"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629781779.4475987/events.out.tfevents.1629781779.c435e1c5ee04.920.51 +3 -0
- model-bin/finetune/base/log/1629782419.3571353/events.out.tfevents.1629782419.c435e1c5ee04.920.53 +3 -0
- model-bin/finetune/base/log/1629783054.1571443/events.out.tfevents.1629783054.c435e1c5ee04.920.55 +3 -0
- model-bin/finetune/base/log/1629783688.327236/events.out.tfevents.1629783688.c435e1c5ee04.920.57 +3 -0
- model-bin/finetune/base/log/1629784348.923546/events.out.tfevents.1629784348.c435e1c5ee04.920.59 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629781779.c435e1c5ee04.920.50 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629782419.c435e1c5ee04.920.52 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629783054.c435e1c5ee04.920.54 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629783688.c435e1c5ee04.920.56 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629784348.c435e1c5ee04.920.58 +3 -0
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/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 sha256:47556b2cec3463a6aa4269367dc245dd435f1863efc92d8dabf24b30bdc036ae
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/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 sha256:285882c24655340e01efed5bb5bf8eebe8bd8039085781e3d58bdb13b7291eec
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a2864ab6a170c024ff628fb561ae76b593148414870cfa7c0af4616b1a7eda9c
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/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 sha256:2d877dcfa2522ca7c375508f4cd4b04d7f59de6050f1a9c446b6a443ff36e7fc
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/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 sha256:1bebd2cd6e83993d92c9c725e6d02f8c57f184f96d8edaea1a482e44faea70b7
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -171876,11 +171876,806 @@
|
|
| 171876 |
"eval_steps_per_second": 0.666,
|
| 171877 |
"eval_wer": 0.1988261188554659,
|
| 171878 |
"step": 56994
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 171879 |
}
|
| 171880 |
],
|
| 171881 |
-
"max_steps":
|
| 171882 |
"num_train_epochs": 5000,
|
| 171883 |
-
"total_flos": 1.
|
| 171884 |
"trial_name": null,
|
| 171885 |
"trial_params": null
|
| 171886 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
+
"epoch": 463.99598393574297,
|
| 5 |
+
"global_step": 57616,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 171876 |
"eval_steps_per_second": 0.666,
|
| 171877 |
"eval_wer": 0.1988261188554659,
|
| 171878 |
"step": 56994
|
| 171879 |
+
},
|
| 171880 |
+
{
|
| 171881 |
+
"epoch": 459.01,
|
| 171882 |
+
"learning_rate": 9.10298076923077e-06,
|
| 171883 |
+
"loss": 0.4379,
|
| 171884 |
+
"step": 56995
|
| 171885 |
+
},
|
| 171886 |
+
{
|
| 171887 |
+
"epoch": 459.05,
|
| 171888 |
+
"learning_rate": 9.102900641025641e-06,
|
| 171889 |
+
"loss": 0.351,
|
| 171890 |
+
"step": 57000
|
| 171891 |
+
},
|
| 171892 |
+
{
|
| 171893 |
+
"epoch": 459.09,
|
| 171894 |
+
"learning_rate": 9.102820512820514e-06,
|
| 171895 |
+
"loss": 0.3083,
|
| 171896 |
+
"step": 57005
|
| 171897 |
+
},
|
| 171898 |
+
{
|
| 171899 |
+
"epoch": 459.13,
|
| 171900 |
+
"learning_rate": 9.102740384615386e-06,
|
| 171901 |
+
"loss": 0.3943,
|
| 171902 |
+
"step": 57010
|
| 171903 |
+
},
|
| 171904 |
+
{
|
| 171905 |
+
"epoch": 459.17,
|
| 171906 |
+
"learning_rate": 9.102660256410257e-06,
|
| 171907 |
+
"loss": 0.5645,
|
| 171908 |
+
"step": 57015
|
| 171909 |
+
},
|
| 171910 |
+
{
|
| 171911 |
+
"epoch": 459.21,
|
| 171912 |
+
"learning_rate": 9.102580128205128e-06,
|
| 171913 |
+
"loss": 0.9915,
|
| 171914 |
+
"step": 57020
|
| 171915 |
+
},
|
| 171916 |
+
{
|
| 171917 |
+
"epoch": 459.25,
|
| 171918 |
+
"learning_rate": 9.102500000000001e-06,
|
| 171919 |
+
"loss": 0.4694,
|
| 171920 |
+
"step": 57025
|
| 171921 |
+
},
|
| 171922 |
+
{
|
| 171923 |
+
"epoch": 459.29,
|
| 171924 |
+
"learning_rate": 9.102419871794873e-06,
|
| 171925 |
+
"loss": 0.2995,
|
| 171926 |
+
"step": 57030
|
| 171927 |
+
},
|
| 171928 |
+
{
|
| 171929 |
+
"epoch": 459.33,
|
| 171930 |
+
"learning_rate": 9.102339743589744e-06,
|
| 171931 |
+
"loss": 0.3401,
|
| 171932 |
+
"step": 57035
|
| 171933 |
+
},
|
| 171934 |
+
{
|
| 171935 |
+
"epoch": 459.37,
|
| 171936 |
+
"learning_rate": 9.102259615384617e-06,
|
| 171937 |
+
"loss": 0.6487,
|
| 171938 |
+
"step": 57040
|
| 171939 |
+
},
|
| 171940 |
+
{
|
| 171941 |
+
"epoch": 459.41,
|
| 171942 |
+
"learning_rate": 9.102179487179488e-06,
|
| 171943 |
+
"loss": 1.127,
|
| 171944 |
+
"step": 57045
|
| 171945 |
+
},
|
| 171946 |
+
{
|
| 171947 |
+
"epoch": 459.45,
|
| 171948 |
+
"learning_rate": 9.10209935897436e-06,
|
| 171949 |
+
"loss": 0.381,
|
| 171950 |
+
"step": 57050
|
| 171951 |
+
},
|
| 171952 |
+
{
|
| 171953 |
+
"epoch": 459.49,
|
| 171954 |
+
"learning_rate": 9.102019230769231e-06,
|
| 171955 |
+
"loss": 0.2994,
|
| 171956 |
+
"step": 57055
|
| 171957 |
+
},
|
| 171958 |
+
{
|
| 171959 |
+
"epoch": 459.53,
|
| 171960 |
+
"learning_rate": 9.101939102564104e-06,
|
| 171961 |
+
"loss": 0.4006,
|
| 171962 |
+
"step": 57060
|
| 171963 |
+
},
|
| 171964 |
+
{
|
| 171965 |
+
"epoch": 459.57,
|
| 171966 |
+
"learning_rate": 9.101858974358974e-06,
|
| 171967 |
+
"loss": 0.5035,
|
| 171968 |
+
"step": 57065
|
| 171969 |
+
},
|
| 171970 |
+
{
|
| 171971 |
+
"epoch": 459.61,
|
| 171972 |
+
"learning_rate": 9.101778846153847e-06,
|
| 171973 |
+
"loss": 1.0705,
|
| 171974 |
+
"step": 57070
|
| 171975 |
+
},
|
| 171976 |
+
{
|
| 171977 |
+
"epoch": 459.65,
|
| 171978 |
+
"learning_rate": 9.101698717948718e-06,
|
| 171979 |
+
"loss": 0.3249,
|
| 171980 |
+
"step": 57075
|
| 171981 |
+
},
|
| 171982 |
+
{
|
| 171983 |
+
"epoch": 459.69,
|
| 171984 |
+
"learning_rate": 9.10161858974359e-06,
|
| 171985 |
+
"loss": 0.3996,
|
| 171986 |
+
"step": 57080
|
| 171987 |
+
},
|
| 171988 |
+
{
|
| 171989 |
+
"epoch": 459.73,
|
| 171990 |
+
"learning_rate": 9.101538461538463e-06,
|
| 171991 |
+
"loss": 0.3953,
|
| 171992 |
+
"step": 57085
|
| 171993 |
+
},
|
| 171994 |
+
{
|
| 171995 |
+
"epoch": 459.77,
|
| 171996 |
+
"learning_rate": 9.101458333333334e-06,
|
| 171997 |
+
"loss": 0.6121,
|
| 171998 |
+
"step": 57090
|
| 171999 |
+
},
|
| 172000 |
+
{
|
| 172001 |
+
"epoch": 459.81,
|
| 172002 |
+
"learning_rate": 9.101378205128205e-06,
|
| 172003 |
+
"loss": 1.126,
|
| 172004 |
+
"step": 57095
|
| 172005 |
+
},
|
| 172006 |
+
{
|
| 172007 |
+
"epoch": 459.85,
|
| 172008 |
+
"learning_rate": 9.101298076923077e-06,
|
| 172009 |
+
"loss": 0.3517,
|
| 172010 |
+
"step": 57100
|
| 172011 |
+
},
|
| 172012 |
+
{
|
| 172013 |
+
"epoch": 459.89,
|
| 172014 |
+
"learning_rate": 9.10121794871795e-06,
|
| 172015 |
+
"loss": 0.3572,
|
| 172016 |
+
"step": 57105
|
| 172017 |
+
},
|
| 172018 |
+
{
|
| 172019 |
+
"epoch": 459.93,
|
| 172020 |
+
"learning_rate": 9.101137820512821e-06,
|
| 172021 |
+
"loss": 0.4526,
|
| 172022 |
+
"step": 57110
|
| 172023 |
+
},
|
| 172024 |
+
{
|
| 172025 |
+
"epoch": 459.97,
|
| 172026 |
+
"learning_rate": 9.101057692307693e-06,
|
| 172027 |
+
"loss": 0.5939,
|
| 172028 |
+
"step": 57115
|
| 172029 |
+
},
|
| 172030 |
+
{
|
| 172031 |
+
"epoch": 460.0,
|
| 172032 |
+
"eval_loss": 0.47473660111427307,
|
| 172033 |
+
"eval_runtime": 39.5768,
|
| 172034 |
+
"eval_samples_per_second": 21.225,
|
| 172035 |
+
"eval_steps_per_second": 0.682,
|
| 172036 |
+
"eval_wer": 0.19522174326002775,
|
| 172037 |
+
"step": 57118
|
| 172038 |
+
},
|
| 172039 |
+
{
|
| 172040 |
+
"epoch": 460.02,
|
| 172041 |
+
"learning_rate": 9.100977564102564e-06,
|
| 172042 |
+
"loss": 0.4356,
|
| 172043 |
+
"step": 57120
|
| 172044 |
+
},
|
| 172045 |
+
{
|
| 172046 |
+
"epoch": 460.06,
|
| 172047 |
+
"learning_rate": 9.100897435897437e-06,
|
| 172048 |
+
"loss": 0.3095,
|
| 172049 |
+
"step": 57125
|
| 172050 |
+
},
|
| 172051 |
+
{
|
| 172052 |
+
"epoch": 460.1,
|
| 172053 |
+
"learning_rate": 9.100817307692308e-06,
|
| 172054 |
+
"loss": 0.3426,
|
| 172055 |
+
"step": 57130
|
| 172056 |
+
},
|
| 172057 |
+
{
|
| 172058 |
+
"epoch": 460.14,
|
| 172059 |
+
"learning_rate": 9.10073717948718e-06,
|
| 172060 |
+
"loss": 0.3603,
|
| 172061 |
+
"step": 57135
|
| 172062 |
+
},
|
| 172063 |
+
{
|
| 172064 |
+
"epoch": 460.18,
|
| 172065 |
+
"learning_rate": 9.100657051282053e-06,
|
| 172066 |
+
"loss": 0.6415,
|
| 172067 |
+
"step": 57140
|
| 172068 |
+
},
|
| 172069 |
+
{
|
| 172070 |
+
"epoch": 460.22,
|
| 172071 |
+
"learning_rate": 9.100576923076924e-06,
|
| 172072 |
+
"loss": 1.1359,
|
| 172073 |
+
"step": 57145
|
| 172074 |
+
},
|
| 172075 |
+
{
|
| 172076 |
+
"epoch": 460.26,
|
| 172077 |
+
"learning_rate": 9.100496794871795e-06,
|
| 172078 |
+
"loss": 0.4338,
|
| 172079 |
+
"step": 57150
|
| 172080 |
+
},
|
| 172081 |
+
{
|
| 172082 |
+
"epoch": 460.3,
|
| 172083 |
+
"learning_rate": 9.100416666666667e-06,
|
| 172084 |
+
"loss": 0.3151,
|
| 172085 |
+
"step": 57155
|
| 172086 |
+
},
|
| 172087 |
+
{
|
| 172088 |
+
"epoch": 460.34,
|
| 172089 |
+
"learning_rate": 9.10033653846154e-06,
|
| 172090 |
+
"loss": 0.4653,
|
| 172091 |
+
"step": 57160
|
| 172092 |
+
},
|
| 172093 |
+
{
|
| 172094 |
+
"epoch": 460.38,
|
| 172095 |
+
"learning_rate": 9.100256410256411e-06,
|
| 172096 |
+
"loss": 0.8546,
|
| 172097 |
+
"step": 57165
|
| 172098 |
+
},
|
| 172099 |
+
{
|
| 172100 |
+
"epoch": 460.42,
|
| 172101 |
+
"learning_rate": 9.100176282051283e-06,
|
| 172102 |
+
"loss": 1.2716,
|
| 172103 |
+
"step": 57170
|
| 172104 |
+
},
|
| 172105 |
+
{
|
| 172106 |
+
"epoch": 460.46,
|
| 172107 |
+
"learning_rate": 9.100096153846154e-06,
|
| 172108 |
+
"loss": 0.3919,
|
| 172109 |
+
"step": 57175
|
| 172110 |
+
},
|
| 172111 |
+
{
|
| 172112 |
+
"epoch": 460.5,
|
| 172113 |
+
"learning_rate": 9.100016025641027e-06,
|
| 172114 |
+
"loss": 0.3562,
|
| 172115 |
+
"step": 57180
|
| 172116 |
+
},
|
| 172117 |
+
{
|
| 172118 |
+
"epoch": 460.54,
|
| 172119 |
+
"learning_rate": 9.099935897435898e-06,
|
| 172120 |
+
"loss": 0.4361,
|
| 172121 |
+
"step": 57185
|
| 172122 |
+
},
|
| 172123 |
+
{
|
| 172124 |
+
"epoch": 460.58,
|
| 172125 |
+
"learning_rate": 9.09985576923077e-06,
|
| 172126 |
+
"loss": 0.7023,
|
| 172127 |
+
"step": 57190
|
| 172128 |
+
},
|
| 172129 |
+
{
|
| 172130 |
+
"epoch": 460.62,
|
| 172131 |
+
"learning_rate": 9.099775641025643e-06,
|
| 172132 |
+
"loss": 1.0559,
|
| 172133 |
+
"step": 57195
|
| 172134 |
+
},
|
| 172135 |
+
{
|
| 172136 |
+
"epoch": 460.66,
|
| 172137 |
+
"learning_rate": 9.099695512820514e-06,
|
| 172138 |
+
"loss": 0.2995,
|
| 172139 |
+
"step": 57200
|
| 172140 |
+
},
|
| 172141 |
+
{
|
| 172142 |
+
"epoch": 460.7,
|
| 172143 |
+
"learning_rate": 9.099615384615386e-06,
|
| 172144 |
+
"loss": 0.3347,
|
| 172145 |
+
"step": 57205
|
| 172146 |
+
},
|
| 172147 |
+
{
|
| 172148 |
+
"epoch": 460.74,
|
| 172149 |
+
"learning_rate": 9.099535256410257e-06,
|
| 172150 |
+
"loss": 0.3746,
|
| 172151 |
+
"step": 57210
|
| 172152 |
+
},
|
| 172153 |
+
{
|
| 172154 |
+
"epoch": 460.78,
|
| 172155 |
+
"learning_rate": 9.09945512820513e-06,
|
| 172156 |
+
"loss": 0.7339,
|
| 172157 |
+
"step": 57215
|
| 172158 |
+
},
|
| 172159 |
+
{
|
| 172160 |
+
"epoch": 460.82,
|
| 172161 |
+
"learning_rate": 9.099375e-06,
|
| 172162 |
+
"loss": 1.0458,
|
| 172163 |
+
"step": 57220
|
| 172164 |
+
},
|
| 172165 |
+
{
|
| 172166 |
+
"epoch": 460.86,
|
| 172167 |
+
"learning_rate": 9.099294871794873e-06,
|
| 172168 |
+
"loss": 0.2959,
|
| 172169 |
+
"step": 57225
|
| 172170 |
+
},
|
| 172171 |
+
{
|
| 172172 |
+
"epoch": 460.9,
|
| 172173 |
+
"learning_rate": 9.099214743589746e-06,
|
| 172174 |
+
"loss": 0.3632,
|
| 172175 |
+
"step": 57230
|
| 172176 |
+
},
|
| 172177 |
+
{
|
| 172178 |
+
"epoch": 460.94,
|
| 172179 |
+
"learning_rate": 9.099134615384615e-06,
|
| 172180 |
+
"loss": 0.4225,
|
| 172181 |
+
"step": 57235
|
| 172182 |
+
},
|
| 172183 |
+
{
|
| 172184 |
+
"epoch": 460.98,
|
| 172185 |
+
"learning_rate": 9.099054487179488e-06,
|
| 172186 |
+
"loss": 0.69,
|
| 172187 |
+
"step": 57240
|
| 172188 |
+
},
|
| 172189 |
+
{
|
| 172190 |
+
"epoch": 461.0,
|
| 172191 |
+
"eval_loss": 0.40402549505233765,
|
| 172192 |
+
"eval_runtime": 38.9779,
|
| 172193 |
+
"eval_samples_per_second": 21.551,
|
| 172194 |
+
"eval_steps_per_second": 0.693,
|
| 172195 |
+
"eval_wer": 0.19619686800894853,
|
| 172196 |
+
"step": 57242
|
| 172197 |
+
},
|
| 172198 |
+
{
|
| 172199 |
+
"epoch": 457.02,
|
| 172200 |
+
"learning_rate": 9.09897435897436e-06,
|
| 172201 |
+
"loss": 0.341,
|
| 172202 |
+
"step": 57245
|
| 172203 |
+
},
|
| 172204 |
+
{
|
| 172205 |
+
"epoch": 457.06,
|
| 172206 |
+
"learning_rate": 9.098894230769231e-06,
|
| 172207 |
+
"loss": 0.4011,
|
| 172208 |
+
"step": 57250
|
| 172209 |
+
},
|
| 172210 |
+
{
|
| 172211 |
+
"epoch": 457.1,
|
| 172212 |
+
"learning_rate": 9.098814102564102e-06,
|
| 172213 |
+
"loss": 0.3644,
|
| 172214 |
+
"step": 57255
|
| 172215 |
+
},
|
| 172216 |
+
{
|
| 172217 |
+
"epoch": 457.14,
|
| 172218 |
+
"learning_rate": 9.098733974358976e-06,
|
| 172219 |
+
"loss": 0.5498,
|
| 172220 |
+
"step": 57260
|
| 172221 |
+
},
|
| 172222 |
+
{
|
| 172223 |
+
"epoch": 457.18,
|
| 172224 |
+
"learning_rate": 9.098653846153847e-06,
|
| 172225 |
+
"loss": 0.8031,
|
| 172226 |
+
"step": 57265
|
| 172227 |
+
},
|
| 172228 |
+
{
|
| 172229 |
+
"epoch": 457.22,
|
| 172230 |
+
"learning_rate": 9.098573717948718e-06,
|
| 172231 |
+
"loss": 0.9761,
|
| 172232 |
+
"step": 57270
|
| 172233 |
+
},
|
| 172234 |
+
{
|
| 172235 |
+
"epoch": 457.26,
|
| 172236 |
+
"learning_rate": 9.09849358974359e-06,
|
| 172237 |
+
"loss": 0.3063,
|
| 172238 |
+
"step": 57275
|
| 172239 |
+
},
|
| 172240 |
+
{
|
| 172241 |
+
"epoch": 457.3,
|
| 172242 |
+
"learning_rate": 9.098413461538463e-06,
|
| 172243 |
+
"loss": 0.3693,
|
| 172244 |
+
"step": 57280
|
| 172245 |
+
},
|
| 172246 |
+
{
|
| 172247 |
+
"epoch": 457.34,
|
| 172248 |
+
"learning_rate": 9.098333333333334e-06,
|
| 172249 |
+
"loss": 0.3899,
|
| 172250 |
+
"step": 57285
|
| 172251 |
+
},
|
| 172252 |
+
{
|
| 172253 |
+
"epoch": 457.38,
|
| 172254 |
+
"learning_rate": 9.098253205128205e-06,
|
| 172255 |
+
"loss": 0.8442,
|
| 172256 |
+
"step": 57290
|
| 172257 |
+
},
|
| 172258 |
+
{
|
| 172259 |
+
"epoch": 457.42,
|
| 172260 |
+
"learning_rate": 9.098173076923078e-06,
|
| 172261 |
+
"loss": 0.9218,
|
| 172262 |
+
"step": 57295
|
| 172263 |
+
},
|
| 172264 |
+
{
|
| 172265 |
+
"epoch": 457.46,
|
| 172266 |
+
"learning_rate": 9.09809294871795e-06,
|
| 172267 |
+
"loss": 0.3124,
|
| 172268 |
+
"step": 57300
|
| 172269 |
+
},
|
| 172270 |
+
{
|
| 172271 |
+
"epoch": 457.5,
|
| 172272 |
+
"learning_rate": 9.098012820512821e-06,
|
| 172273 |
+
"loss": 0.3694,
|
| 172274 |
+
"step": 57305
|
| 172275 |
+
},
|
| 172276 |
+
{
|
| 172277 |
+
"epoch": 457.54,
|
| 172278 |
+
"learning_rate": 9.097932692307693e-06,
|
| 172279 |
+
"loss": 0.4006,
|
| 172280 |
+
"step": 57310
|
| 172281 |
+
},
|
| 172282 |
+
{
|
| 172283 |
+
"epoch": 457.58,
|
| 172284 |
+
"learning_rate": 9.097852564102566e-06,
|
| 172285 |
+
"loss": 0.781,
|
| 172286 |
+
"step": 57315
|
| 172287 |
+
},
|
| 172288 |
+
{
|
| 172289 |
+
"epoch": 457.62,
|
| 172290 |
+
"learning_rate": 9.097772435897437e-06,
|
| 172291 |
+
"loss": 0.9623,
|
| 172292 |
+
"step": 57320
|
| 172293 |
+
},
|
| 172294 |
+
{
|
| 172295 |
+
"epoch": 457.66,
|
| 172296 |
+
"learning_rate": 9.097692307692308e-06,
|
| 172297 |
+
"loss": 0.3865,
|
| 172298 |
+
"step": 57325
|
| 172299 |
+
},
|
| 172300 |
+
{
|
| 172301 |
+
"epoch": 457.7,
|
| 172302 |
+
"learning_rate": 9.097612179487181e-06,
|
| 172303 |
+
"loss": 0.3595,
|
| 172304 |
+
"step": 57330
|
| 172305 |
+
},
|
| 172306 |
+
{
|
| 172307 |
+
"epoch": 457.74,
|
| 172308 |
+
"learning_rate": 9.097532051282053e-06,
|
| 172309 |
+
"loss": 0.4494,
|
| 172310 |
+
"step": 57335
|
| 172311 |
+
},
|
| 172312 |
+
{
|
| 172313 |
+
"epoch": 457.78,
|
| 172314 |
+
"learning_rate": 9.097451923076924e-06,
|
| 172315 |
+
"loss": 0.7866,
|
| 172316 |
+
"step": 57340
|
| 172317 |
+
},
|
| 172318 |
+
{
|
| 172319 |
+
"epoch": 457.82,
|
| 172320 |
+
"learning_rate": 9.097371794871795e-06,
|
| 172321 |
+
"loss": 0.9645,
|
| 172322 |
+
"step": 57345
|
| 172323 |
+
},
|
| 172324 |
+
{
|
| 172325 |
+
"epoch": 457.86,
|
| 172326 |
+
"learning_rate": 9.097291666666668e-06,
|
| 172327 |
+
"loss": 0.3257,
|
| 172328 |
+
"step": 57350
|
| 172329 |
+
},
|
| 172330 |
+
{
|
| 172331 |
+
"epoch": 457.9,
|
| 172332 |
+
"learning_rate": 9.097211538461538e-06,
|
| 172333 |
+
"loss": 0.6479,
|
| 172334 |
+
"step": 57355
|
| 172335 |
+
},
|
| 172336 |
+
{
|
| 172337 |
+
"epoch": 457.94,
|
| 172338 |
+
"learning_rate": 9.097131410256411e-06,
|
| 172339 |
+
"loss": 0.4004,
|
| 172340 |
+
"step": 57360
|
| 172341 |
+
},
|
| 172342 |
+
{
|
| 172343 |
+
"epoch": 457.98,
|
| 172344 |
+
"learning_rate": 9.097051282051283e-06,
|
| 172345 |
+
"loss": 0.8598,
|
| 172346 |
+
"step": 57365
|
| 172347 |
+
},
|
| 172348 |
+
{
|
| 172349 |
+
"epoch": 458.0,
|
| 172350 |
+
"eval_loss": 0.4067958891391754,
|
| 172351 |
+
"eval_runtime": 39.9405,
|
| 172352 |
+
"eval_samples_per_second": 21.031,
|
| 172353 |
+
"eval_steps_per_second": 0.676,
|
| 172354 |
+
"eval_wer": 0.2078081106389109,
|
| 172355 |
+
"step": 57367
|
| 172356 |
+
},
|
| 172357 |
+
{
|
| 172358 |
+
"epoch": 458.02,
|
| 172359 |
+
"learning_rate": 9.096971153846154e-06,
|
| 172360 |
+
"loss": 0.3758,
|
| 172361 |
+
"step": 57370
|
| 172362 |
+
},
|
| 172363 |
+
{
|
| 172364 |
+
"epoch": 458.06,
|
| 172365 |
+
"learning_rate": 9.096891025641025e-06,
|
| 172366 |
+
"loss": 0.3565,
|
| 172367 |
+
"step": 57375
|
| 172368 |
+
},
|
| 172369 |
+
{
|
| 172370 |
+
"epoch": 458.1,
|
| 172371 |
+
"learning_rate": 9.096810897435898e-06,
|
| 172372 |
+
"loss": 0.3734,
|
| 172373 |
+
"step": 57380
|
| 172374 |
+
},
|
| 172375 |
+
{
|
| 172376 |
+
"epoch": 458.14,
|
| 172377 |
+
"learning_rate": 9.09673076923077e-06,
|
| 172378 |
+
"loss": 0.3895,
|
| 172379 |
+
"step": 57385
|
| 172380 |
+
},
|
| 172381 |
+
{
|
| 172382 |
+
"epoch": 458.18,
|
| 172383 |
+
"learning_rate": 9.096650641025641e-06,
|
| 172384 |
+
"loss": 0.7967,
|
| 172385 |
+
"step": 57390
|
| 172386 |
+
},
|
| 172387 |
+
{
|
| 172388 |
+
"epoch": 458.22,
|
| 172389 |
+
"learning_rate": 9.096570512820514e-06,
|
| 172390 |
+
"loss": 0.9086,
|
| 172391 |
+
"step": 57395
|
| 172392 |
+
},
|
| 172393 |
+
{
|
| 172394 |
+
"epoch": 458.26,
|
| 172395 |
+
"learning_rate": 9.096490384615385e-06,
|
| 172396 |
+
"loss": 0.364,
|
| 172397 |
+
"step": 57400
|
| 172398 |
+
},
|
| 172399 |
+
{
|
| 172400 |
+
"epoch": 458.3,
|
| 172401 |
+
"learning_rate": 9.096410256410257e-06,
|
| 172402 |
+
"loss": 0.299,
|
| 172403 |
+
"step": 57405
|
| 172404 |
+
},
|
| 172405 |
+
{
|
| 172406 |
+
"epoch": 458.34,
|
| 172407 |
+
"learning_rate": 9.096330128205128e-06,
|
| 172408 |
+
"loss": 0.4314,
|
| 172409 |
+
"step": 57410
|
| 172410 |
+
},
|
| 172411 |
+
{
|
| 172412 |
+
"epoch": 458.38,
|
| 172413 |
+
"learning_rate": 9.096250000000001e-06,
|
| 172414 |
+
"loss": 0.8097,
|
| 172415 |
+
"step": 57415
|
| 172416 |
+
},
|
| 172417 |
+
{
|
| 172418 |
+
"epoch": 458.42,
|
| 172419 |
+
"learning_rate": 9.096169871794873e-06,
|
| 172420 |
+
"loss": 1.0123,
|
| 172421 |
+
"step": 57420
|
| 172422 |
+
},
|
| 172423 |
+
{
|
| 172424 |
+
"epoch": 458.46,
|
| 172425 |
+
"learning_rate": 9.096089743589744e-06,
|
| 172426 |
+
"loss": 0.3041,
|
| 172427 |
+
"step": 57425
|
| 172428 |
+
},
|
| 172429 |
+
{
|
| 172430 |
+
"epoch": 458.5,
|
| 172431 |
+
"learning_rate": 9.096009615384615e-06,
|
| 172432 |
+
"loss": 0.3947,
|
| 172433 |
+
"step": 57430
|
| 172434 |
+
},
|
| 172435 |
+
{
|
| 172436 |
+
"epoch": 458.54,
|
| 172437 |
+
"learning_rate": 9.095929487179488e-06,
|
| 172438 |
+
"loss": 0.4739,
|
| 172439 |
+
"step": 57435
|
| 172440 |
+
},
|
| 172441 |
+
{
|
| 172442 |
+
"epoch": 458.58,
|
| 172443 |
+
"learning_rate": 9.09584935897436e-06,
|
| 172444 |
+
"loss": 0.7613,
|
| 172445 |
+
"step": 57440
|
| 172446 |
+
},
|
| 172447 |
+
{
|
| 172448 |
+
"epoch": 458.62,
|
| 172449 |
+
"learning_rate": 9.095769230769231e-06,
|
| 172450 |
+
"loss": 0.9429,
|
| 172451 |
+
"step": 57445
|
| 172452 |
+
},
|
| 172453 |
+
{
|
| 172454 |
+
"epoch": 458.66,
|
| 172455 |
+
"learning_rate": 9.095689102564104e-06,
|
| 172456 |
+
"loss": 0.355,
|
| 172457 |
+
"step": 57450
|
| 172458 |
+
},
|
| 172459 |
+
{
|
| 172460 |
+
"epoch": 458.7,
|
| 172461 |
+
"learning_rate": 9.095608974358975e-06,
|
| 172462 |
+
"loss": 0.3264,
|
| 172463 |
+
"step": 57455
|
| 172464 |
+
},
|
| 172465 |
+
{
|
| 172466 |
+
"epoch": 458.74,
|
| 172467 |
+
"learning_rate": 9.095528846153847e-06,
|
| 172468 |
+
"loss": 0.4523,
|
| 172469 |
+
"step": 57460
|
| 172470 |
+
},
|
| 172471 |
+
{
|
| 172472 |
+
"epoch": 458.78,
|
| 172473 |
+
"learning_rate": 9.095448717948718e-06,
|
| 172474 |
+
"loss": 0.7704,
|
| 172475 |
+
"step": 57465
|
| 172476 |
+
},
|
| 172477 |
+
{
|
| 172478 |
+
"epoch": 458.82,
|
| 172479 |
+
"learning_rate": 9.095368589743591e-06,
|
| 172480 |
+
"loss": 0.9098,
|
| 172481 |
+
"step": 57470
|
| 172482 |
+
},
|
| 172483 |
+
{
|
| 172484 |
+
"epoch": 458.86,
|
| 172485 |
+
"learning_rate": 9.095288461538463e-06,
|
| 172486 |
+
"loss": 0.3971,
|
| 172487 |
+
"step": 57475
|
| 172488 |
+
},
|
| 172489 |
+
{
|
| 172490 |
+
"epoch": 458.9,
|
| 172491 |
+
"learning_rate": 9.095208333333334e-06,
|
| 172492 |
+
"loss": 0.3481,
|
| 172493 |
+
"step": 57480
|
| 172494 |
+
},
|
| 172495 |
+
{
|
| 172496 |
+
"epoch": 458.94,
|
| 172497 |
+
"learning_rate": 9.095128205128207e-06,
|
| 172498 |
+
"loss": 0.4587,
|
| 172499 |
+
"step": 57485
|
| 172500 |
+
},
|
| 172501 |
+
{
|
| 172502 |
+
"epoch": 458.98,
|
| 172503 |
+
"learning_rate": 9.095048076923078e-06,
|
| 172504 |
+
"loss": 0.787,
|
| 172505 |
+
"step": 57490
|
| 172506 |
+
},
|
| 172507 |
+
{
|
| 172508 |
+
"epoch": 459.0,
|
| 172509 |
+
"eval_loss": 0.39386695623397827,
|
| 172510 |
+
"eval_runtime": 39.4488,
|
| 172511 |
+
"eval_samples_per_second": 21.293,
|
| 172512 |
+
"eval_steps_per_second": 0.684,
|
| 172513 |
+
"eval_wer": 0.18808523188085233,
|
| 172514 |
+
"step": 57492
|
| 172515 |
+
},
|
| 172516 |
+
{
|
| 172517 |
+
"epoch": 463.02,
|
| 172518 |
+
"learning_rate": 9.09496794871795e-06,
|
| 172519 |
+
"loss": 0.4252,
|
| 172520 |
+
"step": 57495
|
| 172521 |
+
},
|
| 172522 |
+
{
|
| 172523 |
+
"epoch": 463.06,
|
| 172524 |
+
"learning_rate": 9.094887820512821e-06,
|
| 172525 |
+
"loss": 0.3075,
|
| 172526 |
+
"step": 57500
|
| 172527 |
+
},
|
| 172528 |
+
{
|
| 172529 |
+
"epoch": 463.1,
|
| 172530 |
+
"learning_rate": 9.094807692307694e-06,
|
| 172531 |
+
"loss": 0.3454,
|
| 172532 |
+
"step": 57505
|
| 172533 |
+
},
|
| 172534 |
+
{
|
| 172535 |
+
"epoch": 463.14,
|
| 172536 |
+
"learning_rate": 9.094727564102564e-06,
|
| 172537 |
+
"loss": 0.3353,
|
| 172538 |
+
"step": 57510
|
| 172539 |
+
},
|
| 172540 |
+
{
|
| 172541 |
+
"epoch": 463.18,
|
| 172542 |
+
"learning_rate": 9.094647435897437e-06,
|
| 172543 |
+
"loss": 0.7697,
|
| 172544 |
+
"step": 57515
|
| 172545 |
+
},
|
| 172546 |
+
{
|
| 172547 |
+
"epoch": 463.22,
|
| 172548 |
+
"learning_rate": 9.094567307692308e-06,
|
| 172549 |
+
"loss": 0.8605,
|
| 172550 |
+
"step": 57520
|
| 172551 |
+
},
|
| 172552 |
+
{
|
| 172553 |
+
"epoch": 463.27,
|
| 172554 |
+
"learning_rate": 9.09448717948718e-06,
|
| 172555 |
+
"loss": 0.4164,
|
| 172556 |
+
"step": 57525
|
| 172557 |
+
},
|
| 172558 |
+
{
|
| 172559 |
+
"epoch": 463.31,
|
| 172560 |
+
"learning_rate": 9.094407051282051e-06,
|
| 172561 |
+
"loss": 0.3367,
|
| 172562 |
+
"step": 57530
|
| 172563 |
+
},
|
| 172564 |
+
{
|
| 172565 |
+
"epoch": 463.35,
|
| 172566 |
+
"learning_rate": 9.094326923076924e-06,
|
| 172567 |
+
"loss": 0.4312,
|
| 172568 |
+
"step": 57535
|
| 172569 |
+
},
|
| 172570 |
+
{
|
| 172571 |
+
"epoch": 463.39,
|
| 172572 |
+
"learning_rate": 9.094246794871795e-06,
|
| 172573 |
+
"loss": 0.8302,
|
| 172574 |
+
"step": 57540
|
| 172575 |
+
},
|
| 172576 |
+
{
|
| 172577 |
+
"epoch": 463.43,
|
| 172578 |
+
"learning_rate": 9.094166666666667e-06,
|
| 172579 |
+
"loss": 0.9646,
|
| 172580 |
+
"step": 57545
|
| 172581 |
+
},
|
| 172582 |
+
{
|
| 172583 |
+
"epoch": 463.47,
|
| 172584 |
+
"learning_rate": 9.09408653846154e-06,
|
| 172585 |
+
"loss": 0.3868,
|
| 172586 |
+
"step": 57550
|
| 172587 |
+
},
|
| 172588 |
+
{
|
| 172589 |
+
"epoch": 463.51,
|
| 172590 |
+
"learning_rate": 9.094006410256411e-06,
|
| 172591 |
+
"loss": 0.3463,
|
| 172592 |
+
"step": 57555
|
| 172593 |
+
},
|
| 172594 |
+
{
|
| 172595 |
+
"epoch": 463.55,
|
| 172596 |
+
"learning_rate": 9.093926282051282e-06,
|
| 172597 |
+
"loss": 0.4389,
|
| 172598 |
+
"step": 57560
|
| 172599 |
+
},
|
| 172600 |
+
{
|
| 172601 |
+
"epoch": 463.59,
|
| 172602 |
+
"learning_rate": 9.093846153846154e-06,
|
| 172603 |
+
"loss": 0.8334,
|
| 172604 |
+
"step": 57565
|
| 172605 |
+
},
|
| 172606 |
+
{
|
| 172607 |
+
"epoch": 463.63,
|
| 172608 |
+
"learning_rate": 9.093766025641027e-06,
|
| 172609 |
+
"loss": 1.0065,
|
| 172610 |
+
"step": 57570
|
| 172611 |
+
},
|
| 172612 |
+
{
|
| 172613 |
+
"epoch": 463.67,
|
| 172614 |
+
"learning_rate": 9.093685897435898e-06,
|
| 172615 |
+
"loss": 0.3759,
|
| 172616 |
+
"step": 57575
|
| 172617 |
+
},
|
| 172618 |
+
{
|
| 172619 |
+
"epoch": 463.71,
|
| 172620 |
+
"learning_rate": 9.09360576923077e-06,
|
| 172621 |
+
"loss": 0.3655,
|
| 172622 |
+
"step": 57580
|
| 172623 |
+
},
|
| 172624 |
+
{
|
| 172625 |
+
"epoch": 463.75,
|
| 172626 |
+
"learning_rate": 9.093525641025643e-06,
|
| 172627 |
+
"loss": 0.3777,
|
| 172628 |
+
"step": 57585
|
| 172629 |
+
},
|
| 172630 |
+
{
|
| 172631 |
+
"epoch": 463.79,
|
| 172632 |
+
"learning_rate": 9.093445512820514e-06,
|
| 172633 |
+
"loss": 0.837,
|
| 172634 |
+
"step": 57590
|
| 172635 |
+
},
|
| 172636 |
+
{
|
| 172637 |
+
"epoch": 463.83,
|
| 172638 |
+
"learning_rate": 9.093365384615385e-06,
|
| 172639 |
+
"loss": 0.8211,
|
| 172640 |
+
"step": 57595
|
| 172641 |
+
},
|
| 172642 |
+
{
|
| 172643 |
+
"epoch": 463.87,
|
| 172644 |
+
"learning_rate": 9.093285256410257e-06,
|
| 172645 |
+
"loss": 0.3356,
|
| 172646 |
+
"step": 57600
|
| 172647 |
+
},
|
| 172648 |
+
{
|
| 172649 |
+
"epoch": 463.91,
|
| 172650 |
+
"learning_rate": 9.09320512820513e-06,
|
| 172651 |
+
"loss": 0.408,
|
| 172652 |
+
"step": 57605
|
| 172653 |
+
},
|
| 172654 |
+
{
|
| 172655 |
+
"epoch": 463.95,
|
| 172656 |
+
"learning_rate": 9.093125000000001e-06,
|
| 172657 |
+
"loss": 0.4394,
|
| 172658 |
+
"step": 57610
|
| 172659 |
+
},
|
| 172660 |
+
{
|
| 172661 |
+
"epoch": 463.99,
|
| 172662 |
+
"learning_rate": 9.093044871794872e-06,
|
| 172663 |
+
"loss": 0.9514,
|
| 172664 |
+
"step": 57615
|
| 172665 |
+
},
|
| 172666 |
+
{
|
| 172667 |
+
"epoch": 464.0,
|
| 172668 |
+
"eval_loss": 0.4048291742801666,
|
| 172669 |
+
"eval_runtime": 40.3898,
|
| 172670 |
+
"eval_samples_per_second": 20.797,
|
| 172671 |
+
"eval_steps_per_second": 0.668,
|
| 172672 |
+
"eval_wer": 0.19942978287886542,
|
| 172673 |
+
"step": 57616
|
| 172674 |
}
|
| 172675 |
],
|
| 172676 |
+
"max_steps": 620000,
|
| 172677 |
"num_train_epochs": 5000,
|
| 172678 |
+
"total_flos": 1.6214899429546325e+20,
|
| 172679 |
"trial_name": null,
|
| 172680 |
"trial_params": null
|
| 172681 |
}
|
model-bin/finetune/base/{checkpoint-56994 → checkpoint-57616}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629781779.4475987/events.out.tfevents.1629781779.c435e1c5ee04.920.51
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e4fa7ec7e7fc74d96a7680850aca9e47a52cbf04b371667efc567ead28769a8b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629782419.3571353/events.out.tfevents.1629782419.c435e1c5ee04.920.53
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dd3b72e68666f085c9d7aa31887c4568226dd7567a4d846bc9d63b782ed226e3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629783054.1571443/events.out.tfevents.1629783054.c435e1c5ee04.920.55
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:941dfe4a232cdf5cf8b6e2ed28613d17e86ef56fc9802127517ac9560f83ec25
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629783688.327236/events.out.tfevents.1629783688.c435e1c5ee04.920.57
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1ad6a9ff3c18357bd629de27ee1599648dcfcabd514c572ee371fcc4a729c846
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629784348.923546/events.out.tfevents.1629784348.c435e1c5ee04.920.59
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:22be810563e9f97e360b14bb795f6637bbef94085806470ef097d987b7c8b76c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629781779.c435e1c5ee04.920.50
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:185dd0c868c2090ff5110b59c4c07e107f54f49839b0792faa7601ac2a5d7040
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629782419.c435e1c5ee04.920.52
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de839938a1e7aa3b5a4d3a8895fcfdd63d4912277ed40943b38b2db577178fc6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629783054.c435e1c5ee04.920.54
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9d4680f084a260ef27acb963188e10e426a0462bae4cc7d776c797bd882e129
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629783688.c435e1c5ee04.920.56
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7b974bf34d146c9d049a412a31c00d639426cb2011c9c035cdf6b742351dc5a4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629784348.c435e1c5ee04.920.58
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5d8f93695ebacf9367acecc7dab20fa78e9f619884bc95755cc2a376f5ae981a
|
| 3 |
+
size 8622
|