"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630085091.0178604/events.out.tfevents.1630085091.dea8124df033.1391.17 +3 -0
- model-bin/finetune/base/log/1630085509.6253698/events.out.tfevents.1630085509.dea8124df033.1391.19 +3 -0
- model-bin/finetune/base/log/1630085938.0146723/events.out.tfevents.1630085938.dea8124df033.1391.21 +3 -0
- model-bin/finetune/base/log/1630086359.258517/events.out.tfevents.1630086359.dea8124df033.1391.23 +3 -0
- model-bin/finetune/base/log/1630086783.2838607/events.out.tfevents.1630086783.dea8124df033.1391.25 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630085091.dea8124df033.1391.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630085509.dea8124df033.1391.18 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630085938.dea8124df033.1391.20 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630086359.dea8124df033.1391.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630086783.dea8124df033.1391.24 +3 -0
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/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 sha256:0f4f9e0ee9471c71831eaec61932733e15a611cf26ed1b7f271e6474f3f01c9f
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/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 sha256:24e4b1f38df5421de8e487702e1e4f0709dc5d00e9a56d78221675e881e8d441
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/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:eb8c86e3a58cf24e873bdae44ba1346a463db062b7dde9639caebc5bcd4b49b7
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/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 sha256:1c561868acf735ac6c9b5bb98ac6184be244d99c25305b71d752f027d4438d76
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/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:e9ec5eb205fb1186456b27df8a3a22735db56014ea3ec653c0f4bf9495c4e307
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -243609,11 +243609,800 @@
|
|
| 243609 |
"eval_steps_per_second": 0.678,
|
| 243610 |
"eval_wer": 0.18508635386553998,
|
| 243611 |
"step": 113370
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 243612 |
}
|
| 243613 |
],
|
| 243614 |
-
"max_steps":
|
| 243615 |
"num_train_epochs": 5000,
|
| 243616 |
-
"total_flos": 3.
|
| 243617 |
"trial_name": null,
|
| 243618 |
"trial_params": null
|
| 243619 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 911.0,
|
| 5 |
+
"global_step": 113992,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 243609 |
"eval_steps_per_second": 0.678,
|
| 243610 |
"eval_wer": 0.18508635386553998,
|
| 243611 |
"step": 113370
|
| 243612 |
+
},
|
| 243613 |
+
{
|
| 243614 |
+
"epoch": 914.04,
|
| 243615 |
+
"learning_rate": 8.18529886914378e-06,
|
| 243616 |
+
"loss": 0.3386,
|
| 243617 |
+
"step": 113375
|
| 243618 |
+
},
|
| 243619 |
+
{
|
| 243620 |
+
"epoch": 914.08,
|
| 243621 |
+
"learning_rate": 8.185218093699517e-06,
|
| 243622 |
+
"loss": 0.3232,
|
| 243623 |
+
"step": 113380
|
| 243624 |
+
},
|
| 243625 |
+
{
|
| 243626 |
+
"epoch": 914.12,
|
| 243627 |
+
"learning_rate": 8.18513731825525e-06,
|
| 243628 |
+
"loss": 0.3823,
|
| 243629 |
+
"step": 113385
|
| 243630 |
+
},
|
| 243631 |
+
{
|
| 243632 |
+
"epoch": 914.16,
|
| 243633 |
+
"learning_rate": 8.185056542810987e-06,
|
| 243634 |
+
"loss": 0.4816,
|
| 243635 |
+
"step": 113390
|
| 243636 |
+
},
|
| 243637 |
+
{
|
| 243638 |
+
"epoch": 914.2,
|
| 243639 |
+
"learning_rate": 8.18497576736672e-06,
|
| 243640 |
+
"loss": 1.133,
|
| 243641 |
+
"step": 113395
|
| 243642 |
+
},
|
| 243643 |
+
{
|
| 243644 |
+
"epoch": 914.24,
|
| 243645 |
+
"learning_rate": 8.184894991922456e-06,
|
| 243646 |
+
"loss": 0.3591,
|
| 243647 |
+
"step": 113400
|
| 243648 |
+
},
|
| 243649 |
+
{
|
| 243650 |
+
"epoch": 914.28,
|
| 243651 |
+
"learning_rate": 8.184814216478192e-06,
|
| 243652 |
+
"loss": 0.298,
|
| 243653 |
+
"step": 113405
|
| 243654 |
+
},
|
| 243655 |
+
{
|
| 243656 |
+
"epoch": 914.32,
|
| 243657 |
+
"learning_rate": 8.184733441033926e-06,
|
| 243658 |
+
"loss": 0.3137,
|
| 243659 |
+
"step": 113410
|
| 243660 |
+
},
|
| 243661 |
+
{
|
| 243662 |
+
"epoch": 914.36,
|
| 243663 |
+
"learning_rate": 8.184652665589662e-06,
|
| 243664 |
+
"loss": 0.4768,
|
| 243665 |
+
"step": 113415
|
| 243666 |
+
},
|
| 243667 |
+
{
|
| 243668 |
+
"epoch": 914.4,
|
| 243669 |
+
"learning_rate": 8.184571890145396e-06,
|
| 243670 |
+
"loss": 1.1675,
|
| 243671 |
+
"step": 113420
|
| 243672 |
+
},
|
| 243673 |
+
{
|
| 243674 |
+
"epoch": 914.44,
|
| 243675 |
+
"learning_rate": 8.184491114701132e-06,
|
| 243676 |
+
"loss": 0.3729,
|
| 243677 |
+
"step": 113425
|
| 243678 |
+
},
|
| 243679 |
+
{
|
| 243680 |
+
"epoch": 914.48,
|
| 243681 |
+
"learning_rate": 8.184410339256866e-06,
|
| 243682 |
+
"loss": 0.285,
|
| 243683 |
+
"step": 113430
|
| 243684 |
+
},
|
| 243685 |
+
{
|
| 243686 |
+
"epoch": 914.52,
|
| 243687 |
+
"learning_rate": 8.184329563812602e-06,
|
| 243688 |
+
"loss": 0.2975,
|
| 243689 |
+
"step": 113435
|
| 243690 |
+
},
|
| 243691 |
+
{
|
| 243692 |
+
"epoch": 914.56,
|
| 243693 |
+
"learning_rate": 8.184248788368336e-06,
|
| 243694 |
+
"loss": 0.4192,
|
| 243695 |
+
"step": 113440
|
| 243696 |
+
},
|
| 243697 |
+
{
|
| 243698 |
+
"epoch": 914.6,
|
| 243699 |
+
"learning_rate": 8.184168012924072e-06,
|
| 243700 |
+
"loss": 1.1731,
|
| 243701 |
+
"step": 113445
|
| 243702 |
+
},
|
| 243703 |
+
{
|
| 243704 |
+
"epoch": 914.65,
|
| 243705 |
+
"learning_rate": 8.184087237479806e-06,
|
| 243706 |
+
"loss": 0.2925,
|
| 243707 |
+
"step": 113450
|
| 243708 |
+
},
|
| 243709 |
+
{
|
| 243710 |
+
"epoch": 914.69,
|
| 243711 |
+
"learning_rate": 8.184006462035542e-06,
|
| 243712 |
+
"loss": 0.2705,
|
| 243713 |
+
"step": 113455
|
| 243714 |
+
},
|
| 243715 |
+
{
|
| 243716 |
+
"epoch": 914.73,
|
| 243717 |
+
"learning_rate": 8.183925686591276e-06,
|
| 243718 |
+
"loss": 0.303,
|
| 243719 |
+
"step": 113460
|
| 243720 |
+
},
|
| 243721 |
+
{
|
| 243722 |
+
"epoch": 914.77,
|
| 243723 |
+
"learning_rate": 8.183844911147012e-06,
|
| 243724 |
+
"loss": 0.443,
|
| 243725 |
+
"step": 113465
|
| 243726 |
+
},
|
| 243727 |
+
{
|
| 243728 |
+
"epoch": 914.81,
|
| 243729 |
+
"learning_rate": 8.183764135702748e-06,
|
| 243730 |
+
"loss": 1.1133,
|
| 243731 |
+
"step": 113470
|
| 243732 |
+
},
|
| 243733 |
+
{
|
| 243734 |
+
"epoch": 914.85,
|
| 243735 |
+
"learning_rate": 8.183683360258482e-06,
|
| 243736 |
+
"loss": 0.3293,
|
| 243737 |
+
"step": 113475
|
| 243738 |
+
},
|
| 243739 |
+
{
|
| 243740 |
+
"epoch": 914.89,
|
| 243741 |
+
"learning_rate": 8.183602584814218e-06,
|
| 243742 |
+
"loss": 0.2448,
|
| 243743 |
+
"step": 113480
|
| 243744 |
+
},
|
| 243745 |
+
{
|
| 243746 |
+
"epoch": 914.93,
|
| 243747 |
+
"learning_rate": 8.183521809369952e-06,
|
| 243748 |
+
"loss": 0.4102,
|
| 243749 |
+
"step": 113485
|
| 243750 |
+
},
|
| 243751 |
+
{
|
| 243752 |
+
"epoch": 914.97,
|
| 243753 |
+
"learning_rate": 8.183441033925688e-06,
|
| 243754 |
+
"loss": 0.5599,
|
| 243755 |
+
"step": 113490
|
| 243756 |
+
},
|
| 243757 |
+
{
|
| 243758 |
+
"epoch": 915.0,
|
| 243759 |
+
"eval_loss": 0.3741353750228882,
|
| 243760 |
+
"eval_runtime": 40.0199,
|
| 243761 |
+
"eval_samples_per_second": 20.915,
|
| 243762 |
+
"eval_steps_per_second": 0.675,
|
| 243763 |
+
"eval_wer": 0.1836039431719339,
|
| 243764 |
+
"step": 113494
|
| 243765 |
+
},
|
| 243766 |
+
{
|
| 243767 |
+
"epoch": 915.01,
|
| 243768 |
+
"learning_rate": 8.183360258481422e-06,
|
| 243769 |
+
"loss": 0.3819,
|
| 243770 |
+
"step": 113495
|
| 243771 |
+
},
|
| 243772 |
+
{
|
| 243773 |
+
"epoch": 915.05,
|
| 243774 |
+
"learning_rate": 8.183279483037158e-06,
|
| 243775 |
+
"loss": 0.2852,
|
| 243776 |
+
"step": 113500
|
| 243777 |
+
},
|
| 243778 |
+
{
|
| 243779 |
+
"epoch": 915.09,
|
| 243780 |
+
"learning_rate": 8.183198707592892e-06,
|
| 243781 |
+
"loss": 0.3385,
|
| 243782 |
+
"step": 113505
|
| 243783 |
+
},
|
| 243784 |
+
{
|
| 243785 |
+
"epoch": 915.13,
|
| 243786 |
+
"learning_rate": 8.183117932148628e-06,
|
| 243787 |
+
"loss": 0.3233,
|
| 243788 |
+
"step": 113510
|
| 243789 |
+
},
|
| 243790 |
+
{
|
| 243791 |
+
"epoch": 915.17,
|
| 243792 |
+
"learning_rate": 8.183037156704362e-06,
|
| 243793 |
+
"loss": 0.5698,
|
| 243794 |
+
"step": 113515
|
| 243795 |
+
},
|
| 243796 |
+
{
|
| 243797 |
+
"epoch": 915.21,
|
| 243798 |
+
"learning_rate": 8.182956381260098e-06,
|
| 243799 |
+
"loss": 1.2247,
|
| 243800 |
+
"step": 113520
|
| 243801 |
+
},
|
| 243802 |
+
{
|
| 243803 |
+
"epoch": 915.25,
|
| 243804 |
+
"learning_rate": 8.182875605815834e-06,
|
| 243805 |
+
"loss": 0.3235,
|
| 243806 |
+
"step": 113525
|
| 243807 |
+
},
|
| 243808 |
+
{
|
| 243809 |
+
"epoch": 915.29,
|
| 243810 |
+
"learning_rate": 8.182794830371568e-06,
|
| 243811 |
+
"loss": 0.3095,
|
| 243812 |
+
"step": 113530
|
| 243813 |
+
},
|
| 243814 |
+
{
|
| 243815 |
+
"epoch": 915.33,
|
| 243816 |
+
"learning_rate": 8.182714054927304e-06,
|
| 243817 |
+
"loss": 0.323,
|
| 243818 |
+
"step": 113535
|
| 243819 |
+
},
|
| 243820 |
+
{
|
| 243821 |
+
"epoch": 915.37,
|
| 243822 |
+
"learning_rate": 8.182633279483038e-06,
|
| 243823 |
+
"loss": 0.4947,
|
| 243824 |
+
"step": 113540
|
| 243825 |
+
},
|
| 243826 |
+
{
|
| 243827 |
+
"epoch": 915.41,
|
| 243828 |
+
"learning_rate": 8.182552504038774e-06,
|
| 243829 |
+
"loss": 1.2059,
|
| 243830 |
+
"step": 113545
|
| 243831 |
+
},
|
| 243832 |
+
{
|
| 243833 |
+
"epoch": 915.45,
|
| 243834 |
+
"learning_rate": 8.182471728594508e-06,
|
| 243835 |
+
"loss": 0.2672,
|
| 243836 |
+
"step": 113550
|
| 243837 |
+
},
|
| 243838 |
+
{
|
| 243839 |
+
"epoch": 915.49,
|
| 243840 |
+
"learning_rate": 8.182390953150244e-06,
|
| 243841 |
+
"loss": 0.3121,
|
| 243842 |
+
"step": 113555
|
| 243843 |
+
},
|
| 243844 |
+
{
|
| 243845 |
+
"epoch": 915.53,
|
| 243846 |
+
"learning_rate": 8.182310177705978e-06,
|
| 243847 |
+
"loss": 0.2868,
|
| 243848 |
+
"step": 113560
|
| 243849 |
+
},
|
| 243850 |
+
{
|
| 243851 |
+
"epoch": 915.57,
|
| 243852 |
+
"learning_rate": 8.182229402261714e-06,
|
| 243853 |
+
"loss": 0.5263,
|
| 243854 |
+
"step": 113565
|
| 243855 |
+
},
|
| 243856 |
+
{
|
| 243857 |
+
"epoch": 915.61,
|
| 243858 |
+
"learning_rate": 8.182148626817448e-06,
|
| 243859 |
+
"loss": 1.0193,
|
| 243860 |
+
"step": 113570
|
| 243861 |
+
},
|
| 243862 |
+
{
|
| 243863 |
+
"epoch": 915.65,
|
| 243864 |
+
"learning_rate": 8.182067851373184e-06,
|
| 243865 |
+
"loss": 0.3044,
|
| 243866 |
+
"step": 113575
|
| 243867 |
+
},
|
| 243868 |
+
{
|
| 243869 |
+
"epoch": 915.69,
|
| 243870 |
+
"learning_rate": 8.18198707592892e-06,
|
| 243871 |
+
"loss": 0.2556,
|
| 243872 |
+
"step": 113580
|
| 243873 |
+
},
|
| 243874 |
+
{
|
| 243875 |
+
"epoch": 915.73,
|
| 243876 |
+
"learning_rate": 8.181906300484654e-06,
|
| 243877 |
+
"loss": 0.3336,
|
| 243878 |
+
"step": 113585
|
| 243879 |
+
},
|
| 243880 |
+
{
|
| 243881 |
+
"epoch": 915.77,
|
| 243882 |
+
"learning_rate": 8.18182552504039e-06,
|
| 243883 |
+
"loss": 0.5224,
|
| 243884 |
+
"step": 113590
|
| 243885 |
+
},
|
| 243886 |
+
{
|
| 243887 |
+
"epoch": 915.81,
|
| 243888 |
+
"learning_rate": 8.181744749596124e-06,
|
| 243889 |
+
"loss": 1.2317,
|
| 243890 |
+
"step": 113595
|
| 243891 |
+
},
|
| 243892 |
+
{
|
| 243893 |
+
"epoch": 915.85,
|
| 243894 |
+
"learning_rate": 8.18166397415186e-06,
|
| 243895 |
+
"loss": 0.303,
|
| 243896 |
+
"step": 113600
|
| 243897 |
+
},
|
| 243898 |
+
{
|
| 243899 |
+
"epoch": 915.89,
|
| 243900 |
+
"learning_rate": 8.181583198707594e-06,
|
| 243901 |
+
"loss": 0.2783,
|
| 243902 |
+
"step": 113605
|
| 243903 |
+
},
|
| 243904 |
+
{
|
| 243905 |
+
"epoch": 915.93,
|
| 243906 |
+
"learning_rate": 8.18150242326333e-06,
|
| 243907 |
+
"loss": 0.3852,
|
| 243908 |
+
"step": 113610
|
| 243909 |
+
},
|
| 243910 |
+
{
|
| 243911 |
+
"epoch": 915.97,
|
| 243912 |
+
"learning_rate": 8.181421647819063e-06,
|
| 243913 |
+
"loss": 0.4534,
|
| 243914 |
+
"step": 113615
|
| 243915 |
+
},
|
| 243916 |
+
{
|
| 243917 |
+
"epoch": 916.0,
|
| 243918 |
+
"eval_loss": 0.466743141412735,
|
| 243919 |
+
"eval_runtime": 39.7713,
|
| 243920 |
+
"eval_samples_per_second": 21.02,
|
| 243921 |
+
"eval_steps_per_second": 0.679,
|
| 243922 |
+
"eval_wer": 0.18429758141586414,
|
| 243923 |
+
"step": 113618
|
| 243924 |
+
},
|
| 243925 |
+
{
|
| 243926 |
+
"epoch": 908.02,
|
| 243927 |
+
"learning_rate": 8.1813408723748e-06,
|
| 243928 |
+
"loss": 0.3404,
|
| 243929 |
+
"step": 113620
|
| 243930 |
+
},
|
| 243931 |
+
{
|
| 243932 |
+
"epoch": 908.06,
|
| 243933 |
+
"learning_rate": 8.181260096930533e-06,
|
| 243934 |
+
"loss": 0.3143,
|
| 243935 |
+
"step": 113625
|
| 243936 |
+
},
|
| 243937 |
+
{
|
| 243938 |
+
"epoch": 908.1,
|
| 243939 |
+
"learning_rate": 8.18117932148627e-06,
|
| 243940 |
+
"loss": 0.2869,
|
| 243941 |
+
"step": 113630
|
| 243942 |
+
},
|
| 243943 |
+
{
|
| 243944 |
+
"epoch": 908.14,
|
| 243945 |
+
"learning_rate": 8.181098546042003e-06,
|
| 243946 |
+
"loss": 0.3377,
|
| 243947 |
+
"step": 113635
|
| 243948 |
+
},
|
| 243949 |
+
{
|
| 243950 |
+
"epoch": 908.18,
|
| 243951 |
+
"learning_rate": 8.18101777059774e-06,
|
| 243952 |
+
"loss": 0.5821,
|
| 243953 |
+
"step": 113640
|
| 243954 |
+
},
|
| 243955 |
+
{
|
| 243956 |
+
"epoch": 908.22,
|
| 243957 |
+
"learning_rate": 8.180936995153475e-06,
|
| 243958 |
+
"loss": 1.1251,
|
| 243959 |
+
"step": 113645
|
| 243960 |
+
},
|
| 243961 |
+
{
|
| 243962 |
+
"epoch": 908.26,
|
| 243963 |
+
"learning_rate": 8.18085621970921e-06,
|
| 243964 |
+
"loss": 0.327,
|
| 243965 |
+
"step": 113650
|
| 243966 |
+
},
|
| 243967 |
+
{
|
| 243968 |
+
"epoch": 908.3,
|
| 243969 |
+
"learning_rate": 8.180775444264945e-06,
|
| 243970 |
+
"loss": 0.2657,
|
| 243971 |
+
"step": 113655
|
| 243972 |
+
},
|
| 243973 |
+
{
|
| 243974 |
+
"epoch": 908.34,
|
| 243975 |
+
"learning_rate": 8.18069466882068e-06,
|
| 243976 |
+
"loss": 0.3922,
|
| 243977 |
+
"step": 113660
|
| 243978 |
+
},
|
| 243979 |
+
{
|
| 243980 |
+
"epoch": 908.38,
|
| 243981 |
+
"learning_rate": 8.180613893376415e-06,
|
| 243982 |
+
"loss": 0.5555,
|
| 243983 |
+
"step": 113665
|
| 243984 |
+
},
|
| 243985 |
+
{
|
| 243986 |
+
"epoch": 908.42,
|
| 243987 |
+
"learning_rate": 8.18053311793215e-06,
|
| 243988 |
+
"loss": 0.8309,
|
| 243989 |
+
"step": 113670
|
| 243990 |
+
},
|
| 243991 |
+
{
|
| 243992 |
+
"epoch": 908.46,
|
| 243993 |
+
"learning_rate": 8.180452342487885e-06,
|
| 243994 |
+
"loss": 0.3719,
|
| 243995 |
+
"step": 113675
|
| 243996 |
+
},
|
| 243997 |
+
{
|
| 243998 |
+
"epoch": 908.5,
|
| 243999 |
+
"learning_rate": 8.18037156704362e-06,
|
| 244000 |
+
"loss": 0.354,
|
| 244001 |
+
"step": 113680
|
| 244002 |
+
},
|
| 244003 |
+
{
|
| 244004 |
+
"epoch": 908.54,
|
| 244005 |
+
"learning_rate": 8.180290791599355e-06,
|
| 244006 |
+
"loss": 0.3629,
|
| 244007 |
+
"step": 113685
|
| 244008 |
+
},
|
| 244009 |
+
{
|
| 244010 |
+
"epoch": 908.58,
|
| 244011 |
+
"learning_rate": 8.180210016155089e-06,
|
| 244012 |
+
"loss": 0.7741,
|
| 244013 |
+
"step": 113690
|
| 244014 |
+
},
|
| 244015 |
+
{
|
| 244016 |
+
"epoch": 908.62,
|
| 244017 |
+
"learning_rate": 8.180129240710825e-06,
|
| 244018 |
+
"loss": 0.8774,
|
| 244019 |
+
"step": 113695
|
| 244020 |
+
},
|
| 244021 |
+
{
|
| 244022 |
+
"epoch": 908.66,
|
| 244023 |
+
"learning_rate": 8.18004846526656e-06,
|
| 244024 |
+
"loss": 0.2846,
|
| 244025 |
+
"step": 113700
|
| 244026 |
+
},
|
| 244027 |
+
{
|
| 244028 |
+
"epoch": 908.7,
|
| 244029 |
+
"learning_rate": 8.179967689822295e-06,
|
| 244030 |
+
"loss": 0.2622,
|
| 244031 |
+
"step": 113705
|
| 244032 |
+
},
|
| 244033 |
+
{
|
| 244034 |
+
"epoch": 908.74,
|
| 244035 |
+
"learning_rate": 8.17988691437803e-06,
|
| 244036 |
+
"loss": 0.4547,
|
| 244037 |
+
"step": 113710
|
| 244038 |
+
},
|
| 244039 |
+
{
|
| 244040 |
+
"epoch": 908.78,
|
| 244041 |
+
"learning_rate": 8.179806138933765e-06,
|
| 244042 |
+
"loss": 0.6174,
|
| 244043 |
+
"step": 113715
|
| 244044 |
+
},
|
| 244045 |
+
{
|
| 244046 |
+
"epoch": 908.82,
|
| 244047 |
+
"learning_rate": 8.1797253634895e-06,
|
| 244048 |
+
"loss": 1.027,
|
| 244049 |
+
"step": 113720
|
| 244050 |
+
},
|
| 244051 |
+
{
|
| 244052 |
+
"epoch": 908.86,
|
| 244053 |
+
"learning_rate": 8.179644588045235e-06,
|
| 244054 |
+
"loss": 0.2858,
|
| 244055 |
+
"step": 113725
|
| 244056 |
+
},
|
| 244057 |
+
{
|
| 244058 |
+
"epoch": 908.9,
|
| 244059 |
+
"learning_rate": 8.17956381260097e-06,
|
| 244060 |
+
"loss": 0.2602,
|
| 244061 |
+
"step": 113730
|
| 244062 |
+
},
|
| 244063 |
+
{
|
| 244064 |
+
"epoch": 908.94,
|
| 244065 |
+
"learning_rate": 8.179483037156705e-06,
|
| 244066 |
+
"loss": 0.347,
|
| 244067 |
+
"step": 113735
|
| 244068 |
+
},
|
| 244069 |
+
{
|
| 244070 |
+
"epoch": 908.98,
|
| 244071 |
+
"learning_rate": 8.17940226171244e-06,
|
| 244072 |
+
"loss": 0.6511,
|
| 244073 |
+
"step": 113740
|
| 244074 |
+
},
|
| 244075 |
+
{
|
| 244076 |
+
"epoch": 909.0,
|
| 244077 |
+
"eval_loss": 0.4806082248687744,
|
| 244078 |
+
"eval_runtime": 40.3137,
|
| 244079 |
+
"eval_samples_per_second": 20.737,
|
| 244080 |
+
"eval_steps_per_second": 0.67,
|
| 244081 |
+
"eval_wer": 0.19115031129558258,
|
| 244082 |
+
"step": 113743
|
| 244083 |
+
},
|
| 244084 |
+
{
|
| 244085 |
+
"epoch": 917.02,
|
| 244086 |
+
"learning_rate": 8.179321486268175e-06,
|
| 244087 |
+
"loss": 0.3522,
|
| 244088 |
+
"step": 113745
|
| 244089 |
+
},
|
| 244090 |
+
{
|
| 244091 |
+
"epoch": 917.06,
|
| 244092 |
+
"learning_rate": 8.17924071082391e-06,
|
| 244093 |
+
"loss": 0.2977,
|
| 244094 |
+
"step": 113750
|
| 244095 |
+
},
|
| 244096 |
+
{
|
| 244097 |
+
"epoch": 917.1,
|
| 244098 |
+
"learning_rate": 8.179159935379647e-06,
|
| 244099 |
+
"loss": 0.3204,
|
| 244100 |
+
"step": 113755
|
| 244101 |
+
},
|
| 244102 |
+
{
|
| 244103 |
+
"epoch": 917.14,
|
| 244104 |
+
"learning_rate": 8.17907915993538e-06,
|
| 244105 |
+
"loss": 0.4102,
|
| 244106 |
+
"step": 113760
|
| 244107 |
+
},
|
| 244108 |
+
{
|
| 244109 |
+
"epoch": 917.18,
|
| 244110 |
+
"learning_rate": 8.178998384491117e-06,
|
| 244111 |
+
"loss": 0.6249,
|
| 244112 |
+
"step": 113765
|
| 244113 |
+
},
|
| 244114 |
+
{
|
| 244115 |
+
"epoch": 917.22,
|
| 244116 |
+
"learning_rate": 8.17891760904685e-06,
|
| 244117 |
+
"loss": 0.9684,
|
| 244118 |
+
"step": 113770
|
| 244119 |
+
},
|
| 244120 |
+
{
|
| 244121 |
+
"epoch": 917.26,
|
| 244122 |
+
"learning_rate": 8.178836833602586e-06,
|
| 244123 |
+
"loss": 0.3437,
|
| 244124 |
+
"step": 113775
|
| 244125 |
+
},
|
| 244126 |
+
{
|
| 244127 |
+
"epoch": 917.3,
|
| 244128 |
+
"learning_rate": 8.17875605815832e-06,
|
| 244129 |
+
"loss": 0.2964,
|
| 244130 |
+
"step": 113780
|
| 244131 |
+
},
|
| 244132 |
+
{
|
| 244133 |
+
"epoch": 917.34,
|
| 244134 |
+
"learning_rate": 8.178675282714056e-06,
|
| 244135 |
+
"loss": 0.3275,
|
| 244136 |
+
"step": 113785
|
| 244137 |
+
},
|
| 244138 |
+
{
|
| 244139 |
+
"epoch": 917.38,
|
| 244140 |
+
"learning_rate": 8.17859450726979e-06,
|
| 244141 |
+
"loss": 0.6473,
|
| 244142 |
+
"step": 113790
|
| 244143 |
+
},
|
| 244144 |
+
{
|
| 244145 |
+
"epoch": 917.42,
|
| 244146 |
+
"learning_rate": 8.178513731825526e-06,
|
| 244147 |
+
"loss": 1.1345,
|
| 244148 |
+
"step": 113795
|
| 244149 |
+
},
|
| 244150 |
+
{
|
| 244151 |
+
"epoch": 917.46,
|
| 244152 |
+
"learning_rate": 8.17843295638126e-06,
|
| 244153 |
+
"loss": 0.2581,
|
| 244154 |
+
"step": 113800
|
| 244155 |
+
},
|
| 244156 |
+
{
|
| 244157 |
+
"epoch": 917.5,
|
| 244158 |
+
"learning_rate": 8.178352180936996e-06,
|
| 244159 |
+
"loss": 0.3575,
|
| 244160 |
+
"step": 113805
|
| 244161 |
+
},
|
| 244162 |
+
{
|
| 244163 |
+
"epoch": 917.54,
|
| 244164 |
+
"learning_rate": 8.17827140549273e-06,
|
| 244165 |
+
"loss": 0.3293,
|
| 244166 |
+
"step": 113810
|
| 244167 |
+
},
|
| 244168 |
+
{
|
| 244169 |
+
"epoch": 917.58,
|
| 244170 |
+
"learning_rate": 8.178190630048466e-06,
|
| 244171 |
+
"loss": 0.5812,
|
| 244172 |
+
"step": 113815
|
| 244173 |
+
},
|
| 244174 |
+
{
|
| 244175 |
+
"epoch": 917.62,
|
| 244176 |
+
"learning_rate": 8.178109854604202e-06,
|
| 244177 |
+
"loss": 1.0217,
|
| 244178 |
+
"step": 113820
|
| 244179 |
+
},
|
| 244180 |
+
{
|
| 244181 |
+
"epoch": 917.66,
|
| 244182 |
+
"learning_rate": 8.178029079159936e-06,
|
| 244183 |
+
"loss": 0.3646,
|
| 244184 |
+
"step": 113825
|
| 244185 |
+
},
|
| 244186 |
+
{
|
| 244187 |
+
"epoch": 917.7,
|
| 244188 |
+
"learning_rate": 8.177948303715672e-06,
|
| 244189 |
+
"loss": 0.2918,
|
| 244190 |
+
"step": 113830
|
| 244191 |
+
},
|
| 244192 |
+
{
|
| 244193 |
+
"epoch": 917.74,
|
| 244194 |
+
"learning_rate": 8.177867528271406e-06,
|
| 244195 |
+
"loss": 0.3763,
|
| 244196 |
+
"step": 113835
|
| 244197 |
+
},
|
| 244198 |
+
{
|
| 244199 |
+
"epoch": 917.78,
|
| 244200 |
+
"learning_rate": 8.177786752827142e-06,
|
| 244201 |
+
"loss": 0.4992,
|
| 244202 |
+
"step": 113840
|
| 244203 |
+
},
|
| 244204 |
+
{
|
| 244205 |
+
"epoch": 917.82,
|
| 244206 |
+
"learning_rate": 8.177705977382876e-06,
|
| 244207 |
+
"loss": 0.8907,
|
| 244208 |
+
"step": 113845
|
| 244209 |
+
},
|
| 244210 |
+
{
|
| 244211 |
+
"epoch": 917.86,
|
| 244212 |
+
"learning_rate": 8.177625201938612e-06,
|
| 244213 |
+
"loss": 0.2465,
|
| 244214 |
+
"step": 113850
|
| 244215 |
+
},
|
| 244216 |
+
{
|
| 244217 |
+
"epoch": 917.9,
|
| 244218 |
+
"learning_rate": 8.177544426494346e-06,
|
| 244219 |
+
"loss": 0.2878,
|
| 244220 |
+
"step": 113855
|
| 244221 |
+
},
|
| 244222 |
+
{
|
| 244223 |
+
"epoch": 917.94,
|
| 244224 |
+
"learning_rate": 8.177463651050082e-06,
|
| 244225 |
+
"loss": 0.3729,
|
| 244226 |
+
"step": 113860
|
| 244227 |
+
},
|
| 244228 |
+
{
|
| 244229 |
+
"epoch": 917.98,
|
| 244230 |
+
"learning_rate": 8.177382875605816e-06,
|
| 244231 |
+
"loss": 0.6637,
|
| 244232 |
+
"step": 113865
|
| 244233 |
+
},
|
| 244234 |
+
{
|
| 244235 |
+
"epoch": 918.0,
|
| 244236 |
+
"eval_loss": 0.38021937012672424,
|
| 244237 |
+
"eval_runtime": 38.5081,
|
| 244238 |
+
"eval_samples_per_second": 21.71,
|
| 244239 |
+
"eval_steps_per_second": 0.701,
|
| 244240 |
+
"eval_wer": 0.18388491353607633,
|
| 244241 |
+
"step": 113867
|
| 244242 |
+
},
|
| 244243 |
+
{
|
| 244244 |
+
"epoch": 910.02,
|
| 244245 |
+
"learning_rate": 8.177302100161552e-06,
|
| 244246 |
+
"loss": 0.35,
|
| 244247 |
+
"step": 113870
|
| 244248 |
+
},
|
| 244249 |
+
{
|
| 244250 |
+
"epoch": 910.06,
|
| 244251 |
+
"learning_rate": 8.177221324717288e-06,
|
| 244252 |
+
"loss": 0.2942,
|
| 244253 |
+
"step": 113875
|
| 244254 |
+
},
|
| 244255 |
+
{
|
| 244256 |
+
"epoch": 910.1,
|
| 244257 |
+
"learning_rate": 8.177140549273022e-06,
|
| 244258 |
+
"loss": 0.2951,
|
| 244259 |
+
"step": 113880
|
| 244260 |
+
},
|
| 244261 |
+
{
|
| 244262 |
+
"epoch": 910.14,
|
| 244263 |
+
"learning_rate": 8.177059773828758e-06,
|
| 244264 |
+
"loss": 0.3673,
|
| 244265 |
+
"step": 113885
|
| 244266 |
+
},
|
| 244267 |
+
{
|
| 244268 |
+
"epoch": 910.18,
|
| 244269 |
+
"learning_rate": 8.176978998384492e-06,
|
| 244270 |
+
"loss": 0.8055,
|
| 244271 |
+
"step": 113890
|
| 244272 |
+
},
|
| 244273 |
+
{
|
| 244274 |
+
"epoch": 910.22,
|
| 244275 |
+
"learning_rate": 8.176898222940228e-06,
|
| 244276 |
+
"loss": 0.8096,
|
| 244277 |
+
"step": 113895
|
| 244278 |
+
},
|
| 244279 |
+
{
|
| 244280 |
+
"epoch": 910.26,
|
| 244281 |
+
"learning_rate": 8.176817447495962e-06,
|
| 244282 |
+
"loss": 0.2766,
|
| 244283 |
+
"step": 113900
|
| 244284 |
+
},
|
| 244285 |
+
{
|
| 244286 |
+
"epoch": 910.3,
|
| 244287 |
+
"learning_rate": 8.176736672051698e-06,
|
| 244288 |
+
"loss": 0.3135,
|
| 244289 |
+
"step": 113905
|
| 244290 |
+
},
|
| 244291 |
+
{
|
| 244292 |
+
"epoch": 910.34,
|
| 244293 |
+
"learning_rate": 8.176655896607432e-06,
|
| 244294 |
+
"loss": 0.3225,
|
| 244295 |
+
"step": 113910
|
| 244296 |
+
},
|
| 244297 |
+
{
|
| 244298 |
+
"epoch": 910.38,
|
| 244299 |
+
"learning_rate": 8.176575121163168e-06,
|
| 244300 |
+
"loss": 0.684,
|
| 244301 |
+
"step": 113915
|
| 244302 |
+
},
|
| 244303 |
+
{
|
| 244304 |
+
"epoch": 910.42,
|
| 244305 |
+
"learning_rate": 8.176494345718902e-06,
|
| 244306 |
+
"loss": 0.9405,
|
| 244307 |
+
"step": 113920
|
| 244308 |
+
},
|
| 244309 |
+
{
|
| 244310 |
+
"epoch": 910.46,
|
| 244311 |
+
"learning_rate": 8.176413570274638e-06,
|
| 244312 |
+
"loss": 0.3006,
|
| 244313 |
+
"step": 113925
|
| 244314 |
+
},
|
| 244315 |
+
{
|
| 244316 |
+
"epoch": 910.5,
|
| 244317 |
+
"learning_rate": 8.176332794830374e-06,
|
| 244318 |
+
"loss": 0.2658,
|
| 244319 |
+
"step": 113930
|
| 244320 |
+
},
|
| 244321 |
+
{
|
| 244322 |
+
"epoch": 910.54,
|
| 244323 |
+
"learning_rate": 8.176252019386108e-06,
|
| 244324 |
+
"loss": 0.36,
|
| 244325 |
+
"step": 113935
|
| 244326 |
+
},
|
| 244327 |
+
{
|
| 244328 |
+
"epoch": 910.58,
|
| 244329 |
+
"learning_rate": 8.176171243941844e-06,
|
| 244330 |
+
"loss": 0.7644,
|
| 244331 |
+
"step": 113940
|
| 244332 |
+
},
|
| 244333 |
+
{
|
| 244334 |
+
"epoch": 910.62,
|
| 244335 |
+
"learning_rate": 8.176090468497578e-06,
|
| 244336 |
+
"loss": 0.8812,
|
| 244337 |
+
"step": 113945
|
| 244338 |
+
},
|
| 244339 |
+
{
|
| 244340 |
+
"epoch": 910.66,
|
| 244341 |
+
"learning_rate": 8.176009693053314e-06,
|
| 244342 |
+
"loss": 0.2433,
|
| 244343 |
+
"step": 113950
|
| 244344 |
+
},
|
| 244345 |
+
{
|
| 244346 |
+
"epoch": 910.7,
|
| 244347 |
+
"learning_rate": 8.175928917609048e-06,
|
| 244348 |
+
"loss": 0.3455,
|
| 244349 |
+
"step": 113955
|
| 244350 |
+
},
|
| 244351 |
+
{
|
| 244352 |
+
"epoch": 910.74,
|
| 244353 |
+
"learning_rate": 8.175848142164784e-06,
|
| 244354 |
+
"loss": 0.3529,
|
| 244355 |
+
"step": 113960
|
| 244356 |
+
},
|
| 244357 |
+
{
|
| 244358 |
+
"epoch": 910.78,
|
| 244359 |
+
"learning_rate": 8.175767366720518e-06,
|
| 244360 |
+
"loss": 0.7352,
|
| 244361 |
+
"step": 113965
|
| 244362 |
+
},
|
| 244363 |
+
{
|
| 244364 |
+
"epoch": 910.82,
|
| 244365 |
+
"learning_rate": 8.175686591276254e-06,
|
| 244366 |
+
"loss": 0.9688,
|
| 244367 |
+
"step": 113970
|
| 244368 |
+
},
|
| 244369 |
+
{
|
| 244370 |
+
"epoch": 910.86,
|
| 244371 |
+
"learning_rate": 8.175605815831988e-06,
|
| 244372 |
+
"loss": 0.258,
|
| 244373 |
+
"step": 113975
|
| 244374 |
+
},
|
| 244375 |
+
{
|
| 244376 |
+
"epoch": 910.9,
|
| 244377 |
+
"learning_rate": 8.175525040387724e-06,
|
| 244378 |
+
"loss": 0.3258,
|
| 244379 |
+
"step": 113980
|
| 244380 |
+
},
|
| 244381 |
+
{
|
| 244382 |
+
"epoch": 910.94,
|
| 244383 |
+
"learning_rate": 8.175444264943458e-06,
|
| 244384 |
+
"loss": 0.352,
|
| 244385 |
+
"step": 113985
|
| 244386 |
+
},
|
| 244387 |
+
{
|
| 244388 |
+
"epoch": 910.98,
|
| 244389 |
+
"learning_rate": 8.175363489499193e-06,
|
| 244390 |
+
"loss": 0.7014,
|
| 244391 |
+
"step": 113990
|
| 244392 |
+
},
|
| 244393 |
+
{
|
| 244394 |
+
"epoch": 911.0,
|
| 244395 |
+
"eval_loss": 0.3746826946735382,
|
| 244396 |
+
"eval_runtime": 40.8605,
|
| 244397 |
+
"eval_samples_per_second": 20.46,
|
| 244398 |
+
"eval_steps_per_second": 0.661,
|
| 244399 |
+
"eval_wer": 0.18297995933778682,
|
| 244400 |
+
"step": 113992
|
| 244401 |
}
|
| 244402 |
],
|
| 244403 |
+
"max_steps": 625000,
|
| 244404 |
"num_train_epochs": 5000,
|
| 244405 |
+
"total_flos": 3.208038482122093e+20,
|
| 244406 |
"trial_name": null,
|
| 244407 |
"trial_params": null
|
| 244408 |
}
|
model-bin/finetune/base/{checkpoint-113370 → checkpoint-113992}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630085091.0178604/events.out.tfevents.1630085091.dea8124df033.1391.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:724fa55c26a31b6861c29464b4aa44e1153ca653da766200cd07b69901843734
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630085509.6253698/events.out.tfevents.1630085509.dea8124df033.1391.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e19ed8f3c9103b8c5207f08be039cd324847ae27fba4ec71b43af202f23eecda
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630085938.0146723/events.out.tfevents.1630085938.dea8124df033.1391.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6648f0f003792cbf7efb5a8b9943d6758b19291935312169c29dabc59e082d74
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630086359.258517/events.out.tfevents.1630086359.dea8124df033.1391.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bf37a00de48407cdbb55345e5c8f388ec1f64d0df3334d272a8bac197445a8ef
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630086783.2838607/events.out.tfevents.1630086783.dea8124df033.1391.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c61e3cea3725234da7bb2057146da95fae81d7e869f1e0696bdbae131acc1f8d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630085091.dea8124df033.1391.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bd6c54ecfce5fc0db85f947ad564150626be2eb0a47260e5df8d743898f17576
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630085509.dea8124df033.1391.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:58560bed31557c072a19c26fd8177a2df7bd1cb03a2f4fabaf39733f27b49500
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630085938.dea8124df033.1391.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6902e15a463f0d14113fe267221727da5c9630167708808ac6153c2c032d4f05
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630086359.dea8124df033.1391.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6bc157a93f6a0c1fbb99622f41cd7f704f39c3c0398f6a8cd5729338879cad51
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630086783.dea8124df033.1391.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c2a20f8962da390a8f29ada856c0d68c828b0ed1141d6c57fca8ba02669d1cf1
|
| 3 |
+
size 8622
|