.94,
|
| 185481 |
+
"learning_rate": 8.931650641025642e-06,
|
| 185482 |
+
"loss": 0.4479,
|
| 185483 |
+
"step": 67690
|
| 185484 |
+
},
|
| 185485 |
+
{
|
| 185486 |
+
"epoch": 540.98,
|
| 185487 |
+
"learning_rate": 8.931570512820513e-06,
|
| 185488 |
+
"loss": 0.7604,
|
| 185489 |
+
"step": 67695
|
| 185490 |
+
},
|
| 185491 |
+
{
|
| 185492 |
+
"epoch": 541.0,
|
| 185493 |
+
"eval_loss": 0.44643455743789673,
|
| 185494 |
+
"eval_runtime": 38.3882,
|
| 185495 |
+
"eval_samples_per_second": 21.856,
|
| 185496 |
+
"eval_steps_per_second": 0.703,
|
| 185497 |
+
"eval_wer": 0.1991250455705432,
|
| 185498 |
+
"step": 67697
|
| 185499 |
+
},
|
| 185500 |
+
{
|
| 185501 |
+
"epoch": 541.02,
|
| 185502 |
+
"learning_rate": 8.931490384615386e-06,
|
| 185503 |
+
"loss": 0.3752,
|
| 185504 |
+
"step": 67700
|
| 185505 |
+
},
|
| 185506 |
+
{
|
| 185507 |
+
"epoch": 541.06,
|
| 185508 |
+
"learning_rate": 8.931410256410258e-06,
|
| 185509 |
+
"loss": 0.2947,
|
| 185510 |
+
"step": 67705
|
| 185511 |
+
},
|
| 185512 |
+
{
|
| 185513 |
+
"epoch": 541.1,
|
| 185514 |
+
"learning_rate": 8.931330128205129e-06,
|
| 185515 |
+
"loss": 0.3437,
|
| 185516 |
+
"step": 67710
|
| 185517 |
+
},
|
| 185518 |
+
{
|
| 185519 |
+
"epoch": 541.14,
|
| 185520 |
+
"learning_rate": 8.93125e-06,
|
| 185521 |
+
"loss": 0.3961,
|
| 185522 |
+
"step": 67715
|
| 185523 |
+
},
|
| 185524 |
+
{
|
| 185525 |
+
"epoch": 541.18,
|
| 185526 |
+
"learning_rate": 8.931169871794873e-06,
|
| 185527 |
+
"loss": 0.7735,
|
| 185528 |
+
"step": 67720
|
| 185529 |
+
},
|
| 185530 |
+
{
|
| 185531 |
+
"epoch": 541.22,
|
| 185532 |
+
"learning_rate": 8.931089743589745e-06,
|
| 185533 |
+
"loss": 1.0214,
|
| 185534 |
+
"step": 67725
|
| 185535 |
+
},
|
| 185536 |
+
{
|
| 185537 |
+
"epoch": 541.26,
|
| 185538 |
+
"learning_rate": 8.931009615384616e-06,
|
| 185539 |
+
"loss": 0.2891,
|
| 185540 |
+
"step": 67730
|
| 185541 |
+
},
|
| 185542 |
+
{
|
| 185543 |
+
"epoch": 541.3,
|
| 185544 |
+
"learning_rate": 8.930929487179489e-06,
|
| 185545 |
+
"loss": 0.2835,
|
| 185546 |
+
"step": 67735
|
| 185547 |
+
},
|
| 185548 |
+
{
|
| 185549 |
+
"epoch": 541.34,
|
| 185550 |
+
"learning_rate": 8.930849358974359e-06,
|
| 185551 |
+
"loss": 0.3882,
|
| 185552 |
+
"step": 67740
|
| 185553 |
+
},
|
| 185554 |
+
{
|
| 185555 |
+
"epoch": 541.38,
|
| 185556 |
+
"learning_rate": 8.930769230769232e-06,
|
| 185557 |
+
"loss": 0.769,
|
| 185558 |
+
"step": 67745
|
| 185559 |
+
},
|
| 185560 |
+
{
|
| 185561 |
+
"epoch": 541.42,
|
| 185562 |
+
"learning_rate": 8.930689102564103e-06,
|
| 185563 |
+
"loss": 0.7533,
|
| 185564 |
+
"step": 67750
|
| 185565 |
+
},
|
| 185566 |
+
{
|
| 185567 |
+
"epoch": 541.46,
|
| 185568 |
+
"learning_rate": 8.930608974358975e-06,
|
| 185569 |
+
"loss": 0.2833,
|
| 185570 |
+
"step": 67755
|
| 185571 |
+
},
|
| 185572 |
+
{
|
| 185573 |
+
"epoch": 541.5,
|
| 185574 |
+
"learning_rate": 8.930528846153846e-06,
|
| 185575 |
+
"loss": 0.2899,
|
| 185576 |
+
"step": 67760
|
| 185577 |
+
},
|
| 185578 |
+
{
|
| 185579 |
+
"epoch": 541.54,
|
| 185580 |
+
"learning_rate": 8.930448717948719e-06,
|
| 185581 |
+
"loss": 0.4753,
|
| 185582 |
+
"step": 67765
|
| 185583 |
+
},
|
| 185584 |
+
{
|
| 185585 |
+
"epoch": 541.58,
|
| 185586 |
+
"learning_rate": 8.93036858974359e-06,
|
| 185587 |
+
"loss": 0.7755,
|
| 185588 |
+
"step": 67770
|
| 185589 |
+
},
|
| 185590 |
+
{
|
| 185591 |
+
"epoch": 541.62,
|
| 185592 |
+
"learning_rate": 8.930288461538462e-06,
|
| 185593 |
+
"loss": 0.8648,
|
| 185594 |
+
"step": 67775
|
| 185595 |
+
},
|
| 185596 |
+
{
|
| 185597 |
+
"epoch": 541.66,
|
| 185598 |
+
"learning_rate": 8.930208333333335e-06,
|
| 185599 |
+
"loss": 0.2927,
|
| 185600 |
+
"step": 67780
|
| 185601 |
+
},
|
| 185602 |
+
{
|
| 185603 |
+
"epoch": 541.7,
|
| 185604 |
+
"learning_rate": 8.930128205128206e-06,
|
| 185605 |
+
"loss": 0.3099,
|
| 185606 |
+
"step": 67785
|
| 185607 |
+
},
|
| 185608 |
+
{
|
| 185609 |
+
"epoch": 541.74,
|
| 185610 |
+
"learning_rate": 8.930048076923077e-06,
|
| 185611 |
+
"loss": 0.3615,
|
| 185612 |
+
"step": 67790
|
| 185613 |
+
},
|
| 185614 |
+
{
|
| 185615 |
+
"epoch": 541.78,
|
| 185616 |
+
"learning_rate": 8.929967948717949e-06,
|
| 185617 |
+
"loss": 0.8285,
|
| 185618 |
+
"step": 67795
|
| 185619 |
+
},
|
| 185620 |
+
{
|
| 185621 |
+
"epoch": 541.82,
|
| 185622 |
+
"learning_rate": 8.929887820512822e-06,
|
| 185623 |
+
"loss": 0.8921,
|
| 185624 |
+
"step": 67800
|
| 185625 |
+
},
|
| 185626 |
+
{
|
| 185627 |
+
"epoch": 541.86,
|
| 185628 |
+
"learning_rate": 8.929807692307693e-06,
|
| 185629 |
+
"loss": 0.3251,
|
| 185630 |
+
"step": 67805
|
| 185631 |
+
},
|
| 185632 |
+
{
|
| 185633 |
+
"epoch": 541.9,
|
| 185634 |
+
"learning_rate": 8.929727564102565e-06,
|
| 185635 |
+
"loss": 0.338,
|
| 185636 |
+
"step": 67810
|
| 185637 |
+
},
|
| 185638 |
+
{
|
| 185639 |
+
"epoch": 541.94,
|
| 185640 |
+
"learning_rate": 8.929647435897436e-06,
|
| 185641 |
+
"loss": 0.408,
|
| 185642 |
+
"step": 67815
|
| 185643 |
+
},
|
| 185644 |
+
{
|
| 185645 |
+
"epoch": 541.98,
|
| 185646 |
+
"learning_rate": 8.929567307692309e-06,
|
| 185647 |
+
"loss": 0.7369,
|
| 185648 |
+
"step": 67820
|
| 185649 |
+
},
|
| 185650 |
+
{
|
| 185651 |
+
"epoch": 542.0,
|
| 185652 |
+
"eval_loss": 0.3617897629737854,
|
| 185653 |
+
"eval_runtime": 39.0831,
|
| 185654 |
+
"eval_samples_per_second": 21.441,
|
| 185655 |
+
"eval_steps_per_second": 0.691,
|
| 185656 |
+
"eval_wer": 0.19498813760379596,
|
| 185657 |
+
"step": 67822
|
| 185658 |
+
},
|
| 185659 |
+
{
|
| 185660 |
+
"epoch": 542.02,
|
| 185661 |
+
"learning_rate": 8.92948717948718e-06,
|
| 185662 |
+
"loss": 0.482,
|
| 185663 |
+
"step": 67825
|
| 185664 |
+
},
|
| 185665 |
+
{
|
| 185666 |
+
"epoch": 542.06,
|
| 185667 |
+
"learning_rate": 8.929407051282052e-06,
|
| 185668 |
+
"loss": 0.3239,
|
| 185669 |
+
"step": 67830
|
| 185670 |
+
},
|
| 185671 |
+
{
|
| 185672 |
+
"epoch": 542.1,
|
| 185673 |
+
"learning_rate": 8.929326923076925e-06,
|
| 185674 |
+
"loss": 0.2679,
|
| 185675 |
+
"step": 67835
|
| 185676 |
+
},
|
| 185677 |
+
{
|
| 185678 |
+
"epoch": 542.14,
|
| 185679 |
+
"learning_rate": 8.929246794871796e-06,
|
| 185680 |
+
"loss": 0.3645,
|
| 185681 |
+
"step": 67840
|
| 185682 |
+
},
|
| 185683 |
+
{
|
| 185684 |
+
"epoch": 542.18,
|
| 185685 |
+
"learning_rate": 8.929166666666667e-06,
|
| 185686 |
+
"loss": 0.8296,
|
| 185687 |
+
"step": 67845
|
| 185688 |
+
},
|
| 185689 |
+
{
|
| 185690 |
+
"epoch": 542.22,
|
| 185691 |
+
"learning_rate": 8.929086538461539e-06,
|
| 185692 |
+
"loss": 0.8821,
|
| 185693 |
+
"step": 67850
|
| 185694 |
+
},
|
| 185695 |
+
{
|
| 185696 |
+
"epoch": 542.26,
|
| 185697 |
+
"learning_rate": 8.929006410256412e-06,
|
| 185698 |
+
"loss": 0.3484,
|
| 185699 |
+
"step": 67855
|
| 185700 |
+
},
|
| 185701 |
+
{
|
| 185702 |
+
"epoch": 542.3,
|
| 185703 |
+
"learning_rate": 8.928926282051282e-06,
|
| 185704 |
+
"loss": 0.4008,
|
| 185705 |
+
"step": 67860
|
| 185706 |
+
},
|
| 185707 |
+
{
|
| 185708 |
+
"epoch": 542.34,
|
| 185709 |
+
"learning_rate": 8.928846153846155e-06,
|
| 185710 |
+
"loss": 0.3902,
|
| 185711 |
+
"step": 67865
|
| 185712 |
+
},
|
| 185713 |
+
{
|
| 185714 |
+
"epoch": 542.38,
|
| 185715 |
+
"learning_rate": 8.928766025641028e-06,
|
| 185716 |
+
"loss": 0.7359,
|
| 185717 |
+
"step": 67870
|
| 185718 |
+
},
|
| 185719 |
+
{
|
| 185720 |
+
"epoch": 542.42,
|
| 185721 |
+
"learning_rate": 8.928685897435897e-06,
|
| 185722 |
+
"loss": 1.0072,
|
| 185723 |
+
"step": 67875
|
| 185724 |
+
},
|
| 185725 |
+
{
|
| 185726 |
+
"epoch": 542.46,
|
| 185727 |
+
"learning_rate": 8.92860576923077e-06,
|
| 185728 |
+
"loss": 0.2912,
|
| 185729 |
+
"step": 67880
|
| 185730 |
+
},
|
| 185731 |
+
{
|
| 185732 |
+
"epoch": 542.5,
|
| 185733 |
+
"learning_rate": 8.928525641025642e-06,
|
| 185734 |
+
"loss": 0.3905,
|
| 185735 |
+
"step": 67885
|
| 185736 |
+
},
|
| 185737 |
+
{
|
| 185738 |
+
"epoch": 542.54,
|
| 185739 |
+
"learning_rate": 8.928445512820513e-06,
|
| 185740 |
+
"loss": 0.381,
|
| 185741 |
+
"step": 67890
|
| 185742 |
+
},
|
| 185743 |
+
{
|
| 185744 |
+
"epoch": 542.58,
|
| 185745 |
+
"learning_rate": 8.928365384615384e-06,
|
| 185746 |
+
"loss": 0.7648,
|
| 185747 |
+
"step": 67895
|
| 185748 |
+
},
|
| 185749 |
+
{
|
| 185750 |
+
"epoch": 542.62,
|
| 185751 |
+
"learning_rate": 8.928285256410257e-06,
|
| 185752 |
+
"loss": 0.8711,
|
| 185753 |
+
"step": 67900
|
| 185754 |
+
},
|
| 185755 |
+
{
|
| 185756 |
+
"epoch": 542.66,
|
| 185757 |
+
"learning_rate": 8.928205128205129e-06,
|
| 185758 |
+
"loss": 0.2766,
|
| 185759 |
+
"step": 67905
|
| 185760 |
+
},
|
| 185761 |
+
{
|
| 185762 |
+
"epoch": 542.7,
|
| 185763 |
+
"learning_rate": 8.928125e-06,
|
| 185764 |
+
"loss": 0.3258,
|
| 185765 |
+
"step": 67910
|
| 185766 |
+
},
|
| 185767 |
+
{
|
| 185768 |
+
"epoch": 542.74,
|
| 185769 |
+
"learning_rate": 8.928044871794872e-06,
|
| 185770 |
+
"loss": 0.3602,
|
| 185771 |
+
"step": 67915
|
| 185772 |
+
},
|
| 185773 |
+
{
|
| 185774 |
+
"epoch": 542.78,
|
| 185775 |
+
"learning_rate": 8.927964743589745e-06,
|
| 185776 |
+
"loss": 0.7964,
|
| 185777 |
+
"step": 67920
|
| 185778 |
+
},
|
| 185779 |
+
{
|
| 185780 |
+
"epoch": 542.82,
|
| 185781 |
+
"learning_rate": 8.927884615384616e-06,
|
| 185782 |
+
"loss": 1.025,
|
| 185783 |
+
"step": 67925
|
| 185784 |
+
},
|
| 185785 |
+
{
|
| 185786 |
+
"epoch": 542.86,
|
| 185787 |
+
"learning_rate": 8.927804487179487e-06,
|
| 185788 |
+
"loss": 0.2855,
|
| 185789 |
+
"step": 67930
|
| 185790 |
+
},
|
| 185791 |
+
{
|
| 185792 |
+
"epoch": 542.9,
|
| 185793 |
+
"learning_rate": 8.92772435897436e-06,
|
| 185794 |
+
"loss": 0.3043,
|
| 185795 |
+
"step": 67935
|
| 185796 |
+
},
|
| 185797 |
+
{
|
| 185798 |
+
"epoch": 542.94,
|
| 185799 |
+
"learning_rate": 8.927644230769232e-06,
|
| 185800 |
+
"loss": 0.3756,
|
| 185801 |
+
"step": 67940
|
| 185802 |
+
},
|
| 185803 |
+
{
|
| 185804 |
+
"epoch": 542.98,
|
| 185805 |
+
"learning_rate": 8.927564102564103e-06,
|
| 185806 |
+
"loss": 0.8851,
|
| 185807 |
+
"step": 67945
|
| 185808 |
+
},
|
| 185809 |
+
{
|
| 185810 |
+
"epoch": 543.0,
|
| 185811 |
+
"eval_loss": 0.3958238661289215,
|
| 185812 |
+
"eval_runtime": 39.4545,
|
| 185813 |
+
"eval_samples_per_second": 21.24,
|
| 185814 |
+
"eval_steps_per_second": 0.684,
|
| 185815 |
+
"eval_wer": 0.19044464075382803,
|
| 185816 |
+
"step": 67947
|
| 185817 |
+
},
|
| 185818 |
+
{
|
| 185819 |
+
"epoch": 547.02,
|
| 185820 |
+
"learning_rate": 8.927483974358974e-06,
|
| 185821 |
+
"loss": 0.3812,
|
| 185822 |
+
"step": 67950
|
| 185823 |
+
},
|
| 185824 |
+
{
|
| 185825 |
+
"epoch": 547.06,
|
| 185826 |
+
"learning_rate": 8.927403846153847e-06,
|
| 185827 |
+
"loss": 0.3034,
|
| 185828 |
+
"step": 67955
|
| 185829 |
+
},
|
| 185830 |
+
{
|
| 185831 |
+
"epoch": 547.1,
|
| 185832 |
+
"learning_rate": 8.927323717948719e-06,
|
| 185833 |
+
"loss": 0.3513,
|
| 185834 |
+
"step": 67960
|
| 185835 |
+
},
|
| 185836 |
+
{
|
| 185837 |
+
"epoch": 547.14,
|
| 185838 |
+
"learning_rate": 8.92724358974359e-06,
|
| 185839 |
+
"loss": 0.4248,
|
| 185840 |
+
"step": 67965
|
| 185841 |
+
},
|
| 185842 |
+
{
|
| 185843 |
+
"epoch": 547.18,
|
| 185844 |
+
"learning_rate": 8.927163461538463e-06,
|
| 185845 |
+
"loss": 0.579,
|
| 185846 |
+
"step": 67970
|
| 185847 |
+
},
|
| 185848 |
+
{
|
| 185849 |
+
"epoch": 547.22,
|
| 185850 |
+
"learning_rate": 8.927083333333335e-06,
|
| 185851 |
+
"loss": 0.9471,
|
| 185852 |
+
"step": 67975
|
| 185853 |
+
},
|
| 185854 |
+
{
|
| 185855 |
+
"epoch": 547.27,
|
| 185856 |
+
"learning_rate": 8.927003205128206e-06,
|
| 185857 |
+
"loss": 0.3662,
|
| 185858 |
+
"step": 67980
|
| 185859 |
+
},
|
| 185860 |
+
{
|
| 185861 |
+
"epoch": 547.31,
|
| 185862 |
+
"learning_rate": 8.926923076923077e-06,
|
| 185863 |
+
"loss": 0.3769,
|
| 185864 |
+
"step": 67985
|
| 185865 |
+
},
|
| 185866 |
+
{
|
| 185867 |
+
"epoch": 547.35,
|
| 185868 |
+
"learning_rate": 8.92684294871795e-06,
|
| 185869 |
+
"loss": 0.4178,
|
| 185870 |
+
"step": 67990
|
| 185871 |
+
},
|
| 185872 |
+
{
|
| 185873 |
+
"epoch": 547.39,
|
| 185874 |
+
"learning_rate": 8.926762820512822e-06,
|
| 185875 |
+
"loss": 0.703,
|
| 185876 |
+
"step": 67995
|
| 185877 |
+
},
|
| 185878 |
+
{
|
| 185879 |
+
"epoch": 547.43,
|
| 185880 |
+
"learning_rate": 8.926682692307693e-06,
|
| 185881 |
+
"loss": 1.0705,
|
| 185882 |
+
"step": 68000
|
| 185883 |
+
},
|
| 185884 |
+
{
|
| 185885 |
+
"epoch": 547.47,
|
| 185886 |
+
"learning_rate": 8.926602564102564e-06,
|
| 185887 |
+
"loss": 0.3506,
|
| 185888 |
+
"step": 68005
|
| 185889 |
+
},
|
| 185890 |
+
{
|
| 185891 |
+
"epoch": 547.51,
|
| 185892 |
+
"learning_rate": 8.926522435897438e-06,
|
| 185893 |
+
"loss": 0.3535,
|
| 185894 |
+
"step": 68010
|
| 185895 |
+
},
|
| 185896 |
+
{
|
| 185897 |
+
"epoch": 547.55,
|
| 185898 |
+
"learning_rate": 8.926442307692307e-06,
|
| 185899 |
+
"loss": 0.3949,
|
| 185900 |
+
"step": 68015
|
| 185901 |
+
},
|
| 185902 |
+
{
|
| 185903 |
+
"epoch": 547.59,
|
| 185904 |
+
"learning_rate": 8.92636217948718e-06,
|
| 185905 |
+
"loss": 0.7904,
|
| 185906 |
+
"step": 68020
|
| 185907 |
+
},
|
| 185908 |
+
{
|
| 185909 |
+
"epoch": 547.63,
|
| 185910 |
+
"learning_rate": 8.926282051282053e-06,
|
| 185911 |
+
"loss": 0.9079,
|
| 185912 |
+
"step": 68025
|
| 185913 |
+
},
|
| 185914 |
+
{
|
| 185915 |
+
"epoch": 547.67,
|
| 185916 |
+
"learning_rate": 8.926201923076923e-06,
|
| 185917 |
+
"loss": 0.4741,
|
| 185918 |
+
"step": 68030
|
| 185919 |
+
},
|
| 185920 |
+
{
|
| 185921 |
+
"epoch": 547.71,
|
| 185922 |
+
"learning_rate": 8.926121794871796e-06,
|
| 185923 |
+
"loss": 0.3193,
|
| 185924 |
+
"step": 68035
|
| 185925 |
+
},
|
| 185926 |
+
{
|
| 185927 |
+
"epoch": 547.75,
|
| 185928 |
+
"learning_rate": 8.926041666666667e-06,
|
| 185929 |
+
"loss": 0.3734,
|
| 185930 |
+
"step": 68040
|
| 185931 |
+
},
|
| 185932 |
+
{
|
| 185933 |
+
"epoch": 547.79,
|
| 185934 |
+
"learning_rate": 8.925961538461539e-06,
|
| 185935 |
+
"loss": 0.7489,
|
| 185936 |
+
"step": 68045
|
| 185937 |
+
},
|
| 185938 |
+
{
|
| 185939 |
+
"epoch": 547.83,
|
| 185940 |
+
"learning_rate": 8.92588141025641e-06,
|
| 185941 |
+
"loss": 1.1998,
|
| 185942 |
+
"step": 68050
|
| 185943 |
+
},
|
| 185944 |
+
{
|
| 185945 |
+
"epoch": 547.87,
|
| 185946 |
+
"learning_rate": 8.925801282051283e-06,
|
| 185947 |
+
"loss": 0.321,
|
| 185948 |
+
"step": 68055
|
| 185949 |
+
},
|
| 185950 |
+
{
|
| 185951 |
+
"epoch": 547.91,
|
| 185952 |
+
"learning_rate": 8.925721153846154e-06,
|
| 185953 |
+
"loss": 0.3205,
|
| 185954 |
+
"step": 68060
|
| 185955 |
+
},
|
| 185956 |
+
{
|
| 185957 |
+
"epoch": 547.95,
|
| 185958 |
+
"learning_rate": 8.925641025641026e-06,
|
| 185959 |
+
"loss": 0.4711,
|
| 185960 |
+
"step": 68065
|
| 185961 |
+
},
|
| 185962 |
+
{
|
| 185963 |
+
"epoch": 547.99,
|
| 185964 |
+
"learning_rate": 8.925560897435899e-06,
|
| 185965 |
+
"loss": 0.897,
|
| 185966 |
+
"step": 68070
|
| 185967 |
+
},
|
| 185968 |
+
{
|
| 185969 |
+
"epoch": 548.0,
|
| 185970 |
+
"eval_loss": 0.3655170798301697,
|
| 185971 |
+
"eval_runtime": 38.9264,
|
| 185972 |
+
"eval_samples_per_second": 21.553,
|
| 185973 |
+
"eval_steps_per_second": 0.694,
|
| 185974 |
+
"eval_wer": 0.19745080500894455,
|
| 185975 |
+
"step": 68071
|
| 185976 |
+
},
|
| 185977 |
+
{
|
| 185978 |
+
"epoch": 548.03,
|
| 185979 |
+
"learning_rate": 8.92548076923077e-06,
|
| 185980 |
+
"loss": 0.3636,
|
| 185981 |
+
"step": 68075
|
| 185982 |
+
},
|
| 185983 |
+
{
|
| 185984 |
+
"epoch": 548.07,
|
| 185985 |
+
"learning_rate": 8.925400641025642e-06,
|
| 185986 |
+
"loss": 0.3824,
|
| 185987 |
+
"step": 68080
|
| 185988 |
+
},
|
| 185989 |
+
{
|
| 185990 |
+
"epoch": 548.11,
|
| 185991 |
+
"learning_rate": 8.925320512820513e-06,
|
| 185992 |
+
"loss": 0.3019,
|
| 185993 |
+
"step": 68085
|
| 185994 |
+
},
|
| 185995 |
+
{
|
| 185996 |
+
"epoch": 548.15,
|
| 185997 |
+
"learning_rate": 8.925240384615386e-06,
|
| 185998 |
+
"loss": 0.3797,
|
| 185999 |
+
"step": 68090
|
| 186000 |
+
},
|
| 186001 |
+
{
|
| 186002 |
+
"epoch": 548.19,
|
| 186003 |
+
"learning_rate": 8.925160256410257e-06,
|
| 186004 |
+
"loss": 0.9859,
|
| 186005 |
+
"step": 68095
|
| 186006 |
+
},
|
| 186007 |
+
{
|
| 186008 |
+
"epoch": 548.23,
|
| 186009 |
+
"learning_rate": 8.925080128205129e-06,
|
| 186010 |
+
"loss": 0.7557,
|
| 186011 |
+
"step": 68100
|
| 186012 |
+
},
|
| 186013 |
+
{
|
| 186014 |
+
"epoch": 548.27,
|
| 186015 |
+
"learning_rate": 8.925e-06,
|
| 186016 |
+
"loss": 0.359,
|
| 186017 |
+
"step": 68105
|
| 186018 |
+
},
|
| 186019 |
+
{
|
| 186020 |
+
"epoch": 548.31,
|
| 186021 |
+
"learning_rate": 8.924919871794873e-06,
|
| 186022 |
+
"loss": 0.273,
|
| 186023 |
+
"step": 68110
|
| 186024 |
+
},
|
| 186025 |
+
{
|
| 186026 |
+
"epoch": 548.35,
|
| 186027 |
+
"learning_rate": 8.924839743589745e-06,
|
| 186028 |
+
"loss": 0.4286,
|
| 186029 |
+
"step": 68115
|
| 186030 |
+
},
|
| 186031 |
+
{
|
| 186032 |
+
"epoch": 548.39,
|
| 186033 |
+
"learning_rate": 8.924759615384616e-06,
|
| 186034 |
+
"loss": 0.8112,
|
| 186035 |
+
"step": 68120
|
| 186036 |
+
},
|
| 186037 |
+
{
|
| 186038 |
+
"epoch": 548.43,
|
| 186039 |
+
"learning_rate": 8.924679487179489e-06,
|
| 186040 |
+
"loss": 0.6704,
|
| 186041 |
+
"step": 68125
|
| 186042 |
+
},
|
| 186043 |
+
{
|
| 186044 |
+
"epoch": 548.47,
|
| 186045 |
+
"learning_rate": 8.92459935897436e-06,
|
| 186046 |
+
"loss": 0.3017,
|
| 186047 |
+
"step": 68130
|
| 186048 |
+
},
|
| 186049 |
+
{
|
| 186050 |
+
"epoch": 548.51,
|
| 186051 |
+
"learning_rate": 8.924519230769232e-06,
|
| 186052 |
+
"loss": 0.3403,
|
| 186053 |
+
"step": 68135
|
| 186054 |
+
},
|
| 186055 |
+
{
|
| 186056 |
+
"epoch": 548.55,
|
| 186057 |
+
"learning_rate": 8.924439102564103e-06,
|
| 186058 |
+
"loss": 0.4607,
|
| 186059 |
+
"step": 68140
|
| 186060 |
+
},
|
| 186061 |
+
{
|
| 186062 |
+
"epoch": 548.59,
|
| 186063 |
+
"learning_rate": 8.924358974358976e-06,
|
| 186064 |
+
"loss": 1.0333,
|
| 186065 |
+
"step": 68145
|
| 186066 |
+
},
|
| 186067 |
+
{
|
| 186068 |
+
"epoch": 548.63,
|
| 186069 |
+
"learning_rate": 8.924278846153846e-06,
|
| 186070 |
+
"loss": 0.8082,
|
| 186071 |
+
"step": 68150
|
| 186072 |
+
},
|
| 186073 |
+
{
|
| 186074 |
+
"epoch": 548.67,
|
| 186075 |
+
"learning_rate": 8.924198717948719e-06,
|
| 186076 |
+
"loss": 0.2737,
|
| 186077 |
+
"step": 68155
|
| 186078 |
+
},
|
| 186079 |
+
{
|
| 186080 |
+
"epoch": 548.71,
|
| 186081 |
+
"learning_rate": 8.92411858974359e-06,
|
| 186082 |
+
"loss": 0.283,
|
| 186083 |
+
"step": 68160
|
| 186084 |
+
},
|
| 186085 |
+
{
|
| 186086 |
+
"epoch": 548.76,
|
| 186087 |
+
"learning_rate": 8.924038461538461e-06,
|
| 186088 |
+
"loss": 0.5159,
|
| 186089 |
+
"step": 68165
|
| 186090 |
+
},
|
| 186091 |
+
{
|
| 186092 |
+
"epoch": 548.8,
|
| 186093 |
+
"learning_rate": 8.923958333333335e-06,
|
| 186094 |
+
"loss": 1.1066,
|
| 186095 |
+
"step": 68170
|
| 186096 |
+
},
|
| 186097 |
+
{
|
| 186098 |
+
"epoch": 548.84,
|
| 186099 |
+
"learning_rate": 8.923878205128206e-06,
|
| 186100 |
+
"loss": 0.7521,
|
| 186101 |
+
"step": 68175
|
| 186102 |
+
},
|
| 186103 |
+
{
|
| 186104 |
+
"epoch": 548.88,
|
| 186105 |
+
"learning_rate": 8.923798076923077e-06,
|
| 186106 |
+
"loss": 0.3451,
|
| 186107 |
+
"step": 68180
|
| 186108 |
+
},
|
| 186109 |
+
{
|
| 186110 |
+
"epoch": 548.92,
|
| 186111 |
+
"learning_rate": 8.923717948717949e-06,
|
| 186112 |
+
"loss": 0.3402,
|
| 186113 |
+
"step": 68185
|
| 186114 |
+
},
|
| 186115 |
+
{
|
| 186116 |
+
"epoch": 548.96,
|
| 186117 |
+
"learning_rate": 8.923637820512822e-06,
|
| 186118 |
+
"loss": 0.4594,
|
| 186119 |
+
"step": 68190
|
| 186120 |
+
},
|
| 186121 |
+
{
|
| 186122 |
+
"epoch": 549.0,
|
| 186123 |
+
"learning_rate": 8.923573717948719e-06,
|
| 186124 |
+
"loss": 1.3483,
|
| 186125 |
+
"step": 68195
|
| 186126 |
+
},
|
| 186127 |
+
{
|
| 186128 |
+
"epoch": 549.0,
|
| 186129 |
+
"eval_loss": 0.38058263063430786,
|
| 186130 |
+
"eval_runtime": 39.1105,
|
| 186131 |
+
"eval_samples_per_second": 21.452,
|
| 186132 |
+
"eval_steps_per_second": 0.69,
|
| 186133 |
+
"eval_wer": 0.1986700537955768,
|
| 186134 |
+
"step": 68195
|
| 186135 |
}
|
| 186136 |
],
|
| 186137 |
+
"max_steps": 620000,
|
| 186138 |
"num_train_epochs": 5000,
|
| 186139 |
+
"total_flos": 1.919002060140202e+20,
|
| 186140 |
"trial_name": null,
|
| 186141 |
"trial_params": null
|
| 186142 |
}
|
model-bin/finetune/base/{checkpoint-67572 → checkpoint-68195}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629837488.925602/events.out.tfevents.1629837488.c435e1c5ee04.920.221
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:11a21fc8f147677a8204ea6fd68167f58938d658f3f6c9d85a072bf833a47153
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629838113.24123/events.out.tfevents.1629838113.c435e1c5ee04.920.223
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c87c23309dcbd740181b93f110e946100c7e5c1d6e79d8ed9ec5fdf67c127beb
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629838737.3102496/events.out.tfevents.1629838737.c435e1c5ee04.920.225
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:361864c826c35b2fc42f117bbb96788d0dc929c5248438f1c561cdc439022b27
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629839378.965553/events.out.tfevents.1629839378.c435e1c5ee04.920.227
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38efd8537eb21083b0de52e22cb6825508891dcfc3e1937d91b899af6bf030e4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629840015.2190616/events.out.tfevents.1629840015.c435e1c5ee04.920.229
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d5f973dd5e9cb6c871e2e118ff1419d9451dd9cc081e892d90a6195c7e3a0921
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629837488.c435e1c5ee04.920.220
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:618e987937fec56b32ef0ac288968c32fef6b3f06c07c5895e99df6432aa907f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629838113.c435e1c5ee04.920.222
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f2d243d2392ff1b3a2498ac040ea1fbde1159401225f1325087750bc2a099f77
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629838737.c435e1c5ee04.920.224
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff9e714dc0ae26ecf20d6efa0cf08919b8511cd95c487a9f8cb023380f51d40c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629839378.c435e1c5ee04.920.226
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:97109b02856def0906fd893d520582d1b4c154e4d3264c9f47d105068a5140d5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629840015.c435e1c5ee04.920.228
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2e2e1336f165fb7fa488885abb82e9ab78ead84f17ca64909350e74604045451
|
| 3 |
+
size 8622
|