"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629901744.9403872/events.out.tfevents.1629901744.7e498afd5545.7645.55 +3 -0
- model-bin/finetune/base/log/1629902211.333175/events.out.tfevents.1629902211.7e498afd5545.7645.57 +3 -0
- model-bin/finetune/base/log/1629902676.4913468/events.out.tfevents.1629902676.7e498afd5545.7645.59 +3 -0
- model-bin/finetune/base/log/1629903142.829317/events.out.tfevents.1629903142.7e498afd5545.7645.61 +3 -0
- model-bin/finetune/base/log/1629903613.046621/events.out.tfevents.1629903613.7e498afd5545.7645.63 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629901744.7e498afd5545.7645.54 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629902211.7e498afd5545.7645.56 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629902675.7e498afd5545.7645.58 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629903142.7e498afd5545.7645.60 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629903613.7e498afd5545.7645.62 +3 -0
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/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 sha256:2402c050872a1b6c2132f1354c8235dbc30ffdaec42a7c3e2fe40a3fa3597e6e
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/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 sha256:9ba89a49bd0701ea0db537127d26ed94e4e5fa6eed33bd9ae9d90248a8d473ba
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/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:ea9d23665821b123a85d6ac5512b6a51253cf13d4f317b3d68b6de999fb06ba7
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/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 sha256:2edae0f6397edeecd1aedd6821f5f0301b0061a47a55585acacade7161821b94
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/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 sha256:5ca9b1c1ac3a3f74bba9515f59a166eb16c9179459f7d8dfba604bf995fff1f5
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -200700,11 +200700,800 @@
|
|
| 200700 |
"eval_steps_per_second": 0.643,
|
| 200701 |
"eval_wer": 0.19413337186619464,
|
| 200702 |
"step": 79647
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 200703 |
}
|
| 200704 |
],
|
| 200705 |
-
"max_steps":
|
| 200706 |
"num_train_epochs": 5000,
|
| 200707 |
-
"total_flos": 2.
|
| 200708 |
"trial_name": null,
|
| 200709 |
"trial_params": null
|
| 200710 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1822989349981638,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-79148",
|
| 4 |
+
"epoch": 642.0,
|
| 5 |
+
"global_step": 80269,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 200700 |
"eval_steps_per_second": 0.643,
|
| 200701 |
"eval_wer": 0.19413337186619464,
|
| 200702 |
"step": 79647
|
| 200703 |
+
},
|
| 200704 |
+
{
|
| 200705 |
+
"epoch": 642.02,
|
| 200706 |
+
"learning_rate": 8.729903069466882e-06,
|
| 200707 |
+
"loss": 0.3702,
|
| 200708 |
+
"step": 79650
|
| 200709 |
+
},
|
| 200710 |
+
{
|
| 200711 |
+
"epoch": 642.06,
|
| 200712 |
+
"learning_rate": 8.729822294022618e-06,
|
| 200713 |
+
"loss": 0.3383,
|
| 200714 |
+
"step": 79655
|
| 200715 |
+
},
|
| 200716 |
+
{
|
| 200717 |
+
"epoch": 642.1,
|
| 200718 |
+
"learning_rate": 8.729741518578354e-06,
|
| 200719 |
+
"loss": 0.4167,
|
| 200720 |
+
"step": 79660
|
| 200721 |
+
},
|
| 200722 |
+
{
|
| 200723 |
+
"epoch": 642.14,
|
| 200724 |
+
"learning_rate": 8.729660743134088e-06,
|
| 200725 |
+
"loss": 0.3496,
|
| 200726 |
+
"step": 79665
|
| 200727 |
+
},
|
| 200728 |
+
{
|
| 200729 |
+
"epoch": 642.18,
|
| 200730 |
+
"learning_rate": 8.729579967689824e-06,
|
| 200731 |
+
"loss": 0.7132,
|
| 200732 |
+
"step": 79670
|
| 200733 |
+
},
|
| 200734 |
+
{
|
| 200735 |
+
"epoch": 642.22,
|
| 200736 |
+
"learning_rate": 8.729499192245558e-06,
|
| 200737 |
+
"loss": 0.7956,
|
| 200738 |
+
"step": 79675
|
| 200739 |
+
},
|
| 200740 |
+
{
|
| 200741 |
+
"epoch": 642.27,
|
| 200742 |
+
"learning_rate": 8.729418416801294e-06,
|
| 200743 |
+
"loss": 0.3064,
|
| 200744 |
+
"step": 79680
|
| 200745 |
+
},
|
| 200746 |
+
{
|
| 200747 |
+
"epoch": 642.31,
|
| 200748 |
+
"learning_rate": 8.729337641357028e-06,
|
| 200749 |
+
"loss": 0.3004,
|
| 200750 |
+
"step": 79685
|
| 200751 |
+
},
|
| 200752 |
+
{
|
| 200753 |
+
"epoch": 642.35,
|
| 200754 |
+
"learning_rate": 8.729256865912764e-06,
|
| 200755 |
+
"loss": 0.3636,
|
| 200756 |
+
"step": 79690
|
| 200757 |
+
},
|
| 200758 |
+
{
|
| 200759 |
+
"epoch": 642.39,
|
| 200760 |
+
"learning_rate": 8.729176090468498e-06,
|
| 200761 |
+
"loss": 0.7391,
|
| 200762 |
+
"step": 79695
|
| 200763 |
+
},
|
| 200764 |
+
{
|
| 200765 |
+
"epoch": 642.43,
|
| 200766 |
+
"learning_rate": 8.729095315024234e-06,
|
| 200767 |
+
"loss": 0.7999,
|
| 200768 |
+
"step": 79700
|
| 200769 |
+
},
|
| 200770 |
+
{
|
| 200771 |
+
"epoch": 642.47,
|
| 200772 |
+
"learning_rate": 8.729014539579968e-06,
|
| 200773 |
+
"loss": 0.3359,
|
| 200774 |
+
"step": 79705
|
| 200775 |
+
},
|
| 200776 |
+
{
|
| 200777 |
+
"epoch": 642.51,
|
| 200778 |
+
"learning_rate": 8.728933764135704e-06,
|
| 200779 |
+
"loss": 0.2725,
|
| 200780 |
+
"step": 79710
|
| 200781 |
+
},
|
| 200782 |
+
{
|
| 200783 |
+
"epoch": 642.55,
|
| 200784 |
+
"learning_rate": 8.728852988691438e-06,
|
| 200785 |
+
"loss": 0.3936,
|
| 200786 |
+
"step": 79715
|
| 200787 |
+
},
|
| 200788 |
+
{
|
| 200789 |
+
"epoch": 642.59,
|
| 200790 |
+
"learning_rate": 8.728772213247174e-06,
|
| 200791 |
+
"loss": 0.8775,
|
| 200792 |
+
"step": 79720
|
| 200793 |
+
},
|
| 200794 |
+
{
|
| 200795 |
+
"epoch": 642.63,
|
| 200796 |
+
"learning_rate": 8.72869143780291e-06,
|
| 200797 |
+
"loss": 1.0067,
|
| 200798 |
+
"step": 79725
|
| 200799 |
+
},
|
| 200800 |
+
{
|
| 200801 |
+
"epoch": 642.67,
|
| 200802 |
+
"learning_rate": 8.728610662358644e-06,
|
| 200803 |
+
"loss": 0.3226,
|
| 200804 |
+
"step": 79730
|
| 200805 |
+
},
|
| 200806 |
+
{
|
| 200807 |
+
"epoch": 642.71,
|
| 200808 |
+
"learning_rate": 8.72852988691438e-06,
|
| 200809 |
+
"loss": 0.3211,
|
| 200810 |
+
"step": 79735
|
| 200811 |
+
},
|
| 200812 |
+
{
|
| 200813 |
+
"epoch": 642.75,
|
| 200814 |
+
"learning_rate": 8.728449111470114e-06,
|
| 200815 |
+
"loss": 0.4379,
|
| 200816 |
+
"step": 79740
|
| 200817 |
+
},
|
| 200818 |
+
{
|
| 200819 |
+
"epoch": 642.79,
|
| 200820 |
+
"learning_rate": 8.72836833602585e-06,
|
| 200821 |
+
"loss": 0.7525,
|
| 200822 |
+
"step": 79745
|
| 200823 |
+
},
|
| 200824 |
+
{
|
| 200825 |
+
"epoch": 642.83,
|
| 200826 |
+
"learning_rate": 8.728287560581584e-06,
|
| 200827 |
+
"loss": 0.8896,
|
| 200828 |
+
"step": 79750
|
| 200829 |
+
},
|
| 200830 |
+
{
|
| 200831 |
+
"epoch": 642.87,
|
| 200832 |
+
"learning_rate": 8.72820678513732e-06,
|
| 200833 |
+
"loss": 0.5549,
|
| 200834 |
+
"step": 79755
|
| 200835 |
+
},
|
| 200836 |
+
{
|
| 200837 |
+
"epoch": 642.91,
|
| 200838 |
+
"learning_rate": 8.728126009693054e-06,
|
| 200839 |
+
"loss": 0.3093,
|
| 200840 |
+
"step": 79760
|
| 200841 |
+
},
|
| 200842 |
+
{
|
| 200843 |
+
"epoch": 642.95,
|
| 200844 |
+
"learning_rate": 8.72804523424879e-06,
|
| 200845 |
+
"loss": 0.3845,
|
| 200846 |
+
"step": 79765
|
| 200847 |
+
},
|
| 200848 |
+
{
|
| 200849 |
+
"epoch": 642.99,
|
| 200850 |
+
"learning_rate": 8.727964458804523e-06,
|
| 200851 |
+
"loss": 0.8744,
|
| 200852 |
+
"step": 79770
|
| 200853 |
+
},
|
| 200854 |
+
{
|
| 200855 |
+
"epoch": 643.0,
|
| 200856 |
+
"eval_loss": 0.3736218214035034,
|
| 200857 |
+
"eval_runtime": 43.1561,
|
| 200858 |
+
"eval_samples_per_second": 19.487,
|
| 200859 |
+
"eval_steps_per_second": 0.626,
|
| 200860 |
+
"eval_wer": 0.18469290878551084,
|
| 200861 |
+
"step": 79771
|
| 200862 |
+
},
|
| 200863 |
+
{
|
| 200864 |
+
"epoch": 643.03,
|
| 200865 |
+
"learning_rate": 8.72788368336026e-06,
|
| 200866 |
+
"loss": 0.3615,
|
| 200867 |
+
"step": 79775
|
| 200868 |
+
},
|
| 200869 |
+
{
|
| 200870 |
+
"epoch": 643.07,
|
| 200871 |
+
"learning_rate": 8.727802907915993e-06,
|
| 200872 |
+
"loss": 0.3373,
|
| 200873 |
+
"step": 79780
|
| 200874 |
+
},
|
| 200875 |
+
{
|
| 200876 |
+
"epoch": 643.11,
|
| 200877 |
+
"learning_rate": 8.72772213247173e-06,
|
| 200878 |
+
"loss": 0.3368,
|
| 200879 |
+
"step": 79785
|
| 200880 |
+
},
|
| 200881 |
+
{
|
| 200882 |
+
"epoch": 643.15,
|
| 200883 |
+
"learning_rate": 8.727641357027465e-06,
|
| 200884 |
+
"loss": 0.412,
|
| 200885 |
+
"step": 79790
|
| 200886 |
+
},
|
| 200887 |
+
{
|
| 200888 |
+
"epoch": 643.19,
|
| 200889 |
+
"learning_rate": 8.7275605815832e-06,
|
| 200890 |
+
"loss": 0.9684,
|
| 200891 |
+
"step": 79795
|
| 200892 |
+
},
|
| 200893 |
+
{
|
| 200894 |
+
"epoch": 643.23,
|
| 200895 |
+
"learning_rate": 8.727479806138935e-06,
|
| 200896 |
+
"loss": 0.7489,
|
| 200897 |
+
"step": 79800
|
| 200898 |
+
},
|
| 200899 |
+
{
|
| 200900 |
+
"epoch": 643.27,
|
| 200901 |
+
"learning_rate": 8.72739903069467e-06,
|
| 200902 |
+
"loss": 0.3246,
|
| 200903 |
+
"step": 79805
|
| 200904 |
+
},
|
| 200905 |
+
{
|
| 200906 |
+
"epoch": 643.31,
|
| 200907 |
+
"learning_rate": 8.727318255250405e-06,
|
| 200908 |
+
"loss": 0.3444,
|
| 200909 |
+
"step": 79810
|
| 200910 |
+
},
|
| 200911 |
+
{
|
| 200912 |
+
"epoch": 643.35,
|
| 200913 |
+
"learning_rate": 8.72723747980614e-06,
|
| 200914 |
+
"loss": 0.4126,
|
| 200915 |
+
"step": 79815
|
| 200916 |
+
},
|
| 200917 |
+
{
|
| 200918 |
+
"epoch": 643.39,
|
| 200919 |
+
"learning_rate": 8.727156704361875e-06,
|
| 200920 |
+
"loss": 0.865,
|
| 200921 |
+
"step": 79820
|
| 200922 |
+
},
|
| 200923 |
+
{
|
| 200924 |
+
"epoch": 643.43,
|
| 200925 |
+
"learning_rate": 8.72707592891761e-06,
|
| 200926 |
+
"loss": 0.743,
|
| 200927 |
+
"step": 79825
|
| 200928 |
+
},
|
| 200929 |
+
{
|
| 200930 |
+
"epoch": 643.47,
|
| 200931 |
+
"learning_rate": 8.726995153473345e-06,
|
| 200932 |
+
"loss": 0.3948,
|
| 200933 |
+
"step": 79830
|
| 200934 |
+
},
|
| 200935 |
+
{
|
| 200936 |
+
"epoch": 643.51,
|
| 200937 |
+
"learning_rate": 8.72691437802908e-06,
|
| 200938 |
+
"loss": 0.3934,
|
| 200939 |
+
"step": 79835
|
| 200940 |
+
},
|
| 200941 |
+
{
|
| 200942 |
+
"epoch": 643.55,
|
| 200943 |
+
"learning_rate": 8.726833602584815e-06,
|
| 200944 |
+
"loss": 0.4111,
|
| 200945 |
+
"step": 79840
|
| 200946 |
+
},
|
| 200947 |
+
{
|
| 200948 |
+
"epoch": 643.59,
|
| 200949 |
+
"learning_rate": 8.726752827140549e-06,
|
| 200950 |
+
"loss": 1.0506,
|
| 200951 |
+
"step": 79845
|
| 200952 |
+
},
|
| 200953 |
+
{
|
| 200954 |
+
"epoch": 643.63,
|
| 200955 |
+
"learning_rate": 8.726672051696285e-06,
|
| 200956 |
+
"loss": 0.6835,
|
| 200957 |
+
"step": 79850
|
| 200958 |
+
},
|
| 200959 |
+
{
|
| 200960 |
+
"epoch": 643.67,
|
| 200961 |
+
"learning_rate": 8.726591276252019e-06,
|
| 200962 |
+
"loss": 0.2641,
|
| 200963 |
+
"step": 79855
|
| 200964 |
+
},
|
| 200965 |
+
{
|
| 200966 |
+
"epoch": 643.71,
|
| 200967 |
+
"learning_rate": 8.726510500807755e-06,
|
| 200968 |
+
"loss": 0.7048,
|
| 200969 |
+
"step": 79860
|
| 200970 |
+
},
|
| 200971 |
+
{
|
| 200972 |
+
"epoch": 643.76,
|
| 200973 |
+
"learning_rate": 8.72642972536349e-06,
|
| 200974 |
+
"loss": 0.3964,
|
| 200975 |
+
"step": 79865
|
| 200976 |
+
},
|
| 200977 |
+
{
|
| 200978 |
+
"epoch": 643.8,
|
| 200979 |
+
"learning_rate": 8.726348949919225e-06,
|
| 200980 |
+
"loss": 1.0147,
|
| 200981 |
+
"step": 79870
|
| 200982 |
+
},
|
| 200983 |
+
{
|
| 200984 |
+
"epoch": 643.84,
|
| 200985 |
+
"learning_rate": 8.72626817447496e-06,
|
| 200986 |
+
"loss": 0.7456,
|
| 200987 |
+
"step": 79875
|
| 200988 |
+
},
|
| 200989 |
+
{
|
| 200990 |
+
"epoch": 643.88,
|
| 200991 |
+
"learning_rate": 8.726187399030695e-06,
|
| 200992 |
+
"loss": 0.3131,
|
| 200993 |
+
"step": 79880
|
| 200994 |
+
},
|
| 200995 |
+
{
|
| 200996 |
+
"epoch": 643.92,
|
| 200997 |
+
"learning_rate": 8.72610662358643e-06,
|
| 200998 |
+
"loss": 0.3797,
|
| 200999 |
+
"step": 79885
|
| 201000 |
+
},
|
| 201001 |
+
{
|
| 201002 |
+
"epoch": 643.96,
|
| 201003 |
+
"learning_rate": 8.726025848142165e-06,
|
| 201004 |
+
"loss": 0.4774,
|
| 201005 |
+
"step": 79890
|
| 201006 |
+
},
|
| 201007 |
+
{
|
| 201008 |
+
"epoch": 644.0,
|
| 201009 |
+
"learning_rate": 8.7259450726979e-06,
|
| 201010 |
+
"loss": 1.1313,
|
| 201011 |
+
"step": 79895
|
| 201012 |
+
},
|
| 201013 |
+
{
|
| 201014 |
+
"epoch": 644.0,
|
| 201015 |
+
"eval_loss": 0.4221903383731842,
|
| 201016 |
+
"eval_runtime": 41.5264,
|
| 201017 |
+
"eval_samples_per_second": 20.252,
|
| 201018 |
+
"eval_steps_per_second": 0.65,
|
| 201019 |
+
"eval_wer": 0.1915672351950809,
|
| 201020 |
+
"step": 79895
|
| 201021 |
+
},
|
| 201022 |
+
{
|
| 201023 |
+
"epoch": 639.04,
|
| 201024 |
+
"learning_rate": 8.725864297253635e-06,
|
| 201025 |
+
"loss": 0.3224,
|
| 201026 |
+
"step": 79900
|
| 201027 |
+
},
|
| 201028 |
+
{
|
| 201029 |
+
"epoch": 639.08,
|
| 201030 |
+
"learning_rate": 8.72578352180937e-06,
|
| 201031 |
+
"loss": 0.2609,
|
| 201032 |
+
"step": 79905
|
| 201033 |
+
},
|
| 201034 |
+
{
|
| 201035 |
+
"epoch": 639.12,
|
| 201036 |
+
"learning_rate": 8.725702746365105e-06,
|
| 201037 |
+
"loss": 0.3024,
|
| 201038 |
+
"step": 79910
|
| 201039 |
+
},
|
| 201040 |
+
{
|
| 201041 |
+
"epoch": 639.16,
|
| 201042 |
+
"learning_rate": 8.72562197092084e-06,
|
| 201043 |
+
"loss": 0.4686,
|
| 201044 |
+
"step": 79915
|
| 201045 |
+
},
|
| 201046 |
+
{
|
| 201047 |
+
"epoch": 639.2,
|
| 201048 |
+
"learning_rate": 8.725541195476575e-06,
|
| 201049 |
+
"loss": 1.3208,
|
| 201050 |
+
"step": 79920
|
| 201051 |
+
},
|
| 201052 |
+
{
|
| 201053 |
+
"epoch": 639.24,
|
| 201054 |
+
"learning_rate": 8.72546042003231e-06,
|
| 201055 |
+
"loss": 0.3217,
|
| 201056 |
+
"step": 79925
|
| 201057 |
+
},
|
| 201058 |
+
{
|
| 201059 |
+
"epoch": 639.28,
|
| 201060 |
+
"learning_rate": 8.725379644588046e-06,
|
| 201061 |
+
"loss": 0.3378,
|
| 201062 |
+
"step": 79930
|
| 201063 |
+
},
|
| 201064 |
+
{
|
| 201065 |
+
"epoch": 639.32,
|
| 201066 |
+
"learning_rate": 8.72529886914378e-06,
|
| 201067 |
+
"loss": 0.3434,
|
| 201068 |
+
"step": 79935
|
| 201069 |
+
},
|
| 201070 |
+
{
|
| 201071 |
+
"epoch": 639.36,
|
| 201072 |
+
"learning_rate": 8.725218093699516e-06,
|
| 201073 |
+
"loss": 0.4547,
|
| 201074 |
+
"step": 79940
|
| 201075 |
+
},
|
| 201076 |
+
{
|
| 201077 |
+
"epoch": 639.4,
|
| 201078 |
+
"learning_rate": 8.72513731825525e-06,
|
| 201079 |
+
"loss": 1.2898,
|
| 201080 |
+
"step": 79945
|
| 201081 |
+
},
|
| 201082 |
+
{
|
| 201083 |
+
"epoch": 639.44,
|
| 201084 |
+
"learning_rate": 8.725056542810986e-06,
|
| 201085 |
+
"loss": 0.3957,
|
| 201086 |
+
"step": 79950
|
| 201087 |
+
},
|
| 201088 |
+
{
|
| 201089 |
+
"epoch": 639.48,
|
| 201090 |
+
"learning_rate": 8.72497576736672e-06,
|
| 201091 |
+
"loss": 0.3162,
|
| 201092 |
+
"step": 79955
|
| 201093 |
+
},
|
| 201094 |
+
{
|
| 201095 |
+
"epoch": 639.52,
|
| 201096 |
+
"learning_rate": 8.724894991922456e-06,
|
| 201097 |
+
"loss": 0.3148,
|
| 201098 |
+
"step": 79960
|
| 201099 |
+
},
|
| 201100 |
+
{
|
| 201101 |
+
"epoch": 639.56,
|
| 201102 |
+
"learning_rate": 8.72481421647819e-06,
|
| 201103 |
+
"loss": 0.5403,
|
| 201104 |
+
"step": 79965
|
| 201105 |
+
},
|
| 201106 |
+
{
|
| 201107 |
+
"epoch": 639.6,
|
| 201108 |
+
"learning_rate": 8.724733441033926e-06,
|
| 201109 |
+
"loss": 1.2322,
|
| 201110 |
+
"step": 79970
|
| 201111 |
+
},
|
| 201112 |
+
{
|
| 201113 |
+
"epoch": 639.64,
|
| 201114 |
+
"learning_rate": 8.72465266558966e-06,
|
| 201115 |
+
"loss": 0.3714,
|
| 201116 |
+
"step": 79975
|
| 201117 |
+
},
|
| 201118 |
+
{
|
| 201119 |
+
"epoch": 639.68,
|
| 201120 |
+
"learning_rate": 8.724571890145396e-06,
|
| 201121 |
+
"loss": 0.2936,
|
| 201122 |
+
"step": 79980
|
| 201123 |
+
},
|
| 201124 |
+
{
|
| 201125 |
+
"epoch": 639.72,
|
| 201126 |
+
"learning_rate": 8.72449111470113e-06,
|
| 201127 |
+
"loss": 0.3196,
|
| 201128 |
+
"step": 79985
|
| 201129 |
+
},
|
| 201130 |
+
{
|
| 201131 |
+
"epoch": 639.76,
|
| 201132 |
+
"learning_rate": 8.724410339256866e-06,
|
| 201133 |
+
"loss": 0.4734,
|
| 201134 |
+
"step": 79990
|
| 201135 |
+
},
|
| 201136 |
+
{
|
| 201137 |
+
"epoch": 639.8,
|
| 201138 |
+
"learning_rate": 8.724329563812602e-06,
|
| 201139 |
+
"loss": 1.2245,
|
| 201140 |
+
"step": 79995
|
| 201141 |
+
},
|
| 201142 |
+
{
|
| 201143 |
+
"epoch": 639.84,
|
| 201144 |
+
"learning_rate": 8.724248788368336e-06,
|
| 201145 |
+
"loss": 0.3541,
|
| 201146 |
+
"step": 80000
|
| 201147 |
+
},
|
| 201148 |
+
{
|
| 201149 |
+
"epoch": 639.88,
|
| 201150 |
+
"learning_rate": 8.724168012924072e-06,
|
| 201151 |
+
"loss": 0.3572,
|
| 201152 |
+
"step": 80005
|
| 201153 |
+
},
|
| 201154 |
+
{
|
| 201155 |
+
"epoch": 639.92,
|
| 201156 |
+
"learning_rate": 8.724087237479806e-06,
|
| 201157 |
+
"loss": 0.3127,
|
| 201158 |
+
"step": 80010
|
| 201159 |
+
},
|
| 201160 |
+
{
|
| 201161 |
+
"epoch": 639.96,
|
| 201162 |
+
"learning_rate": 8.724006462035542e-06,
|
| 201163 |
+
"loss": 0.482,
|
| 201164 |
+
"step": 80015
|
| 201165 |
+
},
|
| 201166 |
+
{
|
| 201167 |
+
"epoch": 640.0,
|
| 201168 |
+
"learning_rate": 8.723925686591276e-06,
|
| 201169 |
+
"loss": 1.1057,
|
| 201170 |
+
"step": 80020
|
| 201171 |
+
},
|
| 201172 |
+
{
|
| 201173 |
+
"epoch": 640.0,
|
| 201174 |
+
"eval_loss": 0.40086424350738525,
|
| 201175 |
+
"eval_runtime": 43.143,
|
| 201176 |
+
"eval_samples_per_second": 19.493,
|
| 201177 |
+
"eval_steps_per_second": 0.626,
|
| 201178 |
+
"eval_wer": 0.19523465703971118,
|
| 201179 |
+
"step": 80020
|
| 201180 |
+
},
|
| 201181 |
+
{
|
| 201182 |
+
"epoch": 645.04,
|
| 201183 |
+
"learning_rate": 8.723844911147012e-06,
|
| 201184 |
+
"loss": 0.3555,
|
| 201185 |
+
"step": 80025
|
| 201186 |
+
},
|
| 201187 |
+
{
|
| 201188 |
+
"epoch": 645.08,
|
| 201189 |
+
"learning_rate": 8.723764135702746e-06,
|
| 201190 |
+
"loss": 0.2825,
|
| 201191 |
+
"step": 80030
|
| 201192 |
+
},
|
| 201193 |
+
{
|
| 201194 |
+
"epoch": 645.12,
|
| 201195 |
+
"learning_rate": 8.723683360258482e-06,
|
| 201196 |
+
"loss": 0.3342,
|
| 201197 |
+
"step": 80035
|
| 201198 |
+
},
|
| 201199 |
+
{
|
| 201200 |
+
"epoch": 645.16,
|
| 201201 |
+
"learning_rate": 8.723602584814216e-06,
|
| 201202 |
+
"loss": 0.5168,
|
| 201203 |
+
"step": 80040
|
| 201204 |
+
},
|
| 201205 |
+
{
|
| 201206 |
+
"epoch": 645.2,
|
| 201207 |
+
"learning_rate": 8.723521809369952e-06,
|
| 201208 |
+
"loss": 1.1547,
|
| 201209 |
+
"step": 80045
|
| 201210 |
+
},
|
| 201211 |
+
{
|
| 201212 |
+
"epoch": 645.24,
|
| 201213 |
+
"learning_rate": 8.723441033925686e-06,
|
| 201214 |
+
"loss": 0.331,
|
| 201215 |
+
"step": 80050
|
| 201216 |
+
},
|
| 201217 |
+
{
|
| 201218 |
+
"epoch": 645.28,
|
| 201219 |
+
"learning_rate": 8.723360258481422e-06,
|
| 201220 |
+
"loss": 0.3295,
|
| 201221 |
+
"step": 80055
|
| 201222 |
+
},
|
| 201223 |
+
{
|
| 201224 |
+
"epoch": 645.32,
|
| 201225 |
+
"learning_rate": 8.723279483037156e-06,
|
| 201226 |
+
"loss": 0.3176,
|
| 201227 |
+
"step": 80060
|
| 201228 |
+
},
|
| 201229 |
+
{
|
| 201230 |
+
"epoch": 645.36,
|
| 201231 |
+
"learning_rate": 8.723198707592892e-06,
|
| 201232 |
+
"loss": 0.5193,
|
| 201233 |
+
"step": 80065
|
| 201234 |
+
},
|
| 201235 |
+
{
|
| 201236 |
+
"epoch": 645.4,
|
| 201237 |
+
"learning_rate": 8.723117932148628e-06,
|
| 201238 |
+
"loss": 1.2665,
|
| 201239 |
+
"step": 80070
|
| 201240 |
+
},
|
| 201241 |
+
{
|
| 201242 |
+
"epoch": 645.44,
|
| 201243 |
+
"learning_rate": 8.723037156704362e-06,
|
| 201244 |
+
"loss": 0.306,
|
| 201245 |
+
"step": 80075
|
| 201246 |
+
},
|
| 201247 |
+
{
|
| 201248 |
+
"epoch": 645.48,
|
| 201249 |
+
"learning_rate": 8.722956381260098e-06,
|
| 201250 |
+
"loss": 0.3392,
|
| 201251 |
+
"step": 80080
|
| 201252 |
+
},
|
| 201253 |
+
{
|
| 201254 |
+
"epoch": 645.52,
|
| 201255 |
+
"learning_rate": 8.722875605815832e-06,
|
| 201256 |
+
"loss": 0.3802,
|
| 201257 |
+
"step": 80085
|
| 201258 |
+
},
|
| 201259 |
+
{
|
| 201260 |
+
"epoch": 645.56,
|
| 201261 |
+
"learning_rate": 8.722794830371568e-06,
|
| 201262 |
+
"loss": 0.4104,
|
| 201263 |
+
"step": 80090
|
| 201264 |
+
},
|
| 201265 |
+
{
|
| 201266 |
+
"epoch": 645.6,
|
| 201267 |
+
"learning_rate": 8.722714054927302e-06,
|
| 201268 |
+
"loss": 1.2221,
|
| 201269 |
+
"step": 80095
|
| 201270 |
+
},
|
| 201271 |
+
{
|
| 201272 |
+
"epoch": 645.64,
|
| 201273 |
+
"learning_rate": 8.722633279483038e-06,
|
| 201274 |
+
"loss": 0.3854,
|
| 201275 |
+
"step": 80100
|
| 201276 |
+
},
|
| 201277 |
+
{
|
| 201278 |
+
"epoch": 645.68,
|
| 201279 |
+
"learning_rate": 8.722552504038772e-06,
|
| 201280 |
+
"loss": 0.357,
|
| 201281 |
+
"step": 80105
|
| 201282 |
+
},
|
| 201283 |
+
{
|
| 201284 |
+
"epoch": 645.72,
|
| 201285 |
+
"learning_rate": 8.722471728594508e-06,
|
| 201286 |
+
"loss": 0.5174,
|
| 201287 |
+
"step": 80110
|
| 201288 |
+
},
|
| 201289 |
+
{
|
| 201290 |
+
"epoch": 645.76,
|
| 201291 |
+
"learning_rate": 8.722390953150242e-06,
|
| 201292 |
+
"loss": 0.5564,
|
| 201293 |
+
"step": 80115
|
| 201294 |
+
},
|
| 201295 |
+
{
|
| 201296 |
+
"epoch": 645.8,
|
| 201297 |
+
"learning_rate": 8.722310177705978e-06,
|
| 201298 |
+
"loss": 1.1946,
|
| 201299 |
+
"step": 80120
|
| 201300 |
+
},
|
| 201301 |
+
{
|
| 201302 |
+
"epoch": 645.84,
|
| 201303 |
+
"learning_rate": 8.722229402261712e-06,
|
| 201304 |
+
"loss": 0.3119,
|
| 201305 |
+
"step": 80125
|
| 201306 |
+
},
|
| 201307 |
+
{
|
| 201308 |
+
"epoch": 645.88,
|
| 201309 |
+
"learning_rate": 8.722148626817448e-06,
|
| 201310 |
+
"loss": 0.305,
|
| 201311 |
+
"step": 80130
|
| 201312 |
+
},
|
| 201313 |
+
{
|
| 201314 |
+
"epoch": 645.92,
|
| 201315 |
+
"learning_rate": 8.722067851373184e-06,
|
| 201316 |
+
"loss": 0.4067,
|
| 201317 |
+
"step": 80135
|
| 201318 |
+
},
|
| 201319 |
+
{
|
| 201320 |
+
"epoch": 645.96,
|
| 201321 |
+
"learning_rate": 8.721987075928918e-06,
|
| 201322 |
+
"loss": 0.4924,
|
| 201323 |
+
"step": 80140
|
| 201324 |
+
},
|
| 201325 |
+
{
|
| 201326 |
+
"epoch": 646.0,
|
| 201327 |
+
"eval_loss": 0.40166351199150085,
|
| 201328 |
+
"eval_runtime": 41.5534,
|
| 201329 |
+
"eval_samples_per_second": 20.239,
|
| 201330 |
+
"eval_steps_per_second": 0.65,
|
| 201331 |
+
"eval_wer": 0.1887832837553508,
|
| 201332 |
+
"step": 80144
|
| 201333 |
+
},
|
| 201334 |
+
{
|
| 201335 |
+
"epoch": 641.01,
|
| 201336 |
+
"learning_rate": 8.721906300484653e-06,
|
| 201337 |
+
"loss": 0.3404,
|
| 201338 |
+
"step": 80145
|
| 201339 |
+
},
|
| 201340 |
+
{
|
| 201341 |
+
"epoch": 641.05,
|
| 201342 |
+
"learning_rate": 8.721825525040388e-06,
|
| 201343 |
+
"loss": 0.325,
|
| 201344 |
+
"step": 80150
|
| 201345 |
+
},
|
| 201346 |
+
{
|
| 201347 |
+
"epoch": 641.09,
|
| 201348 |
+
"learning_rate": 8.721744749596123e-06,
|
| 201349 |
+
"loss": 0.3356,
|
| 201350 |
+
"step": 80155
|
| 201351 |
+
},
|
| 201352 |
+
{
|
| 201353 |
+
"epoch": 641.13,
|
| 201354 |
+
"learning_rate": 8.721663974151858e-06,
|
| 201355 |
+
"loss": 0.3636,
|
| 201356 |
+
"step": 80160
|
| 201357 |
+
},
|
| 201358 |
+
{
|
| 201359 |
+
"epoch": 641.17,
|
| 201360 |
+
"learning_rate": 8.721583198707593e-06,
|
| 201361 |
+
"loss": 0.7166,
|
| 201362 |
+
"step": 80165
|
| 201363 |
+
},
|
| 201364 |
+
{
|
| 201365 |
+
"epoch": 641.21,
|
| 201366 |
+
"learning_rate": 8.721502423263328e-06,
|
| 201367 |
+
"loss": 1.2031,
|
| 201368 |
+
"step": 80170
|
| 201369 |
+
},
|
| 201370 |
+
{
|
| 201371 |
+
"epoch": 641.25,
|
| 201372 |
+
"learning_rate": 8.721421647819063e-06,
|
| 201373 |
+
"loss": 0.2894,
|
| 201374 |
+
"step": 80175
|
| 201375 |
+
},
|
| 201376 |
+
{
|
| 201377 |
+
"epoch": 641.29,
|
| 201378 |
+
"learning_rate": 8.721340872374798e-06,
|
| 201379 |
+
"loss": 0.3588,
|
| 201380 |
+
"step": 80180
|
| 201381 |
+
},
|
| 201382 |
+
{
|
| 201383 |
+
"epoch": 641.33,
|
| 201384 |
+
"learning_rate": 8.721260096930533e-06,
|
| 201385 |
+
"loss": 0.344,
|
| 201386 |
+
"step": 80185
|
| 201387 |
+
},
|
| 201388 |
+
{
|
| 201389 |
+
"epoch": 641.37,
|
| 201390 |
+
"learning_rate": 8.721179321486268e-06,
|
| 201391 |
+
"loss": 0.5297,
|
| 201392 |
+
"step": 80190
|
| 201393 |
+
},
|
| 201394 |
+
{
|
| 201395 |
+
"epoch": 641.41,
|
| 201396 |
+
"learning_rate": 8.721098546042003e-06,
|
| 201397 |
+
"loss": 1.0916,
|
| 201398 |
+
"step": 80195
|
| 201399 |
+
},
|
| 201400 |
+
{
|
| 201401 |
+
"epoch": 641.45,
|
| 201402 |
+
"learning_rate": 8.72101777059774e-06,
|
| 201403 |
+
"loss": 0.2758,
|
| 201404 |
+
"step": 80200
|
| 201405 |
+
},
|
| 201406 |
+
{
|
| 201407 |
+
"epoch": 641.49,
|
| 201408 |
+
"learning_rate": 8.720936995153473e-06,
|
| 201409 |
+
"loss": 0.2893,
|
| 201410 |
+
"step": 80205
|
| 201411 |
+
},
|
| 201412 |
+
{
|
| 201413 |
+
"epoch": 641.53,
|
| 201414 |
+
"learning_rate": 8.72085621970921e-06,
|
| 201415 |
+
"loss": 0.3141,
|
| 201416 |
+
"step": 80210
|
| 201417 |
+
},
|
| 201418 |
+
{
|
| 201419 |
+
"epoch": 641.57,
|
| 201420 |
+
"learning_rate": 8.720775444264943e-06,
|
| 201421 |
+
"loss": 0.5601,
|
| 201422 |
+
"step": 80215
|
| 201423 |
+
},
|
| 201424 |
+
{
|
| 201425 |
+
"epoch": 641.61,
|
| 201426 |
+
"learning_rate": 8.720694668820679e-06,
|
| 201427 |
+
"loss": 1.1339,
|
| 201428 |
+
"step": 80220
|
| 201429 |
+
},
|
| 201430 |
+
{
|
| 201431 |
+
"epoch": 641.65,
|
| 201432 |
+
"learning_rate": 8.720613893376413e-06,
|
| 201433 |
+
"loss": 0.2912,
|
| 201434 |
+
"step": 80225
|
| 201435 |
+
},
|
| 201436 |
+
{
|
| 201437 |
+
"epoch": 641.69,
|
| 201438 |
+
"learning_rate": 8.720533117932149e-06,
|
| 201439 |
+
"loss": 0.2561,
|
| 201440 |
+
"step": 80230
|
| 201441 |
+
},
|
| 201442 |
+
{
|
| 201443 |
+
"epoch": 641.73,
|
| 201444 |
+
"learning_rate": 8.720452342487883e-06,
|
| 201445 |
+
"loss": 0.3522,
|
| 201446 |
+
"step": 80235
|
| 201447 |
+
},
|
| 201448 |
+
{
|
| 201449 |
+
"epoch": 641.77,
|
| 201450 |
+
"learning_rate": 8.720371567043619e-06,
|
| 201451 |
+
"loss": 0.5565,
|
| 201452 |
+
"step": 80240
|
| 201453 |
+
},
|
| 201454 |
+
{
|
| 201455 |
+
"epoch": 641.81,
|
| 201456 |
+
"learning_rate": 8.720290791599353e-06,
|
| 201457 |
+
"loss": 1.2682,
|
| 201458 |
+
"step": 80245
|
| 201459 |
+
},
|
| 201460 |
+
{
|
| 201461 |
+
"epoch": 641.85,
|
| 201462 |
+
"learning_rate": 8.720210016155089e-06,
|
| 201463 |
+
"loss": 0.2933,
|
| 201464 |
+
"step": 80250
|
| 201465 |
+
},
|
| 201466 |
+
{
|
| 201467 |
+
"epoch": 641.89,
|
| 201468 |
+
"learning_rate": 8.720129240710825e-06,
|
| 201469 |
+
"loss": 0.3283,
|
| 201470 |
+
"step": 80255
|
| 201471 |
+
},
|
| 201472 |
+
{
|
| 201473 |
+
"epoch": 641.93,
|
| 201474 |
+
"learning_rate": 8.720048465266559e-06,
|
| 201475 |
+
"loss": 0.3409,
|
| 201476 |
+
"step": 80260
|
| 201477 |
+
},
|
| 201478 |
+
{
|
| 201479 |
+
"epoch": 641.97,
|
| 201480 |
+
"learning_rate": 8.719967689822295e-06,
|
| 201481 |
+
"loss": 0.5759,
|
| 201482 |
+
"step": 80265
|
| 201483 |
+
},
|
| 201484 |
+
{
|
| 201485 |
+
"epoch": 642.0,
|
| 201486 |
+
"eval_loss": 0.43214529752731323,
|
| 201487 |
+
"eval_runtime": 43.1264,
|
| 201488 |
+
"eval_samples_per_second": 19.501,
|
| 201489 |
+
"eval_steps_per_second": 0.626,
|
| 201490 |
+
"eval_wer": 0.18905362543335547,
|
| 201491 |
+
"step": 80269
|
| 201492 |
}
|
| 201493 |
],
|
| 201494 |
+
"max_steps": 625000,
|
| 201495 |
"num_train_epochs": 5000,
|
| 201496 |
+
"total_flos": 2.258889246771459e+20,
|
| 201497 |
"trial_name": null,
|
| 201498 |
"trial_params": null
|
| 201499 |
}
|
model-bin/finetune/base/{checkpoint-79647 → checkpoint-80269}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629901744.9403872/events.out.tfevents.1629901744.7e498afd5545.7645.55
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:34dc2a50ea0adcb3166e140d46a097ce8d11caf48de16acd58516039a1925944
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629902211.333175/events.out.tfevents.1629902211.7e498afd5545.7645.57
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5d64ee7baa7bf328f599e5a902ce7aff625fd640180a7b6ae1da505d91481609
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629902676.4913468/events.out.tfevents.1629902676.7e498afd5545.7645.59
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ae7fa1866134c5a1d49d534d7f3268678df59831d6c3683e29b7bf8ce8d37eab
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629903142.829317/events.out.tfevents.1629903142.7e498afd5545.7645.61
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a41a8cf3d2fae8ea06f813a49c44c9057db63837b1fb2c4dff7b5c666089a920
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629903613.046621/events.out.tfevents.1629903613.7e498afd5545.7645.63
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6a89bea88e97e0a0919727e772bdf5bf53103d2ff9e0cb5bee7d87ca28fb28dc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629901744.7e498afd5545.7645.54
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e841d0dda1f95d21f31e00fc53b7c264fb3de7fcf0abbf4dda9d1a368708ec1d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629902211.7e498afd5545.7645.56
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:01089992c000a2e2684bfc1abab5ba7dd58b1ccd453c7a0d87fe2df937261c42
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629902675.7e498afd5545.7645.58
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cdb53525493a599bbbc2df21d433acacc8dd297d8772df0924c31b6d03c9bafd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629903142.7e498afd5545.7645.60
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ad2c155d804bffe59b4cdde03efb02cc4ab51680477529c24ca48a27997986c3
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629903613.7e498afd5545.7645.62
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f3607d09f1eb06a16f8978600368ee506d04af1ecee9ef47d24458f03718a15c
|
| 3 |
+
size 8622
|