"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630040846.5578432/events.out.tfevents.1630040846.52f5c7e305a3.886.51 +3 -0
- model-bin/finetune/base/log/1630041307.6197045/events.out.tfevents.1630041307.52f5c7e305a3.886.53 +3 -0
- model-bin/finetune/base/log/1630041769.276012/events.out.tfevents.1630041769.52f5c7e305a3.886.55 +3 -0
- model-bin/finetune/base/log/1630042227.0612097/events.out.tfevents.1630042227.52f5c7e305a3.886.57 +3 -0
- model-bin/finetune/base/log/1630042684.0997107/events.out.tfevents.1630042684.52f5c7e305a3.886.59 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630040846.52f5c7e305a3.886.50 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630041307.52f5c7e305a3.886.52 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630041769.52f5c7e305a3.886.54 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630042227.52f5c7e305a3.886.56 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630042684.52f5c7e305a3.886.58 +3 -0
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/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 sha256:0e5dc90a70a85e957755cb6202f4c4d80c379d7fd74f040f1b1350f81adac857
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/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 sha256:1088f43ebdbcfe27fd3e1d964f959ea73e3429b6627619ee6b2f4fe2fd90842a
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/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 sha256:410f7dbaa22760785fc91c3fc36ea41ac05e07db17e386037557b7982be23c9a
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/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:ec64d5f24f4ae2327c64f546de1c854fafe6ed45fe3467e4570a1f1e9d84778b
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/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:ae42d03429901dcd1ce598ce8ed8f03436068c6121a373adbc3dbbcf80611550
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/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,
|
|
@@ -238704,11 +238704,800 @@
|
|
| 238704 |
"eval_steps_per_second": 0.648,
|
| 238705 |
"eval_wer": 0.18675573627650305,
|
| 238706 |
"step": 109516
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 238707 |
}
|
| 238708 |
],
|
| 238709 |
-
"max_steps":
|
| 238710 |
"num_train_epochs": 5000,
|
| 238711 |
-
"total_flos": 3.
|
| 238712 |
"trial_name": null,
|
| 238713 |
"trial_params": null
|
| 238714 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 887.995983935743,
|
| 5 |
+
"global_step": 110137,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 238704 |
"eval_steps_per_second": 0.648,
|
| 238705 |
"eval_wer": 0.18675573627650305,
|
| 238706 |
"step": 109516
|
| 238707 |
+
},
|
| 238708 |
+
{
|
| 238709 |
+
"epoch": 883.03,
|
| 238710 |
+
"learning_rate": 8.247544426494346e-06,
|
| 238711 |
+
"loss": 0.3481,
|
| 238712 |
+
"step": 109520
|
| 238713 |
+
},
|
| 238714 |
+
{
|
| 238715 |
+
"epoch": 883.07,
|
| 238716 |
+
"learning_rate": 8.247463651050081e-06,
|
| 238717 |
+
"loss": 0.2659,
|
| 238718 |
+
"step": 109525
|
| 238719 |
+
},
|
| 238720 |
+
{
|
| 238721 |
+
"epoch": 883.11,
|
| 238722 |
+
"learning_rate": 8.247382875605816e-06,
|
| 238723 |
+
"loss": 0.342,
|
| 238724 |
+
"step": 109530
|
| 238725 |
+
},
|
| 238726 |
+
{
|
| 238727 |
+
"epoch": 883.15,
|
| 238728 |
+
"learning_rate": 8.247302100161551e-06,
|
| 238729 |
+
"loss": 0.3842,
|
| 238730 |
+
"step": 109535
|
| 238731 |
+
},
|
| 238732 |
+
{
|
| 238733 |
+
"epoch": 883.19,
|
| 238734 |
+
"learning_rate": 8.247221324717286e-06,
|
| 238735 |
+
"loss": 0.9115,
|
| 238736 |
+
"step": 109540
|
| 238737 |
+
},
|
| 238738 |
+
{
|
| 238739 |
+
"epoch": 883.23,
|
| 238740 |
+
"learning_rate": 8.247140549273021e-06,
|
| 238741 |
+
"loss": 0.7016,
|
| 238742 |
+
"step": 109545
|
| 238743 |
+
},
|
| 238744 |
+
{
|
| 238745 |
+
"epoch": 883.27,
|
| 238746 |
+
"learning_rate": 8.247059773828757e-06,
|
| 238747 |
+
"loss": 0.3301,
|
| 238748 |
+
"step": 109550
|
| 238749 |
+
},
|
| 238750 |
+
{
|
| 238751 |
+
"epoch": 883.31,
|
| 238752 |
+
"learning_rate": 8.246978998384491e-06,
|
| 238753 |
+
"loss": 0.2978,
|
| 238754 |
+
"step": 109555
|
| 238755 |
+
},
|
| 238756 |
+
{
|
| 238757 |
+
"epoch": 883.35,
|
| 238758 |
+
"learning_rate": 8.246898222940227e-06,
|
| 238759 |
+
"loss": 0.3891,
|
| 238760 |
+
"step": 109560
|
| 238761 |
+
},
|
| 238762 |
+
{
|
| 238763 |
+
"epoch": 883.4,
|
| 238764 |
+
"learning_rate": 8.246817447495961e-06,
|
| 238765 |
+
"loss": 0.9619,
|
| 238766 |
+
"step": 109565
|
| 238767 |
+
},
|
| 238768 |
+
{
|
| 238769 |
+
"epoch": 883.44,
|
| 238770 |
+
"learning_rate": 8.246736672051697e-06,
|
| 238771 |
+
"loss": 0.6178,
|
| 238772 |
+
"step": 109570
|
| 238773 |
+
},
|
| 238774 |
+
{
|
| 238775 |
+
"epoch": 883.48,
|
| 238776 |
+
"learning_rate": 8.246655896607431e-06,
|
| 238777 |
+
"loss": 0.2598,
|
| 238778 |
+
"step": 109575
|
| 238779 |
+
},
|
| 238780 |
+
{
|
| 238781 |
+
"epoch": 883.52,
|
| 238782 |
+
"learning_rate": 8.246575121163167e-06,
|
| 238783 |
+
"loss": 0.353,
|
| 238784 |
+
"step": 109580
|
| 238785 |
+
},
|
| 238786 |
+
{
|
| 238787 |
+
"epoch": 883.56,
|
| 238788 |
+
"learning_rate": 8.246494345718901e-06,
|
| 238789 |
+
"loss": 0.3751,
|
| 238790 |
+
"step": 109585
|
| 238791 |
+
},
|
| 238792 |
+
{
|
| 238793 |
+
"epoch": 883.6,
|
| 238794 |
+
"learning_rate": 8.246413570274637e-06,
|
| 238795 |
+
"loss": 0.9872,
|
| 238796 |
+
"step": 109590
|
| 238797 |
+
},
|
| 238798 |
+
{
|
| 238799 |
+
"epoch": 883.64,
|
| 238800 |
+
"learning_rate": 8.246332794830371e-06,
|
| 238801 |
+
"loss": 0.5591,
|
| 238802 |
+
"step": 109595
|
| 238803 |
+
},
|
| 238804 |
+
{
|
| 238805 |
+
"epoch": 883.68,
|
| 238806 |
+
"learning_rate": 8.246252019386107e-06,
|
| 238807 |
+
"loss": 0.2976,
|
| 238808 |
+
"step": 109600
|
| 238809 |
+
},
|
| 238810 |
+
{
|
| 238811 |
+
"epoch": 883.72,
|
| 238812 |
+
"learning_rate": 8.246171243941843e-06,
|
| 238813 |
+
"loss": 0.3517,
|
| 238814 |
+
"step": 109605
|
| 238815 |
+
},
|
| 238816 |
+
{
|
| 238817 |
+
"epoch": 883.76,
|
| 238818 |
+
"learning_rate": 8.246090468497577e-06,
|
| 238819 |
+
"loss": 0.4131,
|
| 238820 |
+
"step": 109610
|
| 238821 |
+
},
|
| 238822 |
+
{
|
| 238823 |
+
"epoch": 883.8,
|
| 238824 |
+
"learning_rate": 8.246009693053313e-06,
|
| 238825 |
+
"loss": 0.9514,
|
| 238826 |
+
"step": 109615
|
| 238827 |
+
},
|
| 238828 |
+
{
|
| 238829 |
+
"epoch": 883.84,
|
| 238830 |
+
"learning_rate": 8.245928917609047e-06,
|
| 238831 |
+
"loss": 0.7815,
|
| 238832 |
+
"step": 109620
|
| 238833 |
+
},
|
| 238834 |
+
{
|
| 238835 |
+
"epoch": 883.88,
|
| 238836 |
+
"learning_rate": 8.245848142164783e-06,
|
| 238837 |
+
"loss": 0.2899,
|
| 238838 |
+
"step": 109625
|
| 238839 |
+
},
|
| 238840 |
+
{
|
| 238841 |
+
"epoch": 883.92,
|
| 238842 |
+
"learning_rate": 8.245767366720517e-06,
|
| 238843 |
+
"loss": 0.3618,
|
| 238844 |
+
"step": 109630
|
| 238845 |
+
},
|
| 238846 |
+
{
|
| 238847 |
+
"epoch": 883.96,
|
| 238848 |
+
"learning_rate": 8.245686591276253e-06,
|
| 238849 |
+
"loss": 0.4588,
|
| 238850 |
+
"step": 109635
|
| 238851 |
+
},
|
| 238852 |
+
{
|
| 238853 |
+
"epoch": 884.0,
|
| 238854 |
+
"learning_rate": 8.245605815831987e-06,
|
| 238855 |
+
"loss": 1.2207,
|
| 238856 |
+
"step": 109640
|
| 238857 |
+
},
|
| 238858 |
+
{
|
| 238859 |
+
"epoch": 884.0,
|
| 238860 |
+
"eval_loss": 0.3926634192466736,
|
| 238861 |
+
"eval_runtime": 42.3731,
|
| 238862 |
+
"eval_samples_per_second": 19.777,
|
| 238863 |
+
"eval_steps_per_second": 0.637,
|
| 238864 |
+
"eval_wer": 0.1874064091045223,
|
| 238865 |
+
"step": 109640
|
| 238866 |
+
},
|
| 238867 |
+
{
|
| 238868 |
+
"epoch": 877.04,
|
| 238869 |
+
"learning_rate": 8.245525040387723e-06,
|
| 238870 |
+
"loss": 0.3087,
|
| 238871 |
+
"step": 109645
|
| 238872 |
+
},
|
| 238873 |
+
{
|
| 238874 |
+
"epoch": 877.08,
|
| 238875 |
+
"learning_rate": 8.245444264943457e-06,
|
| 238876 |
+
"loss": 0.291,
|
| 238877 |
+
"step": 109650
|
| 238878 |
+
},
|
| 238879 |
+
{
|
| 238880 |
+
"epoch": 877.12,
|
| 238881 |
+
"learning_rate": 8.245363489499193e-06,
|
| 238882 |
+
"loss": 0.2647,
|
| 238883 |
+
"step": 109655
|
| 238884 |
+
},
|
| 238885 |
+
{
|
| 238886 |
+
"epoch": 877.16,
|
| 238887 |
+
"learning_rate": 8.245282714054929e-06,
|
| 238888 |
+
"loss": 0.4465,
|
| 238889 |
+
"step": 109660
|
| 238890 |
+
},
|
| 238891 |
+
{
|
| 238892 |
+
"epoch": 877.2,
|
| 238893 |
+
"learning_rate": 8.245201938610663e-06,
|
| 238894 |
+
"loss": 1.1782,
|
| 238895 |
+
"step": 109665
|
| 238896 |
+
},
|
| 238897 |
+
{
|
| 238898 |
+
"epoch": 877.24,
|
| 238899 |
+
"learning_rate": 8.245121163166399e-06,
|
| 238900 |
+
"loss": 0.3699,
|
| 238901 |
+
"step": 109670
|
| 238902 |
+
},
|
| 238903 |
+
{
|
| 238904 |
+
"epoch": 877.28,
|
| 238905 |
+
"learning_rate": 8.245040387722133e-06,
|
| 238906 |
+
"loss": 0.3423,
|
| 238907 |
+
"step": 109675
|
| 238908 |
+
},
|
| 238909 |
+
{
|
| 238910 |
+
"epoch": 877.32,
|
| 238911 |
+
"learning_rate": 8.244959612277869e-06,
|
| 238912 |
+
"loss": 0.3079,
|
| 238913 |
+
"step": 109680
|
| 238914 |
+
},
|
| 238915 |
+
{
|
| 238916 |
+
"epoch": 877.36,
|
| 238917 |
+
"learning_rate": 8.244878836833603e-06,
|
| 238918 |
+
"loss": 0.5391,
|
| 238919 |
+
"step": 109685
|
| 238920 |
+
},
|
| 238921 |
+
{
|
| 238922 |
+
"epoch": 877.4,
|
| 238923 |
+
"learning_rate": 8.244798061389339e-06,
|
| 238924 |
+
"loss": 1.0807,
|
| 238925 |
+
"step": 109690
|
| 238926 |
+
},
|
| 238927 |
+
{
|
| 238928 |
+
"epoch": 877.44,
|
| 238929 |
+
"learning_rate": 8.244717285945073e-06,
|
| 238930 |
+
"loss": 0.4013,
|
| 238931 |
+
"step": 109695
|
| 238932 |
+
},
|
| 238933 |
+
{
|
| 238934 |
+
"epoch": 877.48,
|
| 238935 |
+
"learning_rate": 8.244636510500809e-06,
|
| 238936 |
+
"loss": 0.2937,
|
| 238937 |
+
"step": 109700
|
| 238938 |
+
},
|
| 238939 |
+
{
|
| 238940 |
+
"epoch": 877.52,
|
| 238941 |
+
"learning_rate": 8.244555735056543e-06,
|
| 238942 |
+
"loss": 0.3526,
|
| 238943 |
+
"step": 109705
|
| 238944 |
+
},
|
| 238945 |
+
{
|
| 238946 |
+
"epoch": 877.56,
|
| 238947 |
+
"learning_rate": 8.244474959612279e-06,
|
| 238948 |
+
"loss": 0.4522,
|
| 238949 |
+
"step": 109710
|
| 238950 |
+
},
|
| 238951 |
+
{
|
| 238952 |
+
"epoch": 877.6,
|
| 238953 |
+
"learning_rate": 8.244394184168013e-06,
|
| 238954 |
+
"loss": 1.1569,
|
| 238955 |
+
"step": 109715
|
| 238956 |
+
},
|
| 238957 |
+
{
|
| 238958 |
+
"epoch": 877.64,
|
| 238959 |
+
"learning_rate": 8.244313408723748e-06,
|
| 238960 |
+
"loss": 0.3141,
|
| 238961 |
+
"step": 109720
|
| 238962 |
+
},
|
| 238963 |
+
{
|
| 238964 |
+
"epoch": 877.68,
|
| 238965 |
+
"learning_rate": 8.244232633279484e-06,
|
| 238966 |
+
"loss": 0.3222,
|
| 238967 |
+
"step": 109725
|
| 238968 |
+
},
|
| 238969 |
+
{
|
| 238970 |
+
"epoch": 877.72,
|
| 238971 |
+
"learning_rate": 8.244151857835218e-06,
|
| 238972 |
+
"loss": 0.3052,
|
| 238973 |
+
"step": 109730
|
| 238974 |
+
},
|
| 238975 |
+
{
|
| 238976 |
+
"epoch": 877.76,
|
| 238977 |
+
"learning_rate": 8.244071082390954e-06,
|
| 238978 |
+
"loss": 0.5342,
|
| 238979 |
+
"step": 109735
|
| 238980 |
+
},
|
| 238981 |
+
{
|
| 238982 |
+
"epoch": 877.8,
|
| 238983 |
+
"learning_rate": 8.243990306946688e-06,
|
| 238984 |
+
"loss": 1.289,
|
| 238985 |
+
"step": 109740
|
| 238986 |
+
},
|
| 238987 |
+
{
|
| 238988 |
+
"epoch": 877.84,
|
| 238989 |
+
"learning_rate": 8.243909531502424e-06,
|
| 238990 |
+
"loss": 0.2907,
|
| 238991 |
+
"step": 109745
|
| 238992 |
+
},
|
| 238993 |
+
{
|
| 238994 |
+
"epoch": 877.88,
|
| 238995 |
+
"learning_rate": 8.243828756058158e-06,
|
| 238996 |
+
"loss": 0.2946,
|
| 238997 |
+
"step": 109750
|
| 238998 |
+
},
|
| 238999 |
+
{
|
| 239000 |
+
"epoch": 877.92,
|
| 239001 |
+
"learning_rate": 8.243747980613894e-06,
|
| 239002 |
+
"loss": 0.339,
|
| 239003 |
+
"step": 109755
|
| 239004 |
+
},
|
| 239005 |
+
{
|
| 239006 |
+
"epoch": 877.96,
|
| 239007 |
+
"learning_rate": 8.243667205169628e-06,
|
| 239008 |
+
"loss": 0.3804,
|
| 239009 |
+
"step": 109760
|
| 239010 |
+
},
|
| 239011 |
+
{
|
| 239012 |
+
"epoch": 878.0,
|
| 239013 |
+
"learning_rate": 8.243586429725364e-06,
|
| 239014 |
+
"loss": 1.107,
|
| 239015 |
+
"step": 109765
|
| 239016 |
+
},
|
| 239017 |
+
{
|
| 239018 |
+
"epoch": 878.0,
|
| 239019 |
+
"eval_loss": 0.39036035537719727,
|
| 239020 |
+
"eval_runtime": 40.186,
|
| 239021 |
+
"eval_samples_per_second": 20.828,
|
| 239022 |
+
"eval_steps_per_second": 0.672,
|
| 239023 |
+
"eval_wer": 0.19128113879003558,
|
| 239024 |
+
"step": 109765
|
| 239025 |
+
},
|
| 239026 |
+
{
|
| 239027 |
+
"epoch": 885.04,
|
| 239028 |
+
"learning_rate": 8.243505654281098e-06,
|
| 239029 |
+
"loss": 0.3556,
|
| 239030 |
+
"step": 109770
|
| 239031 |
+
},
|
| 239032 |
+
{
|
| 239033 |
+
"epoch": 885.08,
|
| 239034 |
+
"learning_rate": 8.243424878836834e-06,
|
| 239035 |
+
"loss": 0.2762,
|
| 239036 |
+
"step": 109775
|
| 239037 |
+
},
|
| 239038 |
+
{
|
| 239039 |
+
"epoch": 885.12,
|
| 239040 |
+
"learning_rate": 8.24334410339257e-06,
|
| 239041 |
+
"loss": 0.3401,
|
| 239042 |
+
"step": 109780
|
| 239043 |
+
},
|
| 239044 |
+
{
|
| 239045 |
+
"epoch": 885.16,
|
| 239046 |
+
"learning_rate": 8.243263327948304e-06,
|
| 239047 |
+
"loss": 0.6001,
|
| 239048 |
+
"step": 109785
|
| 239049 |
+
},
|
| 239050 |
+
{
|
| 239051 |
+
"epoch": 885.2,
|
| 239052 |
+
"learning_rate": 8.24318255250404e-06,
|
| 239053 |
+
"loss": 1.3321,
|
| 239054 |
+
"step": 109790
|
| 239055 |
+
},
|
| 239056 |
+
{
|
| 239057 |
+
"epoch": 885.24,
|
| 239058 |
+
"learning_rate": 8.243101777059774e-06,
|
| 239059 |
+
"loss": 0.3579,
|
| 239060 |
+
"step": 109795
|
| 239061 |
+
},
|
| 239062 |
+
{
|
| 239063 |
+
"epoch": 885.28,
|
| 239064 |
+
"learning_rate": 8.24302100161551e-06,
|
| 239065 |
+
"loss": 0.2614,
|
| 239066 |
+
"step": 109800
|
| 239067 |
+
},
|
| 239068 |
+
{
|
| 239069 |
+
"epoch": 885.32,
|
| 239070 |
+
"learning_rate": 8.242940226171244e-06,
|
| 239071 |
+
"loss": 0.3233,
|
| 239072 |
+
"step": 109805
|
| 239073 |
+
},
|
| 239074 |
+
{
|
| 239075 |
+
"epoch": 885.36,
|
| 239076 |
+
"learning_rate": 8.24285945072698e-06,
|
| 239077 |
+
"loss": 0.3893,
|
| 239078 |
+
"step": 109810
|
| 239079 |
+
},
|
| 239080 |
+
{
|
| 239081 |
+
"epoch": 885.4,
|
| 239082 |
+
"learning_rate": 8.242778675282714e-06,
|
| 239083 |
+
"loss": 1.1275,
|
| 239084 |
+
"step": 109815
|
| 239085 |
+
},
|
| 239086 |
+
{
|
| 239087 |
+
"epoch": 885.44,
|
| 239088 |
+
"learning_rate": 8.24269789983845e-06,
|
| 239089 |
+
"loss": 0.2981,
|
| 239090 |
+
"step": 109820
|
| 239091 |
+
},
|
| 239092 |
+
{
|
| 239093 |
+
"epoch": 885.48,
|
| 239094 |
+
"learning_rate": 8.242617124394184e-06,
|
| 239095 |
+
"loss": 0.2639,
|
| 239096 |
+
"step": 109825
|
| 239097 |
+
},
|
| 239098 |
+
{
|
| 239099 |
+
"epoch": 885.52,
|
| 239100 |
+
"learning_rate": 8.24253634894992e-06,
|
| 239101 |
+
"loss": 0.2683,
|
| 239102 |
+
"step": 109830
|
| 239103 |
+
},
|
| 239104 |
+
{
|
| 239105 |
+
"epoch": 885.56,
|
| 239106 |
+
"learning_rate": 8.242455573505656e-06,
|
| 239107 |
+
"loss": 0.4767,
|
| 239108 |
+
"step": 109835
|
| 239109 |
+
},
|
| 239110 |
+
{
|
| 239111 |
+
"epoch": 885.6,
|
| 239112 |
+
"learning_rate": 8.24237479806139e-06,
|
| 239113 |
+
"loss": 1.2656,
|
| 239114 |
+
"step": 109840
|
| 239115 |
+
},
|
| 239116 |
+
{
|
| 239117 |
+
"epoch": 885.64,
|
| 239118 |
+
"learning_rate": 8.242294022617126e-06,
|
| 239119 |
+
"loss": 0.3519,
|
| 239120 |
+
"step": 109845
|
| 239121 |
+
},
|
| 239122 |
+
{
|
| 239123 |
+
"epoch": 885.68,
|
| 239124 |
+
"learning_rate": 8.24221324717286e-06,
|
| 239125 |
+
"loss": 0.3075,
|
| 239126 |
+
"step": 109850
|
| 239127 |
+
},
|
| 239128 |
+
{
|
| 239129 |
+
"epoch": 885.72,
|
| 239130 |
+
"learning_rate": 8.242132471728596e-06,
|
| 239131 |
+
"loss": 0.2999,
|
| 239132 |
+
"step": 109855
|
| 239133 |
+
},
|
| 239134 |
+
{
|
| 239135 |
+
"epoch": 885.76,
|
| 239136 |
+
"learning_rate": 8.24205169628433e-06,
|
| 239137 |
+
"loss": 0.5727,
|
| 239138 |
+
"step": 109860
|
| 239139 |
+
},
|
| 239140 |
+
{
|
| 239141 |
+
"epoch": 885.8,
|
| 239142 |
+
"learning_rate": 8.241970920840066e-06,
|
| 239143 |
+
"loss": 1.2419,
|
| 239144 |
+
"step": 109865
|
| 239145 |
+
},
|
| 239146 |
+
{
|
| 239147 |
+
"epoch": 885.84,
|
| 239148 |
+
"learning_rate": 8.2418901453958e-06,
|
| 239149 |
+
"loss": 0.3246,
|
| 239150 |
+
"step": 109870
|
| 239151 |
+
},
|
| 239152 |
+
{
|
| 239153 |
+
"epoch": 885.88,
|
| 239154 |
+
"learning_rate": 8.241809369951536e-06,
|
| 239155 |
+
"loss": 0.5255,
|
| 239156 |
+
"step": 109875
|
| 239157 |
+
},
|
| 239158 |
+
{
|
| 239159 |
+
"epoch": 885.92,
|
| 239160 |
+
"learning_rate": 8.24172859450727e-06,
|
| 239161 |
+
"loss": 0.3365,
|
| 239162 |
+
"step": 109880
|
| 239163 |
+
},
|
| 239164 |
+
{
|
| 239165 |
+
"epoch": 885.96,
|
| 239166 |
+
"learning_rate": 8.241647819063006e-06,
|
| 239167 |
+
"loss": 0.467,
|
| 239168 |
+
"step": 109885
|
| 239169 |
+
},
|
| 239170 |
+
{
|
| 239171 |
+
"epoch": 886.0,
|
| 239172 |
+
"eval_loss": 0.37404191493988037,
|
| 239173 |
+
"eval_runtime": 41.5225,
|
| 239174 |
+
"eval_samples_per_second": 20.158,
|
| 239175 |
+
"eval_steps_per_second": 0.65,
|
| 239176 |
+
"eval_wer": 0.17856338662463409,
|
| 239177 |
+
"step": 109889
|
| 239178 |
+
},
|
| 239179 |
+
{
|
| 239180 |
+
"epoch": 886.01,
|
| 239181 |
+
"learning_rate": 8.24156704361874e-06,
|
| 239182 |
+
"loss": 0.4168,
|
| 239183 |
+
"step": 109890
|
| 239184 |
+
},
|
| 239185 |
+
{
|
| 239186 |
+
"epoch": 886.05,
|
| 239187 |
+
"learning_rate": 8.241486268174476e-06,
|
| 239188 |
+
"loss": 0.2944,
|
| 239189 |
+
"step": 109895
|
| 239190 |
+
},
|
| 239191 |
+
{
|
| 239192 |
+
"epoch": 886.09,
|
| 239193 |
+
"learning_rate": 8.241405492730211e-06,
|
| 239194 |
+
"loss": 0.2789,
|
| 239195 |
+
"step": 109900
|
| 239196 |
+
},
|
| 239197 |
+
{
|
| 239198 |
+
"epoch": 886.13,
|
| 239199 |
+
"learning_rate": 8.241324717285946e-06,
|
| 239200 |
+
"loss": 0.3298,
|
| 239201 |
+
"step": 109905
|
| 239202 |
+
},
|
| 239203 |
+
{
|
| 239204 |
+
"epoch": 886.17,
|
| 239205 |
+
"learning_rate": 8.241243941841681e-06,
|
| 239206 |
+
"loss": 0.5337,
|
| 239207 |
+
"step": 109910
|
| 239208 |
+
},
|
| 239209 |
+
{
|
| 239210 |
+
"epoch": 886.21,
|
| 239211 |
+
"learning_rate": 8.241163166397416e-06,
|
| 239212 |
+
"loss": 1.0091,
|
| 239213 |
+
"step": 109915
|
| 239214 |
+
},
|
| 239215 |
+
{
|
| 239216 |
+
"epoch": 886.25,
|
| 239217 |
+
"learning_rate": 8.241082390953151e-06,
|
| 239218 |
+
"loss": 0.3106,
|
| 239219 |
+
"step": 109920
|
| 239220 |
+
},
|
| 239221 |
+
{
|
| 239222 |
+
"epoch": 886.29,
|
| 239223 |
+
"learning_rate": 8.241001615508886e-06,
|
| 239224 |
+
"loss": 0.2778,
|
| 239225 |
+
"step": 109925
|
| 239226 |
+
},
|
| 239227 |
+
{
|
| 239228 |
+
"epoch": 886.33,
|
| 239229 |
+
"learning_rate": 8.240920840064621e-06,
|
| 239230 |
+
"loss": 0.2931,
|
| 239231 |
+
"step": 109930
|
| 239232 |
+
},
|
| 239233 |
+
{
|
| 239234 |
+
"epoch": 886.37,
|
| 239235 |
+
"learning_rate": 8.240840064620356e-06,
|
| 239236 |
+
"loss": 0.506,
|
| 239237 |
+
"step": 109935
|
| 239238 |
+
},
|
| 239239 |
+
{
|
| 239240 |
+
"epoch": 886.41,
|
| 239241 |
+
"learning_rate": 8.240759289176091e-06,
|
| 239242 |
+
"loss": 1.1223,
|
| 239243 |
+
"step": 109940
|
| 239244 |
+
},
|
| 239245 |
+
{
|
| 239246 |
+
"epoch": 886.45,
|
| 239247 |
+
"learning_rate": 8.240678513731825e-06,
|
| 239248 |
+
"loss": 0.3279,
|
| 239249 |
+
"step": 109945
|
| 239250 |
+
},
|
| 239251 |
+
{
|
| 239252 |
+
"epoch": 886.49,
|
| 239253 |
+
"learning_rate": 8.240597738287561e-06,
|
| 239254 |
+
"loss": 0.2782,
|
| 239255 |
+
"step": 109950
|
| 239256 |
+
},
|
| 239257 |
+
{
|
| 239258 |
+
"epoch": 886.53,
|
| 239259 |
+
"learning_rate": 8.240516962843297e-06,
|
| 239260 |
+
"loss": 0.3676,
|
| 239261 |
+
"step": 109955
|
| 239262 |
+
},
|
| 239263 |
+
{
|
| 239264 |
+
"epoch": 886.57,
|
| 239265 |
+
"learning_rate": 8.240436187399031e-06,
|
| 239266 |
+
"loss": 0.5922,
|
| 239267 |
+
"step": 109960
|
| 239268 |
+
},
|
| 239269 |
+
{
|
| 239270 |
+
"epoch": 886.61,
|
| 239271 |
+
"learning_rate": 8.240355411954767e-06,
|
| 239272 |
+
"loss": 1.0638,
|
| 239273 |
+
"step": 109965
|
| 239274 |
+
},
|
| 239275 |
+
{
|
| 239276 |
+
"epoch": 886.65,
|
| 239277 |
+
"learning_rate": 8.240274636510501e-06,
|
| 239278 |
+
"loss": 0.3115,
|
| 239279 |
+
"step": 109970
|
| 239280 |
+
},
|
| 239281 |
+
{
|
| 239282 |
+
"epoch": 886.69,
|
| 239283 |
+
"learning_rate": 8.240193861066237e-06,
|
| 239284 |
+
"loss": 0.3667,
|
| 239285 |
+
"step": 109975
|
| 239286 |
+
},
|
| 239287 |
+
{
|
| 239288 |
+
"epoch": 886.73,
|
| 239289 |
+
"learning_rate": 8.240113085621971e-06,
|
| 239290 |
+
"loss": 0.3261,
|
| 239291 |
+
"step": 109980
|
| 239292 |
+
},
|
| 239293 |
+
{
|
| 239294 |
+
"epoch": 886.77,
|
| 239295 |
+
"learning_rate": 8.240032310177707e-06,
|
| 239296 |
+
"loss": 0.5334,
|
| 239297 |
+
"step": 109985
|
| 239298 |
+
},
|
| 239299 |
+
{
|
| 239300 |
+
"epoch": 886.81,
|
| 239301 |
+
"learning_rate": 8.239951534733441e-06,
|
| 239302 |
+
"loss": 1.115,
|
| 239303 |
+
"step": 109990
|
| 239304 |
+
},
|
| 239305 |
+
{
|
| 239306 |
+
"epoch": 886.85,
|
| 239307 |
+
"learning_rate": 8.239870759289177e-06,
|
| 239308 |
+
"loss": 0.3048,
|
| 239309 |
+
"step": 109995
|
| 239310 |
+
},
|
| 239311 |
+
{
|
| 239312 |
+
"epoch": 886.89,
|
| 239313 |
+
"learning_rate": 8.239789983844911e-06,
|
| 239314 |
+
"loss": 0.3563,
|
| 239315 |
+
"step": 110000
|
| 239316 |
+
},
|
| 239317 |
+
{
|
| 239318 |
+
"epoch": 886.93,
|
| 239319 |
+
"learning_rate": 8.239709208400647e-06,
|
| 239320 |
+
"loss": 0.2975,
|
| 239321 |
+
"step": 110005
|
| 239322 |
+
},
|
| 239323 |
+
{
|
| 239324 |
+
"epoch": 886.97,
|
| 239325 |
+
"learning_rate": 8.239628432956383e-06,
|
| 239326 |
+
"loss": 0.6596,
|
| 239327 |
+
"step": 110010
|
| 239328 |
+
},
|
| 239329 |
+
{
|
| 239330 |
+
"epoch": 887.0,
|
| 239331 |
+
"eval_loss": 0.3918881118297577,
|
| 239332 |
+
"eval_runtime": 42.624,
|
| 239333 |
+
"eval_samples_per_second": 19.637,
|
| 239334 |
+
"eval_steps_per_second": 0.633,
|
| 239335 |
+
"eval_wer": 0.19029392648227053,
|
| 239336 |
+
"step": 110013
|
| 239337 |
+
},
|
| 239338 |
+
{
|
| 239339 |
+
"epoch": 887.02,
|
| 239340 |
+
"learning_rate": 8.239547657512117e-06,
|
| 239341 |
+
"loss": 0.3341,
|
| 239342 |
+
"step": 110015
|
| 239343 |
+
},
|
| 239344 |
+
{
|
| 239345 |
+
"epoch": 887.06,
|
| 239346 |
+
"learning_rate": 8.239466882067853e-06,
|
| 239347 |
+
"loss": 0.2999,
|
| 239348 |
+
"step": 110020
|
| 239349 |
+
},
|
| 239350 |
+
{
|
| 239351 |
+
"epoch": 887.1,
|
| 239352 |
+
"learning_rate": 8.239386106623587e-06,
|
| 239353 |
+
"loss": 0.2885,
|
| 239354 |
+
"step": 110025
|
| 239355 |
+
},
|
| 239356 |
+
{
|
| 239357 |
+
"epoch": 887.14,
|
| 239358 |
+
"learning_rate": 8.239305331179323e-06,
|
| 239359 |
+
"loss": 0.3031,
|
| 239360 |
+
"step": 110030
|
| 239361 |
+
},
|
| 239362 |
+
{
|
| 239363 |
+
"epoch": 887.18,
|
| 239364 |
+
"learning_rate": 8.239224555735057e-06,
|
| 239365 |
+
"loss": 0.6225,
|
| 239366 |
+
"step": 110035
|
| 239367 |
+
},
|
| 239368 |
+
{
|
| 239369 |
+
"epoch": 887.22,
|
| 239370 |
+
"learning_rate": 8.239143780290793e-06,
|
| 239371 |
+
"loss": 0.8657,
|
| 239372 |
+
"step": 110040
|
| 239373 |
+
},
|
| 239374 |
+
{
|
| 239375 |
+
"epoch": 887.26,
|
| 239376 |
+
"learning_rate": 8.239063004846527e-06,
|
| 239377 |
+
"loss": 0.3394,
|
| 239378 |
+
"step": 110045
|
| 239379 |
+
},
|
| 239380 |
+
{
|
| 239381 |
+
"epoch": 887.3,
|
| 239382 |
+
"learning_rate": 8.238982229402263e-06,
|
| 239383 |
+
"loss": 0.2662,
|
| 239384 |
+
"step": 110050
|
| 239385 |
+
},
|
| 239386 |
+
{
|
| 239387 |
+
"epoch": 887.34,
|
| 239388 |
+
"learning_rate": 8.238901453957997e-06,
|
| 239389 |
+
"loss": 0.4114,
|
| 239390 |
+
"step": 110055
|
| 239391 |
+
},
|
| 239392 |
+
{
|
| 239393 |
+
"epoch": 887.38,
|
| 239394 |
+
"learning_rate": 8.238820678513733e-06,
|
| 239395 |
+
"loss": 0.6495,
|
| 239396 |
+
"step": 110060
|
| 239397 |
+
},
|
| 239398 |
+
{
|
| 239399 |
+
"epoch": 887.42,
|
| 239400 |
+
"learning_rate": 8.238739903069467e-06,
|
| 239401 |
+
"loss": 0.991,
|
| 239402 |
+
"step": 110065
|
| 239403 |
+
},
|
| 239404 |
+
{
|
| 239405 |
+
"epoch": 887.46,
|
| 239406 |
+
"learning_rate": 8.238659127625203e-06,
|
| 239407 |
+
"loss": 0.283,
|
| 239408 |
+
"step": 110070
|
| 239409 |
+
},
|
| 239410 |
+
{
|
| 239411 |
+
"epoch": 887.5,
|
| 239412 |
+
"learning_rate": 8.238578352180939e-06,
|
| 239413 |
+
"loss": 0.3429,
|
| 239414 |
+
"step": 110075
|
| 239415 |
+
},
|
| 239416 |
+
{
|
| 239417 |
+
"epoch": 887.54,
|
| 239418 |
+
"learning_rate": 8.238497576736673e-06,
|
| 239419 |
+
"loss": 0.375,
|
| 239420 |
+
"step": 110080
|
| 239421 |
+
},
|
| 239422 |
+
{
|
| 239423 |
+
"epoch": 887.58,
|
| 239424 |
+
"learning_rate": 8.238416801292409e-06,
|
| 239425 |
+
"loss": 0.6509,
|
| 239426 |
+
"step": 110085
|
| 239427 |
+
},
|
| 239428 |
+
{
|
| 239429 |
+
"epoch": 887.62,
|
| 239430 |
+
"learning_rate": 8.238336025848143e-06,
|
| 239431 |
+
"loss": 1.1195,
|
| 239432 |
+
"step": 110090
|
| 239433 |
+
},
|
| 239434 |
+
{
|
| 239435 |
+
"epoch": 887.66,
|
| 239436 |
+
"learning_rate": 8.238255250403878e-06,
|
| 239437 |
+
"loss": 0.2644,
|
| 239438 |
+
"step": 110095
|
| 239439 |
+
},
|
| 239440 |
+
{
|
| 239441 |
+
"epoch": 887.7,
|
| 239442 |
+
"learning_rate": 8.238174474959613e-06,
|
| 239443 |
+
"loss": 0.3223,
|
| 239444 |
+
"step": 110100
|
| 239445 |
+
},
|
| 239446 |
+
{
|
| 239447 |
+
"epoch": 887.74,
|
| 239448 |
+
"learning_rate": 8.238093699515348e-06,
|
| 239449 |
+
"loss": 0.3652,
|
| 239450 |
+
"step": 110105
|
| 239451 |
+
},
|
| 239452 |
+
{
|
| 239453 |
+
"epoch": 887.78,
|
| 239454 |
+
"learning_rate": 8.238012924071083e-06,
|
| 239455 |
+
"loss": 0.6565,
|
| 239456 |
+
"step": 110110
|
| 239457 |
+
},
|
| 239458 |
+
{
|
| 239459 |
+
"epoch": 887.82,
|
| 239460 |
+
"learning_rate": 8.237932148626818e-06,
|
| 239461 |
+
"loss": 1.0233,
|
| 239462 |
+
"step": 110115
|
| 239463 |
+
},
|
| 239464 |
+
{
|
| 239465 |
+
"epoch": 887.86,
|
| 239466 |
+
"learning_rate": 8.237851373182553e-06,
|
| 239467 |
+
"loss": 0.2563,
|
| 239468 |
+
"step": 110120
|
| 239469 |
+
},
|
| 239470 |
+
{
|
| 239471 |
+
"epoch": 887.9,
|
| 239472 |
+
"learning_rate": 8.237770597738288e-06,
|
| 239473 |
+
"loss": 0.2787,
|
| 239474 |
+
"step": 110125
|
| 239475 |
+
},
|
| 239476 |
+
{
|
| 239477 |
+
"epoch": 887.94,
|
| 239478 |
+
"learning_rate": 8.237689822294024e-06,
|
| 239479 |
+
"loss": 0.4034,
|
| 239480 |
+
"step": 110130
|
| 239481 |
+
},
|
| 239482 |
+
{
|
| 239483 |
+
"epoch": 887.98,
|
| 239484 |
+
"learning_rate": 8.237609046849758e-06,
|
| 239485 |
+
"loss": 0.7561,
|
| 239486 |
+
"step": 110135
|
| 239487 |
+
},
|
| 239488 |
+
{
|
| 239489 |
+
"epoch": 888.0,
|
| 239490 |
+
"eval_loss": 0.4561518132686615,
|
| 239491 |
+
"eval_runtime": 43.1073,
|
| 239492 |
+
"eval_samples_per_second": 19.44,
|
| 239493 |
+
"eval_steps_per_second": 0.626,
|
| 239494 |
+
"eval_wer": 0.17663838502047982,
|
| 239495 |
+
"step": 110137
|
| 239496 |
}
|
| 239497 |
],
|
| 239498 |
+
"max_steps": 620000,
|
| 239499 |
"num_train_epochs": 5000,
|
| 239500 |
+
"total_flos": 3.0994849246666765e+20,
|
| 239501 |
"trial_name": null,
|
| 239502 |
"trial_params": null
|
| 239503 |
}
|
model-bin/finetune/base/{checkpoint-109516 → checkpoint-110137}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630040846.5578432/events.out.tfevents.1630040846.52f5c7e305a3.886.51
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e5ca07b144d8a6c23ece37f4697392c280796cd9a65f6d5df772a9a060fabe72
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630041307.6197045/events.out.tfevents.1630041307.52f5c7e305a3.886.53
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9d0781fa9c26422f88f6c672d9db6401640a173514306d1255c59b4fc256cf0e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630041769.276012/events.out.tfevents.1630041769.52f5c7e305a3.886.55
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:82b7eb418836bba011cb377604df8429941e851112cbba4447df18e5bd44464a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630042227.0612097/events.out.tfevents.1630042227.52f5c7e305a3.886.57
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f23fff0c34f9499854ffa000acc2cbae9ee951cde63d5b93cac9ac4eeeeafe5e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630042684.0997107/events.out.tfevents.1630042684.52f5c7e305a3.886.59
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cbaabe387f806c14a23f2afd6d4781fe3ebfa7d269801b774733c8a1347bee96
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630040846.52f5c7e305a3.886.50
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0f6fbe9947d479c3df90fd8bbb97b2bbc4300b17612c4958a14d437ae7f4ae74
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630041307.52f5c7e305a3.886.52
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cd7ff4dc19b683f07e1893cf83ec51d1941671fc6e27db63d6bceb0cfecbf89d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630041769.52f5c7e305a3.886.54
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:314032838069bb5f6176d6ccf91a45d2f870d99a120242103230b911e1d7ee93
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630042227.52f5c7e305a3.886.56
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f4391e1c62d68e301ddd49cd237123586baff02db3bb46af9b2b2641735dd1f9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630042684.52f5c7e305a3.886.58
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9ac5c4e083e9979300c50231c40d20b872e9ed1a6b8f1b1ed4815d30977a6122
|
| 3 |
+
size 8622
|