"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629768310.8799858/events.out.tfevents.1629768310.c435e1c5ee04.920.11 +3 -0
- model-bin/finetune/base/log/1629768958.858682/events.out.tfevents.1629768958.c435e1c5ee04.920.13 +3 -0
- model-bin/finetune/base/log/1629769702.6550722/events.out.tfevents.1629769702.c435e1c5ee04.920.15 +3 -0
- model-bin/finetune/base/log/1629770366.8003614/events.out.tfevents.1629770366.c435e1c5ee04.920.17 +3 -0
- model-bin/finetune/base/log/1629771017.73505/events.out.tfevents.1629771017.c435e1c5ee04.920.19 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629768310.c435e1c5ee04.920.10 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629768958.c435e1c5ee04.920.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629769702.c435e1c5ee04.920.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629770366.c435e1c5ee04.920.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629771017.c435e1c5ee04.920.18 +3 -0
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/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 sha256:9a127b791c0d77b579a5e491584452669368364880890f5818cdc61e22e15559
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/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:f0790fa6fb916d8cbb7e32fc2c62611b3f881e48d0f4439e53ce8ba15b93bfa8
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14567
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2e24db58e87b89cd5a8e4c2bb7f104406fb2f99771af6ab8029b177aa98488cd
|
| 3 |
size 14567
|
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/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 sha256:1ca3c56116107871279dfa4f94323a31be7725479169e3fda63b787a410c07df
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/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:e9bca6cf836e457ed4761e4f9061cf21b21bdda6e37eae835e5df9405f49e54a
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18992848189928482,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-52515",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -168714,11 +168714,800 @@
|
|
| 168714 |
"eval_steps_per_second": 0.692,
|
| 168715 |
"eval_wer": 0.19587552664646316,
|
| 168716 |
"step": 54505
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 168717 |
}
|
| 168718 |
],
|
| 168719 |
-
"max_steps":
|
| 168720 |
"num_train_epochs": 5000,
|
| 168721 |
-
"total_flos": 1.
|
| 168722 |
"trial_name": null,
|
| 168723 |
"trial_params": null
|
| 168724 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18992848189928482,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-52515",
|
| 4 |
+
"epoch": 441.0,
|
| 5 |
+
"global_step": 55127,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 168714 |
"eval_steps_per_second": 0.692,
|
| 168715 |
"eval_wer": 0.19587552664646316,
|
| 168716 |
"step": 54505
|
| 168717 |
+
},
|
| 168718 |
+
{
|
| 168719 |
+
"epoch": 439.04,
|
| 168720 |
+
"learning_rate": 9.142804487179488e-06,
|
| 168721 |
+
"loss": 0.3981,
|
| 168722 |
+
"step": 54510
|
| 168723 |
+
},
|
| 168724 |
+
{
|
| 168725 |
+
"epoch": 439.08,
|
| 168726 |
+
"learning_rate": 9.142724358974359e-06,
|
| 168727 |
+
"loss": 0.3357,
|
| 168728 |
+
"step": 54515
|
| 168729 |
+
},
|
| 168730 |
+
{
|
| 168731 |
+
"epoch": 439.12,
|
| 168732 |
+
"learning_rate": 9.142644230769232e-06,
|
| 168733 |
+
"loss": 0.3933,
|
| 168734 |
+
"step": 54520
|
| 168735 |
+
},
|
| 168736 |
+
{
|
| 168737 |
+
"epoch": 439.16,
|
| 168738 |
+
"learning_rate": 9.142564102564104e-06,
|
| 168739 |
+
"loss": 0.464,
|
| 168740 |
+
"step": 54525
|
| 168741 |
+
},
|
| 168742 |
+
{
|
| 168743 |
+
"epoch": 439.2,
|
| 168744 |
+
"learning_rate": 9.142483974358975e-06,
|
| 168745 |
+
"loss": 1.2545,
|
| 168746 |
+
"step": 54530
|
| 168747 |
+
},
|
| 168748 |
+
{
|
| 168749 |
+
"epoch": 439.24,
|
| 168750 |
+
"learning_rate": 9.142403846153846e-06,
|
| 168751 |
+
"loss": 0.4474,
|
| 168752 |
+
"step": 54535
|
| 168753 |
+
},
|
| 168754 |
+
{
|
| 168755 |
+
"epoch": 439.28,
|
| 168756 |
+
"learning_rate": 9.14232371794872e-06,
|
| 168757 |
+
"loss": 0.2817,
|
| 168758 |
+
"step": 54540
|
| 168759 |
+
},
|
| 168760 |
+
{
|
| 168761 |
+
"epoch": 439.32,
|
| 168762 |
+
"learning_rate": 9.14224358974359e-06,
|
| 168763 |
+
"loss": 0.3687,
|
| 168764 |
+
"step": 54545
|
| 168765 |
+
},
|
| 168766 |
+
{
|
| 168767 |
+
"epoch": 439.36,
|
| 168768 |
+
"learning_rate": 9.142163461538462e-06,
|
| 168769 |
+
"loss": 0.5966,
|
| 168770 |
+
"step": 54550
|
| 168771 |
+
},
|
| 168772 |
+
{
|
| 168773 |
+
"epoch": 439.4,
|
| 168774 |
+
"learning_rate": 9.142083333333333e-06,
|
| 168775 |
+
"loss": 1.4357,
|
| 168776 |
+
"step": 54555
|
| 168777 |
+
},
|
| 168778 |
+
{
|
| 168779 |
+
"epoch": 439.44,
|
| 168780 |
+
"learning_rate": 9.142003205128206e-06,
|
| 168781 |
+
"loss": 0.3903,
|
| 168782 |
+
"step": 54560
|
| 168783 |
+
},
|
| 168784 |
+
{
|
| 168785 |
+
"epoch": 439.48,
|
| 168786 |
+
"learning_rate": 9.141923076923078e-06,
|
| 168787 |
+
"loss": 0.3566,
|
| 168788 |
+
"step": 54565
|
| 168789 |
+
},
|
| 168790 |
+
{
|
| 168791 |
+
"epoch": 439.52,
|
| 168792 |
+
"learning_rate": 9.14184294871795e-06,
|
| 168793 |
+
"loss": 0.4125,
|
| 168794 |
+
"step": 54570
|
| 168795 |
+
},
|
| 168796 |
+
{
|
| 168797 |
+
"epoch": 439.56,
|
| 168798 |
+
"learning_rate": 9.141762820512822e-06,
|
| 168799 |
+
"loss": 0.5697,
|
| 168800 |
+
"step": 54575
|
| 168801 |
+
},
|
| 168802 |
+
{
|
| 168803 |
+
"epoch": 439.6,
|
| 168804 |
+
"learning_rate": 9.141682692307694e-06,
|
| 168805 |
+
"loss": 1.262,
|
| 168806 |
+
"step": 54580
|
| 168807 |
+
},
|
| 168808 |
+
{
|
| 168809 |
+
"epoch": 439.64,
|
| 168810 |
+
"learning_rate": 9.141602564102565e-06,
|
| 168811 |
+
"loss": 0.3247,
|
| 168812 |
+
"step": 54585
|
| 168813 |
+
},
|
| 168814 |
+
{
|
| 168815 |
+
"epoch": 439.68,
|
| 168816 |
+
"learning_rate": 9.141522435897436e-06,
|
| 168817 |
+
"loss": 0.2981,
|
| 168818 |
+
"step": 54590
|
| 168819 |
+
},
|
| 168820 |
+
{
|
| 168821 |
+
"epoch": 439.72,
|
| 168822 |
+
"learning_rate": 9.14144230769231e-06,
|
| 168823 |
+
"loss": 0.4221,
|
| 168824 |
+
"step": 54595
|
| 168825 |
+
},
|
| 168826 |
+
{
|
| 168827 |
+
"epoch": 439.76,
|
| 168828 |
+
"learning_rate": 9.141362179487179e-06,
|
| 168829 |
+
"loss": 0.5187,
|
| 168830 |
+
"step": 54600
|
| 168831 |
+
},
|
| 168832 |
+
{
|
| 168833 |
+
"epoch": 439.8,
|
| 168834 |
+
"learning_rate": 9.141282051282052e-06,
|
| 168835 |
+
"loss": 1.5213,
|
| 168836 |
+
"step": 54605
|
| 168837 |
+
},
|
| 168838 |
+
{
|
| 168839 |
+
"epoch": 439.84,
|
| 168840 |
+
"learning_rate": 9.141201923076923e-06,
|
| 168841 |
+
"loss": 0.4041,
|
| 168842 |
+
"step": 54610
|
| 168843 |
+
},
|
| 168844 |
+
{
|
| 168845 |
+
"epoch": 439.88,
|
| 168846 |
+
"learning_rate": 9.141121794871795e-06,
|
| 168847 |
+
"loss": 0.261,
|
| 168848 |
+
"step": 54615
|
| 168849 |
+
},
|
| 168850 |
+
{
|
| 168851 |
+
"epoch": 439.92,
|
| 168852 |
+
"learning_rate": 9.141041666666668e-06,
|
| 168853 |
+
"loss": 0.4434,
|
| 168854 |
+
"step": 54620
|
| 168855 |
+
},
|
| 168856 |
+
{
|
| 168857 |
+
"epoch": 439.96,
|
| 168858 |
+
"learning_rate": 9.14096153846154e-06,
|
| 168859 |
+
"loss": 0.5753,
|
| 168860 |
+
"step": 54625
|
| 168861 |
+
},
|
| 168862 |
+
{
|
| 168863 |
+
"epoch": 440.0,
|
| 168864 |
+
"eval_loss": 0.4085885286331177,
|
| 168865 |
+
"eval_runtime": 40.9337,
|
| 168866 |
+
"eval_samples_per_second": 20.497,
|
| 168867 |
+
"eval_steps_per_second": 0.66,
|
| 168868 |
+
"eval_wer": 0.2093006224826071,
|
| 168869 |
+
"step": 54629
|
| 168870 |
+
},
|
| 168871 |
+
{
|
| 168872 |
+
"epoch": 440.01,
|
| 168873 |
+
"learning_rate": 9.140881410256412e-06,
|
| 168874 |
+
"loss": 0.418,
|
| 168875 |
+
"step": 54630
|
| 168876 |
+
},
|
| 168877 |
+
{
|
| 168878 |
+
"epoch": 440.05,
|
| 168879 |
+
"learning_rate": 9.140801282051282e-06,
|
| 168880 |
+
"loss": 0.3297,
|
| 168881 |
+
"step": 54635
|
| 168882 |
+
},
|
| 168883 |
+
{
|
| 168884 |
+
"epoch": 440.09,
|
| 168885 |
+
"learning_rate": 9.140721153846155e-06,
|
| 168886 |
+
"loss": 0.3332,
|
| 168887 |
+
"step": 54640
|
| 168888 |
+
},
|
| 168889 |
+
{
|
| 168890 |
+
"epoch": 440.13,
|
| 168891 |
+
"learning_rate": 9.140641025641026e-06,
|
| 168892 |
+
"loss": 0.3913,
|
| 168893 |
+
"step": 54645
|
| 168894 |
+
},
|
| 168895 |
+
{
|
| 168896 |
+
"epoch": 440.17,
|
| 168897 |
+
"learning_rate": 9.140560897435898e-06,
|
| 168898 |
+
"loss": 0.5579,
|
| 168899 |
+
"step": 54650
|
| 168900 |
+
},
|
| 168901 |
+
{
|
| 168902 |
+
"epoch": 440.21,
|
| 168903 |
+
"learning_rate": 9.140480769230769e-06,
|
| 168904 |
+
"loss": 1.2676,
|
| 168905 |
+
"step": 54655
|
| 168906 |
+
},
|
| 168907 |
+
{
|
| 168908 |
+
"epoch": 440.25,
|
| 168909 |
+
"learning_rate": 9.140400641025642e-06,
|
| 168910 |
+
"loss": 0.3456,
|
| 168911 |
+
"step": 54660
|
| 168912 |
+
},
|
| 168913 |
+
{
|
| 168914 |
+
"epoch": 440.29,
|
| 168915 |
+
"learning_rate": 9.140320512820513e-06,
|
| 168916 |
+
"loss": 0.3053,
|
| 168917 |
+
"step": 54665
|
| 168918 |
+
},
|
| 168919 |
+
{
|
| 168920 |
+
"epoch": 440.33,
|
| 168921 |
+
"learning_rate": 9.140240384615385e-06,
|
| 168922 |
+
"loss": 0.3536,
|
| 168923 |
+
"step": 54670
|
| 168924 |
+
},
|
| 168925 |
+
{
|
| 168926 |
+
"epoch": 440.37,
|
| 168927 |
+
"learning_rate": 9.140160256410258e-06,
|
| 168928 |
+
"loss": 0.6171,
|
| 168929 |
+
"step": 54675
|
| 168930 |
+
},
|
| 168931 |
+
{
|
| 168932 |
+
"epoch": 440.41,
|
| 168933 |
+
"learning_rate": 9.14008012820513e-06,
|
| 168934 |
+
"loss": 1.2114,
|
| 168935 |
+
"step": 54680
|
| 168936 |
+
},
|
| 168937 |
+
{
|
| 168938 |
+
"epoch": 440.45,
|
| 168939 |
+
"learning_rate": 9.14e-06,
|
| 168940 |
+
"loss": 0.4063,
|
| 168941 |
+
"step": 54685
|
| 168942 |
+
},
|
| 168943 |
+
{
|
| 168944 |
+
"epoch": 440.49,
|
| 168945 |
+
"learning_rate": 9.139919871794872e-06,
|
| 168946 |
+
"loss": 0.2761,
|
| 168947 |
+
"step": 54690
|
| 168948 |
+
},
|
| 168949 |
+
{
|
| 168950 |
+
"epoch": 440.53,
|
| 168951 |
+
"learning_rate": 9.139839743589745e-06,
|
| 168952 |
+
"loss": 0.3836,
|
| 168953 |
+
"step": 54695
|
| 168954 |
+
},
|
| 168955 |
+
{
|
| 168956 |
+
"epoch": 440.57,
|
| 168957 |
+
"learning_rate": 9.139759615384616e-06,
|
| 168958 |
+
"loss": 0.6551,
|
| 168959 |
+
"step": 54700
|
| 168960 |
+
},
|
| 168961 |
+
{
|
| 168962 |
+
"epoch": 440.61,
|
| 168963 |
+
"learning_rate": 9.139679487179488e-06,
|
| 168964 |
+
"loss": 1.2129,
|
| 168965 |
+
"step": 54705
|
| 168966 |
+
},
|
| 168967 |
+
{
|
| 168968 |
+
"epoch": 440.65,
|
| 168969 |
+
"learning_rate": 9.139599358974359e-06,
|
| 168970 |
+
"loss": 0.4,
|
| 168971 |
+
"step": 54710
|
| 168972 |
+
},
|
| 168973 |
+
{
|
| 168974 |
+
"epoch": 440.69,
|
| 168975 |
+
"learning_rate": 9.139519230769232e-06,
|
| 168976 |
+
"loss": 0.3875,
|
| 168977 |
+
"step": 54715
|
| 168978 |
+
},
|
| 168979 |
+
{
|
| 168980 |
+
"epoch": 440.73,
|
| 168981 |
+
"learning_rate": 9.139439102564103e-06,
|
| 168982 |
+
"loss": 0.3592,
|
| 168983 |
+
"step": 54720
|
| 168984 |
+
},
|
| 168985 |
+
{
|
| 168986 |
+
"epoch": 440.77,
|
| 168987 |
+
"learning_rate": 9.139358974358975e-06,
|
| 168988 |
+
"loss": 0.6077,
|
| 168989 |
+
"step": 54725
|
| 168990 |
+
},
|
| 168991 |
+
{
|
| 168992 |
+
"epoch": 440.81,
|
| 168993 |
+
"learning_rate": 9.139278846153848e-06,
|
| 168994 |
+
"loss": 1.1306,
|
| 168995 |
+
"step": 54730
|
| 168996 |
+
},
|
| 168997 |
+
{
|
| 168998 |
+
"epoch": 440.85,
|
| 168999 |
+
"learning_rate": 9.13919871794872e-06,
|
| 169000 |
+
"loss": 0.4137,
|
| 169001 |
+
"step": 54735
|
| 169002 |
+
},
|
| 169003 |
+
{
|
| 169004 |
+
"epoch": 440.9,
|
| 169005 |
+
"learning_rate": 9.13911858974359e-06,
|
| 169006 |
+
"loss": 0.3306,
|
| 169007 |
+
"step": 54740
|
| 169008 |
+
},
|
| 169009 |
+
{
|
| 169010 |
+
"epoch": 440.94,
|
| 169011 |
+
"learning_rate": 9.139038461538462e-06,
|
| 169012 |
+
"loss": 0.3185,
|
| 169013 |
+
"step": 54745
|
| 169014 |
+
},
|
| 169015 |
+
{
|
| 169016 |
+
"epoch": 440.98,
|
| 169017 |
+
"learning_rate": 9.138958333333335e-06,
|
| 169018 |
+
"loss": 0.5769,
|
| 169019 |
+
"step": 54750
|
| 169020 |
+
},
|
| 169021 |
+
{
|
| 169022 |
+
"epoch": 441.0,
|
| 169023 |
+
"eval_loss": 0.4726785123348236,
|
| 169024 |
+
"eval_runtime": 40.5592,
|
| 169025 |
+
"eval_samples_per_second": 20.686,
|
| 169026 |
+
"eval_steps_per_second": 0.666,
|
| 169027 |
+
"eval_wer": 0.2004497968659315,
|
| 169028 |
+
"step": 54753
|
| 169029 |
+
},
|
| 169030 |
+
{
|
| 169031 |
+
"epoch": 441.02,
|
| 169032 |
+
"learning_rate": 9.138878205128205e-06,
|
| 169033 |
+
"loss": 0.4078,
|
| 169034 |
+
"step": 54755
|
| 169035 |
+
},
|
| 169036 |
+
{
|
| 169037 |
+
"epoch": 441.06,
|
| 169038 |
+
"learning_rate": 9.138798076923078e-06,
|
| 169039 |
+
"loss": 0.3861,
|
| 169040 |
+
"step": 54760
|
| 169041 |
+
},
|
| 169042 |
+
{
|
| 169043 |
+
"epoch": 441.1,
|
| 169044 |
+
"learning_rate": 9.138717948717949e-06,
|
| 169045 |
+
"loss": 0.3216,
|
| 169046 |
+
"step": 54765
|
| 169047 |
+
},
|
| 169048 |
+
{
|
| 169049 |
+
"epoch": 441.14,
|
| 169050 |
+
"learning_rate": 9.13863782051282e-06,
|
| 169051 |
+
"loss": 0.3816,
|
| 169052 |
+
"step": 54770
|
| 169053 |
+
},
|
| 169054 |
+
{
|
| 169055 |
+
"epoch": 441.18,
|
| 169056 |
+
"learning_rate": 9.138557692307694e-06,
|
| 169057 |
+
"loss": 0.5851,
|
| 169058 |
+
"step": 54775
|
| 169059 |
+
},
|
| 169060 |
+
{
|
| 169061 |
+
"epoch": 441.22,
|
| 169062 |
+
"learning_rate": 9.138477564102565e-06,
|
| 169063 |
+
"loss": 0.9708,
|
| 169064 |
+
"step": 54780
|
| 169065 |
+
},
|
| 169066 |
+
{
|
| 169067 |
+
"epoch": 441.26,
|
| 169068 |
+
"learning_rate": 9.138397435897436e-06,
|
| 169069 |
+
"loss": 0.3298,
|
| 169070 |
+
"step": 54785
|
| 169071 |
+
},
|
| 169072 |
+
{
|
| 169073 |
+
"epoch": 441.3,
|
| 169074 |
+
"learning_rate": 9.138317307692308e-06,
|
| 169075 |
+
"loss": 0.3666,
|
| 169076 |
+
"step": 54790
|
| 169077 |
+
},
|
| 169078 |
+
{
|
| 169079 |
+
"epoch": 441.34,
|
| 169080 |
+
"learning_rate": 9.13823717948718e-06,
|
| 169081 |
+
"loss": 0.4135,
|
| 169082 |
+
"step": 54795
|
| 169083 |
+
},
|
| 169084 |
+
{
|
| 169085 |
+
"epoch": 441.38,
|
| 169086 |
+
"learning_rate": 9.138157051282052e-06,
|
| 169087 |
+
"loss": 0.7108,
|
| 169088 |
+
"step": 54800
|
| 169089 |
+
},
|
| 169090 |
+
{
|
| 169091 |
+
"epoch": 441.42,
|
| 169092 |
+
"learning_rate": 9.138076923076923e-06,
|
| 169093 |
+
"loss": 1.1512,
|
| 169094 |
+
"step": 54805
|
| 169095 |
+
},
|
| 169096 |
+
{
|
| 169097 |
+
"epoch": 441.46,
|
| 169098 |
+
"learning_rate": 9.137996794871795e-06,
|
| 169099 |
+
"loss": 0.3209,
|
| 169100 |
+
"step": 54810
|
| 169101 |
+
},
|
| 169102 |
+
{
|
| 169103 |
+
"epoch": 441.5,
|
| 169104 |
+
"learning_rate": 9.137916666666668e-06,
|
| 169105 |
+
"loss": 0.3077,
|
| 169106 |
+
"step": 54815
|
| 169107 |
+
},
|
| 169108 |
+
{
|
| 169109 |
+
"epoch": 441.54,
|
| 169110 |
+
"learning_rate": 9.137836538461539e-06,
|
| 169111 |
+
"loss": 0.363,
|
| 169112 |
+
"step": 54820
|
| 169113 |
+
},
|
| 169114 |
+
{
|
| 169115 |
+
"epoch": 441.58,
|
| 169116 |
+
"learning_rate": 9.13775641025641e-06,
|
| 169117 |
+
"loss": 0.854,
|
| 169118 |
+
"step": 54825
|
| 169119 |
+
},
|
| 169120 |
+
{
|
| 169121 |
+
"epoch": 441.62,
|
| 169122 |
+
"learning_rate": 9.137676282051284e-06,
|
| 169123 |
+
"loss": 1.161,
|
| 169124 |
+
"step": 54830
|
| 169125 |
+
},
|
| 169126 |
+
{
|
| 169127 |
+
"epoch": 441.66,
|
| 169128 |
+
"learning_rate": 9.137596153846155e-06,
|
| 169129 |
+
"loss": 0.2903,
|
| 169130 |
+
"step": 54835
|
| 169131 |
+
},
|
| 169132 |
+
{
|
| 169133 |
+
"epoch": 441.7,
|
| 169134 |
+
"learning_rate": 9.137516025641026e-06,
|
| 169135 |
+
"loss": 0.2767,
|
| 169136 |
+
"step": 54840
|
| 169137 |
+
},
|
| 169138 |
+
{
|
| 169139 |
+
"epoch": 441.74,
|
| 169140 |
+
"learning_rate": 9.137435897435898e-06,
|
| 169141 |
+
"loss": 0.4821,
|
| 169142 |
+
"step": 54845
|
| 169143 |
+
},
|
| 169144 |
+
{
|
| 169145 |
+
"epoch": 441.78,
|
| 169146 |
+
"learning_rate": 9.13735576923077e-06,
|
| 169147 |
+
"loss": 0.6793,
|
| 169148 |
+
"step": 54850
|
| 169149 |
+
},
|
| 169150 |
+
{
|
| 169151 |
+
"epoch": 441.82,
|
| 169152 |
+
"learning_rate": 9.137275641025642e-06,
|
| 169153 |
+
"loss": 1.1459,
|
| 169154 |
+
"step": 54855
|
| 169155 |
+
},
|
| 169156 |
+
{
|
| 169157 |
+
"epoch": 441.86,
|
| 169158 |
+
"learning_rate": 9.137195512820513e-06,
|
| 169159 |
+
"loss": 0.3631,
|
| 169160 |
+
"step": 54860
|
| 169161 |
+
},
|
| 169162 |
+
{
|
| 169163 |
+
"epoch": 441.9,
|
| 169164 |
+
"learning_rate": 9.137115384615385e-06,
|
| 169165 |
+
"loss": 0.3231,
|
| 169166 |
+
"step": 54865
|
| 169167 |
+
},
|
| 169168 |
+
{
|
| 169169 |
+
"epoch": 441.94,
|
| 169170 |
+
"learning_rate": 9.137035256410258e-06,
|
| 169171 |
+
"loss": 0.3946,
|
| 169172 |
+
"step": 54870
|
| 169173 |
+
},
|
| 169174 |
+
{
|
| 169175 |
+
"epoch": 441.98,
|
| 169176 |
+
"learning_rate": 9.136955128205129e-06,
|
| 169177 |
+
"loss": 0.7559,
|
| 169178 |
+
"step": 54875
|
| 169179 |
+
},
|
| 169180 |
+
{
|
| 169181 |
+
"epoch": 442.0,
|
| 169182 |
+
"eval_loss": 0.4403437077999115,
|
| 169183 |
+
"eval_runtime": 38.7472,
|
| 169184 |
+
"eval_samples_per_second": 21.653,
|
| 169185 |
+
"eval_steps_per_second": 0.697,
|
| 169186 |
+
"eval_wer": 0.20528028933092224,
|
| 169187 |
+
"step": 54877
|
| 169188 |
+
},
|
| 169189 |
+
{
|
| 169190 |
+
"epoch": 439.02,
|
| 169191 |
+
"learning_rate": 9.136875e-06,
|
| 169192 |
+
"loss": 0.3878,
|
| 169193 |
+
"step": 54880
|
| 169194 |
+
},
|
| 169195 |
+
{
|
| 169196 |
+
"epoch": 439.06,
|
| 169197 |
+
"learning_rate": 9.136794871794874e-06,
|
| 169198 |
+
"loss": 0.4139,
|
| 169199 |
+
"step": 54885
|
| 169200 |
+
},
|
| 169201 |
+
{
|
| 169202 |
+
"epoch": 439.1,
|
| 169203 |
+
"learning_rate": 9.136714743589745e-06,
|
| 169204 |
+
"loss": 0.3717,
|
| 169205 |
+
"step": 54890
|
| 169206 |
+
},
|
| 169207 |
+
{
|
| 169208 |
+
"epoch": 439.14,
|
| 169209 |
+
"learning_rate": 9.136634615384616e-06,
|
| 169210 |
+
"loss": 0.4635,
|
| 169211 |
+
"step": 54895
|
| 169212 |
+
},
|
| 169213 |
+
{
|
| 169214 |
+
"epoch": 439.18,
|
| 169215 |
+
"learning_rate": 9.136554487179488e-06,
|
| 169216 |
+
"loss": 0.8805,
|
| 169217 |
+
"step": 54900
|
| 169218 |
+
},
|
| 169219 |
+
{
|
| 169220 |
+
"epoch": 439.22,
|
| 169221 |
+
"learning_rate": 9.13647435897436e-06,
|
| 169222 |
+
"loss": 0.7647,
|
| 169223 |
+
"step": 54905
|
| 169224 |
+
},
|
| 169225 |
+
{
|
| 169226 |
+
"epoch": 439.26,
|
| 169227 |
+
"learning_rate": 9.13639423076923e-06,
|
| 169228 |
+
"loss": 0.3908,
|
| 169229 |
+
"step": 54910
|
| 169230 |
+
},
|
| 169231 |
+
{
|
| 169232 |
+
"epoch": 439.3,
|
| 169233 |
+
"learning_rate": 9.136314102564103e-06,
|
| 169234 |
+
"loss": 0.353,
|
| 169235 |
+
"step": 54915
|
| 169236 |
+
},
|
| 169237 |
+
{
|
| 169238 |
+
"epoch": 439.34,
|
| 169239 |
+
"learning_rate": 9.136233974358976e-06,
|
| 169240 |
+
"loss": 0.4116,
|
| 169241 |
+
"step": 54920
|
| 169242 |
+
},
|
| 169243 |
+
{
|
| 169244 |
+
"epoch": 439.38,
|
| 169245 |
+
"learning_rate": 9.136153846153846e-06,
|
| 169246 |
+
"loss": 0.851,
|
| 169247 |
+
"step": 54925
|
| 169248 |
+
},
|
| 169249 |
+
{
|
| 169250 |
+
"epoch": 439.42,
|
| 169251 |
+
"learning_rate": 9.13607371794872e-06,
|
| 169252 |
+
"loss": 0.9931,
|
| 169253 |
+
"step": 54930
|
| 169254 |
+
},
|
| 169255 |
+
{
|
| 169256 |
+
"epoch": 439.46,
|
| 169257 |
+
"learning_rate": 9.13599358974359e-06,
|
| 169258 |
+
"loss": 0.3066,
|
| 169259 |
+
"step": 54935
|
| 169260 |
+
},
|
| 169261 |
+
{
|
| 169262 |
+
"epoch": 439.5,
|
| 169263 |
+
"learning_rate": 9.135913461538462e-06,
|
| 169264 |
+
"loss": 0.2852,
|
| 169265 |
+
"step": 54940
|
| 169266 |
+
},
|
| 169267 |
+
{
|
| 169268 |
+
"epoch": 439.54,
|
| 169269 |
+
"learning_rate": 9.135833333333333e-06,
|
| 169270 |
+
"loss": 0.3729,
|
| 169271 |
+
"step": 54945
|
| 169272 |
+
},
|
| 169273 |
+
{
|
| 169274 |
+
"epoch": 439.58,
|
| 169275 |
+
"learning_rate": 9.135753205128206e-06,
|
| 169276 |
+
"loss": 0.7223,
|
| 169277 |
+
"step": 54950
|
| 169278 |
+
},
|
| 169279 |
+
{
|
| 169280 |
+
"epoch": 439.62,
|
| 169281 |
+
"learning_rate": 9.135673076923078e-06,
|
| 169282 |
+
"loss": 0.8935,
|
| 169283 |
+
"step": 54955
|
| 169284 |
+
},
|
| 169285 |
+
{
|
| 169286 |
+
"epoch": 439.66,
|
| 169287 |
+
"learning_rate": 9.135592948717949e-06,
|
| 169288 |
+
"loss": 0.3228,
|
| 169289 |
+
"step": 54960
|
| 169290 |
+
},
|
| 169291 |
+
{
|
| 169292 |
+
"epoch": 439.7,
|
| 169293 |
+
"learning_rate": 9.13551282051282e-06,
|
| 169294 |
+
"loss": 0.3639,
|
| 169295 |
+
"step": 54965
|
| 169296 |
+
},
|
| 169297 |
+
{
|
| 169298 |
+
"epoch": 439.74,
|
| 169299 |
+
"learning_rate": 9.135432692307693e-06,
|
| 169300 |
+
"loss": 0.3532,
|
| 169301 |
+
"step": 54970
|
| 169302 |
+
},
|
| 169303 |
+
{
|
| 169304 |
+
"epoch": 439.78,
|
| 169305 |
+
"learning_rate": 9.135352564102565e-06,
|
| 169306 |
+
"loss": 0.9238,
|
| 169307 |
+
"step": 54975
|
| 169308 |
+
},
|
| 169309 |
+
{
|
| 169310 |
+
"epoch": 439.82,
|
| 169311 |
+
"learning_rate": 9.135272435897436e-06,
|
| 169312 |
+
"loss": 0.96,
|
| 169313 |
+
"step": 54980
|
| 169314 |
+
},
|
| 169315 |
+
{
|
| 169316 |
+
"epoch": 439.86,
|
| 169317 |
+
"learning_rate": 9.13519230769231e-06,
|
| 169318 |
+
"loss": 0.2877,
|
| 169319 |
+
"step": 54985
|
| 169320 |
+
},
|
| 169321 |
+
{
|
| 169322 |
+
"epoch": 439.9,
|
| 169323 |
+
"learning_rate": 9.13511217948718e-06,
|
| 169324 |
+
"loss": 0.3879,
|
| 169325 |
+
"step": 54990
|
| 169326 |
+
},
|
| 169327 |
+
{
|
| 169328 |
+
"epoch": 439.94,
|
| 169329 |
+
"learning_rate": 9.135032051282052e-06,
|
| 169330 |
+
"loss": 0.4861,
|
| 169331 |
+
"step": 54995
|
| 169332 |
+
},
|
| 169333 |
+
{
|
| 169334 |
+
"epoch": 439.98,
|
| 169335 |
+
"learning_rate": 9.134951923076923e-06,
|
| 169336 |
+
"loss": 0.9325,
|
| 169337 |
+
"step": 55000
|
| 169338 |
+
},
|
| 169339 |
+
{
|
| 169340 |
+
"epoch": 440.0,
|
| 169341 |
+
"eval_loss": 0.3862856924533844,
|
| 169342 |
+
"eval_runtime": 41.5049,
|
| 169343 |
+
"eval_samples_per_second": 20.214,
|
| 169344 |
+
"eval_steps_per_second": 0.651,
|
| 169345 |
+
"eval_wer": 0.19904622157006602,
|
| 169346 |
+
"step": 55002
|
| 169347 |
+
},
|
| 169348 |
+
{
|
| 169349 |
+
"epoch": 440.02,
|
| 169350 |
+
"learning_rate": 9.134871794871796e-06,
|
| 169351 |
+
"loss": 0.3228,
|
| 169352 |
+
"step": 55005
|
| 169353 |
+
},
|
| 169354 |
+
{
|
| 169355 |
+
"epoch": 440.06,
|
| 169356 |
+
"learning_rate": 9.134791666666668e-06,
|
| 169357 |
+
"loss": 0.3744,
|
| 169358 |
+
"step": 55010
|
| 169359 |
+
},
|
| 169360 |
+
{
|
| 169361 |
+
"epoch": 440.1,
|
| 169362 |
+
"learning_rate": 9.134711538461539e-06,
|
| 169363 |
+
"loss": 0.3578,
|
| 169364 |
+
"step": 55015
|
| 169365 |
+
},
|
| 169366 |
+
{
|
| 169367 |
+
"epoch": 440.14,
|
| 169368 |
+
"learning_rate": 9.134631410256412e-06,
|
| 169369 |
+
"loss": 0.4476,
|
| 169370 |
+
"step": 55020
|
| 169371 |
+
},
|
| 169372 |
+
{
|
| 169373 |
+
"epoch": 440.18,
|
| 169374 |
+
"learning_rate": 9.134551282051283e-06,
|
| 169375 |
+
"loss": 0.7147,
|
| 169376 |
+
"step": 55025
|
| 169377 |
+
},
|
| 169378 |
+
{
|
| 169379 |
+
"epoch": 440.22,
|
| 169380 |
+
"learning_rate": 9.134471153846155e-06,
|
| 169381 |
+
"loss": 0.8546,
|
| 169382 |
+
"step": 55030
|
| 169383 |
+
},
|
| 169384 |
+
{
|
| 169385 |
+
"epoch": 440.26,
|
| 169386 |
+
"learning_rate": 9.134391025641026e-06,
|
| 169387 |
+
"loss": 0.3208,
|
| 169388 |
+
"step": 55035
|
| 169389 |
+
},
|
| 169390 |
+
{
|
| 169391 |
+
"epoch": 440.3,
|
| 169392 |
+
"learning_rate": 9.1343108974359e-06,
|
| 169393 |
+
"loss": 0.4333,
|
| 169394 |
+
"step": 55040
|
| 169395 |
+
},
|
| 169396 |
+
{
|
| 169397 |
+
"epoch": 440.34,
|
| 169398 |
+
"learning_rate": 9.134230769230769e-06,
|
| 169399 |
+
"loss": 0.4132,
|
| 169400 |
+
"step": 55045
|
| 169401 |
+
},
|
| 169402 |
+
{
|
| 169403 |
+
"epoch": 440.38,
|
| 169404 |
+
"learning_rate": 9.134150641025642e-06,
|
| 169405 |
+
"loss": 0.7658,
|
| 169406 |
+
"step": 55050
|
| 169407 |
+
},
|
| 169408 |
+
{
|
| 169409 |
+
"epoch": 440.42,
|
| 169410 |
+
"learning_rate": 9.134070512820513e-06,
|
| 169411 |
+
"loss": 0.7724,
|
| 169412 |
+
"step": 55055
|
| 169413 |
+
},
|
| 169414 |
+
{
|
| 169415 |
+
"epoch": 440.46,
|
| 169416 |
+
"learning_rate": 9.133990384615385e-06,
|
| 169417 |
+
"loss": 0.3624,
|
| 169418 |
+
"step": 55060
|
| 169419 |
+
},
|
| 169420 |
+
{
|
| 169421 |
+
"epoch": 440.5,
|
| 169422 |
+
"learning_rate": 9.133910256410256e-06,
|
| 169423 |
+
"loss": 0.3578,
|
| 169424 |
+
"step": 55065
|
| 169425 |
+
},
|
| 169426 |
+
{
|
| 169427 |
+
"epoch": 440.54,
|
| 169428 |
+
"learning_rate": 9.133830128205129e-06,
|
| 169429 |
+
"loss": 0.4768,
|
| 169430 |
+
"step": 55070
|
| 169431 |
+
},
|
| 169432 |
+
{
|
| 169433 |
+
"epoch": 440.58,
|
| 169434 |
+
"learning_rate": 9.13375e-06,
|
| 169435 |
+
"loss": 0.8875,
|
| 169436 |
+
"step": 55075
|
| 169437 |
+
},
|
| 169438 |
+
{
|
| 169439 |
+
"epoch": 440.62,
|
| 169440 |
+
"learning_rate": 9.133669871794872e-06,
|
| 169441 |
+
"loss": 1.0683,
|
| 169442 |
+
"step": 55080
|
| 169443 |
+
},
|
| 169444 |
+
{
|
| 169445 |
+
"epoch": 440.66,
|
| 169446 |
+
"learning_rate": 9.133589743589745e-06,
|
| 169447 |
+
"loss": 0.3035,
|
| 169448 |
+
"step": 55085
|
| 169449 |
+
},
|
| 169450 |
+
{
|
| 169451 |
+
"epoch": 440.7,
|
| 169452 |
+
"learning_rate": 9.133509615384616e-06,
|
| 169453 |
+
"loss": 0.3714,
|
| 169454 |
+
"step": 55090
|
| 169455 |
+
},
|
| 169456 |
+
{
|
| 169457 |
+
"epoch": 440.74,
|
| 169458 |
+
"learning_rate": 9.133429487179488e-06,
|
| 169459 |
+
"loss": 0.4341,
|
| 169460 |
+
"step": 55095
|
| 169461 |
+
},
|
| 169462 |
+
{
|
| 169463 |
+
"epoch": 440.78,
|
| 169464 |
+
"learning_rate": 9.133349358974359e-06,
|
| 169465 |
+
"loss": 0.7635,
|
| 169466 |
+
"step": 55100
|
| 169467 |
+
},
|
| 169468 |
+
{
|
| 169469 |
+
"epoch": 440.82,
|
| 169470 |
+
"learning_rate": 9.133269230769232e-06,
|
| 169471 |
+
"loss": 1.1058,
|
| 169472 |
+
"step": 55105
|
| 169473 |
+
},
|
| 169474 |
+
{
|
| 169475 |
+
"epoch": 440.86,
|
| 169476 |
+
"learning_rate": 9.133189102564103e-06,
|
| 169477 |
+
"loss": 0.2956,
|
| 169478 |
+
"step": 55110
|
| 169479 |
+
},
|
| 169480 |
+
{
|
| 169481 |
+
"epoch": 440.9,
|
| 169482 |
+
"learning_rate": 9.133108974358975e-06,
|
| 169483 |
+
"loss": 0.3866,
|
| 169484 |
+
"step": 55115
|
| 169485 |
+
},
|
| 169486 |
+
{
|
| 169487 |
+
"epoch": 440.94,
|
| 169488 |
+
"learning_rate": 9.133028846153848e-06,
|
| 169489 |
+
"loss": 0.4264,
|
| 169490 |
+
"step": 55120
|
| 169491 |
+
},
|
| 169492 |
+
{
|
| 169493 |
+
"epoch": 440.98,
|
| 169494 |
+
"learning_rate": 9.132948717948719e-06,
|
| 169495 |
+
"loss": 0.8004,
|
| 169496 |
+
"step": 55125
|
| 169497 |
+
},
|
| 169498 |
+
{
|
| 169499 |
+
"epoch": 441.0,
|
| 169500 |
+
"eval_loss": 0.4621056914329529,
|
| 169501 |
+
"eval_runtime": 39.1804,
|
| 169502 |
+
"eval_samples_per_second": 21.439,
|
| 169503 |
+
"eval_steps_per_second": 0.689,
|
| 169504 |
+
"eval_wer": 0.19953240301015562,
|
| 169505 |
+
"step": 55127
|
| 169506 |
}
|
| 169507 |
],
|
| 169508 |
+
"max_steps": 625000,
|
| 169509 |
"num_train_epochs": 5000,
|
| 169510 |
+
"total_flos": 1.551324177518058e+20,
|
| 169511 |
"trial_name": null,
|
| 169512 |
"trial_params": null
|
| 169513 |
}
|
model-bin/finetune/base/{checkpoint-54505 → checkpoint-55127}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629768310.8799858/events.out.tfevents.1629768310.c435e1c5ee04.920.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a9125efb6ff064809a5bade0ff688d2bcec8f7e566c134bd054df89bf6b9e2bf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629768958.858682/events.out.tfevents.1629768958.c435e1c5ee04.920.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a2cdb6d743680c4140da57043fbd7725d14aefedfe5c8f0a0348f066a18f4234
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629769702.6550722/events.out.tfevents.1629769702.c435e1c5ee04.920.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:895b746ca583907807040c845d68cc1ee428009b3f864e953e7f9d9409662a95
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629770366.8003614/events.out.tfevents.1629770366.c435e1c5ee04.920.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da03f0dd87bb7ab4e7ca6c50b3bea1177fbe54ffe6f36ba36c687a2322bb7dd1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629771017.73505/events.out.tfevents.1629771017.c435e1c5ee04.920.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ebc8eeb71b11b51bc2058cc498c706dded1c5501c609ee630113b5cc1f6f31b3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629768310.c435e1c5ee04.920.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d3bd19c5b0461beaf793e8cc66705dd78aea759224a193849076d41ae15b9ac7
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629768958.c435e1c5ee04.920.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b3573ec4bab3f091a838abe323168018c05387ab095fd9f49b0ff46d29b2a62c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629769702.c435e1c5ee04.920.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:29ec26d67558433c844cd8bea4641c9b7c0bade38e32301e53d3d3aa866a4767
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629770366.c435e1c5ee04.920.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a4d964d4ba080c6622c67c57ecc8502c0d227803f3aa1f15df59338224fa149e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629771017.c435e1c5ee04.920.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2c0132cc77d0c766f6aa03134626151ec95bd29793ba83634dfec58f7992bae4
|
| 3 |
+
size 8622
|