"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630113694.5899777/events.out.tfevents.1630113694.86bb0ddabf9b.1042.31 +3 -0
- model-bin/finetune/base/log/1630114108.4351602/events.out.tfevents.1630114108.86bb0ddabf9b.1042.33 +3 -0
- model-bin/finetune/base/log/1630114521.4023952/events.out.tfevents.1630114521.86bb0ddabf9b.1042.35 +3 -0
- model-bin/finetune/base/log/1630114935.8797095/events.out.tfevents.1630114935.86bb0ddabf9b.1042.37 +3 -0
- model-bin/finetune/base/log/1630115344.0735083/events.out.tfevents.1630115344.86bb0ddabf9b.1042.39 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630113694.86bb0ddabf9b.1042.30 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630114108.86bb0ddabf9b.1042.32 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630114521.86bb0ddabf9b.1042.34 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630114935.86bb0ddabf9b.1042.36 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630115344.86bb0ddabf9b.1042.38 +3 -0
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/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:ed6a4a1b60032720f1c3b30726a6595b0807d4670f546ca8940edf524a4e7cd6
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/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:e2bf7f96504e1c0a6174b9f2d5915d9873c10b1506cc00abe352873e28517867
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/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 sha256:796e263a25f4a0a7434c18d8024159eaa3656ad1fa20d7a25b7eae15f6439f16
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/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:f8ec6882b7592c0ac7b9cd7d7fe3da25ac3c6fd19100a89c73fb078e31a1ddc3
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/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 sha256:1cd22da4f6bd733f8bb533d0e36af7533566eafa7ac974fe7b8b64dcc9e30a39
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/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,
|
|
@@ -246771,11 +246771,800 @@
|
|
| 246771 |
"eval_steps_per_second": 0.696,
|
| 246772 |
"eval_wer": 0.18200635765506026,
|
| 246773 |
"step": 115856
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 246774 |
}
|
| 246775 |
],
|
| 246776 |
-
"max_steps":
|
| 246777 |
"num_train_epochs": 5000,
|
| 246778 |
-
"total_flos": 3.
|
| 246779 |
"trial_name": null,
|
| 246780 |
"trial_params": null
|
| 246781 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 931.0,
|
| 5 |
+
"global_step": 116478,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 246771 |
"eval_steps_per_second": 0.696,
|
| 246772 |
"eval_wer": 0.18200635765506026,
|
| 246773 |
"step": 115856
|
| 246774 |
+
},
|
| 246775 |
+
{
|
| 246776 |
+
"epoch": 934.03,
|
| 246777 |
+
"learning_rate": 8.160048076923078e-06,
|
| 246778 |
+
"loss": 0.3382,
|
| 246779 |
+
"step": 115860
|
| 246780 |
+
},
|
| 246781 |
+
{
|
| 246782 |
+
"epoch": 934.07,
|
| 246783 |
+
"learning_rate": 8.15996794871795e-06,
|
| 246784 |
+
"loss": 0.2679,
|
| 246785 |
+
"step": 115865
|
| 246786 |
+
},
|
| 246787 |
+
{
|
| 246788 |
+
"epoch": 934.11,
|
| 246789 |
+
"learning_rate": 8.15988782051282e-06,
|
| 246790 |
+
"loss": 0.277,
|
| 246791 |
+
"step": 115870
|
| 246792 |
+
},
|
| 246793 |
+
{
|
| 246794 |
+
"epoch": 934.15,
|
| 246795 |
+
"learning_rate": 8.159807692307694e-06,
|
| 246796 |
+
"loss": 0.4102,
|
| 246797 |
+
"step": 115875
|
| 246798 |
+
},
|
| 246799 |
+
{
|
| 246800 |
+
"epoch": 934.19,
|
| 246801 |
+
"learning_rate": 8.159727564102565e-06,
|
| 246802 |
+
"loss": 1.1364,
|
| 246803 |
+
"step": 115880
|
| 246804 |
+
},
|
| 246805 |
+
{
|
| 246806 |
+
"epoch": 934.23,
|
| 246807 |
+
"learning_rate": 8.159647435897436e-06,
|
| 246808 |
+
"loss": 0.627,
|
| 246809 |
+
"step": 115885
|
| 246810 |
+
},
|
| 246811 |
+
{
|
| 246812 |
+
"epoch": 934.27,
|
| 246813 |
+
"learning_rate": 8.15956730769231e-06,
|
| 246814 |
+
"loss": 0.2331,
|
| 246815 |
+
"step": 115890
|
| 246816 |
+
},
|
| 246817 |
+
{
|
| 246818 |
+
"epoch": 934.31,
|
| 246819 |
+
"learning_rate": 8.15948717948718e-06,
|
| 246820 |
+
"loss": 0.2426,
|
| 246821 |
+
"step": 115895
|
| 246822 |
+
},
|
| 246823 |
+
{
|
| 246824 |
+
"epoch": 934.35,
|
| 246825 |
+
"learning_rate": 8.159407051282052e-06,
|
| 246826 |
+
"loss": 0.3757,
|
| 246827 |
+
"step": 115900
|
| 246828 |
+
},
|
| 246829 |
+
{
|
| 246830 |
+
"epoch": 934.39,
|
| 246831 |
+
"learning_rate": 8.159326923076924e-06,
|
| 246832 |
+
"loss": 0.8871,
|
| 246833 |
+
"step": 115905
|
| 246834 |
+
},
|
| 246835 |
+
{
|
| 246836 |
+
"epoch": 934.43,
|
| 246837 |
+
"learning_rate": 8.159246794871795e-06,
|
| 246838 |
+
"loss": 0.6067,
|
| 246839 |
+
"step": 115910
|
| 246840 |
+
},
|
| 246841 |
+
{
|
| 246842 |
+
"epoch": 934.47,
|
| 246843 |
+
"learning_rate": 8.159166666666666e-06,
|
| 246844 |
+
"loss": 0.3254,
|
| 246845 |
+
"step": 115915
|
| 246846 |
+
},
|
| 246847 |
+
{
|
| 246848 |
+
"epoch": 934.51,
|
| 246849 |
+
"learning_rate": 8.15908653846154e-06,
|
| 246850 |
+
"loss": 0.3118,
|
| 246851 |
+
"step": 115920
|
| 246852 |
+
},
|
| 246853 |
+
{
|
| 246854 |
+
"epoch": 934.55,
|
| 246855 |
+
"learning_rate": 8.15900641025641e-06,
|
| 246856 |
+
"loss": 0.425,
|
| 246857 |
+
"step": 115925
|
| 246858 |
+
},
|
| 246859 |
+
{
|
| 246860 |
+
"epoch": 934.59,
|
| 246861 |
+
"learning_rate": 8.158926282051282e-06,
|
| 246862 |
+
"loss": 0.9399,
|
| 246863 |
+
"step": 115930
|
| 246864 |
+
},
|
| 246865 |
+
{
|
| 246866 |
+
"epoch": 934.63,
|
| 246867 |
+
"learning_rate": 8.158846153846155e-06,
|
| 246868 |
+
"loss": 0.7605,
|
| 246869 |
+
"step": 115935
|
| 246870 |
+
},
|
| 246871 |
+
{
|
| 246872 |
+
"epoch": 934.67,
|
| 246873 |
+
"learning_rate": 8.158766025641027e-06,
|
| 246874 |
+
"loss": 0.2649,
|
| 246875 |
+
"step": 115940
|
| 246876 |
+
},
|
| 246877 |
+
{
|
| 246878 |
+
"epoch": 934.71,
|
| 246879 |
+
"learning_rate": 8.158685897435898e-06,
|
| 246880 |
+
"loss": 0.3038,
|
| 246881 |
+
"step": 115945
|
| 246882 |
+
},
|
| 246883 |
+
{
|
| 246884 |
+
"epoch": 934.76,
|
| 246885 |
+
"learning_rate": 8.15860576923077e-06,
|
| 246886 |
+
"loss": 0.3894,
|
| 246887 |
+
"step": 115950
|
| 246888 |
+
},
|
| 246889 |
+
{
|
| 246890 |
+
"epoch": 934.8,
|
| 246891 |
+
"learning_rate": 8.158525641025642e-06,
|
| 246892 |
+
"loss": 0.8665,
|
| 246893 |
+
"step": 115955
|
| 246894 |
+
},
|
| 246895 |
+
{
|
| 246896 |
+
"epoch": 934.84,
|
| 246897 |
+
"learning_rate": 8.158445512820514e-06,
|
| 246898 |
+
"loss": 0.6454,
|
| 246899 |
+
"step": 115960
|
| 246900 |
+
},
|
| 246901 |
+
{
|
| 246902 |
+
"epoch": 934.88,
|
| 246903 |
+
"learning_rate": 8.158365384615385e-06,
|
| 246904 |
+
"loss": 0.2806,
|
| 246905 |
+
"step": 115965
|
| 246906 |
+
},
|
| 246907 |
+
{
|
| 246908 |
+
"epoch": 934.92,
|
| 246909 |
+
"learning_rate": 8.158285256410256e-06,
|
| 246910 |
+
"loss": 0.2804,
|
| 246911 |
+
"step": 115970
|
| 246912 |
+
},
|
| 246913 |
+
{
|
| 246914 |
+
"epoch": 934.96,
|
| 246915 |
+
"learning_rate": 8.15820512820513e-06,
|
| 246916 |
+
"loss": 0.4341,
|
| 246917 |
+
"step": 115975
|
| 246918 |
+
},
|
| 246919 |
+
{
|
| 246920 |
+
"epoch": 935.0,
|
| 246921 |
+
"learning_rate": 8.158125e-06,
|
| 246922 |
+
"loss": 0.9677,
|
| 246923 |
+
"step": 115980
|
| 246924 |
+
},
|
| 246925 |
+
{
|
| 246926 |
+
"epoch": 935.0,
|
| 246927 |
+
"eval_loss": 0.3683865964412689,
|
| 246928 |
+
"eval_runtime": 39.1683,
|
| 246929 |
+
"eval_samples_per_second": 21.523,
|
| 246930 |
+
"eval_steps_per_second": 0.689,
|
| 246931 |
+
"eval_wer": 0.18747801618009144,
|
| 246932 |
+
"step": 115980
|
| 246933 |
+
},
|
| 246934 |
+
{
|
| 246935 |
+
"epoch": 935.04,
|
| 246936 |
+
"learning_rate": 8.158044871794872e-06,
|
| 246937 |
+
"loss": 0.2787,
|
| 246938 |
+
"step": 115985
|
| 246939 |
+
},
|
| 246940 |
+
{
|
| 246941 |
+
"epoch": 935.08,
|
| 246942 |
+
"learning_rate": 8.157964743589745e-06,
|
| 246943 |
+
"loss": 0.292,
|
| 246944 |
+
"step": 115990
|
| 246945 |
+
},
|
| 246946 |
+
{
|
| 246947 |
+
"epoch": 935.12,
|
| 246948 |
+
"learning_rate": 8.157884615384617e-06,
|
| 246949 |
+
"loss": 0.3054,
|
| 246950 |
+
"step": 115995
|
| 246951 |
+
},
|
| 246952 |
+
{
|
| 246953 |
+
"epoch": 935.16,
|
| 246954 |
+
"learning_rate": 8.157804487179488e-06,
|
| 246955 |
+
"loss": 0.5644,
|
| 246956 |
+
"step": 116000
|
| 246957 |
+
},
|
| 246958 |
+
{
|
| 246959 |
+
"epoch": 935.2,
|
| 246960 |
+
"learning_rate": 8.15772435897436e-06,
|
| 246961 |
+
"loss": 1.306,
|
| 246962 |
+
"step": 116005
|
| 246963 |
+
},
|
| 246964 |
+
{
|
| 246965 |
+
"epoch": 935.24,
|
| 246966 |
+
"learning_rate": 8.157644230769232e-06,
|
| 246967 |
+
"loss": 0.3061,
|
| 246968 |
+
"step": 116010
|
| 246969 |
+
},
|
| 246970 |
+
{
|
| 246971 |
+
"epoch": 935.28,
|
| 246972 |
+
"learning_rate": 8.157564102564102e-06,
|
| 246973 |
+
"loss": 0.2666,
|
| 246974 |
+
"step": 116015
|
| 246975 |
+
},
|
| 246976 |
+
{
|
| 246977 |
+
"epoch": 935.32,
|
| 246978 |
+
"learning_rate": 8.157483974358975e-06,
|
| 246979 |
+
"loss": 0.3729,
|
| 246980 |
+
"step": 116020
|
| 246981 |
+
},
|
| 246982 |
+
{
|
| 246983 |
+
"epoch": 935.36,
|
| 246984 |
+
"learning_rate": 8.157403846153846e-06,
|
| 246985 |
+
"loss": 0.4313,
|
| 246986 |
+
"step": 116025
|
| 246987 |
+
},
|
| 246988 |
+
{
|
| 246989 |
+
"epoch": 935.4,
|
| 246990 |
+
"learning_rate": 8.157323717948718e-06,
|
| 246991 |
+
"loss": 1.1394,
|
| 246992 |
+
"step": 116030
|
| 246993 |
+
},
|
| 246994 |
+
{
|
| 246995 |
+
"epoch": 935.44,
|
| 246996 |
+
"learning_rate": 8.15724358974359e-06,
|
| 246997 |
+
"loss": 0.3986,
|
| 246998 |
+
"step": 116035
|
| 246999 |
+
},
|
| 247000 |
+
{
|
| 247001 |
+
"epoch": 935.48,
|
| 247002 |
+
"learning_rate": 8.157163461538462e-06,
|
| 247003 |
+
"loss": 0.2467,
|
| 247004 |
+
"step": 116040
|
| 247005 |
+
},
|
| 247006 |
+
{
|
| 247007 |
+
"epoch": 935.52,
|
| 247008 |
+
"learning_rate": 8.157083333333334e-06,
|
| 247009 |
+
"loss": 0.2822,
|
| 247010 |
+
"step": 116045
|
| 247011 |
+
},
|
| 247012 |
+
{
|
| 247013 |
+
"epoch": 935.56,
|
| 247014 |
+
"learning_rate": 8.157003205128205e-06,
|
| 247015 |
+
"loss": 0.4465,
|
| 247016 |
+
"step": 116050
|
| 247017 |
+
},
|
| 247018 |
+
{
|
| 247019 |
+
"epoch": 935.6,
|
| 247020 |
+
"learning_rate": 8.156923076923078e-06,
|
| 247021 |
+
"loss": 1.1917,
|
| 247022 |
+
"step": 116055
|
| 247023 |
+
},
|
| 247024 |
+
{
|
| 247025 |
+
"epoch": 935.64,
|
| 247026 |
+
"learning_rate": 8.15684294871795e-06,
|
| 247027 |
+
"loss": 0.3459,
|
| 247028 |
+
"step": 116060
|
| 247029 |
+
},
|
| 247030 |
+
{
|
| 247031 |
+
"epoch": 935.68,
|
| 247032 |
+
"learning_rate": 8.15676282051282e-06,
|
| 247033 |
+
"loss": 0.2845,
|
| 247034 |
+
"step": 116065
|
| 247035 |
+
},
|
| 247036 |
+
{
|
| 247037 |
+
"epoch": 935.72,
|
| 247038 |
+
"learning_rate": 8.156682692307692e-06,
|
| 247039 |
+
"loss": 0.2771,
|
| 247040 |
+
"step": 116070
|
| 247041 |
+
},
|
| 247042 |
+
{
|
| 247043 |
+
"epoch": 935.76,
|
| 247044 |
+
"learning_rate": 8.156602564102565e-06,
|
| 247045 |
+
"loss": 0.4098,
|
| 247046 |
+
"step": 116075
|
| 247047 |
+
},
|
| 247048 |
+
{
|
| 247049 |
+
"epoch": 935.8,
|
| 247050 |
+
"learning_rate": 8.156522435897436e-06,
|
| 247051 |
+
"loss": 1.3603,
|
| 247052 |
+
"step": 116080
|
| 247053 |
+
},
|
| 247054 |
+
{
|
| 247055 |
+
"epoch": 935.84,
|
| 247056 |
+
"learning_rate": 8.156442307692308e-06,
|
| 247057 |
+
"loss": 0.301,
|
| 247058 |
+
"step": 116085
|
| 247059 |
+
},
|
| 247060 |
+
{
|
| 247061 |
+
"epoch": 935.88,
|
| 247062 |
+
"learning_rate": 8.15636217948718e-06,
|
| 247063 |
+
"loss": 0.2724,
|
| 247064 |
+
"step": 116090
|
| 247065 |
+
},
|
| 247066 |
+
{
|
| 247067 |
+
"epoch": 935.92,
|
| 247068 |
+
"learning_rate": 8.156282051282052e-06,
|
| 247069 |
+
"loss": 0.3323,
|
| 247070 |
+
"step": 116095
|
| 247071 |
+
},
|
| 247072 |
+
{
|
| 247073 |
+
"epoch": 935.96,
|
| 247074 |
+
"learning_rate": 8.156201923076924e-06,
|
| 247075 |
+
"loss": 0.4345,
|
| 247076 |
+
"step": 116100
|
| 247077 |
+
},
|
| 247078 |
+
{
|
| 247079 |
+
"epoch": 936.0,
|
| 247080 |
+
"eval_loss": 0.3591098189353943,
|
| 247081 |
+
"eval_runtime": 40.0141,
|
| 247082 |
+
"eval_samples_per_second": 21.068,
|
| 247083 |
+
"eval_steps_per_second": 0.675,
|
| 247084 |
+
"eval_wer": 0.1825556353194544,
|
| 247085 |
+
"step": 116104
|
| 247086 |
+
},
|
| 247087 |
+
{
|
| 247088 |
+
"epoch": 936.01,
|
| 247089 |
+
"learning_rate": 8.156121794871795e-06,
|
| 247090 |
+
"loss": 0.3954,
|
| 247091 |
+
"step": 116105
|
| 247092 |
+
},
|
| 247093 |
+
{
|
| 247094 |
+
"epoch": 936.05,
|
| 247095 |
+
"learning_rate": 8.156041666666668e-06,
|
| 247096 |
+
"loss": 0.3009,
|
| 247097 |
+
"step": 116110
|
| 247098 |
+
},
|
| 247099 |
+
{
|
| 247100 |
+
"epoch": 936.09,
|
| 247101 |
+
"learning_rate": 8.15596153846154e-06,
|
| 247102 |
+
"loss": 0.3602,
|
| 247103 |
+
"step": 116115
|
| 247104 |
+
},
|
| 247105 |
+
{
|
| 247106 |
+
"epoch": 936.13,
|
| 247107 |
+
"learning_rate": 8.15588141025641e-06,
|
| 247108 |
+
"loss": 0.279,
|
| 247109 |
+
"step": 116120
|
| 247110 |
+
},
|
| 247111 |
+
{
|
| 247112 |
+
"epoch": 936.17,
|
| 247113 |
+
"learning_rate": 8.155801282051282e-06,
|
| 247114 |
+
"loss": 0.5056,
|
| 247115 |
+
"step": 116125
|
| 247116 |
+
},
|
| 247117 |
+
{
|
| 247118 |
+
"epoch": 936.21,
|
| 247119 |
+
"learning_rate": 8.155721153846155e-06,
|
| 247120 |
+
"loss": 1.0783,
|
| 247121 |
+
"step": 116130
|
| 247122 |
+
},
|
| 247123 |
+
{
|
| 247124 |
+
"epoch": 936.25,
|
| 247125 |
+
"learning_rate": 8.155641025641026e-06,
|
| 247126 |
+
"loss": 0.3115,
|
| 247127 |
+
"step": 116135
|
| 247128 |
+
},
|
| 247129 |
+
{
|
| 247130 |
+
"epoch": 936.29,
|
| 247131 |
+
"learning_rate": 8.155560897435898e-06,
|
| 247132 |
+
"loss": 0.2519,
|
| 247133 |
+
"step": 116140
|
| 247134 |
+
},
|
| 247135 |
+
{
|
| 247136 |
+
"epoch": 936.33,
|
| 247137 |
+
"learning_rate": 8.155480769230771e-06,
|
| 247138 |
+
"loss": 0.361,
|
| 247139 |
+
"step": 116145
|
| 247140 |
+
},
|
| 247141 |
+
{
|
| 247142 |
+
"epoch": 936.37,
|
| 247143 |
+
"learning_rate": 8.155400641025642e-06,
|
| 247144 |
+
"loss": 0.5359,
|
| 247145 |
+
"step": 116150
|
| 247146 |
+
},
|
| 247147 |
+
{
|
| 247148 |
+
"epoch": 936.41,
|
| 247149 |
+
"learning_rate": 8.155320512820514e-06,
|
| 247150 |
+
"loss": 1.0703,
|
| 247151 |
+
"step": 116155
|
| 247152 |
+
},
|
| 247153 |
+
{
|
| 247154 |
+
"epoch": 936.45,
|
| 247155 |
+
"learning_rate": 8.155240384615385e-06,
|
| 247156 |
+
"loss": 0.3052,
|
| 247157 |
+
"step": 116160
|
| 247158 |
+
},
|
| 247159 |
+
{
|
| 247160 |
+
"epoch": 936.49,
|
| 247161 |
+
"learning_rate": 8.155160256410258e-06,
|
| 247162 |
+
"loss": 0.2976,
|
| 247163 |
+
"step": 116165
|
| 247164 |
+
},
|
| 247165 |
+
{
|
| 247166 |
+
"epoch": 936.53,
|
| 247167 |
+
"learning_rate": 8.155080128205128e-06,
|
| 247168 |
+
"loss": 0.316,
|
| 247169 |
+
"step": 116170
|
| 247170 |
+
},
|
| 247171 |
+
{
|
| 247172 |
+
"epoch": 936.57,
|
| 247173 |
+
"learning_rate": 8.155e-06,
|
| 247174 |
+
"loss": 0.5374,
|
| 247175 |
+
"step": 116175
|
| 247176 |
+
},
|
| 247177 |
+
{
|
| 247178 |
+
"epoch": 936.61,
|
| 247179 |
+
"learning_rate": 8.154919871794874e-06,
|
| 247180 |
+
"loss": 0.9842,
|
| 247181 |
+
"step": 116180
|
| 247182 |
+
},
|
| 247183 |
+
{
|
| 247184 |
+
"epoch": 936.65,
|
| 247185 |
+
"learning_rate": 8.154839743589743e-06,
|
| 247186 |
+
"loss": 0.2637,
|
| 247187 |
+
"step": 116185
|
| 247188 |
+
},
|
| 247189 |
+
{
|
| 247190 |
+
"epoch": 936.69,
|
| 247191 |
+
"learning_rate": 8.154759615384616e-06,
|
| 247192 |
+
"loss": 0.2658,
|
| 247193 |
+
"step": 116190
|
| 247194 |
+
},
|
| 247195 |
+
{
|
| 247196 |
+
"epoch": 936.73,
|
| 247197 |
+
"learning_rate": 8.154679487179488e-06,
|
| 247198 |
+
"loss": 0.4132,
|
| 247199 |
+
"step": 116195
|
| 247200 |
+
},
|
| 247201 |
+
{
|
| 247202 |
+
"epoch": 936.77,
|
| 247203 |
+
"learning_rate": 8.15459935897436e-06,
|
| 247204 |
+
"loss": 0.4989,
|
| 247205 |
+
"step": 116200
|
| 247206 |
+
},
|
| 247207 |
+
{
|
| 247208 |
+
"epoch": 936.81,
|
| 247209 |
+
"learning_rate": 8.15451923076923e-06,
|
| 247210 |
+
"loss": 1.1565,
|
| 247211 |
+
"step": 116205
|
| 247212 |
+
},
|
| 247213 |
+
{
|
| 247214 |
+
"epoch": 936.85,
|
| 247215 |
+
"learning_rate": 8.154439102564104e-06,
|
| 247216 |
+
"loss": 0.37,
|
| 247217 |
+
"step": 116210
|
| 247218 |
+
},
|
| 247219 |
+
{
|
| 247220 |
+
"epoch": 936.89,
|
| 247221 |
+
"learning_rate": 8.154358974358975e-06,
|
| 247222 |
+
"loss": 0.4002,
|
| 247223 |
+
"step": 116215
|
| 247224 |
+
},
|
| 247225 |
+
{
|
| 247226 |
+
"epoch": 936.93,
|
| 247227 |
+
"learning_rate": 8.154278846153846e-06,
|
| 247228 |
+
"loss": 1.0027,
|
| 247229 |
+
"step": 116220
|
| 247230 |
+
},
|
| 247231 |
+
{
|
| 247232 |
+
"epoch": 936.97,
|
| 247233 |
+
"learning_rate": 8.154198717948718e-06,
|
| 247234 |
+
"loss": 0.5068,
|
| 247235 |
+
"step": 116225
|
| 247236 |
+
},
|
| 247237 |
+
{
|
| 247238 |
+
"epoch": 937.0,
|
| 247239 |
+
"eval_loss": 0.37429821491241455,
|
| 247240 |
+
"eval_runtime": 38.8986,
|
| 247241 |
+
"eval_samples_per_second": 21.697,
|
| 247242 |
+
"eval_steps_per_second": 0.694,
|
| 247243 |
+
"eval_wer": 0.18772589806004278,
|
| 247244 |
+
"step": 116228
|
| 247245 |
+
},
|
| 247246 |
+
{
|
| 247247 |
+
"epoch": 929.02,
|
| 247248 |
+
"learning_rate": 8.15411858974359e-06,
|
| 247249 |
+
"loss": 0.3236,
|
| 247250 |
+
"step": 116230
|
| 247251 |
+
},
|
| 247252 |
+
{
|
| 247253 |
+
"epoch": 929.06,
|
| 247254 |
+
"learning_rate": 8.154038461538462e-06,
|
| 247255 |
+
"loss": 0.3169,
|
| 247256 |
+
"step": 116235
|
| 247257 |
+
},
|
| 247258 |
+
{
|
| 247259 |
+
"epoch": 929.1,
|
| 247260 |
+
"learning_rate": 8.153958333333333e-06,
|
| 247261 |
+
"loss": 0.303,
|
| 247262 |
+
"step": 116240
|
| 247263 |
+
},
|
| 247264 |
+
{
|
| 247265 |
+
"epoch": 929.14,
|
| 247266 |
+
"learning_rate": 8.153878205128206e-06,
|
| 247267 |
+
"loss": 0.3253,
|
| 247268 |
+
"step": 116245
|
| 247269 |
+
},
|
| 247270 |
+
{
|
| 247271 |
+
"epoch": 929.18,
|
| 247272 |
+
"learning_rate": 8.153798076923078e-06,
|
| 247273 |
+
"loss": 0.6763,
|
| 247274 |
+
"step": 116250
|
| 247275 |
+
},
|
| 247276 |
+
{
|
| 247277 |
+
"epoch": 929.22,
|
| 247278 |
+
"learning_rate": 8.15371794871795e-06,
|
| 247279 |
+
"loss": 1.0281,
|
| 247280 |
+
"step": 116255
|
| 247281 |
+
},
|
| 247282 |
+
{
|
| 247283 |
+
"epoch": 929.26,
|
| 247284 |
+
"learning_rate": 8.15363782051282e-06,
|
| 247285 |
+
"loss": 0.2937,
|
| 247286 |
+
"step": 116260
|
| 247287 |
+
},
|
| 247288 |
+
{
|
| 247289 |
+
"epoch": 929.3,
|
| 247290 |
+
"learning_rate": 8.153557692307694e-06,
|
| 247291 |
+
"loss": 0.28,
|
| 247292 |
+
"step": 116265
|
| 247293 |
+
},
|
| 247294 |
+
{
|
| 247295 |
+
"epoch": 929.34,
|
| 247296 |
+
"learning_rate": 8.153477564102565e-06,
|
| 247297 |
+
"loss": 0.282,
|
| 247298 |
+
"step": 116270
|
| 247299 |
+
},
|
| 247300 |
+
{
|
| 247301 |
+
"epoch": 929.38,
|
| 247302 |
+
"learning_rate": 8.153397435897436e-06,
|
| 247303 |
+
"loss": 0.6766,
|
| 247304 |
+
"step": 116275
|
| 247305 |
+
},
|
| 247306 |
+
{
|
| 247307 |
+
"epoch": 929.42,
|
| 247308 |
+
"learning_rate": 8.15331730769231e-06,
|
| 247309 |
+
"loss": 0.8845,
|
| 247310 |
+
"step": 116280
|
| 247311 |
+
},
|
| 247312 |
+
{
|
| 247313 |
+
"epoch": 929.46,
|
| 247314 |
+
"learning_rate": 8.15323717948718e-06,
|
| 247315 |
+
"loss": 0.2686,
|
| 247316 |
+
"step": 116285
|
| 247317 |
+
},
|
| 247318 |
+
{
|
| 247319 |
+
"epoch": 929.5,
|
| 247320 |
+
"learning_rate": 8.153157051282052e-06,
|
| 247321 |
+
"loss": 0.2538,
|
| 247322 |
+
"step": 116290
|
| 247323 |
+
},
|
| 247324 |
+
{
|
| 247325 |
+
"epoch": 929.54,
|
| 247326 |
+
"learning_rate": 8.153076923076923e-06,
|
| 247327 |
+
"loss": 0.2814,
|
| 247328 |
+
"step": 116295
|
| 247329 |
+
},
|
| 247330 |
+
{
|
| 247331 |
+
"epoch": 929.58,
|
| 247332 |
+
"learning_rate": 8.152996794871797e-06,
|
| 247333 |
+
"loss": 0.5815,
|
| 247334 |
+
"step": 116300
|
| 247335 |
+
},
|
| 247336 |
+
{
|
| 247337 |
+
"epoch": 929.62,
|
| 247338 |
+
"learning_rate": 8.152916666666668e-06,
|
| 247339 |
+
"loss": 1.0262,
|
| 247340 |
+
"step": 116305
|
| 247341 |
+
},
|
| 247342 |
+
{
|
| 247343 |
+
"epoch": 929.66,
|
| 247344 |
+
"learning_rate": 8.15283653846154e-06,
|
| 247345 |
+
"loss": 0.2782,
|
| 247346 |
+
"step": 116310
|
| 247347 |
+
},
|
| 247348 |
+
{
|
| 247349 |
+
"epoch": 929.7,
|
| 247350 |
+
"learning_rate": 8.15275641025641e-06,
|
| 247351 |
+
"loss": 0.3347,
|
| 247352 |
+
"step": 116315
|
| 247353 |
+
},
|
| 247354 |
+
{
|
| 247355 |
+
"epoch": 929.74,
|
| 247356 |
+
"learning_rate": 8.152676282051284e-06,
|
| 247357 |
+
"loss": 0.3231,
|
| 247358 |
+
"step": 116320
|
| 247359 |
+
},
|
| 247360 |
+
{
|
| 247361 |
+
"epoch": 929.78,
|
| 247362 |
+
"learning_rate": 8.152596153846153e-06,
|
| 247363 |
+
"loss": 0.5842,
|
| 247364 |
+
"step": 116325
|
| 247365 |
+
},
|
| 247366 |
+
{
|
| 247367 |
+
"epoch": 929.82,
|
| 247368 |
+
"learning_rate": 8.152516025641026e-06,
|
| 247369 |
+
"loss": 1.0521,
|
| 247370 |
+
"step": 116330
|
| 247371 |
+
},
|
| 247372 |
+
{
|
| 247373 |
+
"epoch": 929.86,
|
| 247374 |
+
"learning_rate": 8.1524358974359e-06,
|
| 247375 |
+
"loss": 0.2646,
|
| 247376 |
+
"step": 116335
|
| 247377 |
+
},
|
| 247378 |
+
{
|
| 247379 |
+
"epoch": 929.9,
|
| 247380 |
+
"learning_rate": 8.152355769230769e-06,
|
| 247381 |
+
"loss": 0.3342,
|
| 247382 |
+
"step": 116340
|
| 247383 |
+
},
|
| 247384 |
+
{
|
| 247385 |
+
"epoch": 929.94,
|
| 247386 |
+
"learning_rate": 8.152275641025642e-06,
|
| 247387 |
+
"loss": 0.4099,
|
| 247388 |
+
"step": 116345
|
| 247389 |
+
},
|
| 247390 |
+
{
|
| 247391 |
+
"epoch": 929.98,
|
| 247392 |
+
"learning_rate": 8.152195512820513e-06,
|
| 247393 |
+
"loss": 0.6846,
|
| 247394 |
+
"step": 116350
|
| 247395 |
+
},
|
| 247396 |
+
{
|
| 247397 |
+
"epoch": 930.0,
|
| 247398 |
+
"eval_loss": 0.42502668499946594,
|
| 247399 |
+
"eval_runtime": 39.0001,
|
| 247400 |
+
"eval_samples_per_second": 21.615,
|
| 247401 |
+
"eval_steps_per_second": 0.692,
|
| 247402 |
+
"eval_wer": 0.1844075759800323,
|
| 247403 |
+
"step": 116353
|
| 247404 |
+
},
|
| 247405 |
+
{
|
| 247406 |
+
"epoch": 930.02,
|
| 247407 |
+
"learning_rate": 8.152115384615385e-06,
|
| 247408 |
+
"loss": 0.4367,
|
| 247409 |
+
"step": 116355
|
| 247410 |
+
},
|
| 247411 |
+
{
|
| 247412 |
+
"epoch": 930.06,
|
| 247413 |
+
"learning_rate": 8.152035256410256e-06,
|
| 247414 |
+
"loss": 0.2788,
|
| 247415 |
+
"step": 116360
|
| 247416 |
+
},
|
| 247417 |
+
{
|
| 247418 |
+
"epoch": 930.1,
|
| 247419 |
+
"learning_rate": 8.15195512820513e-06,
|
| 247420 |
+
"loss": 0.289,
|
| 247421 |
+
"step": 116365
|
| 247422 |
+
},
|
| 247423 |
+
{
|
| 247424 |
+
"epoch": 930.14,
|
| 247425 |
+
"learning_rate": 8.151875e-06,
|
| 247426 |
+
"loss": 0.364,
|
| 247427 |
+
"step": 116370
|
| 247428 |
+
},
|
| 247429 |
+
{
|
| 247430 |
+
"epoch": 930.18,
|
| 247431 |
+
"learning_rate": 8.151794871794872e-06,
|
| 247432 |
+
"loss": 0.5893,
|
| 247433 |
+
"step": 116375
|
| 247434 |
+
},
|
| 247435 |
+
{
|
| 247436 |
+
"epoch": 930.22,
|
| 247437 |
+
"learning_rate": 8.151714743589745e-06,
|
| 247438 |
+
"loss": 1.0691,
|
| 247439 |
+
"step": 116380
|
| 247440 |
+
},
|
| 247441 |
+
{
|
| 247442 |
+
"epoch": 930.26,
|
| 247443 |
+
"learning_rate": 8.151634615384616e-06,
|
| 247444 |
+
"loss": 0.2525,
|
| 247445 |
+
"step": 116385
|
| 247446 |
+
},
|
| 247447 |
+
{
|
| 247448 |
+
"epoch": 930.3,
|
| 247449 |
+
"learning_rate": 8.151554487179488e-06,
|
| 247450 |
+
"loss": 0.2818,
|
| 247451 |
+
"step": 116390
|
| 247452 |
+
},
|
| 247453 |
+
{
|
| 247454 |
+
"epoch": 930.34,
|
| 247455 |
+
"learning_rate": 8.151474358974359e-06,
|
| 247456 |
+
"loss": 0.3698,
|
| 247457 |
+
"step": 116395
|
| 247458 |
+
},
|
| 247459 |
+
{
|
| 247460 |
+
"epoch": 930.38,
|
| 247461 |
+
"learning_rate": 8.151394230769232e-06,
|
| 247462 |
+
"loss": 0.6021,
|
| 247463 |
+
"step": 116400
|
| 247464 |
+
},
|
| 247465 |
+
{
|
| 247466 |
+
"epoch": 930.42,
|
| 247467 |
+
"learning_rate": 8.151314102564104e-06,
|
| 247468 |
+
"loss": 1.0208,
|
| 247469 |
+
"step": 116405
|
| 247470 |
+
},
|
| 247471 |
+
{
|
| 247472 |
+
"epoch": 930.46,
|
| 247473 |
+
"learning_rate": 8.151233974358975e-06,
|
| 247474 |
+
"loss": 0.2736,
|
| 247475 |
+
"step": 116410
|
| 247476 |
+
},
|
| 247477 |
+
{
|
| 247478 |
+
"epoch": 930.5,
|
| 247479 |
+
"learning_rate": 8.151153846153846e-06,
|
| 247480 |
+
"loss": 0.3401,
|
| 247481 |
+
"step": 116415
|
| 247482 |
+
},
|
| 247483 |
+
{
|
| 247484 |
+
"epoch": 930.54,
|
| 247485 |
+
"learning_rate": 8.15107371794872e-06,
|
| 247486 |
+
"loss": 0.3492,
|
| 247487 |
+
"step": 116420
|
| 247488 |
+
},
|
| 247489 |
+
{
|
| 247490 |
+
"epoch": 930.58,
|
| 247491 |
+
"learning_rate": 8.15099358974359e-06,
|
| 247492 |
+
"loss": 0.7202,
|
| 247493 |
+
"step": 116425
|
| 247494 |
+
},
|
| 247495 |
+
{
|
| 247496 |
+
"epoch": 930.62,
|
| 247497 |
+
"learning_rate": 8.150913461538462e-06,
|
| 247498 |
+
"loss": 1.0942,
|
| 247499 |
+
"step": 116430
|
| 247500 |
+
},
|
| 247501 |
+
{
|
| 247502 |
+
"epoch": 930.66,
|
| 247503 |
+
"learning_rate": 8.150833333333335e-06,
|
| 247504 |
+
"loss": 0.2699,
|
| 247505 |
+
"step": 116435
|
| 247506 |
+
},
|
| 247507 |
+
{
|
| 247508 |
+
"epoch": 930.7,
|
| 247509 |
+
"learning_rate": 8.150753205128206e-06,
|
| 247510 |
+
"loss": 0.2856,
|
| 247511 |
+
"step": 116440
|
| 247512 |
+
},
|
| 247513 |
+
{
|
| 247514 |
+
"epoch": 930.74,
|
| 247515 |
+
"learning_rate": 8.150673076923078e-06,
|
| 247516 |
+
"loss": 0.3806,
|
| 247517 |
+
"step": 116445
|
| 247518 |
+
},
|
| 247519 |
+
{
|
| 247520 |
+
"epoch": 930.78,
|
| 247521 |
+
"learning_rate": 8.150592948717949e-06,
|
| 247522 |
+
"loss": 0.6902,
|
| 247523 |
+
"step": 116450
|
| 247524 |
+
},
|
| 247525 |
+
{
|
| 247526 |
+
"epoch": 930.82,
|
| 247527 |
+
"learning_rate": 8.150512820512822e-06,
|
| 247528 |
+
"loss": 1.0085,
|
| 247529 |
+
"step": 116455
|
| 247530 |
+
},
|
| 247531 |
+
{
|
| 247532 |
+
"epoch": 930.86,
|
| 247533 |
+
"learning_rate": 8.150432692307692e-06,
|
| 247534 |
+
"loss": 0.2629,
|
| 247535 |
+
"step": 116460
|
| 247536 |
+
},
|
| 247537 |
+
{
|
| 247538 |
+
"epoch": 930.9,
|
| 247539 |
+
"learning_rate": 8.150352564102565e-06,
|
| 247540 |
+
"loss": 0.2615,
|
| 247541 |
+
"step": 116465
|
| 247542 |
+
},
|
| 247543 |
+
{
|
| 247544 |
+
"epoch": 930.94,
|
| 247545 |
+
"learning_rate": 8.150272435897436e-06,
|
| 247546 |
+
"loss": 0.3081,
|
| 247547 |
+
"step": 116470
|
| 247548 |
+
},
|
| 247549 |
+
{
|
| 247550 |
+
"epoch": 930.98,
|
| 247551 |
+
"learning_rate": 8.150192307692308e-06,
|
| 247552 |
+
"loss": 0.5409,
|
| 247553 |
+
"step": 116475
|
| 247554 |
+
},
|
| 247555 |
+
{
|
| 247556 |
+
"epoch": 931.0,
|
| 247557 |
+
"eval_loss": 0.427347868680954,
|
| 247558 |
+
"eval_runtime": 39.3615,
|
| 247559 |
+
"eval_samples_per_second": 21.417,
|
| 247560 |
+
"eval_steps_per_second": 0.686,
|
| 247561 |
+
"eval_wer": 0.18822154026011984,
|
| 247562 |
+
"step": 116478
|
| 247563 |
}
|
| 247564 |
],
|
| 247565 |
+
"max_steps": 625000,
|
| 247566 |
"num_train_epochs": 5000,
|
| 247567 |
+
"total_flos": 3.2779879386977056e+20,
|
| 247568 |
"trial_name": null,
|
| 247569 |
"trial_params": null
|
| 247570 |
}
|
model-bin/finetune/base/{checkpoint-115856 → checkpoint-116478}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630113694.5899777/events.out.tfevents.1630113694.86bb0ddabf9b.1042.31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f2bafd466c630fd1f651cfc6bbaac47320578e8cbe124912388e189cdd02f369
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630114108.4351602/events.out.tfevents.1630114108.86bb0ddabf9b.1042.33
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:07a0a4953561629052e2d0bc1801579b24c34391dc125a4716a5f56a2a3045fa
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630114521.4023952/events.out.tfevents.1630114521.86bb0ddabf9b.1042.35
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f29c80132da81c5a6dd7313800425f0fb049072d5f15e5425c2f62b5e3e9619b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630114935.8797095/events.out.tfevents.1630114935.86bb0ddabf9b.1042.37
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:840305c14ee3814d1fdf3db5aca2eb7cc8afa2665aa999949f65e2e0cb1dee27
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630115344.0735083/events.out.tfevents.1630115344.86bb0ddabf9b.1042.39
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:755d20a6b48427f9c481163078b8315b0f5e80a84035f6c92eaf75a877a3c1bf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630113694.86bb0ddabf9b.1042.30
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0bce1d69ec0c9a328677eb09f0ad0770a749ad9c3fea932b09d28be9fbd1a70
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630114108.86bb0ddabf9b.1042.32
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ba95ea0c440eb81e806d98f8334fc2cca8bd634a1c8b8264996852ee87ee21d0
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630114521.86bb0ddabf9b.1042.34
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3cebd1f50853e7c5b4a05ca51952ae5a5a5f3c76be45e7ddf63b91f9739d919c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630114935.86bb0ddabf9b.1042.36
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:683c8af391b89ef17e13a05e99e3522534eda937abf38257a90c3d8c661bdd3e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630115344.86bb0ddabf9b.1042.38
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:830c5154d8a6e363532f21d0a69517dce85a978e3fa9e0cb2b35a98412204cb1
|
| 3 |
+
size 8622
|