"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630148780.943968/events.out.tfevents.1630148780.86bb0ddabf9b.4092.131 +3 -0
- model-bin/finetune/base/log/1630149168.480171/events.out.tfevents.1630149168.86bb0ddabf9b.4092.133 +3 -0
- model-bin/finetune/base/log/1630149553.0608108/events.out.tfevents.1630149553.86bb0ddabf9b.4092.135 +3 -0
- model-bin/finetune/base/log/1630149942.7762349/events.out.tfevents.1630149942.86bb0ddabf9b.4092.137 +3 -0
- model-bin/finetune/base/log/1630150333.0672135/events.out.tfevents.1630150333.86bb0ddabf9b.4092.139 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630148780.86bb0ddabf9b.4092.130 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630149168.86bb0ddabf9b.4092.132 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630149553.86bb0ddabf9b.4092.134 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630149942.86bb0ddabf9b.4092.136 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630150333.86bb0ddabf9b.4092.138 +3 -0
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/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 sha256:3845dc63451d0a1c44498d69f0b827119c42d6e557cf97a7f9735c52705644bd
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/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 sha256:69b6d80c964bf8059c1020726fed10f992c3b777c5f499f82bdede9d7ca4ced7
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/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 sha256:0dfcfe608f0a02dd951bb37028335198611a4d183fa2e67bc1b1aed79918029b
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/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 sha256:5f7595729f26a675659d761b684c35ca3b092bd8547ff8d0cfbbdd1d52fc1473
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/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:e4e46dd1d0e522e3ca1340174ec5347c27a555d6146d2d14b88e929a99348346
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -259761,11 +259761,800 @@
|
|
| 259761 |
"eval_steps_per_second": 0.739,
|
| 259762 |
"eval_wer": 0.18030087568295786,
|
| 259763 |
"step": 126068
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 259764 |
}
|
| 259765 |
],
|
| 259766 |
-
"max_steps":
|
| 259767 |
"num_train_epochs": 5000,
|
| 259768 |
-
"total_flos": 3.
|
| 259769 |
"trial_name": null,
|
| 259770 |
"trial_params": null
|
| 259771 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
+
"epoch": 1020.995983935743,
|
| 5 |
+
"global_step": 126689,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 259761 |
"eval_steps_per_second": 0.739,
|
| 259762 |
"eval_wer": 0.18030087568295786,
|
| 259763 |
"step": 126068
|
| 259764 |
+
},
|
| 259765 |
+
{
|
| 259766 |
+
"epoch": 1016.02,
|
| 259767 |
+
"learning_rate": 7.996458333333334e-06,
|
| 259768 |
+
"loss": 0.3939,
|
| 259769 |
+
"step": 126070
|
| 259770 |
+
},
|
| 259771 |
+
{
|
| 259772 |
+
"epoch": 1016.06,
|
| 259773 |
+
"learning_rate": 7.996378205128205e-06,
|
| 259774 |
+
"loss": 0.2661,
|
| 259775 |
+
"step": 126075
|
| 259776 |
+
},
|
| 259777 |
+
{
|
| 259778 |
+
"epoch": 1016.1,
|
| 259779 |
+
"learning_rate": 7.996298076923078e-06,
|
| 259780 |
+
"loss": 0.2543,
|
| 259781 |
+
"step": 126080
|
| 259782 |
+
},
|
| 259783 |
+
{
|
| 259784 |
+
"epoch": 1016.14,
|
| 259785 |
+
"learning_rate": 7.99621794871795e-06,
|
| 259786 |
+
"loss": 0.3273,
|
| 259787 |
+
"step": 126085
|
| 259788 |
+
},
|
| 259789 |
+
{
|
| 259790 |
+
"epoch": 1016.18,
|
| 259791 |
+
"learning_rate": 7.99613782051282e-06,
|
| 259792 |
+
"loss": 0.5625,
|
| 259793 |
+
"step": 126090
|
| 259794 |
+
},
|
| 259795 |
+
{
|
| 259796 |
+
"epoch": 1016.22,
|
| 259797 |
+
"learning_rate": 7.996057692307692e-06,
|
| 259798 |
+
"loss": 1.0146,
|
| 259799 |
+
"step": 126095
|
| 259800 |
+
},
|
| 259801 |
+
{
|
| 259802 |
+
"epoch": 1016.26,
|
| 259803 |
+
"learning_rate": 7.995977564102565e-06,
|
| 259804 |
+
"loss": 0.4505,
|
| 259805 |
+
"step": 126100
|
| 259806 |
+
},
|
| 259807 |
+
{
|
| 259808 |
+
"epoch": 1016.3,
|
| 259809 |
+
"learning_rate": 7.995897435897437e-06,
|
| 259810 |
+
"loss": 0.3086,
|
| 259811 |
+
"step": 126105
|
| 259812 |
+
},
|
| 259813 |
+
{
|
| 259814 |
+
"epoch": 1016.34,
|
| 259815 |
+
"learning_rate": 7.995817307692308e-06,
|
| 259816 |
+
"loss": 0.3847,
|
| 259817 |
+
"step": 126110
|
| 259818 |
+
},
|
| 259819 |
+
{
|
| 259820 |
+
"epoch": 1016.38,
|
| 259821 |
+
"learning_rate": 7.99573717948718e-06,
|
| 259822 |
+
"loss": 0.5578,
|
| 259823 |
+
"step": 126115
|
| 259824 |
+
},
|
| 259825 |
+
{
|
| 259826 |
+
"epoch": 1016.42,
|
| 259827 |
+
"learning_rate": 7.995657051282052e-06,
|
| 259828 |
+
"loss": 0.9584,
|
| 259829 |
+
"step": 126120
|
| 259830 |
+
},
|
| 259831 |
+
{
|
| 259832 |
+
"epoch": 1016.46,
|
| 259833 |
+
"learning_rate": 7.995576923076924e-06,
|
| 259834 |
+
"loss": 0.307,
|
| 259835 |
+
"step": 126125
|
| 259836 |
+
},
|
| 259837 |
+
{
|
| 259838 |
+
"epoch": 1016.5,
|
| 259839 |
+
"learning_rate": 7.995496794871795e-06,
|
| 259840 |
+
"loss": 0.267,
|
| 259841 |
+
"step": 126130
|
| 259842 |
+
},
|
| 259843 |
+
{
|
| 259844 |
+
"epoch": 1016.54,
|
| 259845 |
+
"learning_rate": 7.995416666666668e-06,
|
| 259846 |
+
"loss": 0.2988,
|
| 259847 |
+
"step": 126135
|
| 259848 |
+
},
|
| 259849 |
+
{
|
| 259850 |
+
"epoch": 1016.58,
|
| 259851 |
+
"learning_rate": 7.99533653846154e-06,
|
| 259852 |
+
"loss": 0.6387,
|
| 259853 |
+
"step": 126140
|
| 259854 |
+
},
|
| 259855 |
+
{
|
| 259856 |
+
"epoch": 1016.62,
|
| 259857 |
+
"learning_rate": 7.99525641025641e-06,
|
| 259858 |
+
"loss": 1.0086,
|
| 259859 |
+
"step": 126145
|
| 259860 |
+
},
|
| 259861 |
+
{
|
| 259862 |
+
"epoch": 1016.66,
|
| 259863 |
+
"learning_rate": 7.995176282051282e-06,
|
| 259864 |
+
"loss": 0.2825,
|
| 259865 |
+
"step": 126150
|
| 259866 |
+
},
|
| 259867 |
+
{
|
| 259868 |
+
"epoch": 1016.7,
|
| 259869 |
+
"learning_rate": 7.995096153846155e-06,
|
| 259870 |
+
"loss": 0.2735,
|
| 259871 |
+
"step": 126155
|
| 259872 |
+
},
|
| 259873 |
+
{
|
| 259874 |
+
"epoch": 1016.74,
|
| 259875 |
+
"learning_rate": 7.995016025641027e-06,
|
| 259876 |
+
"loss": 0.3921,
|
| 259877 |
+
"step": 126160
|
| 259878 |
+
},
|
| 259879 |
+
{
|
| 259880 |
+
"epoch": 1016.78,
|
| 259881 |
+
"learning_rate": 7.994935897435898e-06,
|
| 259882 |
+
"loss": 0.5328,
|
| 259883 |
+
"step": 126165
|
| 259884 |
+
},
|
| 259885 |
+
{
|
| 259886 |
+
"epoch": 1016.82,
|
| 259887 |
+
"learning_rate": 7.99485576923077e-06,
|
| 259888 |
+
"loss": 0.9327,
|
| 259889 |
+
"step": 126170
|
| 259890 |
+
},
|
| 259891 |
+
{
|
| 259892 |
+
"epoch": 1016.86,
|
| 259893 |
+
"learning_rate": 7.994775641025642e-06,
|
| 259894 |
+
"loss": 0.2889,
|
| 259895 |
+
"step": 126175
|
| 259896 |
+
},
|
| 259897 |
+
{
|
| 259898 |
+
"epoch": 1016.9,
|
| 259899 |
+
"learning_rate": 7.994695512820514e-06,
|
| 259900 |
+
"loss": 0.2814,
|
| 259901 |
+
"step": 126180
|
| 259902 |
+
},
|
| 259903 |
+
{
|
| 259904 |
+
"epoch": 1016.94,
|
| 259905 |
+
"learning_rate": 7.994615384615385e-06,
|
| 259906 |
+
"loss": 0.3506,
|
| 259907 |
+
"step": 126185
|
| 259908 |
+
},
|
| 259909 |
+
{
|
| 259910 |
+
"epoch": 1016.98,
|
| 259911 |
+
"learning_rate": 7.994535256410258e-06,
|
| 259912 |
+
"loss": 0.7309,
|
| 259913 |
+
"step": 126190
|
| 259914 |
+
},
|
| 259915 |
+
{
|
| 259916 |
+
"epoch": 1017.0,
|
| 259917 |
+
"eval_loss": 0.31492879986763,
|
| 259918 |
+
"eval_runtime": 36.0348,
|
| 259919 |
+
"eval_samples_per_second": 23.339,
|
| 259920 |
+
"eval_steps_per_second": 0.749,
|
| 259921 |
+
"eval_wer": 0.17529711375212223,
|
| 259922 |
+
"step": 126192
|
| 259923 |
+
},
|
| 259924 |
+
{
|
| 259925 |
+
"epoch": 1017.02,
|
| 259926 |
+
"learning_rate": 7.994455128205128e-06,
|
| 259927 |
+
"loss": 0.3466,
|
| 259928 |
+
"step": 126195
|
| 259929 |
+
},
|
| 259930 |
+
{
|
| 259931 |
+
"epoch": 1017.06,
|
| 259932 |
+
"learning_rate": 7.994375000000001e-06,
|
| 259933 |
+
"loss": 0.2535,
|
| 259934 |
+
"step": 126200
|
| 259935 |
+
},
|
| 259936 |
+
{
|
| 259937 |
+
"epoch": 1017.1,
|
| 259938 |
+
"learning_rate": 7.994294871794872e-06,
|
| 259939 |
+
"loss": 0.2802,
|
| 259940 |
+
"step": 126205
|
| 259941 |
+
},
|
| 259942 |
+
{
|
| 259943 |
+
"epoch": 1017.14,
|
| 259944 |
+
"learning_rate": 7.994214743589744e-06,
|
| 259945 |
+
"loss": 0.3687,
|
| 259946 |
+
"step": 126210
|
| 259947 |
+
},
|
| 259948 |
+
{
|
| 259949 |
+
"epoch": 1017.18,
|
| 259950 |
+
"learning_rate": 7.994134615384615e-06,
|
| 259951 |
+
"loss": 0.811,
|
| 259952 |
+
"step": 126215
|
| 259953 |
+
},
|
| 259954 |
+
{
|
| 259955 |
+
"epoch": 1017.22,
|
| 259956 |
+
"learning_rate": 7.994054487179488e-06,
|
| 259957 |
+
"loss": 0.7224,
|
| 259958 |
+
"step": 126220
|
| 259959 |
+
},
|
| 259960 |
+
{
|
| 259961 |
+
"epoch": 1017.27,
|
| 259962 |
+
"learning_rate": 7.99397435897436e-06,
|
| 259963 |
+
"loss": 0.2839,
|
| 259964 |
+
"step": 126225
|
| 259965 |
+
},
|
| 259966 |
+
{
|
| 259967 |
+
"epoch": 1017.31,
|
| 259968 |
+
"learning_rate": 7.99389423076923e-06,
|
| 259969 |
+
"loss": 0.2759,
|
| 259970 |
+
"step": 126230
|
| 259971 |
+
},
|
| 259972 |
+
{
|
| 259973 |
+
"epoch": 1017.35,
|
| 259974 |
+
"learning_rate": 7.993814102564104e-06,
|
| 259975 |
+
"loss": 0.4042,
|
| 259976 |
+
"step": 126235
|
| 259977 |
+
},
|
| 259978 |
+
{
|
| 259979 |
+
"epoch": 1017.39,
|
| 259980 |
+
"learning_rate": 7.993733974358975e-06,
|
| 259981 |
+
"loss": 0.7433,
|
| 259982 |
+
"step": 126240
|
| 259983 |
+
},
|
| 259984 |
+
{
|
| 259985 |
+
"epoch": 1017.43,
|
| 259986 |
+
"learning_rate": 7.993653846153846e-06,
|
| 259987 |
+
"loss": 0.84,
|
| 259988 |
+
"step": 126245
|
| 259989 |
+
},
|
| 259990 |
+
{
|
| 259991 |
+
"epoch": 1017.47,
|
| 259992 |
+
"learning_rate": 7.993573717948718e-06,
|
| 259993 |
+
"loss": 0.2865,
|
| 259994 |
+
"step": 126250
|
| 259995 |
+
},
|
| 259996 |
+
{
|
| 259997 |
+
"epoch": 1017.51,
|
| 259998 |
+
"learning_rate": 7.993493589743591e-06,
|
| 259999 |
+
"loss": 0.3096,
|
| 260000 |
+
"step": 126255
|
| 260001 |
+
},
|
| 260002 |
+
{
|
| 260003 |
+
"epoch": 1017.55,
|
| 260004 |
+
"learning_rate": 7.993413461538462e-06,
|
| 260005 |
+
"loss": 0.4599,
|
| 260006 |
+
"step": 126260
|
| 260007 |
+
},
|
| 260008 |
+
{
|
| 260009 |
+
"epoch": 1017.59,
|
| 260010 |
+
"learning_rate": 7.993333333333334e-06,
|
| 260011 |
+
"loss": 0.7373,
|
| 260012 |
+
"step": 126265
|
| 260013 |
+
},
|
| 260014 |
+
{
|
| 260015 |
+
"epoch": 1017.63,
|
| 260016 |
+
"learning_rate": 7.993253205128205e-06,
|
| 260017 |
+
"loss": 0.9646,
|
| 260018 |
+
"step": 126270
|
| 260019 |
+
},
|
| 260020 |
+
{
|
| 260021 |
+
"epoch": 1017.67,
|
| 260022 |
+
"learning_rate": 7.993173076923078e-06,
|
| 260023 |
+
"loss": 0.2937,
|
| 260024 |
+
"step": 126275
|
| 260025 |
+
},
|
| 260026 |
+
{
|
| 260027 |
+
"epoch": 1017.71,
|
| 260028 |
+
"learning_rate": 7.99309294871795e-06,
|
| 260029 |
+
"loss": 0.2543,
|
| 260030 |
+
"step": 126280
|
| 260031 |
+
},
|
| 260032 |
+
{
|
| 260033 |
+
"epoch": 1017.75,
|
| 260034 |
+
"learning_rate": 7.99301282051282e-06,
|
| 260035 |
+
"loss": 0.3509,
|
| 260036 |
+
"step": 126285
|
| 260037 |
+
},
|
| 260038 |
+
{
|
| 260039 |
+
"epoch": 1017.79,
|
| 260040 |
+
"learning_rate": 7.992932692307694e-06,
|
| 260041 |
+
"loss": 0.7394,
|
| 260042 |
+
"step": 126290
|
| 260043 |
+
},
|
| 260044 |
+
{
|
| 260045 |
+
"epoch": 1017.83,
|
| 260046 |
+
"learning_rate": 7.992852564102565e-06,
|
| 260047 |
+
"loss": 0.8648,
|
| 260048 |
+
"step": 126295
|
| 260049 |
+
},
|
| 260050 |
+
{
|
| 260051 |
+
"epoch": 1017.87,
|
| 260052 |
+
"learning_rate": 7.992772435897436e-06,
|
| 260053 |
+
"loss": 0.4316,
|
| 260054 |
+
"step": 126300
|
| 260055 |
+
},
|
| 260056 |
+
{
|
| 260057 |
+
"epoch": 1017.91,
|
| 260058 |
+
"learning_rate": 7.992692307692308e-06,
|
| 260059 |
+
"loss": 0.307,
|
| 260060 |
+
"step": 126305
|
| 260061 |
+
},
|
| 260062 |
+
{
|
| 260063 |
+
"epoch": 1017.95,
|
| 260064 |
+
"learning_rate": 7.992612179487181e-06,
|
| 260065 |
+
"loss": 0.4597,
|
| 260066 |
+
"step": 126310
|
| 260067 |
+
},
|
| 260068 |
+
{
|
| 260069 |
+
"epoch": 1017.99,
|
| 260070 |
+
"learning_rate": 7.992532051282052e-06,
|
| 260071 |
+
"loss": 0.8193,
|
| 260072 |
+
"step": 126315
|
| 260073 |
+
},
|
| 260074 |
+
{
|
| 260075 |
+
"epoch": 1018.0,
|
| 260076 |
+
"eval_loss": 0.4346490502357483,
|
| 260077 |
+
"eval_runtime": 35.2379,
|
| 260078 |
+
"eval_samples_per_second": 23.866,
|
| 260079 |
+
"eval_steps_per_second": 0.766,
|
| 260080 |
+
"eval_wer": 0.18682317304051496,
|
| 260081 |
+
"step": 126316
|
| 260082 |
+
},
|
| 260083 |
+
{
|
| 260084 |
+
"epoch": 1018.03,
|
| 260085 |
+
"learning_rate": 7.992451923076924e-06,
|
| 260086 |
+
"loss": 0.3603,
|
| 260087 |
+
"step": 126320
|
| 260088 |
+
},
|
| 260089 |
+
{
|
| 260090 |
+
"epoch": 1018.07,
|
| 260091 |
+
"learning_rate": 7.992371794871797e-06,
|
| 260092 |
+
"loss": 0.3009,
|
| 260093 |
+
"step": 126325
|
| 260094 |
+
},
|
| 260095 |
+
{
|
| 260096 |
+
"epoch": 1018.11,
|
| 260097 |
+
"learning_rate": 7.992291666666668e-06,
|
| 260098 |
+
"loss": 0.3143,
|
| 260099 |
+
"step": 126330
|
| 260100 |
+
},
|
| 260101 |
+
{
|
| 260102 |
+
"epoch": 1018.15,
|
| 260103 |
+
"learning_rate": 7.99221153846154e-06,
|
| 260104 |
+
"loss": 0.3757,
|
| 260105 |
+
"step": 126335
|
| 260106 |
+
},
|
| 260107 |
+
{
|
| 260108 |
+
"epoch": 1018.19,
|
| 260109 |
+
"learning_rate": 7.99213141025641e-06,
|
| 260110 |
+
"loss": 0.8593,
|
| 260111 |
+
"step": 126340
|
| 260112 |
+
},
|
| 260113 |
+
{
|
| 260114 |
+
"epoch": 1018.23,
|
| 260115 |
+
"learning_rate": 7.992051282051284e-06,
|
| 260116 |
+
"loss": 0.644,
|
| 260117 |
+
"step": 126345
|
| 260118 |
+
},
|
| 260119 |
+
{
|
| 260120 |
+
"epoch": 1018.27,
|
| 260121 |
+
"learning_rate": 7.991971153846153e-06,
|
| 260122 |
+
"loss": 0.377,
|
| 260123 |
+
"step": 126350
|
| 260124 |
+
},
|
| 260125 |
+
{
|
| 260126 |
+
"epoch": 1018.31,
|
| 260127 |
+
"learning_rate": 7.991891025641027e-06,
|
| 260128 |
+
"loss": 0.3072,
|
| 260129 |
+
"step": 126355
|
| 260130 |
+
},
|
| 260131 |
+
{
|
| 260132 |
+
"epoch": 1018.35,
|
| 260133 |
+
"learning_rate": 7.991810897435898e-06,
|
| 260134 |
+
"loss": 0.3897,
|
| 260135 |
+
"step": 126360
|
| 260136 |
+
},
|
| 260137 |
+
{
|
| 260138 |
+
"epoch": 1018.39,
|
| 260139 |
+
"learning_rate": 7.99173076923077e-06,
|
| 260140 |
+
"loss": 0.9711,
|
| 260141 |
+
"step": 126365
|
| 260142 |
+
},
|
| 260143 |
+
{
|
| 260144 |
+
"epoch": 1018.43,
|
| 260145 |
+
"learning_rate": 7.99165064102564e-06,
|
| 260146 |
+
"loss": 0.6407,
|
| 260147 |
+
"step": 126370
|
| 260148 |
+
},
|
| 260149 |
+
{
|
| 260150 |
+
"epoch": 1018.47,
|
| 260151 |
+
"learning_rate": 7.991570512820514e-06,
|
| 260152 |
+
"loss": 0.3422,
|
| 260153 |
+
"step": 126375
|
| 260154 |
+
},
|
| 260155 |
+
{
|
| 260156 |
+
"epoch": 1018.51,
|
| 260157 |
+
"learning_rate": 7.991490384615385e-06,
|
| 260158 |
+
"loss": 0.3255,
|
| 260159 |
+
"step": 126380
|
| 260160 |
+
},
|
| 260161 |
+
{
|
| 260162 |
+
"epoch": 1018.55,
|
| 260163 |
+
"learning_rate": 7.991410256410256e-06,
|
| 260164 |
+
"loss": 0.4262,
|
| 260165 |
+
"step": 126385
|
| 260166 |
+
},
|
| 260167 |
+
{
|
| 260168 |
+
"epoch": 1018.59,
|
| 260169 |
+
"learning_rate": 7.99133012820513e-06,
|
| 260170 |
+
"loss": 1.022,
|
| 260171 |
+
"step": 126390
|
| 260172 |
+
},
|
| 260173 |
+
{
|
| 260174 |
+
"epoch": 1018.63,
|
| 260175 |
+
"learning_rate": 7.99125e-06,
|
| 260176 |
+
"loss": 0.6429,
|
| 260177 |
+
"step": 126395
|
| 260178 |
+
},
|
| 260179 |
+
{
|
| 260180 |
+
"epoch": 1018.67,
|
| 260181 |
+
"learning_rate": 7.991169871794872e-06,
|
| 260182 |
+
"loss": 0.2113,
|
| 260183 |
+
"step": 126400
|
| 260184 |
+
},
|
| 260185 |
+
{
|
| 260186 |
+
"epoch": 1018.71,
|
| 260187 |
+
"learning_rate": 7.991089743589743e-06,
|
| 260188 |
+
"loss": 0.3158,
|
| 260189 |
+
"step": 126405
|
| 260190 |
+
},
|
| 260191 |
+
{
|
| 260192 |
+
"epoch": 1018.76,
|
| 260193 |
+
"learning_rate": 7.991009615384617e-06,
|
| 260194 |
+
"loss": 0.4313,
|
| 260195 |
+
"step": 126410
|
| 260196 |
+
},
|
| 260197 |
+
{
|
| 260198 |
+
"epoch": 1018.8,
|
| 260199 |
+
"learning_rate": 7.990929487179488e-06,
|
| 260200 |
+
"loss": 0.879,
|
| 260201 |
+
"step": 126415
|
| 260202 |
+
},
|
| 260203 |
+
{
|
| 260204 |
+
"epoch": 1018.84,
|
| 260205 |
+
"learning_rate": 7.99084935897436e-06,
|
| 260206 |
+
"loss": 0.7075,
|
| 260207 |
+
"step": 126420
|
| 260208 |
+
},
|
| 260209 |
+
{
|
| 260210 |
+
"epoch": 1018.88,
|
| 260211 |
+
"learning_rate": 7.990769230769232e-06,
|
| 260212 |
+
"loss": 0.262,
|
| 260213 |
+
"step": 126425
|
| 260214 |
+
},
|
| 260215 |
+
{
|
| 260216 |
+
"epoch": 1018.92,
|
| 260217 |
+
"learning_rate": 7.990689102564104e-06,
|
| 260218 |
+
"loss": 0.2746,
|
| 260219 |
+
"step": 126430
|
| 260220 |
+
},
|
| 260221 |
+
{
|
| 260222 |
+
"epoch": 1018.96,
|
| 260223 |
+
"learning_rate": 7.990608974358975e-06,
|
| 260224 |
+
"loss": 0.4399,
|
| 260225 |
+
"step": 126435
|
| 260226 |
+
},
|
| 260227 |
+
{
|
| 260228 |
+
"epoch": 1019.0,
|
| 260229 |
+
"learning_rate": 7.990528846153846e-06,
|
| 260230 |
+
"loss": 0.9204,
|
| 260231 |
+
"step": 126440
|
| 260232 |
+
},
|
| 260233 |
+
{
|
| 260234 |
+
"epoch": 1019.0,
|
| 260235 |
+
"eval_loss": 0.4004644453525543,
|
| 260236 |
+
"eval_runtime": 36.4199,
|
| 260237 |
+
"eval_samples_per_second": 23.092,
|
| 260238 |
+
"eval_steps_per_second": 0.741,
|
| 260239 |
+
"eval_wer": 0.1846421831905937,
|
| 260240 |
+
"step": 126440
|
| 260241 |
+
},
|
| 260242 |
+
{
|
| 260243 |
+
"epoch": 1011.04,
|
| 260244 |
+
"learning_rate": 7.99044871794872e-06,
|
| 260245 |
+
"loss": 0.3103,
|
| 260246 |
+
"step": 126445
|
| 260247 |
+
},
|
| 260248 |
+
{
|
| 260249 |
+
"epoch": 1011.08,
|
| 260250 |
+
"learning_rate": 7.99036858974359e-06,
|
| 260251 |
+
"loss": 0.2886,
|
| 260252 |
+
"step": 126450
|
| 260253 |
+
},
|
| 260254 |
+
{
|
| 260255 |
+
"epoch": 1011.12,
|
| 260256 |
+
"learning_rate": 7.990288461538462e-06,
|
| 260257 |
+
"loss": 0.3135,
|
| 260258 |
+
"step": 126455
|
| 260259 |
+
},
|
| 260260 |
+
{
|
| 260261 |
+
"epoch": 1011.16,
|
| 260262 |
+
"learning_rate": 7.990208333333334e-06,
|
| 260263 |
+
"loss": 0.5276,
|
| 260264 |
+
"step": 126460
|
| 260265 |
+
},
|
| 260266 |
+
{
|
| 260267 |
+
"epoch": 1011.2,
|
| 260268 |
+
"learning_rate": 7.990128205128207e-06,
|
| 260269 |
+
"loss": 1.2226,
|
| 260270 |
+
"step": 126465
|
| 260271 |
+
},
|
| 260272 |
+
{
|
| 260273 |
+
"epoch": 1011.24,
|
| 260274 |
+
"learning_rate": 7.990048076923076e-06,
|
| 260275 |
+
"loss": 0.3239,
|
| 260276 |
+
"step": 126470
|
| 260277 |
+
},
|
| 260278 |
+
{
|
| 260279 |
+
"epoch": 1011.28,
|
| 260280 |
+
"learning_rate": 7.98996794871795e-06,
|
| 260281 |
+
"loss": 0.301,
|
| 260282 |
+
"step": 126475
|
| 260283 |
+
},
|
| 260284 |
+
{
|
| 260285 |
+
"epoch": 1011.32,
|
| 260286 |
+
"learning_rate": 7.989887820512822e-06,
|
| 260287 |
+
"loss": 0.3185,
|
| 260288 |
+
"step": 126480
|
| 260289 |
+
},
|
| 260290 |
+
{
|
| 260291 |
+
"epoch": 1011.36,
|
| 260292 |
+
"learning_rate": 7.989807692307692e-06,
|
| 260293 |
+
"loss": 0.4219,
|
| 260294 |
+
"step": 126485
|
| 260295 |
+
},
|
| 260296 |
+
{
|
| 260297 |
+
"epoch": 1011.4,
|
| 260298 |
+
"learning_rate": 7.989727564102565e-06,
|
| 260299 |
+
"loss": 1.1383,
|
| 260300 |
+
"step": 126490
|
| 260301 |
+
},
|
| 260302 |
+
{
|
| 260303 |
+
"epoch": 1011.44,
|
| 260304 |
+
"learning_rate": 7.989647435897436e-06,
|
| 260305 |
+
"loss": 0.2869,
|
| 260306 |
+
"step": 126495
|
| 260307 |
+
},
|
| 260308 |
+
{
|
| 260309 |
+
"epoch": 1011.48,
|
| 260310 |
+
"learning_rate": 7.989567307692308e-06,
|
| 260311 |
+
"loss": 0.3177,
|
| 260312 |
+
"step": 126500
|
| 260313 |
+
},
|
| 260314 |
+
{
|
| 260315 |
+
"epoch": 1011.52,
|
| 260316 |
+
"learning_rate": 7.989487179487179e-06,
|
| 260317 |
+
"loss": 0.2936,
|
| 260318 |
+
"step": 126505
|
| 260319 |
+
},
|
| 260320 |
+
{
|
| 260321 |
+
"epoch": 1011.56,
|
| 260322 |
+
"learning_rate": 7.989407051282052e-06,
|
| 260323 |
+
"loss": 0.4247,
|
| 260324 |
+
"step": 126510
|
| 260325 |
+
},
|
| 260326 |
+
{
|
| 260327 |
+
"epoch": 1011.6,
|
| 260328 |
+
"learning_rate": 7.989326923076924e-06,
|
| 260329 |
+
"loss": 1.1716,
|
| 260330 |
+
"step": 126515
|
| 260331 |
+
},
|
| 260332 |
+
{
|
| 260333 |
+
"epoch": 1011.64,
|
| 260334 |
+
"learning_rate": 7.989246794871795e-06,
|
| 260335 |
+
"loss": 0.392,
|
| 260336 |
+
"step": 126520
|
| 260337 |
+
},
|
| 260338 |
+
{
|
| 260339 |
+
"epoch": 1011.68,
|
| 260340 |
+
"learning_rate": 7.989166666666668e-06,
|
| 260341 |
+
"loss": 0.2561,
|
| 260342 |
+
"step": 126525
|
| 260343 |
+
},
|
| 260344 |
+
{
|
| 260345 |
+
"epoch": 1011.72,
|
| 260346 |
+
"learning_rate": 7.98908653846154e-06,
|
| 260347 |
+
"loss": 0.2848,
|
| 260348 |
+
"step": 126530
|
| 260349 |
+
},
|
| 260350 |
+
{
|
| 260351 |
+
"epoch": 1011.76,
|
| 260352 |
+
"learning_rate": 7.98900641025641e-06,
|
| 260353 |
+
"loss": 0.5091,
|
| 260354 |
+
"step": 126535
|
| 260355 |
+
},
|
| 260356 |
+
{
|
| 260357 |
+
"epoch": 1011.8,
|
| 260358 |
+
"learning_rate": 7.988926282051282e-06,
|
| 260359 |
+
"loss": 1.127,
|
| 260360 |
+
"step": 126540
|
| 260361 |
+
},
|
| 260362 |
+
{
|
| 260363 |
+
"epoch": 1011.84,
|
| 260364 |
+
"learning_rate": 7.988846153846155e-06,
|
| 260365 |
+
"loss": 0.2767,
|
| 260366 |
+
"step": 126545
|
| 260367 |
+
},
|
| 260368 |
+
{
|
| 260369 |
+
"epoch": 1011.88,
|
| 260370 |
+
"learning_rate": 7.988766025641026e-06,
|
| 260371 |
+
"loss": 0.265,
|
| 260372 |
+
"step": 126550
|
| 260373 |
+
},
|
| 260374 |
+
{
|
| 260375 |
+
"epoch": 1011.92,
|
| 260376 |
+
"learning_rate": 7.988685897435898e-06,
|
| 260377 |
+
"loss": 0.2979,
|
| 260378 |
+
"step": 126555
|
| 260379 |
+
},
|
| 260380 |
+
{
|
| 260381 |
+
"epoch": 1011.96,
|
| 260382 |
+
"learning_rate": 7.98860576923077e-06,
|
| 260383 |
+
"loss": 0.3722,
|
| 260384 |
+
"step": 126560
|
| 260385 |
+
},
|
| 260386 |
+
{
|
| 260387 |
+
"epoch": 1012.0,
|
| 260388 |
+
"learning_rate": 7.988525641025642e-06,
|
| 260389 |
+
"loss": 1.2953,
|
| 260390 |
+
"step": 126565
|
| 260391 |
+
},
|
| 260392 |
+
{
|
| 260393 |
+
"epoch": 1012.0,
|
| 260394 |
+
"eval_loss": 0.3975546061992645,
|
| 260395 |
+
"eval_runtime": 36.063,
|
| 260396 |
+
"eval_samples_per_second": 23.32,
|
| 260397 |
+
"eval_steps_per_second": 0.749,
|
| 260398 |
+
"eval_wer": 0.17799635701275046,
|
| 260399 |
+
"step": 126565
|
| 260400 |
+
},
|
| 260401 |
+
{
|
| 260402 |
+
"epoch": 1020.04,
|
| 260403 |
+
"learning_rate": 7.988445512820514e-06,
|
| 260404 |
+
"loss": 0.2682,
|
| 260405 |
+
"step": 126570
|
| 260406 |
+
},
|
| 260407 |
+
{
|
| 260408 |
+
"epoch": 1020.08,
|
| 260409 |
+
"learning_rate": 7.988365384615385e-06,
|
| 260410 |
+
"loss": 0.3308,
|
| 260411 |
+
"step": 126575
|
| 260412 |
+
},
|
| 260413 |
+
{
|
| 260414 |
+
"epoch": 1020.12,
|
| 260415 |
+
"learning_rate": 7.988285256410258e-06,
|
| 260416 |
+
"loss": 0.306,
|
| 260417 |
+
"step": 126580
|
| 260418 |
+
},
|
| 260419 |
+
{
|
| 260420 |
+
"epoch": 1020.16,
|
| 260421 |
+
"learning_rate": 7.98820512820513e-06,
|
| 260422 |
+
"loss": 0.4581,
|
| 260423 |
+
"step": 126585
|
| 260424 |
+
},
|
| 260425 |
+
{
|
| 260426 |
+
"epoch": 1020.2,
|
| 260427 |
+
"learning_rate": 7.988125e-06,
|
| 260428 |
+
"loss": 1.1796,
|
| 260429 |
+
"step": 126590
|
| 260430 |
+
},
|
| 260431 |
+
{
|
| 260432 |
+
"epoch": 1020.24,
|
| 260433 |
+
"learning_rate": 7.988044871794872e-06,
|
| 260434 |
+
"loss": 0.3497,
|
| 260435 |
+
"step": 126595
|
| 260436 |
+
},
|
| 260437 |
+
{
|
| 260438 |
+
"epoch": 1020.28,
|
| 260439 |
+
"learning_rate": 7.987964743589745e-06,
|
| 260440 |
+
"loss": 0.2608,
|
| 260441 |
+
"step": 126600
|
| 260442 |
+
},
|
| 260443 |
+
{
|
| 260444 |
+
"epoch": 1020.32,
|
| 260445 |
+
"learning_rate": 7.987884615384616e-06,
|
| 260446 |
+
"loss": 0.3339,
|
| 260447 |
+
"step": 126605
|
| 260448 |
+
},
|
| 260449 |
+
{
|
| 260450 |
+
"epoch": 1020.36,
|
| 260451 |
+
"learning_rate": 7.987804487179488e-06,
|
| 260452 |
+
"loss": 0.4798,
|
| 260453 |
+
"step": 126610
|
| 260454 |
+
},
|
| 260455 |
+
{
|
| 260456 |
+
"epoch": 1020.4,
|
| 260457 |
+
"learning_rate": 7.98772435897436e-06,
|
| 260458 |
+
"loss": 1.2394,
|
| 260459 |
+
"step": 126615
|
| 260460 |
+
},
|
| 260461 |
+
{
|
| 260462 |
+
"epoch": 1020.44,
|
| 260463 |
+
"learning_rate": 7.987644230769232e-06,
|
| 260464 |
+
"loss": 0.3557,
|
| 260465 |
+
"step": 126620
|
| 260466 |
+
},
|
| 260467 |
+
{
|
| 260468 |
+
"epoch": 1020.48,
|
| 260469 |
+
"learning_rate": 7.987564102564104e-06,
|
| 260470 |
+
"loss": 0.2716,
|
| 260471 |
+
"step": 126625
|
| 260472 |
+
},
|
| 260473 |
+
{
|
| 260474 |
+
"epoch": 1020.52,
|
| 260475 |
+
"learning_rate": 7.987483974358975e-06,
|
| 260476 |
+
"loss": 0.3483,
|
| 260477 |
+
"step": 126630
|
| 260478 |
+
},
|
| 260479 |
+
{
|
| 260480 |
+
"epoch": 1020.56,
|
| 260481 |
+
"learning_rate": 7.987403846153848e-06,
|
| 260482 |
+
"loss": 0.4007,
|
| 260483 |
+
"step": 126635
|
| 260484 |
+
},
|
| 260485 |
+
{
|
| 260486 |
+
"epoch": 1020.6,
|
| 260487 |
+
"learning_rate": 7.987323717948718e-06,
|
| 260488 |
+
"loss": 1.1896,
|
| 260489 |
+
"step": 126640
|
| 260490 |
+
},
|
| 260491 |
+
{
|
| 260492 |
+
"epoch": 1020.64,
|
| 260493 |
+
"learning_rate": 7.98724358974359e-06,
|
| 260494 |
+
"loss": 0.3013,
|
| 260495 |
+
"step": 126645
|
| 260496 |
+
},
|
| 260497 |
+
{
|
| 260498 |
+
"epoch": 1020.68,
|
| 260499 |
+
"learning_rate": 7.987163461538462e-06,
|
| 260500 |
+
"loss": 0.3619,
|
| 260501 |
+
"step": 126650
|
| 260502 |
+
},
|
| 260503 |
+
{
|
| 260504 |
+
"epoch": 1020.72,
|
| 260505 |
+
"learning_rate": 7.987083333333333e-06,
|
| 260506 |
+
"loss": 0.3159,
|
| 260507 |
+
"step": 126655
|
| 260508 |
+
},
|
| 260509 |
+
{
|
| 260510 |
+
"epoch": 1020.76,
|
| 260511 |
+
"learning_rate": 7.987003205128205e-06,
|
| 260512 |
+
"loss": 0.5123,
|
| 260513 |
+
"step": 126660
|
| 260514 |
+
},
|
| 260515 |
+
{
|
| 260516 |
+
"epoch": 1020.8,
|
| 260517 |
+
"learning_rate": 7.986923076923078e-06,
|
| 260518 |
+
"loss": 1.0823,
|
| 260519 |
+
"step": 126665
|
| 260520 |
+
},
|
| 260521 |
+
{
|
| 260522 |
+
"epoch": 1020.84,
|
| 260523 |
+
"learning_rate": 7.98684294871795e-06,
|
| 260524 |
+
"loss": 0.4015,
|
| 260525 |
+
"step": 126670
|
| 260526 |
+
},
|
| 260527 |
+
{
|
| 260528 |
+
"epoch": 1020.88,
|
| 260529 |
+
"learning_rate": 7.98676282051282e-06,
|
| 260530 |
+
"loss": 0.282,
|
| 260531 |
+
"step": 126675
|
| 260532 |
+
},
|
| 260533 |
+
{
|
| 260534 |
+
"epoch": 1020.92,
|
| 260535 |
+
"learning_rate": 7.986682692307694e-06,
|
| 260536 |
+
"loss": 0.2633,
|
| 260537 |
+
"step": 126680
|
| 260538 |
+
},
|
| 260539 |
+
{
|
| 260540 |
+
"epoch": 1020.96,
|
| 260541 |
+
"learning_rate": 7.986602564102565e-06,
|
| 260542 |
+
"loss": 0.5204,
|
| 260543 |
+
"step": 126685
|
| 260544 |
+
},
|
| 260545 |
+
{
|
| 260546 |
+
"epoch": 1021.0,
|
| 260547 |
+
"eval_loss": 0.48318079113960266,
|
| 260548 |
+
"eval_runtime": 36.3732,
|
| 260549 |
+
"eval_samples_per_second": 23.121,
|
| 260550 |
+
"eval_steps_per_second": 0.742,
|
| 260551 |
+
"eval_wer": 0.17306996599377758,
|
| 260552 |
+
"step": 126689
|
| 260553 |
}
|
| 260554 |
],
|
| 260555 |
+
"max_steps": 620000,
|
| 260556 |
"num_train_epochs": 5000,
|
| 260557 |
+
"total_flos": 3.565038460292019e+20,
|
| 260558 |
"trial_name": null,
|
| 260559 |
"trial_params": null
|
| 260560 |
}
|
model-bin/finetune/base/{checkpoint-126068 → checkpoint-126689}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630148780.943968/events.out.tfevents.1630148780.86bb0ddabf9b.4092.131
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8ab44b6e5c35156b4ad476574b19d4935688b65613900c4cd88a73d613d37011
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630149168.480171/events.out.tfevents.1630149168.86bb0ddabf9b.4092.133
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:94896acafd410a6524def8fbde908e3695d87dc1fdba3d87fdecd2e7bf2d5c97
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630149553.0608108/events.out.tfevents.1630149553.86bb0ddabf9b.4092.135
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:833c89ce3560863ef1517cce9f72e99ae495783f0489457aa4194d5f504edc98
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630149942.7762349/events.out.tfevents.1630149942.86bb0ddabf9b.4092.137
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b58d267d6bdac191c76983bb0f51d62f719d3bbbc94d369e42d273f1bf6d79ef
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630150333.0672135/events.out.tfevents.1630150333.86bb0ddabf9b.4092.139
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0ead56c0eb2b95b220c1eccc5eba92f37d7152145b86b84a7e7c9c0db293c85
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630148780.86bb0ddabf9b.4092.130
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:00048b58100d6130feb62b3377cf7b29b7917adac51c8fbb83895e49947cb01c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630149168.86bb0ddabf9b.4092.132
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74f26fc93c6313721154ee0e726336b59c3cda1c4fc9581c1175daf4005edd28
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630149553.86bb0ddabf9b.4092.134
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:236afb18a29aeb252568902992854ea8ecb721fdfb487077654c8691d1ddb1f6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630149942.86bb0ddabf9b.4092.136
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1294791834390db7b3bc786993a9abcc101b0e705f9079616c8f8f3002ec5659
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630150333.86bb0ddabf9b.4092.138
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de79b511c4f8041a051940b71d66e3d1c442e9d5899f28fb484327a5ec5d45d5
|
| 3 |
+
size 8462
|