"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629955488.5549626/events.out.tfevents.1629955488.8e89bd551565.924.71 +3 -0
- model-bin/finetune/base/log/1629955916.7474144/events.out.tfevents.1629955916.8e89bd551565.924.73 +3 -0
- model-bin/finetune/base/log/1629956348.062922/events.out.tfevents.1629956348.8e89bd551565.924.75 +3 -0
- model-bin/finetune/base/log/1629956783.883363/events.out.tfevents.1629956785.8e89bd551565.924.77 +3 -0
- model-bin/finetune/base/log/1629957212.363884/events.out.tfevents.1629957212.8e89bd551565.924.79 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629955488.8e89bd551565.924.70 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629955916.8e89bd551565.924.72 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629956348.8e89bd551565.924.74 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629956783.8e89bd551565.924.76 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629957212.8e89bd551565.924.78 +3 -0
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/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:dd888f34151939c973bc1b406e70d1518151a52f0f47a7f3b3e4bf6e20a0814c
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/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:db9d3ab6bd57b00e154b150f52824ff898ace57456f314ffd780af9ef1025982
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/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 sha256:48c031af756610829efb4da17d1a6a0df0c7110ff4a69596b5d680c83878235d
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/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 sha256:70fe0de8b811344c7522c0f25fa011a101a624c1ba8b39523e69af7715d8bd97
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/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 sha256:35eb7d62fd453a5a24e6c8356ac861d30761c911bc452442e0b8186f6dd6e9f9
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17637692697401752,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-92342",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -217329,11 +217329,800 @@
|
|
| 217329 |
"eval_steps_per_second": 0.696,
|
| 217330 |
"eval_wer": 0.19165727170236754,
|
| 217331 |
"step": 92716
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 217332 |
}
|
| 217333 |
],
|
| 217334 |
"max_steps": 620000,
|
| 217335 |
"num_train_epochs": 5000,
|
| 217336 |
-
"total_flos": 2.
|
| 217337 |
"trial_name": null,
|
| 217338 |
"trial_params": null
|
| 217339 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17637692697401752,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-92342",
|
| 4 |
+
"epoch": 751.995983935743,
|
| 5 |
+
"global_step": 93337,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 217329 |
"eval_steps_per_second": 0.696,
|
| 217330 |
"eval_wer": 0.19165727170236754,
|
| 217331 |
"step": 92716
|
| 217332 |
+
},
|
| 217333 |
+
{
|
| 217334 |
+
"epoch": 753.03,
|
| 217335 |
+
"learning_rate": 8.530721153846154e-06,
|
| 217336 |
+
"loss": 0.3505,
|
| 217337 |
+
"step": 92720
|
| 217338 |
+
},
|
| 217339 |
+
{
|
| 217340 |
+
"epoch": 753.07,
|
| 217341 |
+
"learning_rate": 8.530641025641027e-06,
|
| 217342 |
+
"loss": 0.295,
|
| 217343 |
+
"step": 92725
|
| 217344 |
+
},
|
| 217345 |
+
{
|
| 217346 |
+
"epoch": 753.11,
|
| 217347 |
+
"learning_rate": 8.530560897435898e-06,
|
| 217348 |
+
"loss": 0.3735,
|
| 217349 |
+
"step": 92730
|
| 217350 |
+
},
|
| 217351 |
+
{
|
| 217352 |
+
"epoch": 753.15,
|
| 217353 |
+
"learning_rate": 8.53048076923077e-06,
|
| 217354 |
+
"loss": 0.4566,
|
| 217355 |
+
"step": 92735
|
| 217356 |
+
},
|
| 217357 |
+
{
|
| 217358 |
+
"epoch": 753.19,
|
| 217359 |
+
"learning_rate": 8.530400641025641e-06,
|
| 217360 |
+
"loss": 0.9224,
|
| 217361 |
+
"step": 92740
|
| 217362 |
+
},
|
| 217363 |
+
{
|
| 217364 |
+
"epoch": 753.23,
|
| 217365 |
+
"learning_rate": 8.530320512820514e-06,
|
| 217366 |
+
"loss": 0.6817,
|
| 217367 |
+
"step": 92745
|
| 217368 |
+
},
|
| 217369 |
+
{
|
| 217370 |
+
"epoch": 753.28,
|
| 217371 |
+
"learning_rate": 8.530240384615386e-06,
|
| 217372 |
+
"loss": 0.3159,
|
| 217373 |
+
"step": 92750
|
| 217374 |
+
},
|
| 217375 |
+
{
|
| 217376 |
+
"epoch": 753.32,
|
| 217377 |
+
"learning_rate": 8.530160256410257e-06,
|
| 217378 |
+
"loss": 0.3457,
|
| 217379 |
+
"step": 92755
|
| 217380 |
+
},
|
| 217381 |
+
{
|
| 217382 |
+
"epoch": 753.36,
|
| 217383 |
+
"learning_rate": 8.53008012820513e-06,
|
| 217384 |
+
"loss": 0.4083,
|
| 217385 |
+
"step": 92760
|
| 217386 |
+
},
|
| 217387 |
+
{
|
| 217388 |
+
"epoch": 753.4,
|
| 217389 |
+
"learning_rate": 8.530000000000001e-06,
|
| 217390 |
+
"loss": 1.0235,
|
| 217391 |
+
"step": 92765
|
| 217392 |
+
},
|
| 217393 |
+
{
|
| 217394 |
+
"epoch": 753.44,
|
| 217395 |
+
"learning_rate": 8.529919871794873e-06,
|
| 217396 |
+
"loss": 0.6671,
|
| 217397 |
+
"step": 92770
|
| 217398 |
+
},
|
| 217399 |
+
{
|
| 217400 |
+
"epoch": 753.48,
|
| 217401 |
+
"learning_rate": 8.529839743589744e-06,
|
| 217402 |
+
"loss": 0.3144,
|
| 217403 |
+
"step": 92775
|
| 217404 |
+
},
|
| 217405 |
+
{
|
| 217406 |
+
"epoch": 753.52,
|
| 217407 |
+
"learning_rate": 8.529759615384617e-06,
|
| 217408 |
+
"loss": 0.3815,
|
| 217409 |
+
"step": 92780
|
| 217410 |
+
},
|
| 217411 |
+
{
|
| 217412 |
+
"epoch": 753.56,
|
| 217413 |
+
"learning_rate": 8.529679487179487e-06,
|
| 217414 |
+
"loss": 0.4122,
|
| 217415 |
+
"step": 92785
|
| 217416 |
+
},
|
| 217417 |
+
{
|
| 217418 |
+
"epoch": 753.6,
|
| 217419 |
+
"learning_rate": 8.52959935897436e-06,
|
| 217420 |
+
"loss": 0.9775,
|
| 217421 |
+
"step": 92790
|
| 217422 |
+
},
|
| 217423 |
+
{
|
| 217424 |
+
"epoch": 753.64,
|
| 217425 |
+
"learning_rate": 8.529519230769231e-06,
|
| 217426 |
+
"loss": 0.65,
|
| 217427 |
+
"step": 92795
|
| 217428 |
+
},
|
| 217429 |
+
{
|
| 217430 |
+
"epoch": 753.68,
|
| 217431 |
+
"learning_rate": 8.529439102564103e-06,
|
| 217432 |
+
"loss": 0.2865,
|
| 217433 |
+
"step": 92800
|
| 217434 |
+
},
|
| 217435 |
+
{
|
| 217436 |
+
"epoch": 753.72,
|
| 217437 |
+
"learning_rate": 8.529358974358976e-06,
|
| 217438 |
+
"loss": 0.3356,
|
| 217439 |
+
"step": 92805
|
| 217440 |
+
},
|
| 217441 |
+
{
|
| 217442 |
+
"epoch": 753.76,
|
| 217443 |
+
"learning_rate": 8.529278846153847e-06,
|
| 217444 |
+
"loss": 0.3791,
|
| 217445 |
+
"step": 92810
|
| 217446 |
+
},
|
| 217447 |
+
{
|
| 217448 |
+
"epoch": 753.8,
|
| 217449 |
+
"learning_rate": 8.529198717948718e-06,
|
| 217450 |
+
"loss": 0.9281,
|
| 217451 |
+
"step": 92815
|
| 217452 |
+
},
|
| 217453 |
+
{
|
| 217454 |
+
"epoch": 753.84,
|
| 217455 |
+
"learning_rate": 8.52911858974359e-06,
|
| 217456 |
+
"loss": 0.7974,
|
| 217457 |
+
"step": 92820
|
| 217458 |
+
},
|
| 217459 |
+
{
|
| 217460 |
+
"epoch": 753.88,
|
| 217461 |
+
"learning_rate": 8.529038461538463e-06,
|
| 217462 |
+
"loss": 0.343,
|
| 217463 |
+
"step": 92825
|
| 217464 |
+
},
|
| 217465 |
+
{
|
| 217466 |
+
"epoch": 753.92,
|
| 217467 |
+
"learning_rate": 8.528958333333334e-06,
|
| 217468 |
+
"loss": 0.2876,
|
| 217469 |
+
"step": 92830
|
| 217470 |
+
},
|
| 217471 |
+
{
|
| 217472 |
+
"epoch": 753.96,
|
| 217473 |
+
"learning_rate": 8.528878205128205e-06,
|
| 217474 |
+
"loss": 0.4759,
|
| 217475 |
+
"step": 92835
|
| 217476 |
+
},
|
| 217477 |
+
{
|
| 217478 |
+
"epoch": 754.0,
|
| 217479 |
+
"eval_loss": 0.3695593476295471,
|
| 217480 |
+
"eval_runtime": 39.7408,
|
| 217481 |
+
"eval_samples_per_second": 21.087,
|
| 217482 |
+
"eval_steps_per_second": 0.679,
|
| 217483 |
+
"eval_wer": 0.1869410310477714,
|
| 217484 |
+
"step": 92839
|
| 217485 |
+
},
|
| 217486 |
+
{
|
| 217487 |
+
"epoch": 742.01,
|
| 217488 |
+
"learning_rate": 8.528798076923077e-06,
|
| 217489 |
+
"loss": 0.4816,
|
| 217490 |
+
"step": 92840
|
| 217491 |
+
},
|
| 217492 |
+
{
|
| 217493 |
+
"epoch": 742.05,
|
| 217494 |
+
"learning_rate": 8.52871794871795e-06,
|
| 217495 |
+
"loss": 0.356,
|
| 217496 |
+
"step": 92845
|
| 217497 |
+
},
|
| 217498 |
+
{
|
| 217499 |
+
"epoch": 742.09,
|
| 217500 |
+
"learning_rate": 8.528637820512821e-06,
|
| 217501 |
+
"loss": 0.2479,
|
| 217502 |
+
"step": 92850
|
| 217503 |
+
},
|
| 217504 |
+
{
|
| 217505 |
+
"epoch": 742.13,
|
| 217506 |
+
"learning_rate": 8.528557692307693e-06,
|
| 217507 |
+
"loss": 0.3506,
|
| 217508 |
+
"step": 92855
|
| 217509 |
+
},
|
| 217510 |
+
{
|
| 217511 |
+
"epoch": 742.17,
|
| 217512 |
+
"learning_rate": 8.528477564102566e-06,
|
| 217513 |
+
"loss": 0.5059,
|
| 217514 |
+
"step": 92860
|
| 217515 |
+
},
|
| 217516 |
+
{
|
| 217517 |
+
"epoch": 742.21,
|
| 217518 |
+
"learning_rate": 8.528397435897437e-06,
|
| 217519 |
+
"loss": 1.0797,
|
| 217520 |
+
"step": 92865
|
| 217521 |
+
},
|
| 217522 |
+
{
|
| 217523 |
+
"epoch": 742.25,
|
| 217524 |
+
"learning_rate": 8.528317307692308e-06,
|
| 217525 |
+
"loss": 0.3273,
|
| 217526 |
+
"step": 92870
|
| 217527 |
+
},
|
| 217528 |
+
{
|
| 217529 |
+
"epoch": 742.29,
|
| 217530 |
+
"learning_rate": 8.52823717948718e-06,
|
| 217531 |
+
"loss": 0.3208,
|
| 217532 |
+
"step": 92875
|
| 217533 |
+
},
|
| 217534 |
+
{
|
| 217535 |
+
"epoch": 742.33,
|
| 217536 |
+
"learning_rate": 8.528157051282053e-06,
|
| 217537 |
+
"loss": 0.3685,
|
| 217538 |
+
"step": 92880
|
| 217539 |
+
},
|
| 217540 |
+
{
|
| 217541 |
+
"epoch": 742.37,
|
| 217542 |
+
"learning_rate": 8.528076923076924e-06,
|
| 217543 |
+
"loss": 0.553,
|
| 217544 |
+
"step": 92885
|
| 217545 |
+
},
|
| 217546 |
+
{
|
| 217547 |
+
"epoch": 742.41,
|
| 217548 |
+
"learning_rate": 8.527996794871795e-06,
|
| 217549 |
+
"loss": 1.1638,
|
| 217550 |
+
"step": 92890
|
| 217551 |
+
},
|
| 217552 |
+
{
|
| 217553 |
+
"epoch": 742.45,
|
| 217554 |
+
"learning_rate": 8.527916666666667e-06,
|
| 217555 |
+
"loss": 0.3191,
|
| 217556 |
+
"step": 92895
|
| 217557 |
+
},
|
| 217558 |
+
{
|
| 217559 |
+
"epoch": 742.49,
|
| 217560 |
+
"learning_rate": 8.52783653846154e-06,
|
| 217561 |
+
"loss": 0.3565,
|
| 217562 |
+
"step": 92900
|
| 217563 |
+
},
|
| 217564 |
+
{
|
| 217565 |
+
"epoch": 742.53,
|
| 217566 |
+
"learning_rate": 8.527756410256411e-06,
|
| 217567 |
+
"loss": 0.3219,
|
| 217568 |
+
"step": 92905
|
| 217569 |
+
},
|
| 217570 |
+
{
|
| 217571 |
+
"epoch": 742.57,
|
| 217572 |
+
"learning_rate": 8.527676282051283e-06,
|
| 217573 |
+
"loss": 0.4828,
|
| 217574 |
+
"step": 92910
|
| 217575 |
+
},
|
| 217576 |
+
{
|
| 217577 |
+
"epoch": 742.61,
|
| 217578 |
+
"learning_rate": 8.527596153846156e-06,
|
| 217579 |
+
"loss": 1.0651,
|
| 217580 |
+
"step": 92915
|
| 217581 |
+
},
|
| 217582 |
+
{
|
| 217583 |
+
"epoch": 742.65,
|
| 217584 |
+
"learning_rate": 8.527516025641025e-06,
|
| 217585 |
+
"loss": 0.3005,
|
| 217586 |
+
"step": 92920
|
| 217587 |
+
},
|
| 217588 |
+
{
|
| 217589 |
+
"epoch": 742.69,
|
| 217590 |
+
"learning_rate": 8.527435897435898e-06,
|
| 217591 |
+
"loss": 0.2958,
|
| 217592 |
+
"step": 92925
|
| 217593 |
+
},
|
| 217594 |
+
{
|
| 217595 |
+
"epoch": 742.73,
|
| 217596 |
+
"learning_rate": 8.52735576923077e-06,
|
| 217597 |
+
"loss": 0.3653,
|
| 217598 |
+
"step": 92930
|
| 217599 |
+
},
|
| 217600 |
+
{
|
| 217601 |
+
"epoch": 742.77,
|
| 217602 |
+
"learning_rate": 8.527275641025641e-06,
|
| 217603 |
+
"loss": 0.5593,
|
| 217604 |
+
"step": 92935
|
| 217605 |
+
},
|
| 217606 |
+
{
|
| 217607 |
+
"epoch": 742.81,
|
| 217608 |
+
"learning_rate": 8.527195512820512e-06,
|
| 217609 |
+
"loss": 1.1913,
|
| 217610 |
+
"step": 92940
|
| 217611 |
+
},
|
| 217612 |
+
{
|
| 217613 |
+
"epoch": 742.85,
|
| 217614 |
+
"learning_rate": 8.527115384615385e-06,
|
| 217615 |
+
"loss": 0.346,
|
| 217616 |
+
"step": 92945
|
| 217617 |
+
},
|
| 217618 |
+
{
|
| 217619 |
+
"epoch": 742.89,
|
| 217620 |
+
"learning_rate": 8.527035256410257e-06,
|
| 217621 |
+
"loss": 0.3054,
|
| 217622 |
+
"step": 92950
|
| 217623 |
+
},
|
| 217624 |
+
{
|
| 217625 |
+
"epoch": 742.93,
|
| 217626 |
+
"learning_rate": 8.526955128205128e-06,
|
| 217627 |
+
"loss": 0.3304,
|
| 217628 |
+
"step": 92955
|
| 217629 |
+
},
|
| 217630 |
+
{
|
| 217631 |
+
"epoch": 742.97,
|
| 217632 |
+
"learning_rate": 8.526875000000001e-06,
|
| 217633 |
+
"loss": 0.547,
|
| 217634 |
+
"step": 92960
|
| 217635 |
+
},
|
| 217636 |
+
{
|
| 217637 |
+
"epoch": 743.0,
|
| 217638 |
+
"eval_loss": 0.3991490602493286,
|
| 217639 |
+
"eval_runtime": 37.5272,
|
| 217640 |
+
"eval_samples_per_second": 22.33,
|
| 217641 |
+
"eval_steps_per_second": 0.719,
|
| 217642 |
+
"eval_wer": 0.19152162923376906,
|
| 217643 |
+
"step": 92964
|
| 217644 |
+
},
|
| 217645 |
+
{
|
| 217646 |
+
"epoch": 743.01,
|
| 217647 |
+
"learning_rate": 8.526794871794873e-06,
|
| 217648 |
+
"loss": 0.3283,
|
| 217649 |
+
"step": 92965
|
| 217650 |
+
},
|
| 217651 |
+
{
|
| 217652 |
+
"epoch": 743.05,
|
| 217653 |
+
"learning_rate": 8.526714743589744e-06,
|
| 217654 |
+
"loss": 0.2986,
|
| 217655 |
+
"step": 92970
|
| 217656 |
+
},
|
| 217657 |
+
{
|
| 217658 |
+
"epoch": 743.09,
|
| 217659 |
+
"learning_rate": 8.526634615384615e-06,
|
| 217660 |
+
"loss": 0.3763,
|
| 217661 |
+
"step": 92975
|
| 217662 |
+
},
|
| 217663 |
+
{
|
| 217664 |
+
"epoch": 743.13,
|
| 217665 |
+
"learning_rate": 8.526554487179488e-06,
|
| 217666 |
+
"loss": 0.3062,
|
| 217667 |
+
"step": 92980
|
| 217668 |
+
},
|
| 217669 |
+
{
|
| 217670 |
+
"epoch": 743.17,
|
| 217671 |
+
"learning_rate": 8.52647435897436e-06,
|
| 217672 |
+
"loss": 0.5076,
|
| 217673 |
+
"step": 92985
|
| 217674 |
+
},
|
| 217675 |
+
{
|
| 217676 |
+
"epoch": 743.21,
|
| 217677 |
+
"learning_rate": 8.526394230769231e-06,
|
| 217678 |
+
"loss": 1.1239,
|
| 217679 |
+
"step": 92990
|
| 217680 |
+
},
|
| 217681 |
+
{
|
| 217682 |
+
"epoch": 743.25,
|
| 217683 |
+
"learning_rate": 8.526314102564102e-06,
|
| 217684 |
+
"loss": 0.3523,
|
| 217685 |
+
"step": 92995
|
| 217686 |
+
},
|
| 217687 |
+
{
|
| 217688 |
+
"epoch": 743.29,
|
| 217689 |
+
"learning_rate": 8.526233974358976e-06,
|
| 217690 |
+
"loss": 0.2903,
|
| 217691 |
+
"step": 93000
|
| 217692 |
+
},
|
| 217693 |
+
{
|
| 217694 |
+
"epoch": 743.33,
|
| 217695 |
+
"learning_rate": 8.526153846153847e-06,
|
| 217696 |
+
"loss": 0.367,
|
| 217697 |
+
"step": 93005
|
| 217698 |
+
},
|
| 217699 |
+
{
|
| 217700 |
+
"epoch": 743.37,
|
| 217701 |
+
"learning_rate": 8.526073717948718e-06,
|
| 217702 |
+
"loss": 0.6214,
|
| 217703 |
+
"step": 93010
|
| 217704 |
+
},
|
| 217705 |
+
{
|
| 217706 |
+
"epoch": 743.41,
|
| 217707 |
+
"learning_rate": 8.525993589743591e-06,
|
| 217708 |
+
"loss": 1.1313,
|
| 217709 |
+
"step": 93015
|
| 217710 |
+
},
|
| 217711 |
+
{
|
| 217712 |
+
"epoch": 743.45,
|
| 217713 |
+
"learning_rate": 8.525913461538463e-06,
|
| 217714 |
+
"loss": 0.2883,
|
| 217715 |
+
"step": 93020
|
| 217716 |
+
},
|
| 217717 |
+
{
|
| 217718 |
+
"epoch": 743.49,
|
| 217719 |
+
"learning_rate": 8.525833333333334e-06,
|
| 217720 |
+
"loss": 0.2828,
|
| 217721 |
+
"step": 93025
|
| 217722 |
+
},
|
| 217723 |
+
{
|
| 217724 |
+
"epoch": 743.53,
|
| 217725 |
+
"learning_rate": 8.525753205128205e-06,
|
| 217726 |
+
"loss": 0.3696,
|
| 217727 |
+
"step": 93030
|
| 217728 |
+
},
|
| 217729 |
+
{
|
| 217730 |
+
"epoch": 743.57,
|
| 217731 |
+
"learning_rate": 8.525673076923078e-06,
|
| 217732 |
+
"loss": 0.4927,
|
| 217733 |
+
"step": 93035
|
| 217734 |
+
},
|
| 217735 |
+
{
|
| 217736 |
+
"epoch": 743.61,
|
| 217737 |
+
"learning_rate": 8.52559294871795e-06,
|
| 217738 |
+
"loss": 1.1815,
|
| 217739 |
+
"step": 93040
|
| 217740 |
+
},
|
| 217741 |
+
{
|
| 217742 |
+
"epoch": 743.65,
|
| 217743 |
+
"learning_rate": 8.525512820512821e-06,
|
| 217744 |
+
"loss": 0.3192,
|
| 217745 |
+
"step": 93045
|
| 217746 |
+
},
|
| 217747 |
+
{
|
| 217748 |
+
"epoch": 743.69,
|
| 217749 |
+
"learning_rate": 8.525432692307694e-06,
|
| 217750 |
+
"loss": 0.659,
|
| 217751 |
+
"step": 93050
|
| 217752 |
+
},
|
| 217753 |
+
{
|
| 217754 |
+
"epoch": 743.73,
|
| 217755 |
+
"learning_rate": 8.525352564102566e-06,
|
| 217756 |
+
"loss": 0.3991,
|
| 217757 |
+
"step": 93055
|
| 217758 |
+
},
|
| 217759 |
+
{
|
| 217760 |
+
"epoch": 743.77,
|
| 217761 |
+
"learning_rate": 8.525272435897437e-06,
|
| 217762 |
+
"loss": 0.5503,
|
| 217763 |
+
"step": 93060
|
| 217764 |
+
},
|
| 217765 |
+
{
|
| 217766 |
+
"epoch": 743.81,
|
| 217767 |
+
"learning_rate": 8.525192307692308e-06,
|
| 217768 |
+
"loss": 1.1523,
|
| 217769 |
+
"step": 93065
|
| 217770 |
+
},
|
| 217771 |
+
{
|
| 217772 |
+
"epoch": 743.85,
|
| 217773 |
+
"learning_rate": 8.525112179487181e-06,
|
| 217774 |
+
"loss": 0.2851,
|
| 217775 |
+
"step": 93070
|
| 217776 |
+
},
|
| 217777 |
+
{
|
| 217778 |
+
"epoch": 743.89,
|
| 217779 |
+
"learning_rate": 8.525032051282051e-06,
|
| 217780 |
+
"loss": 0.364,
|
| 217781 |
+
"step": 93075
|
| 217782 |
+
},
|
| 217783 |
+
{
|
| 217784 |
+
"epoch": 743.93,
|
| 217785 |
+
"learning_rate": 8.524951923076924e-06,
|
| 217786 |
+
"loss": 0.31,
|
| 217787 |
+
"step": 93080
|
| 217788 |
+
},
|
| 217789 |
+
{
|
| 217790 |
+
"epoch": 743.97,
|
| 217791 |
+
"learning_rate": 8.524871794871795e-06,
|
| 217792 |
+
"loss": 0.5022,
|
| 217793 |
+
"step": 93085
|
| 217794 |
+
},
|
| 217795 |
+
{
|
| 217796 |
+
"epoch": 744.0,
|
| 217797 |
+
"eval_loss": 0.3836924135684967,
|
| 217798 |
+
"eval_runtime": 38.2104,
|
| 217799 |
+
"eval_samples_per_second": 21.931,
|
| 217800 |
+
"eval_steps_per_second": 0.707,
|
| 217801 |
+
"eval_wer": 0.1841402698607403,
|
| 217802 |
+
"step": 93089
|
| 217803 |
+
},
|
| 217804 |
+
{
|
| 217805 |
+
"epoch": 750.01,
|
| 217806 |
+
"learning_rate": 8.524791666666667e-06,
|
| 217807 |
+
"loss": 0.3855,
|
| 217808 |
+
"step": 93090
|
| 217809 |
+
},
|
| 217810 |
+
{
|
| 217811 |
+
"epoch": 750.05,
|
| 217812 |
+
"learning_rate": 8.524711538461538e-06,
|
| 217813 |
+
"loss": 0.3155,
|
| 217814 |
+
"step": 93095
|
| 217815 |
+
},
|
| 217816 |
+
{
|
| 217817 |
+
"epoch": 750.09,
|
| 217818 |
+
"learning_rate": 8.524631410256411e-06,
|
| 217819 |
+
"loss": 0.2896,
|
| 217820 |
+
"step": 93100
|
| 217821 |
+
},
|
| 217822 |
+
{
|
| 217823 |
+
"epoch": 750.13,
|
| 217824 |
+
"learning_rate": 8.524551282051283e-06,
|
| 217825 |
+
"loss": 0.3521,
|
| 217826 |
+
"step": 93105
|
| 217827 |
+
},
|
| 217828 |
+
{
|
| 217829 |
+
"epoch": 750.17,
|
| 217830 |
+
"learning_rate": 8.524471153846154e-06,
|
| 217831 |
+
"loss": 0.5719,
|
| 217832 |
+
"step": 93110
|
| 217833 |
+
},
|
| 217834 |
+
{
|
| 217835 |
+
"epoch": 750.21,
|
| 217836 |
+
"learning_rate": 8.524391025641027e-06,
|
| 217837 |
+
"loss": 1.1404,
|
| 217838 |
+
"step": 93115
|
| 217839 |
+
},
|
| 217840 |
+
{
|
| 217841 |
+
"epoch": 750.25,
|
| 217842 |
+
"learning_rate": 8.524310897435898e-06,
|
| 217843 |
+
"loss": 0.3249,
|
| 217844 |
+
"step": 93120
|
| 217845 |
+
},
|
| 217846 |
+
{
|
| 217847 |
+
"epoch": 750.29,
|
| 217848 |
+
"learning_rate": 8.52423076923077e-06,
|
| 217849 |
+
"loss": 0.2555,
|
| 217850 |
+
"step": 93125
|
| 217851 |
+
},
|
| 217852 |
+
{
|
| 217853 |
+
"epoch": 750.33,
|
| 217854 |
+
"learning_rate": 8.524150641025641e-06,
|
| 217855 |
+
"loss": 0.3787,
|
| 217856 |
+
"step": 93130
|
| 217857 |
+
},
|
| 217858 |
+
{
|
| 217859 |
+
"epoch": 750.37,
|
| 217860 |
+
"learning_rate": 8.524070512820514e-06,
|
| 217861 |
+
"loss": 0.5679,
|
| 217862 |
+
"step": 93135
|
| 217863 |
+
},
|
| 217864 |
+
{
|
| 217865 |
+
"epoch": 750.41,
|
| 217866 |
+
"learning_rate": 8.523990384615385e-06,
|
| 217867 |
+
"loss": 1.1,
|
| 217868 |
+
"step": 93140
|
| 217869 |
+
},
|
| 217870 |
+
{
|
| 217871 |
+
"epoch": 750.45,
|
| 217872 |
+
"learning_rate": 8.523910256410257e-06,
|
| 217873 |
+
"loss": 0.3734,
|
| 217874 |
+
"step": 93145
|
| 217875 |
+
},
|
| 217876 |
+
{
|
| 217877 |
+
"epoch": 750.49,
|
| 217878 |
+
"learning_rate": 8.52383012820513e-06,
|
| 217879 |
+
"loss": 0.278,
|
| 217880 |
+
"step": 93150
|
| 217881 |
+
},
|
| 217882 |
+
{
|
| 217883 |
+
"epoch": 750.53,
|
| 217884 |
+
"learning_rate": 8.523750000000001e-06,
|
| 217885 |
+
"loss": 0.4092,
|
| 217886 |
+
"step": 93155
|
| 217887 |
+
},
|
| 217888 |
+
{
|
| 217889 |
+
"epoch": 750.57,
|
| 217890 |
+
"learning_rate": 8.523669871794873e-06,
|
| 217891 |
+
"loss": 0.6068,
|
| 217892 |
+
"step": 93160
|
| 217893 |
+
},
|
| 217894 |
+
{
|
| 217895 |
+
"epoch": 750.61,
|
| 217896 |
+
"learning_rate": 8.523589743589744e-06,
|
| 217897 |
+
"loss": 1.1341,
|
| 217898 |
+
"step": 93165
|
| 217899 |
+
},
|
| 217900 |
+
{
|
| 217901 |
+
"epoch": 750.65,
|
| 217902 |
+
"learning_rate": 8.523509615384617e-06,
|
| 217903 |
+
"loss": 0.3188,
|
| 217904 |
+
"step": 93170
|
| 217905 |
+
},
|
| 217906 |
+
{
|
| 217907 |
+
"epoch": 750.69,
|
| 217908 |
+
"learning_rate": 8.523429487179488e-06,
|
| 217909 |
+
"loss": 0.2805,
|
| 217910 |
+
"step": 93175
|
| 217911 |
+
},
|
| 217912 |
+
{
|
| 217913 |
+
"epoch": 750.73,
|
| 217914 |
+
"learning_rate": 8.52334935897436e-06,
|
| 217915 |
+
"loss": 0.3427,
|
| 217916 |
+
"step": 93180
|
| 217917 |
+
},
|
| 217918 |
+
{
|
| 217919 |
+
"epoch": 750.77,
|
| 217920 |
+
"learning_rate": 8.523269230769231e-06,
|
| 217921 |
+
"loss": 0.5412,
|
| 217922 |
+
"step": 93185
|
| 217923 |
+
},
|
| 217924 |
+
{
|
| 217925 |
+
"epoch": 750.81,
|
| 217926 |
+
"learning_rate": 8.523189102564104e-06,
|
| 217927 |
+
"loss": 1.1955,
|
| 217928 |
+
"step": 93190
|
| 217929 |
+
},
|
| 217930 |
+
{
|
| 217931 |
+
"epoch": 750.85,
|
| 217932 |
+
"learning_rate": 8.523108974358974e-06,
|
| 217933 |
+
"loss": 0.2784,
|
| 217934 |
+
"step": 93195
|
| 217935 |
+
},
|
| 217936 |
+
{
|
| 217937 |
+
"epoch": 750.89,
|
| 217938 |
+
"learning_rate": 8.523028846153847e-06,
|
| 217939 |
+
"loss": 0.295,
|
| 217940 |
+
"step": 93200
|
| 217941 |
+
},
|
| 217942 |
+
{
|
| 217943 |
+
"epoch": 750.93,
|
| 217944 |
+
"learning_rate": 8.52294871794872e-06,
|
| 217945 |
+
"loss": 0.333,
|
| 217946 |
+
"step": 93205
|
| 217947 |
+
},
|
| 217948 |
+
{
|
| 217949 |
+
"epoch": 750.97,
|
| 217950 |
+
"learning_rate": 8.52286858974359e-06,
|
| 217951 |
+
"loss": 0.5604,
|
| 217952 |
+
"step": 93210
|
| 217953 |
+
},
|
| 217954 |
+
{
|
| 217955 |
+
"epoch": 751.0,
|
| 217956 |
+
"eval_loss": 0.4374641180038452,
|
| 217957 |
+
"eval_runtime": 38.7976,
|
| 217958 |
+
"eval_samples_per_second": 21.599,
|
| 217959 |
+
"eval_steps_per_second": 0.696,
|
| 217960 |
+
"eval_wer": 0.19383097762073026,
|
| 217961 |
+
"step": 93213
|
| 217962 |
+
},
|
| 217963 |
+
{
|
| 217964 |
+
"epoch": 751.02,
|
| 217965 |
+
"learning_rate": 8.522788461538463e-06,
|
| 217966 |
+
"loss": 0.4429,
|
| 217967 |
+
"step": 93215
|
| 217968 |
+
},
|
| 217969 |
+
{
|
| 217970 |
+
"epoch": 751.06,
|
| 217971 |
+
"learning_rate": 8.522708333333334e-06,
|
| 217972 |
+
"loss": 0.277,
|
| 217973 |
+
"step": 93220
|
| 217974 |
+
},
|
| 217975 |
+
{
|
| 217976 |
+
"epoch": 751.1,
|
| 217977 |
+
"learning_rate": 8.522628205128205e-06,
|
| 217978 |
+
"loss": 0.2919,
|
| 217979 |
+
"step": 93225
|
| 217980 |
+
},
|
| 217981 |
+
{
|
| 217982 |
+
"epoch": 751.14,
|
| 217983 |
+
"learning_rate": 8.522548076923077e-06,
|
| 217984 |
+
"loss": 0.4176,
|
| 217985 |
+
"step": 93230
|
| 217986 |
+
},
|
| 217987 |
+
{
|
| 217988 |
+
"epoch": 751.18,
|
| 217989 |
+
"learning_rate": 8.52246794871795e-06,
|
| 217990 |
+
"loss": 0.6155,
|
| 217991 |
+
"step": 93235
|
| 217992 |
+
},
|
| 217993 |
+
{
|
| 217994 |
+
"epoch": 751.22,
|
| 217995 |
+
"learning_rate": 8.522387820512821e-06,
|
| 217996 |
+
"loss": 0.9501,
|
| 217997 |
+
"step": 93240
|
| 217998 |
+
},
|
| 217999 |
+
{
|
| 218000 |
+
"epoch": 751.26,
|
| 218001 |
+
"learning_rate": 8.522307692307692e-06,
|
| 218002 |
+
"loss": 0.311,
|
| 218003 |
+
"step": 93245
|
| 218004 |
+
},
|
| 218005 |
+
{
|
| 218006 |
+
"epoch": 751.3,
|
| 218007 |
+
"learning_rate": 8.522227564102565e-06,
|
| 218008 |
+
"loss": 0.2841,
|
| 218009 |
+
"step": 93250
|
| 218010 |
+
},
|
| 218011 |
+
{
|
| 218012 |
+
"epoch": 751.34,
|
| 218013 |
+
"learning_rate": 8.522147435897437e-06,
|
| 218014 |
+
"loss": 0.356,
|
| 218015 |
+
"step": 93255
|
| 218016 |
+
},
|
| 218017 |
+
{
|
| 218018 |
+
"epoch": 751.38,
|
| 218019 |
+
"learning_rate": 8.522067307692308e-06,
|
| 218020 |
+
"loss": 0.6652,
|
| 218021 |
+
"step": 93260
|
| 218022 |
+
},
|
| 218023 |
+
{
|
| 218024 |
+
"epoch": 751.42,
|
| 218025 |
+
"learning_rate": 8.52198717948718e-06,
|
| 218026 |
+
"loss": 1.0212,
|
| 218027 |
+
"step": 93265
|
| 218028 |
+
},
|
| 218029 |
+
{
|
| 218030 |
+
"epoch": 751.46,
|
| 218031 |
+
"learning_rate": 8.521907051282053e-06,
|
| 218032 |
+
"loss": 0.2716,
|
| 218033 |
+
"step": 93270
|
| 218034 |
+
},
|
| 218035 |
+
{
|
| 218036 |
+
"epoch": 751.5,
|
| 218037 |
+
"learning_rate": 8.521826923076924e-06,
|
| 218038 |
+
"loss": 0.2725,
|
| 218039 |
+
"step": 93275
|
| 218040 |
+
},
|
| 218041 |
+
{
|
| 218042 |
+
"epoch": 751.54,
|
| 218043 |
+
"learning_rate": 8.521746794871795e-06,
|
| 218044 |
+
"loss": 0.3049,
|
| 218045 |
+
"step": 93280
|
| 218046 |
+
},
|
| 218047 |
+
{
|
| 218048 |
+
"epoch": 751.58,
|
| 218049 |
+
"learning_rate": 8.521666666666667e-06,
|
| 218050 |
+
"loss": 0.6016,
|
| 218051 |
+
"step": 93285
|
| 218052 |
+
},
|
| 218053 |
+
{
|
| 218054 |
+
"epoch": 751.62,
|
| 218055 |
+
"learning_rate": 8.52158653846154e-06,
|
| 218056 |
+
"loss": 0.9861,
|
| 218057 |
+
"step": 93290
|
| 218058 |
+
},
|
| 218059 |
+
{
|
| 218060 |
+
"epoch": 751.66,
|
| 218061 |
+
"learning_rate": 8.521506410256411e-06,
|
| 218062 |
+
"loss": 0.3237,
|
| 218063 |
+
"step": 93295
|
| 218064 |
+
},
|
| 218065 |
+
{
|
| 218066 |
+
"epoch": 751.7,
|
| 218067 |
+
"learning_rate": 8.521426282051282e-06,
|
| 218068 |
+
"loss": 0.3375,
|
| 218069 |
+
"step": 93300
|
| 218070 |
+
},
|
| 218071 |
+
{
|
| 218072 |
+
"epoch": 751.74,
|
| 218073 |
+
"learning_rate": 8.521346153846155e-06,
|
| 218074 |
+
"loss": 0.3626,
|
| 218075 |
+
"step": 93305
|
| 218076 |
+
},
|
| 218077 |
+
{
|
| 218078 |
+
"epoch": 751.78,
|
| 218079 |
+
"learning_rate": 8.521266025641027e-06,
|
| 218080 |
+
"loss": 0.5895,
|
| 218081 |
+
"step": 93310
|
| 218082 |
+
},
|
| 218083 |
+
{
|
| 218084 |
+
"epoch": 751.82,
|
| 218085 |
+
"learning_rate": 8.521185897435898e-06,
|
| 218086 |
+
"loss": 1.1215,
|
| 218087 |
+
"step": 93315
|
| 218088 |
+
},
|
| 218089 |
+
{
|
| 218090 |
+
"epoch": 751.86,
|
| 218091 |
+
"learning_rate": 8.52110576923077e-06,
|
| 218092 |
+
"loss": 0.3631,
|
| 218093 |
+
"step": 93320
|
| 218094 |
+
},
|
| 218095 |
+
{
|
| 218096 |
+
"epoch": 751.9,
|
| 218097 |
+
"learning_rate": 8.521025641025643e-06,
|
| 218098 |
+
"loss": 0.3603,
|
| 218099 |
+
"step": 93325
|
| 218100 |
+
},
|
| 218101 |
+
{
|
| 218102 |
+
"epoch": 751.94,
|
| 218103 |
+
"learning_rate": 8.520945512820514e-06,
|
| 218104 |
+
"loss": 0.3875,
|
| 218105 |
+
"step": 93330
|
| 218106 |
+
},
|
| 218107 |
+
{
|
| 218108 |
+
"epoch": 751.98,
|
| 218109 |
+
"learning_rate": 8.520865384615385e-06,
|
| 218110 |
+
"loss": 0.6634,
|
| 218111 |
+
"step": 93335
|
| 218112 |
+
},
|
| 218113 |
+
{
|
| 218114 |
+
"epoch": 752.0,
|
| 218115 |
+
"eval_loss": 0.3540174067020416,
|
| 218116 |
+
"eval_runtime": 38.9251,
|
| 218117 |
+
"eval_samples_per_second": 21.529,
|
| 218118 |
+
"eval_steps_per_second": 0.694,
|
| 218119 |
+
"eval_wer": 0.1765940508069769,
|
| 218120 |
+
"step": 93337
|
| 218121 |
}
|
| 218122 |
],
|
| 218123 |
"max_steps": 620000,
|
| 218124 |
"num_train_epochs": 5000,
|
| 218125 |
+
"total_flos": 2.6266383085344024e+20,
|
| 218126 |
"trial_name": null,
|
| 218127 |
"trial_params": null
|
| 218128 |
}
|
model-bin/finetune/base/{checkpoint-92716 → checkpoint-93337}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629955488.5549626/events.out.tfevents.1629955488.8e89bd551565.924.71
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:04e14dba01eea8fd92a2bfea026c3e2728dad90697dacb6cf8bdf9a5fe28f683
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629955916.7474144/events.out.tfevents.1629955916.8e89bd551565.924.73
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d706d67a1d79166c64761120ed80a2c35cbe807e41eef7b904f3e9efc6cba3d2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629956348.062922/events.out.tfevents.1629956348.8e89bd551565.924.75
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3024b987766fb9e6494299323b13d520005da5063893b63a2917b3333ee6ea1d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629956783.883363/events.out.tfevents.1629956785.8e89bd551565.924.77
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e206c0ce65ed3493f697b631c31b8b694cd65bfc382ec2bbf1f5d6407deba5ab
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629957212.363884/events.out.tfevents.1629957212.8e89bd551565.924.79
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b51e3fe8938a3bb96acd5f946b91fbdadf4befda7ee05f800ea91b0dffdfce88
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629955488.8e89bd551565.924.70
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68ce77205c787ced307650105db2b9356bb578042a5e231de6f9c0511bfacd27
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629955916.8e89bd551565.924.72
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8eee6681c166981704d4093afeea49d356b94dc6dc4d6fa1e2e24f4258f99151
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629956348.8e89bd551565.924.74
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0321e3ba96cb305ffa7b8c1d705148bf6f8dadf3fd53cd47ba5b418fa1c9a18
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629956783.8e89bd551565.924.76
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:15eb9c768c9c0f966ba6276050e09dd19360ca1636680392bcfca588a0873b13
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629957212.8e89bd551565.924.78
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a3c005d93582377cf332fbe534b35049137ee606c4bee90059fbf9e30c25e186
|
| 3 |
+
size 8622
|