"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629785047.895205/events.out.tfevents.1629785047.c435e1c5ee04.920.61 +3 -0
- model-bin/finetune/base/log/1629785696.858617/events.out.tfevents.1629785696.c435e1c5ee04.920.63 +3 -0
- model-bin/finetune/base/log/1629786452.6638298/events.out.tfevents.1629786452.c435e1c5ee04.920.65 +3 -0
- model-bin/finetune/base/log/1629787107.1031253/events.out.tfevents.1629787107.c435e1c5ee04.920.67 +3 -0
- model-bin/finetune/base/log/1629787753.7173839/events.out.tfevents.1629787753.c435e1c5ee04.920.69 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629785047.c435e1c5ee04.920.60 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629785696.c435e1c5ee04.920.62 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629786452.c435e1c5ee04.920.64 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629787107.c435e1c5ee04.920.66 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629787753.c435e1c5ee04.920.68 +3 -0
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e15aae07094715fa3940581269509a90f02f966e57838d740a66a09d2497d9b1
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/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:e48222865b2d6155a6f64120f77f66c5f33da659043bc8746f79e1f0a465e1bd
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/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 sha256:3240d1383b02806c1cba54298be39d75b6ab2e708f15feb5955feff6869092ad
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/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 sha256:60a77c8dd462fc3b4be2ff5da7bab717e71264b6dc82fc18d24d6930b1315c4f
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/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:a9d5be5ad540e619a04127837aff42d25861054b44f9bd177ae0dc86a6ff1bfa
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -172671,11 +172671,806 @@
|
|
| 172671 |
"eval_steps_per_second": 0.668,
|
| 172672 |
"eval_wer": 0.19942978287886542,
|
| 172673 |
"step": 57616
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 172674 |
}
|
| 172675 |
],
|
| 172676 |
-
"max_steps":
|
| 172677 |
"num_train_epochs": 5000,
|
| 172678 |
-
"total_flos": 1.
|
| 172679 |
"trial_name": null,
|
| 172680 |
"trial_params": null
|
| 172681 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
+
"epoch": 465.0,
|
| 5 |
+
"global_step": 58240,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 172671 |
"eval_steps_per_second": 0.668,
|
| 172672 |
"eval_wer": 0.19942978287886542,
|
| 172673 |
"step": 57616
|
| 172674 |
+
},
|
| 172675 |
+
{
|
| 172676 |
+
"epoch": 460.03,
|
| 172677 |
+
"learning_rate": 9.092964743589744e-06,
|
| 172678 |
+
"loss": 0.4239,
|
| 172679 |
+
"step": 57620
|
| 172680 |
+
},
|
| 172681 |
+
{
|
| 172682 |
+
"epoch": 460.07,
|
| 172683 |
+
"learning_rate": 9.092884615384617e-06,
|
| 172684 |
+
"loss": 0.3701,
|
| 172685 |
+
"step": 57625
|
| 172686 |
+
},
|
| 172687 |
+
{
|
| 172688 |
+
"epoch": 460.11,
|
| 172689 |
+
"learning_rate": 9.092804487179487e-06,
|
| 172690 |
+
"loss": 0.2995,
|
| 172691 |
+
"step": 57630
|
| 172692 |
+
},
|
| 172693 |
+
{
|
| 172694 |
+
"epoch": 460.15,
|
| 172695 |
+
"learning_rate": 9.09272435897436e-06,
|
| 172696 |
+
"loss": 0.421,
|
| 172697 |
+
"step": 57635
|
| 172698 |
+
},
|
| 172699 |
+
{
|
| 172700 |
+
"epoch": 460.19,
|
| 172701 |
+
"learning_rate": 9.092644230769233e-06,
|
| 172702 |
+
"loss": 0.9221,
|
| 172703 |
+
"step": 57640
|
| 172704 |
+
},
|
| 172705 |
+
{
|
| 172706 |
+
"epoch": 460.23,
|
| 172707 |
+
"learning_rate": 9.092564102564102e-06,
|
| 172708 |
+
"loss": 0.6804,
|
| 172709 |
+
"step": 57645
|
| 172710 |
+
},
|
| 172711 |
+
{
|
| 172712 |
+
"epoch": 460.27,
|
| 172713 |
+
"learning_rate": 9.092483974358975e-06,
|
| 172714 |
+
"loss": 0.3038,
|
| 172715 |
+
"step": 57650
|
| 172716 |
+
},
|
| 172717 |
+
{
|
| 172718 |
+
"epoch": 460.31,
|
| 172719 |
+
"learning_rate": 9.092403846153847e-06,
|
| 172720 |
+
"loss": 0.3249,
|
| 172721 |
+
"step": 57655
|
| 172722 |
+
},
|
| 172723 |
+
{
|
| 172724 |
+
"epoch": 460.35,
|
| 172725 |
+
"learning_rate": 9.092323717948718e-06,
|
| 172726 |
+
"loss": 0.3898,
|
| 172727 |
+
"step": 57660
|
| 172728 |
+
},
|
| 172729 |
+
{
|
| 172730 |
+
"epoch": 460.39,
|
| 172731 |
+
"learning_rate": 9.09224358974359e-06,
|
| 172732 |
+
"loss": 0.9819,
|
| 172733 |
+
"step": 57665
|
| 172734 |
+
},
|
| 172735 |
+
{
|
| 172736 |
+
"epoch": 460.43,
|
| 172737 |
+
"learning_rate": 9.092163461538463e-06,
|
| 172738 |
+
"loss": 0.7288,
|
| 172739 |
+
"step": 57670
|
| 172740 |
+
},
|
| 172741 |
+
{
|
| 172742 |
+
"epoch": 460.47,
|
| 172743 |
+
"learning_rate": 9.092083333333334e-06,
|
| 172744 |
+
"loss": 0.3039,
|
| 172745 |
+
"step": 57675
|
| 172746 |
+
},
|
| 172747 |
+
{
|
| 172748 |
+
"epoch": 460.51,
|
| 172749 |
+
"learning_rate": 9.092003205128205e-06,
|
| 172750 |
+
"loss": 0.406,
|
| 172751 |
+
"step": 57680
|
| 172752 |
+
},
|
| 172753 |
+
{
|
| 172754 |
+
"epoch": 460.55,
|
| 172755 |
+
"learning_rate": 9.091923076923078e-06,
|
| 172756 |
+
"loss": 0.4621,
|
| 172757 |
+
"step": 57685
|
| 172758 |
+
},
|
| 172759 |
+
{
|
| 172760 |
+
"epoch": 460.59,
|
| 172761 |
+
"learning_rate": 9.09184294871795e-06,
|
| 172762 |
+
"loss": 0.9614,
|
| 172763 |
+
"step": 57690
|
| 172764 |
+
},
|
| 172765 |
+
{
|
| 172766 |
+
"epoch": 460.63,
|
| 172767 |
+
"learning_rate": 9.091762820512821e-06,
|
| 172768 |
+
"loss": 0.6961,
|
| 172769 |
+
"step": 57695
|
| 172770 |
+
},
|
| 172771 |
+
{
|
| 172772 |
+
"epoch": 460.67,
|
| 172773 |
+
"learning_rate": 9.091682692307692e-06,
|
| 172774 |
+
"loss": 0.3913,
|
| 172775 |
+
"step": 57700
|
| 172776 |
+
},
|
| 172777 |
+
{
|
| 172778 |
+
"epoch": 460.71,
|
| 172779 |
+
"learning_rate": 9.091602564102565e-06,
|
| 172780 |
+
"loss": 0.3948,
|
| 172781 |
+
"step": 57705
|
| 172782 |
+
},
|
| 172783 |
+
{
|
| 172784 |
+
"epoch": 460.75,
|
| 172785 |
+
"learning_rate": 9.091522435897437e-06,
|
| 172786 |
+
"loss": 0.4743,
|
| 172787 |
+
"step": 57710
|
| 172788 |
+
},
|
| 172789 |
+
{
|
| 172790 |
+
"epoch": 460.79,
|
| 172791 |
+
"learning_rate": 9.091442307692308e-06,
|
| 172792 |
+
"loss": 1.3435,
|
| 172793 |
+
"step": 57715
|
| 172794 |
+
},
|
| 172795 |
+
{
|
| 172796 |
+
"epoch": 460.83,
|
| 172797 |
+
"learning_rate": 9.09136217948718e-06,
|
| 172798 |
+
"loss": 0.7266,
|
| 172799 |
+
"step": 57720
|
| 172800 |
+
},
|
| 172801 |
+
{
|
| 172802 |
+
"epoch": 460.87,
|
| 172803 |
+
"learning_rate": 9.091282051282053e-06,
|
| 172804 |
+
"loss": 0.306,
|
| 172805 |
+
"step": 57725
|
| 172806 |
+
},
|
| 172807 |
+
{
|
| 172808 |
+
"epoch": 460.91,
|
| 172809 |
+
"learning_rate": 9.091201923076924e-06,
|
| 172810 |
+
"loss": 0.364,
|
| 172811 |
+
"step": 57730
|
| 172812 |
+
},
|
| 172813 |
+
{
|
| 172814 |
+
"epoch": 460.95,
|
| 172815 |
+
"learning_rate": 9.091121794871795e-06,
|
| 172816 |
+
"loss": 0.5054,
|
| 172817 |
+
"step": 57735
|
| 172818 |
+
},
|
| 172819 |
+
{
|
| 172820 |
+
"epoch": 460.99,
|
| 172821 |
+
"learning_rate": 9.091041666666668e-06,
|
| 172822 |
+
"loss": 1.1185,
|
| 172823 |
+
"step": 57740
|
| 172824 |
+
},
|
| 172825 |
+
{
|
| 172826 |
+
"epoch": 461.0,
|
| 172827 |
+
"eval_loss": 0.4637983739376068,
|
| 172828 |
+
"eval_runtime": 39.1206,
|
| 172829 |
+
"eval_samples_per_second": 21.472,
|
| 172830 |
+
"eval_steps_per_second": 0.69,
|
| 172831 |
+
"eval_wer": 0.2032720753025549,
|
| 172832 |
+
"step": 57741
|
| 172833 |
+
},
|
| 172834 |
+
{
|
| 172835 |
+
"epoch": 461.03,
|
| 172836 |
+
"learning_rate": 9.09096153846154e-06,
|
| 172837 |
+
"loss": 0.323,
|
| 172838 |
+
"step": 57745
|
| 172839 |
+
},
|
| 172840 |
+
{
|
| 172841 |
+
"epoch": 461.07,
|
| 172842 |
+
"learning_rate": 9.090881410256411e-06,
|
| 172843 |
+
"loss": 0.2823,
|
| 172844 |
+
"step": 57750
|
| 172845 |
+
},
|
| 172846 |
+
{
|
| 172847 |
+
"epoch": 461.11,
|
| 172848 |
+
"learning_rate": 9.090801282051282e-06,
|
| 172849 |
+
"loss": 0.3398,
|
| 172850 |
+
"step": 57755
|
| 172851 |
+
},
|
| 172852 |
+
{
|
| 172853 |
+
"epoch": 461.15,
|
| 172854 |
+
"learning_rate": 9.090721153846155e-06,
|
| 172855 |
+
"loss": 0.5306,
|
| 172856 |
+
"step": 57760
|
| 172857 |
+
},
|
| 172858 |
+
{
|
| 172859 |
+
"epoch": 461.19,
|
| 172860 |
+
"learning_rate": 9.090641025641027e-06,
|
| 172861 |
+
"loss": 0.9327,
|
| 172862 |
+
"step": 57765
|
| 172863 |
+
},
|
| 172864 |
+
{
|
| 172865 |
+
"epoch": 461.23,
|
| 172866 |
+
"learning_rate": 9.090560897435898e-06,
|
| 172867 |
+
"loss": 1.0262,
|
| 172868 |
+
"step": 57770
|
| 172869 |
+
},
|
| 172870 |
+
{
|
| 172871 |
+
"epoch": 461.27,
|
| 172872 |
+
"learning_rate": 9.09048076923077e-06,
|
| 172873 |
+
"loss": 0.2962,
|
| 172874 |
+
"step": 57775
|
| 172875 |
+
},
|
| 172876 |
+
{
|
| 172877 |
+
"epoch": 461.31,
|
| 172878 |
+
"learning_rate": 9.090400641025643e-06,
|
| 172879 |
+
"loss": 0.3981,
|
| 172880 |
+
"step": 57780
|
| 172881 |
+
},
|
| 172882 |
+
{
|
| 172883 |
+
"epoch": 461.35,
|
| 172884 |
+
"learning_rate": 9.090320512820514e-06,
|
| 172885 |
+
"loss": 0.3824,
|
| 172886 |
+
"step": 57785
|
| 172887 |
+
},
|
| 172888 |
+
{
|
| 172889 |
+
"epoch": 461.39,
|
| 172890 |
+
"learning_rate": 9.090240384615385e-06,
|
| 172891 |
+
"loss": 0.9667,
|
| 172892 |
+
"step": 57790
|
| 172893 |
+
},
|
| 172894 |
+
{
|
| 172895 |
+
"epoch": 461.43,
|
| 172896 |
+
"learning_rate": 9.090160256410258e-06,
|
| 172897 |
+
"loss": 0.7123,
|
| 172898 |
+
"step": 57795
|
| 172899 |
+
},
|
| 172900 |
+
{
|
| 172901 |
+
"epoch": 461.47,
|
| 172902 |
+
"learning_rate": 9.090080128205128e-06,
|
| 172903 |
+
"loss": 0.3426,
|
| 172904 |
+
"step": 57800
|
| 172905 |
+
},
|
| 172906 |
+
{
|
| 172907 |
+
"epoch": 461.51,
|
| 172908 |
+
"learning_rate": 9.090000000000001e-06,
|
| 172909 |
+
"loss": 0.2975,
|
| 172910 |
+
"step": 57805
|
| 172911 |
+
},
|
| 172912 |
+
{
|
| 172913 |
+
"epoch": 461.55,
|
| 172914 |
+
"learning_rate": 9.089919871794872e-06,
|
| 172915 |
+
"loss": 0.3892,
|
| 172916 |
+
"step": 57810
|
| 172917 |
+
},
|
| 172918 |
+
{
|
| 172919 |
+
"epoch": 461.59,
|
| 172920 |
+
"learning_rate": 9.089839743589744e-06,
|
| 172921 |
+
"loss": 1.1198,
|
| 172922 |
+
"step": 57815
|
| 172923 |
+
},
|
| 172924 |
+
{
|
| 172925 |
+
"epoch": 461.63,
|
| 172926 |
+
"learning_rate": 9.089759615384615e-06,
|
| 172927 |
+
"loss": 0.7246,
|
| 172928 |
+
"step": 57820
|
| 172929 |
+
},
|
| 172930 |
+
{
|
| 172931 |
+
"epoch": 461.67,
|
| 172932 |
+
"learning_rate": 9.089679487179488e-06,
|
| 172933 |
+
"loss": 0.3002,
|
| 172934 |
+
"step": 57825
|
| 172935 |
+
},
|
| 172936 |
+
{
|
| 172937 |
+
"epoch": 461.71,
|
| 172938 |
+
"learning_rate": 9.08959935897436e-06,
|
| 172939 |
+
"loss": 0.3869,
|
| 172940 |
+
"step": 57830
|
| 172941 |
+
},
|
| 172942 |
+
{
|
| 172943 |
+
"epoch": 461.75,
|
| 172944 |
+
"learning_rate": 9.089519230769231e-06,
|
| 172945 |
+
"loss": 0.4678,
|
| 172946 |
+
"step": 57835
|
| 172947 |
+
},
|
| 172948 |
+
{
|
| 172949 |
+
"epoch": 461.79,
|
| 172950 |
+
"learning_rate": 9.089439102564104e-06,
|
| 172951 |
+
"loss": 1.04,
|
| 172952 |
+
"step": 57840
|
| 172953 |
+
},
|
| 172954 |
+
{
|
| 172955 |
+
"epoch": 461.83,
|
| 172956 |
+
"learning_rate": 9.089358974358975e-06,
|
| 172957 |
+
"loss": 0.7155,
|
| 172958 |
+
"step": 57845
|
| 172959 |
+
},
|
| 172960 |
+
{
|
| 172961 |
+
"epoch": 461.87,
|
| 172962 |
+
"learning_rate": 9.089278846153847e-06,
|
| 172963 |
+
"loss": 0.2987,
|
| 172964 |
+
"step": 57850
|
| 172965 |
+
},
|
| 172966 |
+
{
|
| 172967 |
+
"epoch": 461.91,
|
| 172968 |
+
"learning_rate": 9.089198717948718e-06,
|
| 172969 |
+
"loss": 0.3488,
|
| 172970 |
+
"step": 57855
|
| 172971 |
+
},
|
| 172972 |
+
{
|
| 172973 |
+
"epoch": 461.95,
|
| 172974 |
+
"learning_rate": 9.089118589743591e-06,
|
| 172975 |
+
"loss": 0.4747,
|
| 172976 |
+
"step": 57860
|
| 172977 |
+
},
|
| 172978 |
+
{
|
| 172979 |
+
"epoch": 461.99,
|
| 172980 |
+
"learning_rate": 9.089038461538462e-06,
|
| 172981 |
+
"loss": 1.099,
|
| 172982 |
+
"step": 57865
|
| 172983 |
+
},
|
| 172984 |
+
{
|
| 172985 |
+
"epoch": 462.0,
|
| 172986 |
+
"eval_loss": 0.45111995935440063,
|
| 172987 |
+
"eval_runtime": 39.3167,
|
| 172988 |
+
"eval_samples_per_second": 21.34,
|
| 172989 |
+
"eval_steps_per_second": 0.687,
|
| 172990 |
+
"eval_wer": 0.20354430379746835,
|
| 172991 |
+
"step": 57866
|
| 172992 |
+
},
|
| 172993 |
+
{
|
| 172994 |
+
"epoch": 466.03,
|
| 172995 |
+
"learning_rate": 9.088958333333334e-06,
|
| 172996 |
+
"loss": 0.4388,
|
| 172997 |
+
"step": 57870
|
| 172998 |
+
},
|
| 172999 |
+
{
|
| 173000 |
+
"epoch": 466.07,
|
| 173001 |
+
"learning_rate": 9.088878205128205e-06,
|
| 173002 |
+
"loss": 0.3305,
|
| 173003 |
+
"step": 57875
|
| 173004 |
+
},
|
| 173005 |
+
{
|
| 173006 |
+
"epoch": 466.11,
|
| 173007 |
+
"learning_rate": 9.088798076923078e-06,
|
| 173008 |
+
"loss": 0.3715,
|
| 173009 |
+
"step": 57880
|
| 173010 |
+
},
|
| 173011 |
+
{
|
| 173012 |
+
"epoch": 466.15,
|
| 173013 |
+
"learning_rate": 9.08871794871795e-06,
|
| 173014 |
+
"loss": 0.3612,
|
| 173015 |
+
"step": 57885
|
| 173016 |
+
},
|
| 173017 |
+
{
|
| 173018 |
+
"epoch": 466.19,
|
| 173019 |
+
"learning_rate": 9.088637820512821e-06,
|
| 173020 |
+
"loss": 1.1466,
|
| 173021 |
+
"step": 57890
|
| 173022 |
+
},
|
| 173023 |
+
{
|
| 173024 |
+
"epoch": 466.23,
|
| 173025 |
+
"learning_rate": 9.088557692307694e-06,
|
| 173026 |
+
"loss": 0.765,
|
| 173027 |
+
"step": 57895
|
| 173028 |
+
},
|
| 173029 |
+
{
|
| 173030 |
+
"epoch": 466.27,
|
| 173031 |
+
"learning_rate": 9.088477564102565e-06,
|
| 173032 |
+
"loss": 0.3008,
|
| 173033 |
+
"step": 57900
|
| 173034 |
+
},
|
| 173035 |
+
{
|
| 173036 |
+
"epoch": 466.31,
|
| 173037 |
+
"learning_rate": 9.088397435897437e-06,
|
| 173038 |
+
"loss": 0.381,
|
| 173039 |
+
"step": 57905
|
| 173040 |
+
},
|
| 173041 |
+
{
|
| 173042 |
+
"epoch": 466.35,
|
| 173043 |
+
"learning_rate": 9.088317307692308e-06,
|
| 173044 |
+
"loss": 0.4673,
|
| 173045 |
+
"step": 57910
|
| 173046 |
+
},
|
| 173047 |
+
{
|
| 173048 |
+
"epoch": 466.39,
|
| 173049 |
+
"learning_rate": 9.088237179487181e-06,
|
| 173050 |
+
"loss": 0.9081,
|
| 173051 |
+
"step": 57915
|
| 173052 |
+
},
|
| 173053 |
+
{
|
| 173054 |
+
"epoch": 466.43,
|
| 173055 |
+
"learning_rate": 9.08815705128205e-06,
|
| 173056 |
+
"loss": 0.6865,
|
| 173057 |
+
"step": 57920
|
| 173058 |
+
},
|
| 173059 |
+
{
|
| 173060 |
+
"epoch": 466.47,
|
| 173061 |
+
"learning_rate": 9.088076923076924e-06,
|
| 173062 |
+
"loss": 0.3149,
|
| 173063 |
+
"step": 57925
|
| 173064 |
+
},
|
| 173065 |
+
{
|
| 173066 |
+
"epoch": 466.51,
|
| 173067 |
+
"learning_rate": 9.087996794871795e-06,
|
| 173068 |
+
"loss": 0.3114,
|
| 173069 |
+
"step": 57930
|
| 173070 |
+
},
|
| 173071 |
+
{
|
| 173072 |
+
"epoch": 466.55,
|
| 173073 |
+
"learning_rate": 9.087916666666667e-06,
|
| 173074 |
+
"loss": 0.4957,
|
| 173075 |
+
"step": 57935
|
| 173076 |
+
},
|
| 173077 |
+
{
|
| 173078 |
+
"epoch": 466.59,
|
| 173079 |
+
"learning_rate": 9.08783653846154e-06,
|
| 173080 |
+
"loss": 1.0383,
|
| 173081 |
+
"step": 57940
|
| 173082 |
+
},
|
| 173083 |
+
{
|
| 173084 |
+
"epoch": 466.63,
|
| 173085 |
+
"learning_rate": 9.087756410256411e-06,
|
| 173086 |
+
"loss": 0.6517,
|
| 173087 |
+
"step": 57945
|
| 173088 |
+
},
|
| 173089 |
+
{
|
| 173090 |
+
"epoch": 466.67,
|
| 173091 |
+
"learning_rate": 9.087676282051282e-06,
|
| 173092 |
+
"loss": 0.3073,
|
| 173093 |
+
"step": 57950
|
| 173094 |
+
},
|
| 173095 |
+
{
|
| 173096 |
+
"epoch": 466.71,
|
| 173097 |
+
"learning_rate": 9.087596153846154e-06,
|
| 173098 |
+
"loss": 0.4306,
|
| 173099 |
+
"step": 57955
|
| 173100 |
+
},
|
| 173101 |
+
{
|
| 173102 |
+
"epoch": 466.76,
|
| 173103 |
+
"learning_rate": 9.087516025641027e-06,
|
| 173104 |
+
"loss": 0.4356,
|
| 173105 |
+
"step": 57960
|
| 173106 |
+
},
|
| 173107 |
+
{
|
| 173108 |
+
"epoch": 466.8,
|
| 173109 |
+
"learning_rate": 9.087435897435898e-06,
|
| 173110 |
+
"loss": 0.9566,
|
| 173111 |
+
"step": 57965
|
| 173112 |
+
},
|
| 173113 |
+
{
|
| 173114 |
+
"epoch": 466.84,
|
| 173115 |
+
"learning_rate": 9.08735576923077e-06,
|
| 173116 |
+
"loss": 0.9934,
|
| 173117 |
+
"step": 57970
|
| 173118 |
+
},
|
| 173119 |
+
{
|
| 173120 |
+
"epoch": 466.88,
|
| 173121 |
+
"learning_rate": 9.08727564102564e-06,
|
| 173122 |
+
"loss": 0.3725,
|
| 173123 |
+
"step": 57975
|
| 173124 |
+
},
|
| 173125 |
+
{
|
| 173126 |
+
"epoch": 466.92,
|
| 173127 |
+
"learning_rate": 9.087195512820514e-06,
|
| 173128 |
+
"loss": 0.4015,
|
| 173129 |
+
"step": 57980
|
| 173130 |
+
},
|
| 173131 |
+
{
|
| 173132 |
+
"epoch": 466.96,
|
| 173133 |
+
"learning_rate": 9.087115384615385e-06,
|
| 173134 |
+
"loss": 0.4956,
|
| 173135 |
+
"step": 57985
|
| 173136 |
+
},
|
| 173137 |
+
{
|
| 173138 |
+
"epoch": 467.0,
|
| 173139 |
+
"learning_rate": 9.087035256410257e-06,
|
| 173140 |
+
"loss": 1.213,
|
| 173141 |
+
"step": 57990
|
| 173142 |
+
},
|
| 173143 |
+
{
|
| 173144 |
+
"epoch": 467.0,
|
| 173145 |
+
"eval_loss": 0.3933221995830536,
|
| 173146 |
+
"eval_runtime": 38.1322,
|
| 173147 |
+
"eval_samples_per_second": 21.976,
|
| 173148 |
+
"eval_steps_per_second": 0.708,
|
| 173149 |
+
"eval_wer": 0.20183753828204753,
|
| 173150 |
+
"step": 57990
|
| 173151 |
+
},
|
| 173152 |
+
{
|
| 173153 |
+
"epoch": 463.04,
|
| 173154 |
+
"learning_rate": 9.08695512820513e-06,
|
| 173155 |
+
"loss": 0.3396,
|
| 173156 |
+
"step": 57995
|
| 173157 |
+
},
|
| 173158 |
+
{
|
| 173159 |
+
"epoch": 463.08,
|
| 173160 |
+
"learning_rate": 9.086875000000001e-06,
|
| 173161 |
+
"loss": 0.3573,
|
| 173162 |
+
"step": 58000
|
| 173163 |
+
},
|
| 173164 |
+
{
|
| 173165 |
+
"epoch": 463.12,
|
| 173166 |
+
"learning_rate": 9.086794871794872e-06,
|
| 173167 |
+
"loss": 0.3316,
|
| 173168 |
+
"step": 58005
|
| 173169 |
+
},
|
| 173170 |
+
{
|
| 173171 |
+
"epoch": 463.16,
|
| 173172 |
+
"learning_rate": 9.086714743589744e-06,
|
| 173173 |
+
"loss": 0.4752,
|
| 173174 |
+
"step": 58010
|
| 173175 |
+
},
|
| 173176 |
+
{
|
| 173177 |
+
"epoch": 463.2,
|
| 173178 |
+
"learning_rate": 9.086634615384617e-06,
|
| 173179 |
+
"loss": 1.2273,
|
| 173180 |
+
"step": 58015
|
| 173181 |
+
},
|
| 173182 |
+
{
|
| 173183 |
+
"epoch": 463.24,
|
| 173184 |
+
"learning_rate": 9.086554487179488e-06,
|
| 173185 |
+
"loss": 0.3782,
|
| 173186 |
+
"step": 58020
|
| 173187 |
+
},
|
| 173188 |
+
{
|
| 173189 |
+
"epoch": 463.28,
|
| 173190 |
+
"learning_rate": 9.08647435897436e-06,
|
| 173191 |
+
"loss": 0.3168,
|
| 173192 |
+
"step": 58025
|
| 173193 |
+
},
|
| 173194 |
+
{
|
| 173195 |
+
"epoch": 463.32,
|
| 173196 |
+
"learning_rate": 9.08639423076923e-06,
|
| 173197 |
+
"loss": 0.3448,
|
| 173198 |
+
"step": 58030
|
| 173199 |
+
},
|
| 173200 |
+
{
|
| 173201 |
+
"epoch": 463.36,
|
| 173202 |
+
"learning_rate": 9.086314102564104e-06,
|
| 173203 |
+
"loss": 0.4827,
|
| 173204 |
+
"step": 58035
|
| 173205 |
+
},
|
| 173206 |
+
{
|
| 173207 |
+
"epoch": 463.4,
|
| 173208 |
+
"learning_rate": 9.086233974358975e-06,
|
| 173209 |
+
"loss": 1.3107,
|
| 173210 |
+
"step": 58040
|
| 173211 |
+
},
|
| 173212 |
+
{
|
| 173213 |
+
"epoch": 463.44,
|
| 173214 |
+
"learning_rate": 9.086153846153847e-06,
|
| 173215 |
+
"loss": 0.3785,
|
| 173216 |
+
"step": 58045
|
| 173217 |
+
},
|
| 173218 |
+
{
|
| 173219 |
+
"epoch": 463.48,
|
| 173220 |
+
"learning_rate": 9.08607371794872e-06,
|
| 173221 |
+
"loss": 0.2864,
|
| 173222 |
+
"step": 58050
|
| 173223 |
+
},
|
| 173224 |
+
{
|
| 173225 |
+
"epoch": 463.52,
|
| 173226 |
+
"learning_rate": 9.085993589743591e-06,
|
| 173227 |
+
"loss": 0.3458,
|
| 173228 |
+
"step": 58055
|
| 173229 |
+
},
|
| 173230 |
+
{
|
| 173231 |
+
"epoch": 463.56,
|
| 173232 |
+
"learning_rate": 9.085913461538462e-06,
|
| 173233 |
+
"loss": 0.4413,
|
| 173234 |
+
"step": 58060
|
| 173235 |
+
},
|
| 173236 |
+
{
|
| 173237 |
+
"epoch": 463.6,
|
| 173238 |
+
"learning_rate": 9.085833333333334e-06,
|
| 173239 |
+
"loss": 1.0885,
|
| 173240 |
+
"step": 58065
|
| 173241 |
+
},
|
| 173242 |
+
{
|
| 173243 |
+
"epoch": 463.64,
|
| 173244 |
+
"learning_rate": 9.085753205128207e-06,
|
| 173245 |
+
"loss": 0.3904,
|
| 173246 |
+
"step": 58070
|
| 173247 |
+
},
|
| 173248 |
+
{
|
| 173249 |
+
"epoch": 463.68,
|
| 173250 |
+
"learning_rate": 9.085673076923076e-06,
|
| 173251 |
+
"loss": 0.328,
|
| 173252 |
+
"step": 58075
|
| 173253 |
+
},
|
| 173254 |
+
{
|
| 173255 |
+
"epoch": 463.72,
|
| 173256 |
+
"learning_rate": 9.08559294871795e-06,
|
| 173257 |
+
"loss": 0.3957,
|
| 173258 |
+
"step": 58080
|
| 173259 |
+
},
|
| 173260 |
+
{
|
| 173261 |
+
"epoch": 463.76,
|
| 173262 |
+
"learning_rate": 9.085512820512823e-06,
|
| 173263 |
+
"loss": 0.5892,
|
| 173264 |
+
"step": 58085
|
| 173265 |
+
},
|
| 173266 |
+
{
|
| 173267 |
+
"epoch": 463.8,
|
| 173268 |
+
"learning_rate": 9.085432692307692e-06,
|
| 173269 |
+
"loss": 1.2747,
|
| 173270 |
+
"step": 58090
|
| 173271 |
+
},
|
| 173272 |
+
{
|
| 173273 |
+
"epoch": 463.84,
|
| 173274 |
+
"learning_rate": 9.085352564102565e-06,
|
| 173275 |
+
"loss": 0.38,
|
| 173276 |
+
"step": 58095
|
| 173277 |
+
},
|
| 173278 |
+
{
|
| 173279 |
+
"epoch": 463.88,
|
| 173280 |
+
"learning_rate": 9.085272435897437e-06,
|
| 173281 |
+
"loss": 0.3201,
|
| 173282 |
+
"step": 58100
|
| 173283 |
+
},
|
| 173284 |
+
{
|
| 173285 |
+
"epoch": 463.92,
|
| 173286 |
+
"learning_rate": 9.085192307692308e-06,
|
| 173287 |
+
"loss": 0.3769,
|
| 173288 |
+
"step": 58105
|
| 173289 |
+
},
|
| 173290 |
+
{
|
| 173291 |
+
"epoch": 463.96,
|
| 173292 |
+
"learning_rate": 9.08511217948718e-06,
|
| 173293 |
+
"loss": 0.527,
|
| 173294 |
+
"step": 58110
|
| 173295 |
+
},
|
| 173296 |
+
{
|
| 173297 |
+
"epoch": 464.0,
|
| 173298 |
+
"learning_rate": 9.085048076923077e-06,
|
| 173299 |
+
"loss": 1.9565,
|
| 173300 |
+
"step": 58115
|
| 173301 |
+
},
|
| 173302 |
+
{
|
| 173303 |
+
"epoch": 464.0,
|
| 173304 |
+
"eval_loss": 0.3823895752429962,
|
| 173305 |
+
"eval_runtime": 39.7863,
|
| 173306 |
+
"eval_samples_per_second": 21.088,
|
| 173307 |
+
"eval_steps_per_second": 0.679,
|
| 173308 |
+
"eval_wer": 0.19843148860609647,
|
| 173309 |
+
"step": 58115
|
| 173310 |
+
},
|
| 173311 |
+
{
|
| 173312 |
+
"epoch": 464.04,
|
| 173313 |
+
"learning_rate": 9.08496794871795e-06,
|
| 173314 |
+
"loss": 0.3675,
|
| 173315 |
+
"step": 58120
|
| 173316 |
+
},
|
| 173317 |
+
{
|
| 173318 |
+
"epoch": 464.08,
|
| 173319 |
+
"learning_rate": 9.084887820512821e-06,
|
| 173320 |
+
"loss": 0.3512,
|
| 173321 |
+
"step": 58125
|
| 173322 |
+
},
|
| 173323 |
+
{
|
| 173324 |
+
"epoch": 464.12,
|
| 173325 |
+
"learning_rate": 9.084807692307693e-06,
|
| 173326 |
+
"loss": 0.3651,
|
| 173327 |
+
"step": 58130
|
| 173328 |
+
},
|
| 173329 |
+
{
|
| 173330 |
+
"epoch": 464.16,
|
| 173331 |
+
"learning_rate": 9.084727564102564e-06,
|
| 173332 |
+
"loss": 0.492,
|
| 173333 |
+
"step": 58135
|
| 173334 |
+
},
|
| 173335 |
+
{
|
| 173336 |
+
"epoch": 464.2,
|
| 173337 |
+
"learning_rate": 9.084647435897437e-06,
|
| 173338 |
+
"loss": 1.3517,
|
| 173339 |
+
"step": 58140
|
| 173340 |
+
},
|
| 173341 |
+
{
|
| 173342 |
+
"epoch": 464.24,
|
| 173343 |
+
"learning_rate": 9.084567307692308e-06,
|
| 173344 |
+
"loss": 0.3766,
|
| 173345 |
+
"step": 58145
|
| 173346 |
+
},
|
| 173347 |
+
{
|
| 173348 |
+
"epoch": 464.28,
|
| 173349 |
+
"learning_rate": 9.08448717948718e-06,
|
| 173350 |
+
"loss": 0.3302,
|
| 173351 |
+
"step": 58150
|
| 173352 |
+
},
|
| 173353 |
+
{
|
| 173354 |
+
"epoch": 464.32,
|
| 173355 |
+
"learning_rate": 9.084407051282053e-06,
|
| 173356 |
+
"loss": 0.4561,
|
| 173357 |
+
"step": 58155
|
| 173358 |
+
},
|
| 173359 |
+
{
|
| 173360 |
+
"epoch": 464.36,
|
| 173361 |
+
"learning_rate": 9.084326923076924e-06,
|
| 173362 |
+
"loss": 0.5971,
|
| 173363 |
+
"step": 58160
|
| 173364 |
+
},
|
| 173365 |
+
{
|
| 173366 |
+
"epoch": 464.4,
|
| 173367 |
+
"learning_rate": 9.084246794871795e-06,
|
| 173368 |
+
"loss": 1.2831,
|
| 173369 |
+
"step": 58165
|
| 173370 |
+
},
|
| 173371 |
+
{
|
| 173372 |
+
"epoch": 464.44,
|
| 173373 |
+
"learning_rate": 9.084166666666667e-06,
|
| 173374 |
+
"loss": 0.3836,
|
| 173375 |
+
"step": 58170
|
| 173376 |
+
},
|
| 173377 |
+
{
|
| 173378 |
+
"epoch": 464.48,
|
| 173379 |
+
"learning_rate": 9.08408653846154e-06,
|
| 173380 |
+
"loss": 0.3569,
|
| 173381 |
+
"step": 58175
|
| 173382 |
+
},
|
| 173383 |
+
{
|
| 173384 |
+
"epoch": 464.52,
|
| 173385 |
+
"learning_rate": 9.084006410256411e-06,
|
| 173386 |
+
"loss": 0.359,
|
| 173387 |
+
"step": 58180
|
| 173388 |
+
},
|
| 173389 |
+
{
|
| 173390 |
+
"epoch": 464.56,
|
| 173391 |
+
"learning_rate": 9.083926282051283e-06,
|
| 173392 |
+
"loss": 0.4872,
|
| 173393 |
+
"step": 58185
|
| 173394 |
+
},
|
| 173395 |
+
{
|
| 173396 |
+
"epoch": 464.6,
|
| 173397 |
+
"learning_rate": 9.083846153846154e-06,
|
| 173398 |
+
"loss": 1.2145,
|
| 173399 |
+
"step": 58190
|
| 173400 |
+
},
|
| 173401 |
+
{
|
| 173402 |
+
"epoch": 464.64,
|
| 173403 |
+
"learning_rate": 9.083766025641027e-06,
|
| 173404 |
+
"loss": 0.3238,
|
| 173405 |
+
"step": 58195
|
| 173406 |
+
},
|
| 173407 |
+
{
|
| 173408 |
+
"epoch": 464.68,
|
| 173409 |
+
"learning_rate": 9.083685897435898e-06,
|
| 173410 |
+
"loss": 0.3827,
|
| 173411 |
+
"step": 58200
|
| 173412 |
+
},
|
| 173413 |
+
{
|
| 173414 |
+
"epoch": 464.72,
|
| 173415 |
+
"learning_rate": 9.08360576923077e-06,
|
| 173416 |
+
"loss": 0.3654,
|
| 173417 |
+
"step": 58205
|
| 173418 |
+
},
|
| 173419 |
+
{
|
| 173420 |
+
"epoch": 464.76,
|
| 173421 |
+
"learning_rate": 9.083525641025643e-06,
|
| 173422 |
+
"loss": 0.8652,
|
| 173423 |
+
"step": 58210
|
| 173424 |
+
},
|
| 173425 |
+
{
|
| 173426 |
+
"epoch": 464.8,
|
| 173427 |
+
"learning_rate": 9.083445512820514e-06,
|
| 173428 |
+
"loss": 1.2272,
|
| 173429 |
+
"step": 58215
|
| 173430 |
+
},
|
| 173431 |
+
{
|
| 173432 |
+
"epoch": 464.84,
|
| 173433 |
+
"learning_rate": 9.083365384615385e-06,
|
| 173434 |
+
"loss": 0.3873,
|
| 173435 |
+
"step": 58220
|
| 173436 |
+
},
|
| 173437 |
+
{
|
| 173438 |
+
"epoch": 464.88,
|
| 173439 |
+
"learning_rate": 9.083285256410257e-06,
|
| 173440 |
+
"loss": 0.3252,
|
| 173441 |
+
"step": 58225
|
| 173442 |
+
},
|
| 173443 |
+
{
|
| 173444 |
+
"epoch": 464.92,
|
| 173445 |
+
"learning_rate": 9.08320512820513e-06,
|
| 173446 |
+
"loss": 0.3194,
|
| 173447 |
+
"step": 58230
|
| 173448 |
+
},
|
| 173449 |
+
{
|
| 173450 |
+
"epoch": 464.96,
|
| 173451 |
+
"learning_rate": 9.083125000000001e-06,
|
| 173452 |
+
"loss": 0.6022,
|
| 173453 |
+
"step": 58235
|
| 173454 |
+
},
|
| 173455 |
+
{
|
| 173456 |
+
"epoch": 465.0,
|
| 173457 |
+
"learning_rate": 9.083044871794873e-06,
|
| 173458 |
+
"loss": 1.5168,
|
| 173459 |
+
"step": 58240
|
| 173460 |
+
},
|
| 173461 |
+
{
|
| 173462 |
+
"epoch": 465.0,
|
| 173463 |
+
"eval_loss": 0.40367698669433594,
|
| 173464 |
+
"eval_runtime": 39.9396,
|
| 173465 |
+
"eval_samples_per_second": 20.982,
|
| 173466 |
+
"eval_steps_per_second": 0.676,
|
| 173467 |
+
"eval_wer": 0.1968630900029317,
|
| 173468 |
+
"step": 58240
|
| 173469 |
}
|
| 173470 |
],
|
| 173471 |
+
"max_steps": 625000,
|
| 173472 |
"num_train_epochs": 5000,
|
| 173473 |
+
"total_flos": 1.6390003887200823e+20,
|
| 173474 |
"trial_name": null,
|
| 173475 |
"trial_params": null
|
| 173476 |
}
|
model-bin/finetune/base/{checkpoint-57616 → checkpoint-58240}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629785047.895205/events.out.tfevents.1629785047.c435e1c5ee04.920.61
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bb0db194d4972093f8c5ded322d388924e9457cf025c30dd40534473c35e16fc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629785696.858617/events.out.tfevents.1629785696.c435e1c5ee04.920.63
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7552fef22df443e2d6f649cea4b3dcbd220d2376e22bd7ef5ed2f3f00dc8c4c4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629786452.6638298/events.out.tfevents.1629786452.c435e1c5ee04.920.65
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0663b0b812fd1fa8bef37c53ff85c81d695899a3e7f5d8b31d8a599aba91220
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629787107.1031253/events.out.tfevents.1629787107.c435e1c5ee04.920.67
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:de88d4c6aef72569e1c54491e68ffe659cae73ab960acdbaea07ad4ffe971d26
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629787753.7173839/events.out.tfevents.1629787753.c435e1c5ee04.920.69
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:554545fa414ed37f1c9e3dfdad446660a7239c4a7c46a5d74d016dd030f91113
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629785047.c435e1c5ee04.920.60
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:220bdf791a40351df2e1f25a63c84ab384315fa87c4fad5f491042dc9853841e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629785696.c435e1c5ee04.920.62
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3c067fb82867c23af6dfeea776b188723bd234e3a27a16e3420da82ef27048cf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629786452.c435e1c5ee04.920.64
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:11e73963bd21cdf1c71106542149b4cddfb987c1fd2d4f35f460b4138c802ae5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629787107.c435e1c5ee04.920.66
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e33f81bf23c8cfdd8b6ef405d5e12e729c3ab2d624b295cc5ad8f42f323608ac
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629787753.c435e1c5ee04.920.68
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e8cbf06b2afeb36be9e9b0abe26fa41bf9f44c9971ba157b6df76cd5fc1015df
|
| 3 |
+
size 8622
|