"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629993468.0041914/events.out.tfevents.1629993468.8e89bd551565.924.241 +3 -0
- model-bin/finetune/base/log/1629993903.8396595/events.out.tfevents.1629993903.8e89bd551565.924.243 +3 -0
- model-bin/finetune/base/log/1629994337.0035079/events.out.tfevents.1629994337.8e89bd551565.924.245 +3 -0
- model-bin/finetune/base/log/1629994779.4147024/events.out.tfevents.1629994779.8e89bd551565.924.247 +3 -0
- model-bin/finetune/base/log/1629995217.6110506/events.out.tfevents.1629995217.8e89bd551565.924.249 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629993467.8e89bd551565.924.240 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629993901.8e89bd551565.924.242 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629994336.8e89bd551565.924.244 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629994779.8e89bd551565.924.246 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629995217.8e89bd551565.924.248 +3 -0
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/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 sha256:7f887ccc9e530275eabb942742b49945da5e396af357fcb4c6216a8b0e38aaa0
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/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:fa66840495953d2dcf8d0f972d70b80e8025406460af3c06559744b0b8e00069
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/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:a1d7ee77a309d4b02f0e0b50a254c8d8994dd350eafd8040d4e8821c92cec5ce
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/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 sha256:7a7e782bf5b4614ad2f9ac21555ba4813411b4db165cdaa4d98d6f042e85cb7e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/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 sha256:1a6409ad26afd7a4d5286ea7d95c3da331c0a58c248fb8e74926edc29a9ff2b5
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/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,
|
|
@@ -230790,11 +230790,800 @@
|
|
| 230790 |
"eval_steps_per_second": 0.684,
|
| 230791 |
"eval_wer": 0.18283473714783618,
|
| 230792 |
"step": 103297
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 230793 |
}
|
| 230794 |
],
|
| 230795 |
-
"max_steps":
|
| 230796 |
"num_train_epochs": 5000,
|
| 230797 |
-
"total_flos": 2.
|
| 230798 |
"trial_name": null,
|
| 230799 |
"trial_params": null
|
| 230800 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 838.0,
|
| 5 |
+
"global_step": 103918,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 230790 |
"eval_steps_per_second": 0.684,
|
| 230791 |
"eval_wer": 0.18283473714783618,
|
| 230792 |
"step": 103297
|
| 230793 |
+
},
|
| 230794 |
+
{
|
| 230795 |
+
"epoch": 833.02,
|
| 230796 |
+
"learning_rate": 8.361233974358976e-06,
|
| 230797 |
+
"loss": 0.351,
|
| 230798 |
+
"step": 103300
|
| 230799 |
+
},
|
| 230800 |
+
{
|
| 230801 |
+
"epoch": 833.06,
|
| 230802 |
+
"learning_rate": 8.361153846153846e-06,
|
| 230803 |
+
"loss": 0.3207,
|
| 230804 |
+
"step": 103305
|
| 230805 |
+
},
|
| 230806 |
+
{
|
| 230807 |
+
"epoch": 833.1,
|
| 230808 |
+
"learning_rate": 8.361073717948719e-06,
|
| 230809 |
+
"loss": 0.2721,
|
| 230810 |
+
"step": 103310
|
| 230811 |
+
},
|
| 230812 |
+
{
|
| 230813 |
+
"epoch": 833.14,
|
| 230814 |
+
"learning_rate": 8.36099358974359e-06,
|
| 230815 |
+
"loss": 0.3887,
|
| 230816 |
+
"step": 103315
|
| 230817 |
+
},
|
| 230818 |
+
{
|
| 230819 |
+
"epoch": 833.18,
|
| 230820 |
+
"learning_rate": 8.360913461538462e-06,
|
| 230821 |
+
"loss": 0.7524,
|
| 230822 |
+
"step": 103320
|
| 230823 |
+
},
|
| 230824 |
+
{
|
| 230825 |
+
"epoch": 833.22,
|
| 230826 |
+
"learning_rate": 8.360833333333333e-06,
|
| 230827 |
+
"loss": 0.915,
|
| 230828 |
+
"step": 103325
|
| 230829 |
+
},
|
| 230830 |
+
{
|
| 230831 |
+
"epoch": 833.27,
|
| 230832 |
+
"learning_rate": 8.360753205128206e-06,
|
| 230833 |
+
"loss": 0.291,
|
| 230834 |
+
"step": 103330
|
| 230835 |
+
},
|
| 230836 |
+
{
|
| 230837 |
+
"epoch": 833.31,
|
| 230838 |
+
"learning_rate": 8.360673076923077e-06,
|
| 230839 |
+
"loss": 0.3147,
|
| 230840 |
+
"step": 103335
|
| 230841 |
+
},
|
| 230842 |
+
{
|
| 230843 |
+
"epoch": 833.35,
|
| 230844 |
+
"learning_rate": 8.360592948717949e-06,
|
| 230845 |
+
"loss": 0.3544,
|
| 230846 |
+
"step": 103340
|
| 230847 |
+
},
|
| 230848 |
+
{
|
| 230849 |
+
"epoch": 833.39,
|
| 230850 |
+
"learning_rate": 8.360512820512822e-06,
|
| 230851 |
+
"loss": 0.7792,
|
| 230852 |
+
"step": 103345
|
| 230853 |
+
},
|
| 230854 |
+
{
|
| 230855 |
+
"epoch": 833.43,
|
| 230856 |
+
"learning_rate": 8.360432692307693e-06,
|
| 230857 |
+
"loss": 0.8214,
|
| 230858 |
+
"step": 103350
|
| 230859 |
+
},
|
| 230860 |
+
{
|
| 230861 |
+
"epoch": 833.47,
|
| 230862 |
+
"learning_rate": 8.360352564102565e-06,
|
| 230863 |
+
"loss": 0.2678,
|
| 230864 |
+
"step": 103355
|
| 230865 |
+
},
|
| 230866 |
+
{
|
| 230867 |
+
"epoch": 833.51,
|
| 230868 |
+
"learning_rate": 8.360272435897436e-06,
|
| 230869 |
+
"loss": 0.288,
|
| 230870 |
+
"step": 103360
|
| 230871 |
+
},
|
| 230872 |
+
{
|
| 230873 |
+
"epoch": 833.55,
|
| 230874 |
+
"learning_rate": 8.360192307692309e-06,
|
| 230875 |
+
"loss": 0.4139,
|
| 230876 |
+
"step": 103365
|
| 230877 |
+
},
|
| 230878 |
+
{
|
| 230879 |
+
"epoch": 833.59,
|
| 230880 |
+
"learning_rate": 8.36011217948718e-06,
|
| 230881 |
+
"loss": 0.7124,
|
| 230882 |
+
"step": 103370
|
| 230883 |
+
},
|
| 230884 |
+
{
|
| 230885 |
+
"epoch": 833.63,
|
| 230886 |
+
"learning_rate": 8.360032051282052e-06,
|
| 230887 |
+
"loss": 0.8996,
|
| 230888 |
+
"step": 103375
|
| 230889 |
+
},
|
| 230890 |
+
{
|
| 230891 |
+
"epoch": 833.67,
|
| 230892 |
+
"learning_rate": 8.359951923076923e-06,
|
| 230893 |
+
"loss": 0.3268,
|
| 230894 |
+
"step": 103380
|
| 230895 |
+
},
|
| 230896 |
+
{
|
| 230897 |
+
"epoch": 833.71,
|
| 230898 |
+
"learning_rate": 8.359871794871796e-06,
|
| 230899 |
+
"loss": 0.599,
|
| 230900 |
+
"step": 103385
|
| 230901 |
+
},
|
| 230902 |
+
{
|
| 230903 |
+
"epoch": 833.75,
|
| 230904 |
+
"learning_rate": 8.359791666666667e-06,
|
| 230905 |
+
"loss": 0.3569,
|
| 230906 |
+
"step": 103390
|
| 230907 |
+
},
|
| 230908 |
+
{
|
| 230909 |
+
"epoch": 833.79,
|
| 230910 |
+
"learning_rate": 8.359711538461539e-06,
|
| 230911 |
+
"loss": 1.1455,
|
| 230912 |
+
"step": 103395
|
| 230913 |
+
},
|
| 230914 |
+
{
|
| 230915 |
+
"epoch": 833.83,
|
| 230916 |
+
"learning_rate": 8.359631410256412e-06,
|
| 230917 |
+
"loss": 0.8055,
|
| 230918 |
+
"step": 103400
|
| 230919 |
+
},
|
| 230920 |
+
{
|
| 230921 |
+
"epoch": 833.87,
|
| 230922 |
+
"learning_rate": 8.359551282051283e-06,
|
| 230923 |
+
"loss": 0.3046,
|
| 230924 |
+
"step": 103405
|
| 230925 |
+
},
|
| 230926 |
+
{
|
| 230927 |
+
"epoch": 833.91,
|
| 230928 |
+
"learning_rate": 8.359471153846155e-06,
|
| 230929 |
+
"loss": 0.3044,
|
| 230930 |
+
"step": 103410
|
| 230931 |
+
},
|
| 230932 |
+
{
|
| 230933 |
+
"epoch": 833.95,
|
| 230934 |
+
"learning_rate": 8.359391025641026e-06,
|
| 230935 |
+
"loss": 0.3412,
|
| 230936 |
+
"step": 103415
|
| 230937 |
+
},
|
| 230938 |
+
{
|
| 230939 |
+
"epoch": 833.99,
|
| 230940 |
+
"learning_rate": 8.359310897435899e-06,
|
| 230941 |
+
"loss": 0.9576,
|
| 230942 |
+
"step": 103420
|
| 230943 |
+
},
|
| 230944 |
+
{
|
| 230945 |
+
"epoch": 834.0,
|
| 230946 |
+
"eval_loss": 0.3438977897167206,
|
| 230947 |
+
"eval_runtime": 40.5967,
|
| 230948 |
+
"eval_samples_per_second": 20.691,
|
| 230949 |
+
"eval_steps_per_second": 0.665,
|
| 230950 |
+
"eval_wer": 0.18366442655638432,
|
| 230951 |
+
"step": 103421
|
| 230952 |
+
},
|
| 230953 |
+
{
|
| 230954 |
+
"epoch": 827.03,
|
| 230955 |
+
"learning_rate": 8.359230769230769e-06,
|
| 230956 |
+
"loss": 0.4528,
|
| 230957 |
+
"step": 103425
|
| 230958 |
+
},
|
| 230959 |
+
{
|
| 230960 |
+
"epoch": 827.07,
|
| 230961 |
+
"learning_rate": 8.359150641025642e-06,
|
| 230962 |
+
"loss": 0.3116,
|
| 230963 |
+
"step": 103430
|
| 230964 |
+
},
|
| 230965 |
+
{
|
| 230966 |
+
"epoch": 827.11,
|
| 230967 |
+
"learning_rate": 8.359070512820513e-06,
|
| 230968 |
+
"loss": 0.3281,
|
| 230969 |
+
"step": 103435
|
| 230970 |
+
},
|
| 230971 |
+
{
|
| 230972 |
+
"epoch": 827.15,
|
| 230973 |
+
"learning_rate": 8.358990384615384e-06,
|
| 230974 |
+
"loss": 0.3707,
|
| 230975 |
+
"step": 103440
|
| 230976 |
+
},
|
| 230977 |
+
{
|
| 230978 |
+
"epoch": 827.19,
|
| 230979 |
+
"learning_rate": 8.358910256410258e-06,
|
| 230980 |
+
"loss": 0.9188,
|
| 230981 |
+
"step": 103445
|
| 230982 |
+
},
|
| 230983 |
+
{
|
| 230984 |
+
"epoch": 827.23,
|
| 230985 |
+
"learning_rate": 8.358830128205129e-06,
|
| 230986 |
+
"loss": 0.6167,
|
| 230987 |
+
"step": 103450
|
| 230988 |
+
},
|
| 230989 |
+
{
|
| 230990 |
+
"epoch": 827.27,
|
| 230991 |
+
"learning_rate": 8.35875e-06,
|
| 230992 |
+
"loss": 0.3096,
|
| 230993 |
+
"step": 103455
|
| 230994 |
+
},
|
| 230995 |
+
{
|
| 230996 |
+
"epoch": 827.31,
|
| 230997 |
+
"learning_rate": 8.358669871794872e-06,
|
| 230998 |
+
"loss": 0.3597,
|
| 230999 |
+
"step": 103460
|
| 231000 |
+
},
|
| 231001 |
+
{
|
| 231002 |
+
"epoch": 827.35,
|
| 231003 |
+
"learning_rate": 8.358589743589745e-06,
|
| 231004 |
+
"loss": 0.3656,
|
| 231005 |
+
"step": 103465
|
| 231006 |
+
},
|
| 231007 |
+
{
|
| 231008 |
+
"epoch": 827.39,
|
| 231009 |
+
"learning_rate": 8.358509615384616e-06,
|
| 231010 |
+
"loss": 0.9181,
|
| 231011 |
+
"step": 103470
|
| 231012 |
+
},
|
| 231013 |
+
{
|
| 231014 |
+
"epoch": 827.43,
|
| 231015 |
+
"learning_rate": 8.358429487179487e-06,
|
| 231016 |
+
"loss": 0.6144,
|
| 231017 |
+
"step": 103475
|
| 231018 |
+
},
|
| 231019 |
+
{
|
| 231020 |
+
"epoch": 827.47,
|
| 231021 |
+
"learning_rate": 8.358349358974359e-06,
|
| 231022 |
+
"loss": 0.3339,
|
| 231023 |
+
"step": 103480
|
| 231024 |
+
},
|
| 231025 |
+
{
|
| 231026 |
+
"epoch": 827.51,
|
| 231027 |
+
"learning_rate": 8.358269230769232e-06,
|
| 231028 |
+
"loss": 0.3008,
|
| 231029 |
+
"step": 103485
|
| 231030 |
+
},
|
| 231031 |
+
{
|
| 231032 |
+
"epoch": 827.55,
|
| 231033 |
+
"learning_rate": 8.358189102564103e-06,
|
| 231034 |
+
"loss": 0.417,
|
| 231035 |
+
"step": 103490
|
| 231036 |
+
},
|
| 231037 |
+
{
|
| 231038 |
+
"epoch": 827.59,
|
| 231039 |
+
"learning_rate": 8.358108974358974e-06,
|
| 231040 |
+
"loss": 0.9079,
|
| 231041 |
+
"step": 103495
|
| 231042 |
+
},
|
| 231043 |
+
{
|
| 231044 |
+
"epoch": 827.63,
|
| 231045 |
+
"learning_rate": 8.358028846153848e-06,
|
| 231046 |
+
"loss": 0.7081,
|
| 231047 |
+
"step": 103500
|
| 231048 |
+
},
|
| 231049 |
+
{
|
| 231050 |
+
"epoch": 827.67,
|
| 231051 |
+
"learning_rate": 8.357948717948719e-06,
|
| 231052 |
+
"loss": 0.2889,
|
| 231053 |
+
"step": 103505
|
| 231054 |
+
},
|
| 231055 |
+
{
|
| 231056 |
+
"epoch": 827.71,
|
| 231057 |
+
"learning_rate": 8.35786858974359e-06,
|
| 231058 |
+
"loss": 0.3271,
|
| 231059 |
+
"step": 103510
|
| 231060 |
+
},
|
| 231061 |
+
{
|
| 231062 |
+
"epoch": 827.75,
|
| 231063 |
+
"learning_rate": 8.357788461538462e-06,
|
| 231064 |
+
"loss": 0.5678,
|
| 231065 |
+
"step": 103515
|
| 231066 |
+
},
|
| 231067 |
+
{
|
| 231068 |
+
"epoch": 827.79,
|
| 231069 |
+
"learning_rate": 8.357708333333335e-06,
|
| 231070 |
+
"loss": 0.9176,
|
| 231071 |
+
"step": 103520
|
| 231072 |
+
},
|
| 231073 |
+
{
|
| 231074 |
+
"epoch": 827.83,
|
| 231075 |
+
"learning_rate": 8.357628205128206e-06,
|
| 231076 |
+
"loss": 0.7331,
|
| 231077 |
+
"step": 103525
|
| 231078 |
+
},
|
| 231079 |
+
{
|
| 231080 |
+
"epoch": 827.87,
|
| 231081 |
+
"learning_rate": 8.357548076923077e-06,
|
| 231082 |
+
"loss": 0.2808,
|
| 231083 |
+
"step": 103530
|
| 231084 |
+
},
|
| 231085 |
+
{
|
| 231086 |
+
"epoch": 827.91,
|
| 231087 |
+
"learning_rate": 8.357467948717949e-06,
|
| 231088 |
+
"loss": 0.3182,
|
| 231089 |
+
"step": 103535
|
| 231090 |
+
},
|
| 231091 |
+
{
|
| 231092 |
+
"epoch": 827.95,
|
| 231093 |
+
"learning_rate": 8.357387820512822e-06,
|
| 231094 |
+
"loss": 0.43,
|
| 231095 |
+
"step": 103540
|
| 231096 |
+
},
|
| 231097 |
+
{
|
| 231098 |
+
"epoch": 827.99,
|
| 231099 |
+
"learning_rate": 8.357307692307693e-06,
|
| 231100 |
+
"loss": 0.8699,
|
| 231101 |
+
"step": 103545
|
| 231102 |
+
},
|
| 231103 |
+
{
|
| 231104 |
+
"epoch": 828.0,
|
| 231105 |
+
"eval_loss": 0.36451485753059387,
|
| 231106 |
+
"eval_runtime": 40.0413,
|
| 231107 |
+
"eval_samples_per_second": 21.003,
|
| 231108 |
+
"eval_steps_per_second": 0.674,
|
| 231109 |
+
"eval_wer": 0.1745266037871697,
|
| 231110 |
+
"step": 103546
|
| 231111 |
+
},
|
| 231112 |
+
{
|
| 231113 |
+
"epoch": 835.03,
|
| 231114 |
+
"learning_rate": 8.357227564102565e-06,
|
| 231115 |
+
"loss": 0.2803,
|
| 231116 |
+
"step": 103550
|
| 231117 |
+
},
|
| 231118 |
+
{
|
| 231119 |
+
"epoch": 835.07,
|
| 231120 |
+
"learning_rate": 8.357147435897438e-06,
|
| 231121 |
+
"loss": 0.3105,
|
| 231122 |
+
"step": 103555
|
| 231123 |
+
},
|
| 231124 |
+
{
|
| 231125 |
+
"epoch": 835.11,
|
| 231126 |
+
"learning_rate": 8.357067307692309e-06,
|
| 231127 |
+
"loss": 0.2909,
|
| 231128 |
+
"step": 103560
|
| 231129 |
+
},
|
| 231130 |
+
{
|
| 231131 |
+
"epoch": 835.15,
|
| 231132 |
+
"learning_rate": 8.35698717948718e-06,
|
| 231133 |
+
"loss": 0.4187,
|
| 231134 |
+
"step": 103565
|
| 231135 |
+
},
|
| 231136 |
+
{
|
| 231137 |
+
"epoch": 835.19,
|
| 231138 |
+
"learning_rate": 8.356907051282052e-06,
|
| 231139 |
+
"loss": 0.9497,
|
| 231140 |
+
"step": 103570
|
| 231141 |
+
},
|
| 231142 |
+
{
|
| 231143 |
+
"epoch": 835.23,
|
| 231144 |
+
"learning_rate": 8.356826923076925e-06,
|
| 231145 |
+
"loss": 0.734,
|
| 231146 |
+
"step": 103575
|
| 231147 |
+
},
|
| 231148 |
+
{
|
| 231149 |
+
"epoch": 835.27,
|
| 231150 |
+
"learning_rate": 8.356746794871794e-06,
|
| 231151 |
+
"loss": 0.3495,
|
| 231152 |
+
"step": 103580
|
| 231153 |
+
},
|
| 231154 |
+
{
|
| 231155 |
+
"epoch": 835.31,
|
| 231156 |
+
"learning_rate": 8.356666666666667e-06,
|
| 231157 |
+
"loss": 0.3019,
|
| 231158 |
+
"step": 103585
|
| 231159 |
+
},
|
| 231160 |
+
{
|
| 231161 |
+
"epoch": 835.35,
|
| 231162 |
+
"learning_rate": 8.35658653846154e-06,
|
| 231163 |
+
"loss": 0.3799,
|
| 231164 |
+
"step": 103590
|
| 231165 |
+
},
|
| 231166 |
+
{
|
| 231167 |
+
"epoch": 835.39,
|
| 231168 |
+
"learning_rate": 8.35650641025641e-06,
|
| 231169 |
+
"loss": 0.8788,
|
| 231170 |
+
"step": 103595
|
| 231171 |
+
},
|
| 231172 |
+
{
|
| 231173 |
+
"epoch": 835.43,
|
| 231174 |
+
"learning_rate": 8.356426282051283e-06,
|
| 231175 |
+
"loss": 0.6812,
|
| 231176 |
+
"step": 103600
|
| 231177 |
+
},
|
| 231178 |
+
{
|
| 231179 |
+
"epoch": 835.47,
|
| 231180 |
+
"learning_rate": 8.356346153846155e-06,
|
| 231181 |
+
"loss": 0.2686,
|
| 231182 |
+
"step": 103605
|
| 231183 |
+
},
|
| 231184 |
+
{
|
| 231185 |
+
"epoch": 835.51,
|
| 231186 |
+
"learning_rate": 8.356266025641026e-06,
|
| 231187 |
+
"loss": 0.3267,
|
| 231188 |
+
"step": 103610
|
| 231189 |
+
},
|
| 231190 |
+
{
|
| 231191 |
+
"epoch": 835.55,
|
| 231192 |
+
"learning_rate": 8.356185897435897e-06,
|
| 231193 |
+
"loss": 0.4017,
|
| 231194 |
+
"step": 103615
|
| 231195 |
+
},
|
| 231196 |
+
{
|
| 231197 |
+
"epoch": 835.59,
|
| 231198 |
+
"learning_rate": 8.35610576923077e-06,
|
| 231199 |
+
"loss": 0.8509,
|
| 231200 |
+
"step": 103620
|
| 231201 |
+
},
|
| 231202 |
+
{
|
| 231203 |
+
"epoch": 835.63,
|
| 231204 |
+
"learning_rate": 8.356025641025642e-06,
|
| 231205 |
+
"loss": 0.6918,
|
| 231206 |
+
"step": 103625
|
| 231207 |
+
},
|
| 231208 |
+
{
|
| 231209 |
+
"epoch": 835.67,
|
| 231210 |
+
"learning_rate": 8.355945512820513e-06,
|
| 231211 |
+
"loss": 0.2692,
|
| 231212 |
+
"step": 103630
|
| 231213 |
+
},
|
| 231214 |
+
{
|
| 231215 |
+
"epoch": 835.71,
|
| 231216 |
+
"learning_rate": 8.355865384615384e-06,
|
| 231217 |
+
"loss": 0.3076,
|
| 231218 |
+
"step": 103635
|
| 231219 |
+
},
|
| 231220 |
+
{
|
| 231221 |
+
"epoch": 835.76,
|
| 231222 |
+
"learning_rate": 8.355785256410257e-06,
|
| 231223 |
+
"loss": 0.3897,
|
| 231224 |
+
"step": 103640
|
| 231225 |
+
},
|
| 231226 |
+
{
|
| 231227 |
+
"epoch": 835.8,
|
| 231228 |
+
"learning_rate": 8.355705128205129e-06,
|
| 231229 |
+
"loss": 0.8999,
|
| 231230 |
+
"step": 103645
|
| 231231 |
+
},
|
| 231232 |
+
{
|
| 231233 |
+
"epoch": 835.84,
|
| 231234 |
+
"learning_rate": 8.355625e-06,
|
| 231235 |
+
"loss": 0.6748,
|
| 231236 |
+
"step": 103650
|
| 231237 |
+
},
|
| 231238 |
+
{
|
| 231239 |
+
"epoch": 835.88,
|
| 231240 |
+
"learning_rate": 8.355544871794873e-06,
|
| 231241 |
+
"loss": 0.3033,
|
| 231242 |
+
"step": 103655
|
| 231243 |
+
},
|
| 231244 |
+
{
|
| 231245 |
+
"epoch": 835.92,
|
| 231246 |
+
"learning_rate": 8.355464743589745e-06,
|
| 231247 |
+
"loss": 0.3177,
|
| 231248 |
+
"step": 103660
|
| 231249 |
+
},
|
| 231250 |
+
{
|
| 231251 |
+
"epoch": 835.96,
|
| 231252 |
+
"learning_rate": 8.355384615384616e-06,
|
| 231253 |
+
"loss": 0.4767,
|
| 231254 |
+
"step": 103665
|
| 231255 |
+
},
|
| 231256 |
+
{
|
| 231257 |
+
"epoch": 836.0,
|
| 231258 |
+
"learning_rate": 8.355304487179487e-06,
|
| 231259 |
+
"loss": 1.0467,
|
| 231260 |
+
"step": 103670
|
| 231261 |
+
},
|
| 231262 |
+
{
|
| 231263 |
+
"epoch": 836.0,
|
| 231264 |
+
"eval_loss": 0.4262886345386505,
|
| 231265 |
+
"eval_runtime": 39.1875,
|
| 231266 |
+
"eval_samples_per_second": 21.435,
|
| 231267 |
+
"eval_steps_per_second": 0.689,
|
| 231268 |
+
"eval_wer": 0.18302426152605733,
|
| 231269 |
+
"step": 103670
|
| 231270 |
+
},
|
| 231271 |
+
{
|
| 231272 |
+
"epoch": 836.04,
|
| 231273 |
+
"learning_rate": 8.35522435897436e-06,
|
| 231274 |
+
"loss": 0.2928,
|
| 231275 |
+
"step": 103675
|
| 231276 |
+
},
|
| 231277 |
+
{
|
| 231278 |
+
"epoch": 836.08,
|
| 231279 |
+
"learning_rate": 8.355144230769232e-06,
|
| 231280 |
+
"loss": 0.2817,
|
| 231281 |
+
"step": 103680
|
| 231282 |
+
},
|
| 231283 |
+
{
|
| 231284 |
+
"epoch": 836.12,
|
| 231285 |
+
"learning_rate": 8.355064102564103e-06,
|
| 231286 |
+
"loss": 0.3598,
|
| 231287 |
+
"step": 103685
|
| 231288 |
+
},
|
| 231289 |
+
{
|
| 231290 |
+
"epoch": 836.16,
|
| 231291 |
+
"learning_rate": 8.354983974358976e-06,
|
| 231292 |
+
"loss": 0.4622,
|
| 231293 |
+
"step": 103690
|
| 231294 |
+
},
|
| 231295 |
+
{
|
| 231296 |
+
"epoch": 836.2,
|
| 231297 |
+
"learning_rate": 8.354903846153847e-06,
|
| 231298 |
+
"loss": 1.2675,
|
| 231299 |
+
"step": 103695
|
| 231300 |
+
},
|
| 231301 |
+
{
|
| 231302 |
+
"epoch": 836.24,
|
| 231303 |
+
"learning_rate": 8.354823717948719e-06,
|
| 231304 |
+
"loss": 0.3431,
|
| 231305 |
+
"step": 103700
|
| 231306 |
+
},
|
| 231307 |
+
{
|
| 231308 |
+
"epoch": 836.28,
|
| 231309 |
+
"learning_rate": 8.35474358974359e-06,
|
| 231310 |
+
"loss": 0.2755,
|
| 231311 |
+
"step": 103705
|
| 231312 |
+
},
|
| 231313 |
+
{
|
| 231314 |
+
"epoch": 836.32,
|
| 231315 |
+
"learning_rate": 8.354663461538463e-06,
|
| 231316 |
+
"loss": 0.3133,
|
| 231317 |
+
"step": 103710
|
| 231318 |
+
},
|
| 231319 |
+
{
|
| 231320 |
+
"epoch": 836.36,
|
| 231321 |
+
"learning_rate": 8.354583333333333e-06,
|
| 231322 |
+
"loss": 0.4812,
|
| 231323 |
+
"step": 103715
|
| 231324 |
+
},
|
| 231325 |
+
{
|
| 231326 |
+
"epoch": 836.4,
|
| 231327 |
+
"learning_rate": 8.354503205128206e-06,
|
| 231328 |
+
"loss": 1.1412,
|
| 231329 |
+
"step": 103720
|
| 231330 |
+
},
|
| 231331 |
+
{
|
| 231332 |
+
"epoch": 836.44,
|
| 231333 |
+
"learning_rate": 8.354423076923077e-06,
|
| 231334 |
+
"loss": 0.294,
|
| 231335 |
+
"step": 103725
|
| 231336 |
+
},
|
| 231337 |
+
{
|
| 231338 |
+
"epoch": 836.48,
|
| 231339 |
+
"learning_rate": 8.354342948717949e-06,
|
| 231340 |
+
"loss": 0.2575,
|
| 231341 |
+
"step": 103730
|
| 231342 |
+
},
|
| 231343 |
+
{
|
| 231344 |
+
"epoch": 836.52,
|
| 231345 |
+
"learning_rate": 8.35426282051282e-06,
|
| 231346 |
+
"loss": 0.3326,
|
| 231347 |
+
"step": 103735
|
| 231348 |
+
},
|
| 231349 |
+
{
|
| 231350 |
+
"epoch": 836.56,
|
| 231351 |
+
"learning_rate": 8.354182692307693e-06,
|
| 231352 |
+
"loss": 0.5133,
|
| 231353 |
+
"step": 103740
|
| 231354 |
+
},
|
| 231355 |
+
{
|
| 231356 |
+
"epoch": 836.6,
|
| 231357 |
+
"learning_rate": 8.354102564102564e-06,
|
| 231358 |
+
"loss": 1.1794,
|
| 231359 |
+
"step": 103745
|
| 231360 |
+
},
|
| 231361 |
+
{
|
| 231362 |
+
"epoch": 836.65,
|
| 231363 |
+
"learning_rate": 8.354022435897436e-06,
|
| 231364 |
+
"loss": 0.307,
|
| 231365 |
+
"step": 103750
|
| 231366 |
+
},
|
| 231367 |
+
{
|
| 231368 |
+
"epoch": 836.69,
|
| 231369 |
+
"learning_rate": 8.353942307692309e-06,
|
| 231370 |
+
"loss": 0.2519,
|
| 231371 |
+
"step": 103755
|
| 231372 |
+
},
|
| 231373 |
+
{
|
| 231374 |
+
"epoch": 836.73,
|
| 231375 |
+
"learning_rate": 8.35386217948718e-06,
|
| 231376 |
+
"loss": 0.3294,
|
| 231377 |
+
"step": 103760
|
| 231378 |
+
},
|
| 231379 |
+
{
|
| 231380 |
+
"epoch": 836.77,
|
| 231381 |
+
"learning_rate": 8.353782051282052e-06,
|
| 231382 |
+
"loss": 0.432,
|
| 231383 |
+
"step": 103765
|
| 231384 |
+
},
|
| 231385 |
+
{
|
| 231386 |
+
"epoch": 836.81,
|
| 231387 |
+
"learning_rate": 8.353701923076923e-06,
|
| 231388 |
+
"loss": 1.1507,
|
| 231389 |
+
"step": 103770
|
| 231390 |
+
},
|
| 231391 |
+
{
|
| 231392 |
+
"epoch": 836.85,
|
| 231393 |
+
"learning_rate": 8.353621794871796e-06,
|
| 231394 |
+
"loss": 0.3704,
|
| 231395 |
+
"step": 103775
|
| 231396 |
+
},
|
| 231397 |
+
{
|
| 231398 |
+
"epoch": 836.89,
|
| 231399 |
+
"learning_rate": 8.353541666666667e-06,
|
| 231400 |
+
"loss": 0.3221,
|
| 231401 |
+
"step": 103780
|
| 231402 |
+
},
|
| 231403 |
+
{
|
| 231404 |
+
"epoch": 836.93,
|
| 231405 |
+
"learning_rate": 8.353461538461539e-06,
|
| 231406 |
+
"loss": 0.304,
|
| 231407 |
+
"step": 103785
|
| 231408 |
+
},
|
| 231409 |
+
{
|
| 231410 |
+
"epoch": 836.97,
|
| 231411 |
+
"learning_rate": 8.353381410256412e-06,
|
| 231412 |
+
"loss": 0.4899,
|
| 231413 |
+
"step": 103790
|
| 231414 |
+
},
|
| 231415 |
+
{
|
| 231416 |
+
"epoch": 837.0,
|
| 231417 |
+
"eval_loss": 0.3803502321243286,
|
| 231418 |
+
"eval_runtime": 39.27,
|
| 231419 |
+
"eval_samples_per_second": 21.39,
|
| 231420 |
+
"eval_steps_per_second": 0.688,
|
| 231421 |
+
"eval_wer": 0.18735172004744957,
|
| 231422 |
+
"step": 103794
|
| 231423 |
+
},
|
| 231424 |
+
{
|
| 231425 |
+
"epoch": 837.01,
|
| 231426 |
+
"learning_rate": 8.353301282051283e-06,
|
| 231427 |
+
"loss": 0.3926,
|
| 231428 |
+
"step": 103795
|
| 231429 |
+
},
|
| 231430 |
+
{
|
| 231431 |
+
"epoch": 837.05,
|
| 231432 |
+
"learning_rate": 8.353221153846154e-06,
|
| 231433 |
+
"loss": 0.284,
|
| 231434 |
+
"step": 103800
|
| 231435 |
+
},
|
| 231436 |
+
{
|
| 231437 |
+
"epoch": 837.09,
|
| 231438 |
+
"learning_rate": 8.353141025641026e-06,
|
| 231439 |
+
"loss": 0.3139,
|
| 231440 |
+
"step": 103805
|
| 231441 |
+
},
|
| 231442 |
+
{
|
| 231443 |
+
"epoch": 837.13,
|
| 231444 |
+
"learning_rate": 8.353060897435899e-06,
|
| 231445 |
+
"loss": 0.3067,
|
| 231446 |
+
"step": 103810
|
| 231447 |
+
},
|
| 231448 |
+
{
|
| 231449 |
+
"epoch": 837.17,
|
| 231450 |
+
"learning_rate": 8.35298076923077e-06,
|
| 231451 |
+
"loss": 0.689,
|
| 231452 |
+
"step": 103815
|
| 231453 |
+
},
|
| 231454 |
+
{
|
| 231455 |
+
"epoch": 837.21,
|
| 231456 |
+
"learning_rate": 8.352900641025642e-06,
|
| 231457 |
+
"loss": 1.1145,
|
| 231458 |
+
"step": 103820
|
| 231459 |
+
},
|
| 231460 |
+
{
|
| 231461 |
+
"epoch": 837.25,
|
| 231462 |
+
"learning_rate": 8.352820512820513e-06,
|
| 231463 |
+
"loss": 0.288,
|
| 231464 |
+
"step": 103825
|
| 231465 |
+
},
|
| 231466 |
+
{
|
| 231467 |
+
"epoch": 837.29,
|
| 231468 |
+
"learning_rate": 8.352740384615386e-06,
|
| 231469 |
+
"loss": 0.2808,
|
| 231470 |
+
"step": 103830
|
| 231471 |
+
},
|
| 231472 |
+
{
|
| 231473 |
+
"epoch": 837.33,
|
| 231474 |
+
"learning_rate": 8.352660256410257e-06,
|
| 231475 |
+
"loss": 0.3067,
|
| 231476 |
+
"step": 103835
|
| 231477 |
+
},
|
| 231478 |
+
{
|
| 231479 |
+
"epoch": 837.37,
|
| 231480 |
+
"learning_rate": 8.352580128205129e-06,
|
| 231481 |
+
"loss": 0.5511,
|
| 231482 |
+
"step": 103840
|
| 231483 |
+
},
|
| 231484 |
+
{
|
| 231485 |
+
"epoch": 837.41,
|
| 231486 |
+
"learning_rate": 8.352500000000002e-06,
|
| 231487 |
+
"loss": 1.3246,
|
| 231488 |
+
"step": 103845
|
| 231489 |
+
},
|
| 231490 |
+
{
|
| 231491 |
+
"epoch": 837.45,
|
| 231492 |
+
"learning_rate": 8.352419871794873e-06,
|
| 231493 |
+
"loss": 0.2766,
|
| 231494 |
+
"step": 103850
|
| 231495 |
+
},
|
| 231496 |
+
{
|
| 231497 |
+
"epoch": 837.49,
|
| 231498 |
+
"learning_rate": 8.352339743589744e-06,
|
| 231499 |
+
"loss": 0.2609,
|
| 231500 |
+
"step": 103855
|
| 231501 |
+
},
|
| 231502 |
+
{
|
| 231503 |
+
"epoch": 837.53,
|
| 231504 |
+
"learning_rate": 8.352259615384616e-06,
|
| 231505 |
+
"loss": 0.3097,
|
| 231506 |
+
"step": 103860
|
| 231507 |
+
},
|
| 231508 |
+
{
|
| 231509 |
+
"epoch": 837.57,
|
| 231510 |
+
"learning_rate": 8.352179487179489e-06,
|
| 231511 |
+
"loss": 0.5214,
|
| 231512 |
+
"step": 103865
|
| 231513 |
+
},
|
| 231514 |
+
{
|
| 231515 |
+
"epoch": 837.61,
|
| 231516 |
+
"learning_rate": 8.352099358974359e-06,
|
| 231517 |
+
"loss": 1.0392,
|
| 231518 |
+
"step": 103870
|
| 231519 |
+
},
|
| 231520 |
+
{
|
| 231521 |
+
"epoch": 837.65,
|
| 231522 |
+
"learning_rate": 8.352019230769232e-06,
|
| 231523 |
+
"loss": 0.3029,
|
| 231524 |
+
"step": 103875
|
| 231525 |
+
},
|
| 231526 |
+
{
|
| 231527 |
+
"epoch": 837.69,
|
| 231528 |
+
"learning_rate": 8.351939102564103e-06,
|
| 231529 |
+
"loss": 0.3948,
|
| 231530 |
+
"step": 103880
|
| 231531 |
+
},
|
| 231532 |
+
{
|
| 231533 |
+
"epoch": 837.73,
|
| 231534 |
+
"learning_rate": 8.351858974358974e-06,
|
| 231535 |
+
"loss": 0.3793,
|
| 231536 |
+
"step": 103885
|
| 231537 |
+
},
|
| 231538 |
+
{
|
| 231539 |
+
"epoch": 837.77,
|
| 231540 |
+
"learning_rate": 8.351778846153847e-06,
|
| 231541 |
+
"loss": 0.5007,
|
| 231542 |
+
"step": 103890
|
| 231543 |
+
},
|
| 231544 |
+
{
|
| 231545 |
+
"epoch": 837.81,
|
| 231546 |
+
"learning_rate": 8.351698717948719e-06,
|
| 231547 |
+
"loss": 1.1322,
|
| 231548 |
+
"step": 103895
|
| 231549 |
+
},
|
| 231550 |
+
{
|
| 231551 |
+
"epoch": 837.85,
|
| 231552 |
+
"learning_rate": 8.35161858974359e-06,
|
| 231553 |
+
"loss": 0.323,
|
| 231554 |
+
"step": 103900
|
| 231555 |
+
},
|
| 231556 |
+
{
|
| 231557 |
+
"epoch": 837.9,
|
| 231558 |
+
"learning_rate": 8.351538461538461e-06,
|
| 231559 |
+
"loss": 0.3,
|
| 231560 |
+
"step": 103905
|
| 231561 |
+
},
|
| 231562 |
+
{
|
| 231563 |
+
"epoch": 837.94,
|
| 231564 |
+
"learning_rate": 8.351458333333335e-06,
|
| 231565 |
+
"loss": 0.3611,
|
| 231566 |
+
"step": 103910
|
| 231567 |
+
},
|
| 231568 |
+
{
|
| 231569 |
+
"epoch": 837.98,
|
| 231570 |
+
"learning_rate": 8.351378205128206e-06,
|
| 231571 |
+
"loss": 0.6142,
|
| 231572 |
+
"step": 103915
|
| 231573 |
+
},
|
| 231574 |
+
{
|
| 231575 |
+
"epoch": 838.0,
|
| 231576 |
+
"eval_loss": 0.4361937344074249,
|
| 231577 |
+
"eval_runtime": 39.334,
|
| 231578 |
+
"eval_samples_per_second": 21.356,
|
| 231579 |
+
"eval_steps_per_second": 0.686,
|
| 231580 |
+
"eval_wer": 0.18322231169021302,
|
| 231581 |
+
"step": 103918
|
| 231582 |
}
|
| 231583 |
],
|
| 231584 |
+
"max_steps": 620000,
|
| 231585 |
"num_train_epochs": 5000,
|
| 231586 |
+
"total_flos": 2.924503505825161e+20,
|
| 231587 |
"trial_name": null,
|
| 231588 |
"trial_params": null
|
| 231589 |
}
|
model-bin/finetune/base/{checkpoint-103297 → checkpoint-103918}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629993468.0041914/events.out.tfevents.1629993468.8e89bd551565.924.241
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:28fba00972f7f5bf570ea4d89510c3990af4becc1debc7d6f73a261a90e24e32
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629993903.8396595/events.out.tfevents.1629993903.8e89bd551565.924.243
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f26487cdcfa5719ab8283301c96055d718938f75c42ab74a4cbfca4503fd9a6f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629994337.0035079/events.out.tfevents.1629994337.8e89bd551565.924.245
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8f1551ae0163f82fa11f10f2d261376555d92ddad08fb73642842db296a3328a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629994779.4147024/events.out.tfevents.1629994779.8e89bd551565.924.247
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:852829eb34335d8026c2ee2937d265a7f770c06c2e7d951fde11f96bb1d64198
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629995217.6110506/events.out.tfevents.1629995217.8e89bd551565.924.249
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3ace0e7105d54f62d60f1d144a5e00d108d00642188814d46ac908f2216fee68
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629993467.8e89bd551565.924.240
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:934eab1a0bf19c364256cdf29ad8b199d2f33cf4c8e3a85df06f64b7a2b16732
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629993901.8e89bd551565.924.242
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3a60899e6d9c9122f1bcfc553fe563a0390463894524f6da806a171562a90249
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629994336.8e89bd551565.924.244
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:993ff42bf365cce8cebae0f498ba6509d70e03cbc2654d64581d3431dacfdb9a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629994779.8e89bd551565.924.246
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:19e4b098e0133780d4710aa6766ec212da6268ae34b59ee4ee0a649ae3f9b240
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629995217.8e89bd551565.924.248
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9a072963f9decedf6ba9cdf71a2019d152e4e70bb933ec60083aecb1b1433ce2
|
| 3 |
+
size 8622
|