"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629686760.2288544/events.out.tfevents.1629686760.74272264b15c.932.23 +3 -0
- model-bin/finetune/base/log/1629687403.1416256/events.out.tfevents.1629687403.74272264b15c.932.25 +3 -0
- model-bin/finetune/base/log/1629688047.4250455/events.out.tfevents.1629688047.74272264b15c.932.27 +3 -0
- model-bin/finetune/base/log/1629688768.5831227/events.out.tfevents.1629688768.74272264b15c.932.29 +3 -0
- model-bin/finetune/base/log/1629689425.810986/events.out.tfevents.1629689425.74272264b15c.932.31 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629686760.74272264b15c.932.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629687403.74272264b15c.932.24 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629688047.74272264b15c.932.26 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629688768.74272264b15c.932.28 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629689425.74272264b15c.932.30 +3 -0
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/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 sha256:4d2ccd43def7c18d44a7ebb1f8ced97b5c0dd9d8e6862fc66e006bbbe63fbd21
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/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 sha256:98f91a1a9c9c182b4643a15a8d4e4030c3692e5556ae3bebaec0a42d1a632acf
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/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:dcb6d3ae9a08223517077daaf8239237d065b9f5fb709ad8f97b7205776b5c62
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/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 sha256:7fd3904a4666a6ab0e43816ee91af91d1805902f406df4cb1d3d0dee455886b8
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/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:c7a8e3a33eafdfdc34ec209f2fef19a4c88166870f16793b585cada922ec4392
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2028974903051145,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-39323",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -149712,11 +149712,806 @@
|
|
| 149712 |
"eval_steps_per_second": 0.688,
|
| 149713 |
"eval_wer": 0.20480646533390046,
|
| 149714 |
"step": 39572
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 149715 |
}
|
| 149716 |
],
|
| 149717 |
-
"max_steps":
|
| 149718 |
"num_train_epochs": 5000,
|
| 149719 |
-
"total_flos": 1.
|
| 149720 |
"trial_name": null,
|
| 149721 |
"trial_params": null
|
| 149722 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2028974903051145,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-39323",
|
| 4 |
+
"epoch": 321.0,
|
| 5 |
+
"global_step": 40195,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 149712 |
"eval_steps_per_second": 0.688,
|
| 149713 |
"eval_wer": 0.20480646533390046,
|
| 149714 |
"step": 39572
|
| 149715 |
+
},
|
| 149716 |
+
{
|
| 149717 |
+
"epoch": 316.02,
|
| 149718 |
+
"learning_rate": 9.382051282051284e-06,
|
| 149719 |
+
"loss": 0.4792,
|
| 149720 |
+
"step": 39575
|
| 149721 |
+
},
|
| 149722 |
+
{
|
| 149723 |
+
"epoch": 316.06,
|
| 149724 |
+
"learning_rate": 9.381971153846154e-06,
|
| 149725 |
+
"loss": 0.3643,
|
| 149726 |
+
"step": 39580
|
| 149727 |
+
},
|
| 149728 |
+
{
|
| 149729 |
+
"epoch": 316.1,
|
| 149730 |
+
"learning_rate": 9.381891025641027e-06,
|
| 149731 |
+
"loss": 0.3956,
|
| 149732 |
+
"step": 39585
|
| 149733 |
+
},
|
| 149734 |
+
{
|
| 149735 |
+
"epoch": 316.14,
|
| 149736 |
+
"learning_rate": 9.381810897435898e-06,
|
| 149737 |
+
"loss": 0.504,
|
| 149738 |
+
"step": 39590
|
| 149739 |
+
},
|
| 149740 |
+
{
|
| 149741 |
+
"epoch": 316.18,
|
| 149742 |
+
"learning_rate": 9.38173076923077e-06,
|
| 149743 |
+
"loss": 0.8401,
|
| 149744 |
+
"step": 39595
|
| 149745 |
+
},
|
| 149746 |
+
{
|
| 149747 |
+
"epoch": 316.22,
|
| 149748 |
+
"learning_rate": 9.38165064102564e-06,
|
| 149749 |
+
"loss": 1.0365,
|
| 149750 |
+
"step": 39600
|
| 149751 |
+
},
|
| 149752 |
+
{
|
| 149753 |
+
"epoch": 316.26,
|
| 149754 |
+
"learning_rate": 9.381570512820514e-06,
|
| 149755 |
+
"loss": 0.399,
|
| 149756 |
+
"step": 39605
|
| 149757 |
+
},
|
| 149758 |
+
{
|
| 149759 |
+
"epoch": 316.3,
|
| 149760 |
+
"learning_rate": 9.381490384615385e-06,
|
| 149761 |
+
"loss": 0.3633,
|
| 149762 |
+
"step": 39610
|
| 149763 |
+
},
|
| 149764 |
+
{
|
| 149765 |
+
"epoch": 316.34,
|
| 149766 |
+
"learning_rate": 9.381410256410257e-06,
|
| 149767 |
+
"loss": 0.3991,
|
| 149768 |
+
"step": 39615
|
| 149769 |
+
},
|
| 149770 |
+
{
|
| 149771 |
+
"epoch": 316.38,
|
| 149772 |
+
"learning_rate": 9.38133012820513e-06,
|
| 149773 |
+
"loss": 0.8245,
|
| 149774 |
+
"step": 39620
|
| 149775 |
+
},
|
| 149776 |
+
{
|
| 149777 |
+
"epoch": 316.42,
|
| 149778 |
+
"learning_rate": 9.381250000000001e-06,
|
| 149779 |
+
"loss": 0.9271,
|
| 149780 |
+
"step": 39625
|
| 149781 |
+
},
|
| 149782 |
+
{
|
| 149783 |
+
"epoch": 316.46,
|
| 149784 |
+
"learning_rate": 9.381169871794872e-06,
|
| 149785 |
+
"loss": 0.4727,
|
| 149786 |
+
"step": 39630
|
| 149787 |
+
},
|
| 149788 |
+
{
|
| 149789 |
+
"epoch": 316.5,
|
| 149790 |
+
"learning_rate": 9.381089743589744e-06,
|
| 149791 |
+
"loss": 0.398,
|
| 149792 |
+
"step": 39635
|
| 149793 |
+
},
|
| 149794 |
+
{
|
| 149795 |
+
"epoch": 316.54,
|
| 149796 |
+
"learning_rate": 9.381009615384617e-06,
|
| 149797 |
+
"loss": 0.4117,
|
| 149798 |
+
"step": 39640
|
| 149799 |
+
},
|
| 149800 |
+
{
|
| 149801 |
+
"epoch": 316.58,
|
| 149802 |
+
"learning_rate": 9.380929487179488e-06,
|
| 149803 |
+
"loss": 0.8217,
|
| 149804 |
+
"step": 39645
|
| 149805 |
+
},
|
| 149806 |
+
{
|
| 149807 |
+
"epoch": 316.62,
|
| 149808 |
+
"learning_rate": 9.38084935897436e-06,
|
| 149809 |
+
"loss": 0.9855,
|
| 149810 |
+
"step": 39650
|
| 149811 |
+
},
|
| 149812 |
+
{
|
| 149813 |
+
"epoch": 316.66,
|
| 149814 |
+
"learning_rate": 9.380769230769231e-06,
|
| 149815 |
+
"loss": 0.3763,
|
| 149816 |
+
"step": 39655
|
| 149817 |
+
},
|
| 149818 |
+
{
|
| 149819 |
+
"epoch": 316.7,
|
| 149820 |
+
"learning_rate": 9.380689102564104e-06,
|
| 149821 |
+
"loss": 0.4198,
|
| 149822 |
+
"step": 39660
|
| 149823 |
+
},
|
| 149824 |
+
{
|
| 149825 |
+
"epoch": 316.74,
|
| 149826 |
+
"learning_rate": 9.380608974358975e-06,
|
| 149827 |
+
"loss": 0.427,
|
| 149828 |
+
"step": 39665
|
| 149829 |
+
},
|
| 149830 |
+
{
|
| 149831 |
+
"epoch": 316.78,
|
| 149832 |
+
"learning_rate": 9.380528846153847e-06,
|
| 149833 |
+
"loss": 0.7916,
|
| 149834 |
+
"step": 39670
|
| 149835 |
+
},
|
| 149836 |
+
{
|
| 149837 |
+
"epoch": 316.82,
|
| 149838 |
+
"learning_rate": 9.38044871794872e-06,
|
| 149839 |
+
"loss": 0.945,
|
| 149840 |
+
"step": 39675
|
| 149841 |
+
},
|
| 149842 |
+
{
|
| 149843 |
+
"epoch": 316.86,
|
| 149844 |
+
"learning_rate": 9.380368589743591e-06,
|
| 149845 |
+
"loss": 0.356,
|
| 149846 |
+
"step": 39680
|
| 149847 |
+
},
|
| 149848 |
+
{
|
| 149849 |
+
"epoch": 316.9,
|
| 149850 |
+
"learning_rate": 9.380288461538462e-06,
|
| 149851 |
+
"loss": 0.4322,
|
| 149852 |
+
"step": 39685
|
| 149853 |
+
},
|
| 149854 |
+
{
|
| 149855 |
+
"epoch": 316.94,
|
| 149856 |
+
"learning_rate": 9.380208333333334e-06,
|
| 149857 |
+
"loss": 0.4505,
|
| 149858 |
+
"step": 39690
|
| 149859 |
+
},
|
| 149860 |
+
{
|
| 149861 |
+
"epoch": 316.98,
|
| 149862 |
+
"learning_rate": 9.380128205128207e-06,
|
| 149863 |
+
"loss": 0.9645,
|
| 149864 |
+
"step": 39695
|
| 149865 |
+
},
|
| 149866 |
+
{
|
| 149867 |
+
"epoch": 317.0,
|
| 149868 |
+
"eval_loss": 0.4265443980693817,
|
| 149869 |
+
"eval_runtime": 40.7698,
|
| 149870 |
+
"eval_samples_per_second": 20.603,
|
| 149871 |
+
"eval_steps_per_second": 0.662,
|
| 149872 |
+
"eval_wer": 0.211847532356706,
|
| 149873 |
+
"step": 39697
|
| 149874 |
+
},
|
| 149875 |
+
{
|
| 149876 |
+
"epoch": 320.02,
|
| 149877 |
+
"learning_rate": 9.380048076923076e-06,
|
| 149878 |
+
"loss": 0.4334,
|
| 149879 |
+
"step": 39700
|
| 149880 |
+
},
|
| 149881 |
+
{
|
| 149882 |
+
"epoch": 320.06,
|
| 149883 |
+
"learning_rate": 9.37996794871795e-06,
|
| 149884 |
+
"loss": 0.3718,
|
| 149885 |
+
"step": 39705
|
| 149886 |
+
},
|
| 149887 |
+
{
|
| 149888 |
+
"epoch": 320.1,
|
| 149889 |
+
"learning_rate": 9.379887820512821e-06,
|
| 149890 |
+
"loss": 0.3352,
|
| 149891 |
+
"step": 39710
|
| 149892 |
+
},
|
| 149893 |
+
{
|
| 149894 |
+
"epoch": 320.15,
|
| 149895 |
+
"learning_rate": 9.379807692307692e-06,
|
| 149896 |
+
"loss": 0.5154,
|
| 149897 |
+
"step": 39715
|
| 149898 |
+
},
|
| 149899 |
+
{
|
| 149900 |
+
"epoch": 320.19,
|
| 149901 |
+
"learning_rate": 9.379727564102565e-06,
|
| 149902 |
+
"loss": 0.8178,
|
| 149903 |
+
"step": 39720
|
| 149904 |
+
},
|
| 149905 |
+
{
|
| 149906 |
+
"epoch": 320.23,
|
| 149907 |
+
"learning_rate": 9.379647435897437e-06,
|
| 149908 |
+
"loss": 0.9318,
|
| 149909 |
+
"step": 39725
|
| 149910 |
+
},
|
| 149911 |
+
{
|
| 149912 |
+
"epoch": 320.27,
|
| 149913 |
+
"learning_rate": 9.379567307692308e-06,
|
| 149914 |
+
"loss": 0.366,
|
| 149915 |
+
"step": 39730
|
| 149916 |
+
},
|
| 149917 |
+
{
|
| 149918 |
+
"epoch": 320.31,
|
| 149919 |
+
"learning_rate": 9.37948717948718e-06,
|
| 149920 |
+
"loss": 0.3614,
|
| 149921 |
+
"step": 39735
|
| 149922 |
+
},
|
| 149923 |
+
{
|
| 149924 |
+
"epoch": 320.35,
|
| 149925 |
+
"learning_rate": 9.379407051282052e-06,
|
| 149926 |
+
"loss": 0.4486,
|
| 149927 |
+
"step": 39740
|
| 149928 |
+
},
|
| 149929 |
+
{
|
| 149930 |
+
"epoch": 320.39,
|
| 149931 |
+
"learning_rate": 9.379326923076924e-06,
|
| 149932 |
+
"loss": 0.8598,
|
| 149933 |
+
"step": 39745
|
| 149934 |
+
},
|
| 149935 |
+
{
|
| 149936 |
+
"epoch": 320.43,
|
| 149937 |
+
"learning_rate": 9.379246794871795e-06,
|
| 149938 |
+
"loss": 1.0338,
|
| 149939 |
+
"step": 39750
|
| 149940 |
+
},
|
| 149941 |
+
{
|
| 149942 |
+
"epoch": 320.47,
|
| 149943 |
+
"learning_rate": 9.379166666666667e-06,
|
| 149944 |
+
"loss": 0.3382,
|
| 149945 |
+
"step": 39755
|
| 149946 |
+
},
|
| 149947 |
+
{
|
| 149948 |
+
"epoch": 320.51,
|
| 149949 |
+
"learning_rate": 9.37908653846154e-06,
|
| 149950 |
+
"loss": 0.34,
|
| 149951 |
+
"step": 39760
|
| 149952 |
+
},
|
| 149953 |
+
{
|
| 149954 |
+
"epoch": 320.55,
|
| 149955 |
+
"learning_rate": 9.379006410256411e-06,
|
| 149956 |
+
"loss": 0.464,
|
| 149957 |
+
"step": 39765
|
| 149958 |
+
},
|
| 149959 |
+
{
|
| 149960 |
+
"epoch": 320.59,
|
| 149961 |
+
"learning_rate": 9.378926282051282e-06,
|
| 149962 |
+
"loss": 0.9151,
|
| 149963 |
+
"step": 39770
|
| 149964 |
+
},
|
| 149965 |
+
{
|
| 149966 |
+
"epoch": 320.63,
|
| 149967 |
+
"learning_rate": 9.378846153846155e-06,
|
| 149968 |
+
"loss": 0.8859,
|
| 149969 |
+
"step": 39775
|
| 149970 |
+
},
|
| 149971 |
+
{
|
| 149972 |
+
"epoch": 320.67,
|
| 149973 |
+
"learning_rate": 9.378766025641027e-06,
|
| 149974 |
+
"loss": 0.4492,
|
| 149975 |
+
"step": 39780
|
| 149976 |
+
},
|
| 149977 |
+
{
|
| 149978 |
+
"epoch": 320.71,
|
| 149979 |
+
"learning_rate": 9.378685897435898e-06,
|
| 149980 |
+
"loss": 0.4145,
|
| 149981 |
+
"step": 39785
|
| 149982 |
+
},
|
| 149983 |
+
{
|
| 149984 |
+
"epoch": 320.75,
|
| 149985 |
+
"learning_rate": 9.37860576923077e-06,
|
| 149986 |
+
"loss": 0.4233,
|
| 149987 |
+
"step": 39790
|
| 149988 |
+
},
|
| 149989 |
+
{
|
| 149990 |
+
"epoch": 320.79,
|
| 149991 |
+
"learning_rate": 9.378525641025642e-06,
|
| 149992 |
+
"loss": 0.9677,
|
| 149993 |
+
"step": 39795
|
| 149994 |
+
},
|
| 149995 |
+
{
|
| 149996 |
+
"epoch": 320.83,
|
| 149997 |
+
"learning_rate": 9.378445512820514e-06,
|
| 149998 |
+
"loss": 0.9935,
|
| 149999 |
+
"step": 39800
|
| 150000 |
+
},
|
| 150001 |
+
{
|
| 150002 |
+
"epoch": 320.87,
|
| 150003 |
+
"learning_rate": 9.378365384615385e-06,
|
| 150004 |
+
"loss": 0.3495,
|
| 150005 |
+
"step": 39805
|
| 150006 |
+
},
|
| 150007 |
+
{
|
| 150008 |
+
"epoch": 320.91,
|
| 150009 |
+
"learning_rate": 9.378285256410257e-06,
|
| 150010 |
+
"loss": 0.4779,
|
| 150011 |
+
"step": 39810
|
| 150012 |
+
},
|
| 150013 |
+
{
|
| 150014 |
+
"epoch": 320.95,
|
| 150015 |
+
"learning_rate": 9.37820512820513e-06,
|
| 150016 |
+
"loss": 0.4441,
|
| 150017 |
+
"step": 39815
|
| 150018 |
+
},
|
| 150019 |
+
{
|
| 150020 |
+
"epoch": 320.99,
|
| 150021 |
+
"learning_rate": 9.378125000000001e-06,
|
| 150022 |
+
"loss": 0.9563,
|
| 150023 |
+
"step": 39820
|
| 150024 |
+
},
|
| 150025 |
+
{
|
| 150026 |
+
"epoch": 321.0,
|
| 150027 |
+
"eval_loss": 0.4913080930709839,
|
| 150028 |
+
"eval_runtime": 40.5625,
|
| 150029 |
+
"eval_samples_per_second": 20.709,
|
| 150030 |
+
"eval_steps_per_second": 0.666,
|
| 150031 |
+
"eval_wer": 0.21510102800425382,
|
| 150032 |
+
"step": 39821
|
| 150033 |
+
},
|
| 150034 |
+
{
|
| 150035 |
+
"epoch": 321.03,
|
| 150036 |
+
"learning_rate": 9.378044871794872e-06,
|
| 150037 |
+
"loss": 0.4117,
|
| 150038 |
+
"step": 39825
|
| 150039 |
+
},
|
| 150040 |
+
{
|
| 150041 |
+
"epoch": 321.07,
|
| 150042 |
+
"learning_rate": 9.377964743589745e-06,
|
| 150043 |
+
"loss": 0.4014,
|
| 150044 |
+
"step": 39830
|
| 150045 |
+
},
|
| 150046 |
+
{
|
| 150047 |
+
"epoch": 321.11,
|
| 150048 |
+
"learning_rate": 9.377884615384617e-06,
|
| 150049 |
+
"loss": 0.3825,
|
| 150050 |
+
"step": 39835
|
| 150051 |
+
},
|
| 150052 |
+
{
|
| 150053 |
+
"epoch": 321.15,
|
| 150054 |
+
"learning_rate": 9.377804487179488e-06,
|
| 150055 |
+
"loss": 0.4884,
|
| 150056 |
+
"step": 39840
|
| 150057 |
+
},
|
| 150058 |
+
{
|
| 150059 |
+
"epoch": 321.19,
|
| 150060 |
+
"learning_rate": 9.37772435897436e-06,
|
| 150061 |
+
"loss": 1.0141,
|
| 150062 |
+
"step": 39845
|
| 150063 |
+
},
|
| 150064 |
+
{
|
| 150065 |
+
"epoch": 321.23,
|
| 150066 |
+
"learning_rate": 9.377644230769232e-06,
|
| 150067 |
+
"loss": 0.8423,
|
| 150068 |
+
"step": 39850
|
| 150069 |
+
},
|
| 150070 |
+
{
|
| 150071 |
+
"epoch": 321.27,
|
| 150072 |
+
"learning_rate": 9.377564102564102e-06,
|
| 150073 |
+
"loss": 0.4353,
|
| 150074 |
+
"step": 39855
|
| 150075 |
+
},
|
| 150076 |
+
{
|
| 150077 |
+
"epoch": 321.31,
|
| 150078 |
+
"learning_rate": 9.377483974358975e-06,
|
| 150079 |
+
"loss": 0.4034,
|
| 150080 |
+
"step": 39860
|
| 150081 |
+
},
|
| 150082 |
+
{
|
| 150083 |
+
"epoch": 321.35,
|
| 150084 |
+
"learning_rate": 9.377403846153848e-06,
|
| 150085 |
+
"loss": 0.4576,
|
| 150086 |
+
"step": 39865
|
| 150087 |
+
},
|
| 150088 |
+
{
|
| 150089 |
+
"epoch": 321.39,
|
| 150090 |
+
"learning_rate": 9.377323717948718e-06,
|
| 150091 |
+
"loss": 1.0148,
|
| 150092 |
+
"step": 39870
|
| 150093 |
+
},
|
| 150094 |
+
{
|
| 150095 |
+
"epoch": 321.43,
|
| 150096 |
+
"learning_rate": 9.377243589743591e-06,
|
| 150097 |
+
"loss": 0.8105,
|
| 150098 |
+
"step": 39875
|
| 150099 |
+
},
|
| 150100 |
+
{
|
| 150101 |
+
"epoch": 321.47,
|
| 150102 |
+
"learning_rate": 9.377163461538462e-06,
|
| 150103 |
+
"loss": 0.3637,
|
| 150104 |
+
"step": 39880
|
| 150105 |
+
},
|
| 150106 |
+
{
|
| 150107 |
+
"epoch": 321.51,
|
| 150108 |
+
"learning_rate": 9.377083333333334e-06,
|
| 150109 |
+
"loss": 0.3656,
|
| 150110 |
+
"step": 39885
|
| 150111 |
+
},
|
| 150112 |
+
{
|
| 150113 |
+
"epoch": 321.55,
|
| 150114 |
+
"learning_rate": 9.377003205128205e-06,
|
| 150115 |
+
"loss": 0.4222,
|
| 150116 |
+
"step": 39890
|
| 150117 |
+
},
|
| 150118 |
+
{
|
| 150119 |
+
"epoch": 321.59,
|
| 150120 |
+
"learning_rate": 9.376923076923078e-06,
|
| 150121 |
+
"loss": 0.9701,
|
| 150122 |
+
"step": 39895
|
| 150123 |
+
},
|
| 150124 |
+
{
|
| 150125 |
+
"epoch": 321.63,
|
| 150126 |
+
"learning_rate": 9.37684294871795e-06,
|
| 150127 |
+
"loss": 0.7682,
|
| 150128 |
+
"step": 39900
|
| 150129 |
+
},
|
| 150130 |
+
{
|
| 150131 |
+
"epoch": 321.67,
|
| 150132 |
+
"learning_rate": 9.37676282051282e-06,
|
| 150133 |
+
"loss": 0.3619,
|
| 150134 |
+
"step": 39905
|
| 150135 |
+
},
|
| 150136 |
+
{
|
| 150137 |
+
"epoch": 321.71,
|
| 150138 |
+
"learning_rate": 9.376682692307692e-06,
|
| 150139 |
+
"loss": 0.3715,
|
| 150140 |
+
"step": 39910
|
| 150141 |
+
},
|
| 150142 |
+
{
|
| 150143 |
+
"epoch": 321.76,
|
| 150144 |
+
"learning_rate": 9.376602564102565e-06,
|
| 150145 |
+
"loss": 0.4064,
|
| 150146 |
+
"step": 39915
|
| 150147 |
+
},
|
| 150148 |
+
{
|
| 150149 |
+
"epoch": 321.8,
|
| 150150 |
+
"learning_rate": 9.376522435897437e-06,
|
| 150151 |
+
"loss": 1.1391,
|
| 150152 |
+
"step": 39920
|
| 150153 |
+
},
|
| 150154 |
+
{
|
| 150155 |
+
"epoch": 321.84,
|
| 150156 |
+
"learning_rate": 9.376442307692308e-06,
|
| 150157 |
+
"loss": 0.7452,
|
| 150158 |
+
"step": 39925
|
| 150159 |
+
},
|
| 150160 |
+
{
|
| 150161 |
+
"epoch": 321.88,
|
| 150162 |
+
"learning_rate": 9.376362179487181e-06,
|
| 150163 |
+
"loss": 0.3457,
|
| 150164 |
+
"step": 39930
|
| 150165 |
+
},
|
| 150166 |
+
{
|
| 150167 |
+
"epoch": 321.92,
|
| 150168 |
+
"learning_rate": 9.376282051282052e-06,
|
| 150169 |
+
"loss": 0.4014,
|
| 150170 |
+
"step": 39935
|
| 150171 |
+
},
|
| 150172 |
+
{
|
| 150173 |
+
"epoch": 321.96,
|
| 150174 |
+
"learning_rate": 9.376201923076924e-06,
|
| 150175 |
+
"loss": 0.5374,
|
| 150176 |
+
"step": 39940
|
| 150177 |
+
},
|
| 150178 |
+
{
|
| 150179 |
+
"epoch": 322.0,
|
| 150180 |
+
"learning_rate": 9.376121794871795e-06,
|
| 150181 |
+
"loss": 1.419,
|
| 150182 |
+
"step": 39945
|
| 150183 |
+
},
|
| 150184 |
+
{
|
| 150185 |
+
"epoch": 322.0,
|
| 150186 |
+
"eval_loss": 0.4495704472064972,
|
| 150187 |
+
"eval_runtime": 39.7624,
|
| 150188 |
+
"eval_samples_per_second": 21.1,
|
| 150189 |
+
"eval_steps_per_second": 0.679,
|
| 150190 |
+
"eval_wer": 0.20808786093534076,
|
| 150191 |
+
"step": 39945
|
| 150192 |
+
},
|
| 150193 |
+
{
|
| 150194 |
+
"epoch": 319.04,
|
| 150195 |
+
"learning_rate": 9.376041666666668e-06,
|
| 150196 |
+
"loss": 0.3993,
|
| 150197 |
+
"step": 39950
|
| 150198 |
+
},
|
| 150199 |
+
{
|
| 150200 |
+
"epoch": 319.08,
|
| 150201 |
+
"learning_rate": 9.37596153846154e-06,
|
| 150202 |
+
"loss": 0.4063,
|
| 150203 |
+
"step": 39955
|
| 150204 |
+
},
|
| 150205 |
+
{
|
| 150206 |
+
"epoch": 319.12,
|
| 150207 |
+
"learning_rate": 9.37588141025641e-06,
|
| 150208 |
+
"loss": 0.3972,
|
| 150209 |
+
"step": 39960
|
| 150210 |
+
},
|
| 150211 |
+
{
|
| 150212 |
+
"epoch": 319.16,
|
| 150213 |
+
"learning_rate": 9.375801282051284e-06,
|
| 150214 |
+
"loss": 0.5071,
|
| 150215 |
+
"step": 39965
|
| 150216 |
+
},
|
| 150217 |
+
{
|
| 150218 |
+
"epoch": 319.2,
|
| 150219 |
+
"learning_rate": 9.375721153846155e-06,
|
| 150220 |
+
"loss": 1.2853,
|
| 150221 |
+
"step": 39970
|
| 150222 |
+
},
|
| 150223 |
+
{
|
| 150224 |
+
"epoch": 319.24,
|
| 150225 |
+
"learning_rate": 9.375641025641027e-06,
|
| 150226 |
+
"loss": 0.4327,
|
| 150227 |
+
"step": 39975
|
| 150228 |
+
},
|
| 150229 |
+
{
|
| 150230 |
+
"epoch": 319.28,
|
| 150231 |
+
"learning_rate": 9.375560897435898e-06,
|
| 150232 |
+
"loss": 0.3403,
|
| 150233 |
+
"step": 39980
|
| 150234 |
+
},
|
| 150235 |
+
{
|
| 150236 |
+
"epoch": 319.32,
|
| 150237 |
+
"learning_rate": 9.375480769230771e-06,
|
| 150238 |
+
"loss": 0.3346,
|
| 150239 |
+
"step": 39985
|
| 150240 |
+
},
|
| 150241 |
+
{
|
| 150242 |
+
"epoch": 319.36,
|
| 150243 |
+
"learning_rate": 9.37540064102564e-06,
|
| 150244 |
+
"loss": 0.5967,
|
| 150245 |
+
"step": 39990
|
| 150246 |
+
},
|
| 150247 |
+
{
|
| 150248 |
+
"epoch": 319.4,
|
| 150249 |
+
"learning_rate": 9.375320512820514e-06,
|
| 150250 |
+
"loss": 1.4268,
|
| 150251 |
+
"step": 39995
|
| 150252 |
+
},
|
| 150253 |
+
{
|
| 150254 |
+
"epoch": 319.44,
|
| 150255 |
+
"learning_rate": 9.375240384615385e-06,
|
| 150256 |
+
"loss": 0.4163,
|
| 150257 |
+
"step": 40000
|
| 150258 |
+
},
|
| 150259 |
+
{
|
| 150260 |
+
"epoch": 319.48,
|
| 150261 |
+
"learning_rate": 9.375160256410256e-06,
|
| 150262 |
+
"loss": 0.3499,
|
| 150263 |
+
"step": 40005
|
| 150264 |
+
},
|
| 150265 |
+
{
|
| 150266 |
+
"epoch": 319.52,
|
| 150267 |
+
"learning_rate": 9.375080128205128e-06,
|
| 150268 |
+
"loss": 0.4167,
|
| 150269 |
+
"step": 40010
|
| 150270 |
+
},
|
| 150271 |
+
{
|
| 150272 |
+
"epoch": 319.56,
|
| 150273 |
+
"learning_rate": 9.375000000000001e-06,
|
| 150274 |
+
"loss": 0.5461,
|
| 150275 |
+
"step": 40015
|
| 150276 |
+
},
|
| 150277 |
+
{
|
| 150278 |
+
"epoch": 319.6,
|
| 150279 |
+
"learning_rate": 9.374919871794872e-06,
|
| 150280 |
+
"loss": 1.5324,
|
| 150281 |
+
"step": 40020
|
| 150282 |
+
},
|
| 150283 |
+
{
|
| 150284 |
+
"epoch": 319.64,
|
| 150285 |
+
"learning_rate": 9.374839743589744e-06,
|
| 150286 |
+
"loss": 0.4764,
|
| 150287 |
+
"step": 40025
|
| 150288 |
+
},
|
| 150289 |
+
{
|
| 150290 |
+
"epoch": 319.68,
|
| 150291 |
+
"learning_rate": 9.374759615384617e-06,
|
| 150292 |
+
"loss": 0.4048,
|
| 150293 |
+
"step": 40030
|
| 150294 |
+
},
|
| 150295 |
+
{
|
| 150296 |
+
"epoch": 319.72,
|
| 150297 |
+
"learning_rate": 9.374679487179488e-06,
|
| 150298 |
+
"loss": 0.3353,
|
| 150299 |
+
"step": 40035
|
| 150300 |
+
},
|
| 150301 |
+
{
|
| 150302 |
+
"epoch": 319.76,
|
| 150303 |
+
"learning_rate": 9.37459935897436e-06,
|
| 150304 |
+
"loss": 0.4721,
|
| 150305 |
+
"step": 40040
|
| 150306 |
+
},
|
| 150307 |
+
{
|
| 150308 |
+
"epoch": 319.8,
|
| 150309 |
+
"learning_rate": 9.37451923076923e-06,
|
| 150310 |
+
"loss": 1.371,
|
| 150311 |
+
"step": 40045
|
| 150312 |
+
},
|
| 150313 |
+
{
|
| 150314 |
+
"epoch": 319.84,
|
| 150315 |
+
"learning_rate": 9.374439102564104e-06,
|
| 150316 |
+
"loss": 0.4472,
|
| 150317 |
+
"step": 40050
|
| 150318 |
+
},
|
| 150319 |
+
{
|
| 150320 |
+
"epoch": 319.88,
|
| 150321 |
+
"learning_rate": 9.374358974358975e-06,
|
| 150322 |
+
"loss": 0.3666,
|
| 150323 |
+
"step": 40055
|
| 150324 |
+
},
|
| 150325 |
+
{
|
| 150326 |
+
"epoch": 319.92,
|
| 150327 |
+
"learning_rate": 9.374278846153846e-06,
|
| 150328 |
+
"loss": 0.347,
|
| 150329 |
+
"step": 40060
|
| 150330 |
+
},
|
| 150331 |
+
{
|
| 150332 |
+
"epoch": 319.96,
|
| 150333 |
+
"learning_rate": 9.374198717948718e-06,
|
| 150334 |
+
"loss": 0.4948,
|
| 150335 |
+
"step": 40065
|
| 150336 |
+
},
|
| 150337 |
+
{
|
| 150338 |
+
"epoch": 320.0,
|
| 150339 |
+
"learning_rate": 9.374118589743591e-06,
|
| 150340 |
+
"loss": 1.298,
|
| 150341 |
+
"step": 40070
|
| 150342 |
+
},
|
| 150343 |
+
{
|
| 150344 |
+
"epoch": 320.0,
|
| 150345 |
+
"eval_loss": 0.46632203459739685,
|
| 150346 |
+
"eval_runtime": 37.4205,
|
| 150347 |
+
"eval_samples_per_second": 22.448,
|
| 150348 |
+
"eval_steps_per_second": 0.722,
|
| 150349 |
+
"eval_wer": 0.211322417211249,
|
| 150350 |
+
"step": 40070
|
| 150351 |
+
},
|
| 150352 |
+
{
|
| 150353 |
+
"epoch": 320.04,
|
| 150354 |
+
"learning_rate": 9.374038461538462e-06,
|
| 150355 |
+
"loss": 0.4296,
|
| 150356 |
+
"step": 40075
|
| 150357 |
+
},
|
| 150358 |
+
{
|
| 150359 |
+
"epoch": 320.08,
|
| 150360 |
+
"learning_rate": 9.373958333333334e-06,
|
| 150361 |
+
"loss": 0.3608,
|
| 150362 |
+
"step": 40080
|
| 150363 |
+
},
|
| 150364 |
+
{
|
| 150365 |
+
"epoch": 320.12,
|
| 150366 |
+
"learning_rate": 9.373878205128207e-06,
|
| 150367 |
+
"loss": 0.3459,
|
| 150368 |
+
"step": 40085
|
| 150369 |
+
},
|
| 150370 |
+
{
|
| 150371 |
+
"epoch": 320.16,
|
| 150372 |
+
"learning_rate": 9.373798076923078e-06,
|
| 150373 |
+
"loss": 0.5596,
|
| 150374 |
+
"step": 40090
|
| 150375 |
+
},
|
| 150376 |
+
{
|
| 150377 |
+
"epoch": 320.2,
|
| 150378 |
+
"learning_rate": 9.37371794871795e-06,
|
| 150379 |
+
"loss": 1.3436,
|
| 150380 |
+
"step": 40095
|
| 150381 |
+
},
|
| 150382 |
+
{
|
| 150383 |
+
"epoch": 320.24,
|
| 150384 |
+
"learning_rate": 9.37363782051282e-06,
|
| 150385 |
+
"loss": 0.4253,
|
| 150386 |
+
"step": 40100
|
| 150387 |
+
},
|
| 150388 |
+
{
|
| 150389 |
+
"epoch": 320.28,
|
| 150390 |
+
"learning_rate": 9.373557692307694e-06,
|
| 150391 |
+
"loss": 0.3333,
|
| 150392 |
+
"step": 40105
|
| 150393 |
+
},
|
| 150394 |
+
{
|
| 150395 |
+
"epoch": 320.32,
|
| 150396 |
+
"learning_rate": 9.373477564102565e-06,
|
| 150397 |
+
"loss": 0.3771,
|
| 150398 |
+
"step": 40110
|
| 150399 |
+
},
|
| 150400 |
+
{
|
| 150401 |
+
"epoch": 320.36,
|
| 150402 |
+
"learning_rate": 9.373397435897437e-06,
|
| 150403 |
+
"loss": 0.5164,
|
| 150404 |
+
"step": 40115
|
| 150405 |
+
},
|
| 150406 |
+
{
|
| 150407 |
+
"epoch": 320.4,
|
| 150408 |
+
"learning_rate": 9.37331730769231e-06,
|
| 150409 |
+
"loss": 1.3631,
|
| 150410 |
+
"step": 40120
|
| 150411 |
+
},
|
| 150412 |
+
{
|
| 150413 |
+
"epoch": 320.44,
|
| 150414 |
+
"learning_rate": 9.373237179487181e-06,
|
| 150415 |
+
"loss": 0.4556,
|
| 150416 |
+
"step": 40125
|
| 150417 |
+
},
|
| 150418 |
+
{
|
| 150419 |
+
"epoch": 320.48,
|
| 150420 |
+
"learning_rate": 9.373157051282052e-06,
|
| 150421 |
+
"loss": 0.368,
|
| 150422 |
+
"step": 40130
|
| 150423 |
+
},
|
| 150424 |
+
{
|
| 150425 |
+
"epoch": 320.52,
|
| 150426 |
+
"learning_rate": 9.373076923076924e-06,
|
| 150427 |
+
"loss": 0.4187,
|
| 150428 |
+
"step": 40135
|
| 150429 |
+
},
|
| 150430 |
+
{
|
| 150431 |
+
"epoch": 320.56,
|
| 150432 |
+
"learning_rate": 9.372996794871797e-06,
|
| 150433 |
+
"loss": 0.5029,
|
| 150434 |
+
"step": 40140
|
| 150435 |
+
},
|
| 150436 |
+
{
|
| 150437 |
+
"epoch": 320.6,
|
| 150438 |
+
"learning_rate": 9.372916666666666e-06,
|
| 150439 |
+
"loss": 1.3927,
|
| 150440 |
+
"step": 40145
|
| 150441 |
+
},
|
| 150442 |
+
{
|
| 150443 |
+
"epoch": 320.64,
|
| 150444 |
+
"learning_rate": 9.37283653846154e-06,
|
| 150445 |
+
"loss": 0.3718,
|
| 150446 |
+
"step": 40150
|
| 150447 |
+
},
|
| 150448 |
+
{
|
| 150449 |
+
"epoch": 320.68,
|
| 150450 |
+
"learning_rate": 9.37275641025641e-06,
|
| 150451 |
+
"loss": 0.3839,
|
| 150452 |
+
"step": 40155
|
| 150453 |
+
},
|
| 150454 |
+
{
|
| 150455 |
+
"epoch": 320.72,
|
| 150456 |
+
"learning_rate": 9.372676282051282e-06,
|
| 150457 |
+
"loss": 0.349,
|
| 150458 |
+
"step": 40160
|
| 150459 |
+
},
|
| 150460 |
+
{
|
| 150461 |
+
"epoch": 320.76,
|
| 150462 |
+
"learning_rate": 9.372596153846153e-06,
|
| 150463 |
+
"loss": 0.5761,
|
| 150464 |
+
"step": 40165
|
| 150465 |
+
},
|
| 150466 |
+
{
|
| 150467 |
+
"epoch": 320.8,
|
| 150468 |
+
"learning_rate": 9.372516025641027e-06,
|
| 150469 |
+
"loss": 1.3147,
|
| 150470 |
+
"step": 40170
|
| 150471 |
+
},
|
| 150472 |
+
{
|
| 150473 |
+
"epoch": 320.84,
|
| 150474 |
+
"learning_rate": 9.372435897435898e-06,
|
| 150475 |
+
"loss": 0.4042,
|
| 150476 |
+
"step": 40175
|
| 150477 |
+
},
|
| 150478 |
+
{
|
| 150479 |
+
"epoch": 320.88,
|
| 150480 |
+
"learning_rate": 9.37235576923077e-06,
|
| 150481 |
+
"loss": 0.3066,
|
| 150482 |
+
"step": 40180
|
| 150483 |
+
},
|
| 150484 |
+
{
|
| 150485 |
+
"epoch": 320.92,
|
| 150486 |
+
"learning_rate": 9.372275641025642e-06,
|
| 150487 |
+
"loss": 0.4208,
|
| 150488 |
+
"step": 40185
|
| 150489 |
+
},
|
| 150490 |
+
{
|
| 150491 |
+
"epoch": 320.96,
|
| 150492 |
+
"learning_rate": 9.372195512820514e-06,
|
| 150493 |
+
"loss": 0.5779,
|
| 150494 |
+
"step": 40190
|
| 150495 |
+
},
|
| 150496 |
+
{
|
| 150497 |
+
"epoch": 321.0,
|
| 150498 |
+
"learning_rate": 9.372115384615385e-06,
|
| 150499 |
+
"loss": 1.7014,
|
| 150500 |
+
"step": 40195
|
| 150501 |
+
},
|
| 150502 |
+
{
|
| 150503 |
+
"epoch": 321.0,
|
| 150504 |
+
"eval_loss": 0.42383575439453125,
|
| 150505 |
+
"eval_runtime": 39.3641,
|
| 150506 |
+
"eval_samples_per_second": 21.365,
|
| 150507 |
+
"eval_steps_per_second": 0.686,
|
| 150508 |
+
"eval_wer": 0.22157434402332363,
|
| 150509 |
+
"step": 40195
|
| 150510 |
}
|
| 150511 |
],
|
| 150512 |
+
"max_steps": 625000,
|
| 150513 |
"num_train_epochs": 5000,
|
| 150514 |
+
"total_flos": 1.1310540615989882e+20,
|
| 150515 |
"trial_name": null,
|
| 150516 |
"trial_params": null
|
| 150517 |
}
|
model-bin/finetune/base/{checkpoint-39572 → checkpoint-40195}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629686760.2288544/events.out.tfevents.1629686760.74272264b15c.932.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ca8f3e3f0e7f23ecf5f6fe67d2c156d59ec801426e03cf01e2f06f111e94b454
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629687403.1416256/events.out.tfevents.1629687403.74272264b15c.932.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:14f76a84f8dd14c406989243078ac6e8e8195fad4723a8279ad95f5f8e56eaf0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629688047.4250455/events.out.tfevents.1629688047.74272264b15c.932.27
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f9cf42fc2c6e35c348cb0ae234b3d24b43a538e4506f4eda6cb0aea319b3ca06
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629688768.5831227/events.out.tfevents.1629688768.74272264b15c.932.29
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd211eb64171f6bf938667c974e6c1215e0f862ab79332c3dd82ee0608941834
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629689425.810986/events.out.tfevents.1629689425.74272264b15c.932.31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:85abd900997510f752162e2be8ad403be6a173684ebb3e17374e594c14175124
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629686760.74272264b15c.932.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:677c9202a9a8c6f4cfc783163ff2916ff30f5cbc18915fe6844fe030dbccfcef
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629687403.74272264b15c.932.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:16a5e92b934672afe30ae6d644dde6cfa6828126a8bb8a92f6f029b2ad6dc734
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629688047.74272264b15c.932.26
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8c0c52f414973b83a0255bdc59fb22dd27882d99f947f7fcb79241453681314b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629688768.74272264b15c.932.28
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e5ca3bc4b8b4b547c540401004b52183f4b136fa66decc26798b10b15e4a973e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629689425.74272264b15c.932.30
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d6c66996844a4339b7f325f891c8b594266cb04f059e4c8c5b9ae87a0c706837
|
| 3 |
+
size 8622
|