"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629872387.216304/events.out.tfevents.1629872387.7e498afd5545.905.63 +3 -0
- model-bin/finetune/base/log/1629873040.9577622/events.out.tfevents.1629873040.7e498afd5545.905.65 +3 -0
- model-bin/finetune/base/log/1629873815.6117036/events.out.tfevents.1629873815.7e498afd5545.905.67 +3 -0
- model-bin/finetune/base/log/1629874460.1476715/events.out.tfevents.1629874460.7e498afd5545.905.69 +3 -0
- model-bin/finetune/base/log/1629875104.9909084/events.out.tfevents.1629875104.7e498afd5545.905.71 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629872387.7e498afd5545.905.62 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629873040.7e498afd5545.905.64 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629873815.7e498afd5545.905.66 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629874460.7e498afd5545.905.68 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629875104.7e498afd5545.905.70 +3 -0
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/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:dfd36b6ee499d562847b385dc3ea784f1e658ef42e11f2762942b6000097d6a6
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/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 sha256:633059da288f26b3e326a5988ea125d0c4538291da52f14c3625a68fb22478fa
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/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:e563717b5c623c6ea44320c2c602e0fe01a69436b866f3fa268004b71402a438
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/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:cb98e84d72fc9445ed9adb6eb66f1ee1e64e6fd672c8cf163f23ed10f15dd0e7
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/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:d5509a63f3d04d5bbd7c28685f03d0c9d6c508593d58909a296e0f060d7e0530
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -192615,11 +192615,806 @@
|
|
| 192615 |
"eval_steps_per_second": 0.654,
|
| 192616 |
"eval_wer": 0.19331724793448857,
|
| 192617 |
"step": 73294
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 192618 |
}
|
| 192619 |
],
|
| 192620 |
-
"max_steps":
|
| 192621 |
"num_train_epochs": 5000,
|
| 192622 |
-
"total_flos": 2.
|
| 192623 |
"trial_name": null,
|
| 192624 |
"trial_params": null
|
| 192625 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 595.995983935743,
|
| 5 |
+
"global_step": 73916,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 192615 |
"eval_steps_per_second": 0.654,
|
| 192616 |
"eval_wer": 0.19331724793448857,
|
| 192617 |
"step": 73294
|
| 192618 |
+
},
|
| 192619 |
+
{
|
| 192620 |
+
"epoch": 591.01,
|
| 192621 |
+
"learning_rate": 8.841858974358975e-06,
|
| 192622 |
+
"loss": 0.3308,
|
| 192623 |
+
"step": 73295
|
| 192624 |
+
},
|
| 192625 |
+
{
|
| 192626 |
+
"epoch": 591.05,
|
| 192627 |
+
"learning_rate": 8.841778846153846e-06,
|
| 192628 |
+
"loss": 0.313,
|
| 192629 |
+
"step": 73300
|
| 192630 |
+
},
|
| 192631 |
+
{
|
| 192632 |
+
"epoch": 591.09,
|
| 192633 |
+
"learning_rate": 8.841698717948719e-06,
|
| 192634 |
+
"loss": 0.279,
|
| 192635 |
+
"step": 73305
|
| 192636 |
+
},
|
| 192637 |
+
{
|
| 192638 |
+
"epoch": 591.13,
|
| 192639 |
+
"learning_rate": 8.84161858974359e-06,
|
| 192640 |
+
"loss": 0.3329,
|
| 192641 |
+
"step": 73310
|
| 192642 |
+
},
|
| 192643 |
+
{
|
| 192644 |
+
"epoch": 591.17,
|
| 192645 |
+
"learning_rate": 8.841538461538462e-06,
|
| 192646 |
+
"loss": 0.6186,
|
| 192647 |
+
"step": 73315
|
| 192648 |
+
},
|
| 192649 |
+
{
|
| 192650 |
+
"epoch": 591.21,
|
| 192651 |
+
"learning_rate": 8.841458333333333e-06,
|
| 192652 |
+
"loss": 1.1625,
|
| 192653 |
+
"step": 73320
|
| 192654 |
+
},
|
| 192655 |
+
{
|
| 192656 |
+
"epoch": 591.25,
|
| 192657 |
+
"learning_rate": 8.841378205128206e-06,
|
| 192658 |
+
"loss": 0.5247,
|
| 192659 |
+
"step": 73325
|
| 192660 |
+
},
|
| 192661 |
+
{
|
| 192662 |
+
"epoch": 591.29,
|
| 192663 |
+
"learning_rate": 8.841298076923078e-06,
|
| 192664 |
+
"loss": 0.2802,
|
| 192665 |
+
"step": 73330
|
| 192666 |
+
},
|
| 192667 |
+
{
|
| 192668 |
+
"epoch": 591.33,
|
| 192669 |
+
"learning_rate": 8.841217948717949e-06,
|
| 192670 |
+
"loss": 0.3827,
|
| 192671 |
+
"step": 73335
|
| 192672 |
+
},
|
| 192673 |
+
{
|
| 192674 |
+
"epoch": 591.37,
|
| 192675 |
+
"learning_rate": 8.841137820512822e-06,
|
| 192676 |
+
"loss": 0.5569,
|
| 192677 |
+
"step": 73340
|
| 192678 |
+
},
|
| 192679 |
+
{
|
| 192680 |
+
"epoch": 591.41,
|
| 192681 |
+
"learning_rate": 8.841057692307693e-06,
|
| 192682 |
+
"loss": 1.2389,
|
| 192683 |
+
"step": 73345
|
| 192684 |
+
},
|
| 192685 |
+
{
|
| 192686 |
+
"epoch": 591.45,
|
| 192687 |
+
"learning_rate": 8.840977564102565e-06,
|
| 192688 |
+
"loss": 0.3074,
|
| 192689 |
+
"step": 73350
|
| 192690 |
+
},
|
| 192691 |
+
{
|
| 192692 |
+
"epoch": 591.49,
|
| 192693 |
+
"learning_rate": 8.840897435897436e-06,
|
| 192694 |
+
"loss": 0.2857,
|
| 192695 |
+
"step": 73355
|
| 192696 |
+
},
|
| 192697 |
+
{
|
| 192698 |
+
"epoch": 591.53,
|
| 192699 |
+
"learning_rate": 8.84081730769231e-06,
|
| 192700 |
+
"loss": 0.3403,
|
| 192701 |
+
"step": 73360
|
| 192702 |
+
},
|
| 192703 |
+
{
|
| 192704 |
+
"epoch": 591.57,
|
| 192705 |
+
"learning_rate": 8.84073717948718e-06,
|
| 192706 |
+
"loss": 0.5408,
|
| 192707 |
+
"step": 73365
|
| 192708 |
+
},
|
| 192709 |
+
{
|
| 192710 |
+
"epoch": 591.61,
|
| 192711 |
+
"learning_rate": 8.840657051282052e-06,
|
| 192712 |
+
"loss": 1.0292,
|
| 192713 |
+
"step": 73370
|
| 192714 |
+
},
|
| 192715 |
+
{
|
| 192716 |
+
"epoch": 591.65,
|
| 192717 |
+
"learning_rate": 8.840576923076923e-06,
|
| 192718 |
+
"loss": 0.3484,
|
| 192719 |
+
"step": 73375
|
| 192720 |
+
},
|
| 192721 |
+
{
|
| 192722 |
+
"epoch": 591.69,
|
| 192723 |
+
"learning_rate": 8.840496794871796e-06,
|
| 192724 |
+
"loss": 0.3111,
|
| 192725 |
+
"step": 73380
|
| 192726 |
+
},
|
| 192727 |
+
{
|
| 192728 |
+
"epoch": 591.73,
|
| 192729 |
+
"learning_rate": 8.840416666666668e-06,
|
| 192730 |
+
"loss": 0.331,
|
| 192731 |
+
"step": 73385
|
| 192732 |
+
},
|
| 192733 |
+
{
|
| 192734 |
+
"epoch": 591.77,
|
| 192735 |
+
"learning_rate": 8.840336538461539e-06,
|
| 192736 |
+
"loss": 0.6024,
|
| 192737 |
+
"step": 73390
|
| 192738 |
+
},
|
| 192739 |
+
{
|
| 192740 |
+
"epoch": 591.81,
|
| 192741 |
+
"learning_rate": 8.840256410256412e-06,
|
| 192742 |
+
"loss": 1.2207,
|
| 192743 |
+
"step": 73395
|
| 192744 |
+
},
|
| 192745 |
+
{
|
| 192746 |
+
"epoch": 591.85,
|
| 192747 |
+
"learning_rate": 8.840176282051283e-06,
|
| 192748 |
+
"loss": 0.3448,
|
| 192749 |
+
"step": 73400
|
| 192750 |
+
},
|
| 192751 |
+
{
|
| 192752 |
+
"epoch": 591.9,
|
| 192753 |
+
"learning_rate": 8.840096153846155e-06,
|
| 192754 |
+
"loss": 0.3189,
|
| 192755 |
+
"step": 73405
|
| 192756 |
+
},
|
| 192757 |
+
{
|
| 192758 |
+
"epoch": 591.94,
|
| 192759 |
+
"learning_rate": 8.840016025641026e-06,
|
| 192760 |
+
"loss": 0.3932,
|
| 192761 |
+
"step": 73410
|
| 192762 |
+
},
|
| 192763 |
+
{
|
| 192764 |
+
"epoch": 591.98,
|
| 192765 |
+
"learning_rate": 8.8399358974359e-06,
|
| 192766 |
+
"loss": 0.6275,
|
| 192767 |
+
"step": 73415
|
| 192768 |
+
},
|
| 192769 |
+
{
|
| 192770 |
+
"epoch": 592.0,
|
| 192771 |
+
"eval_loss": 0.3852952718734741,
|
| 192772 |
+
"eval_runtime": 40.5465,
|
| 192773 |
+
"eval_samples_per_second": 20.717,
|
| 192774 |
+
"eval_steps_per_second": 0.666,
|
| 192775 |
+
"eval_wer": 0.18575874333135745,
|
| 192776 |
+
"step": 73418
|
| 192777 |
+
},
|
| 192778 |
+
{
|
| 192779 |
+
"epoch": 587.02,
|
| 192780 |
+
"learning_rate": 8.839855769230769e-06,
|
| 192781 |
+
"loss": 0.4346,
|
| 192782 |
+
"step": 73420
|
| 192783 |
+
},
|
| 192784 |
+
{
|
| 192785 |
+
"epoch": 587.06,
|
| 192786 |
+
"learning_rate": 8.839775641025642e-06,
|
| 192787 |
+
"loss": 0.3137,
|
| 192788 |
+
"step": 73425
|
| 192789 |
+
},
|
| 192790 |
+
{
|
| 192791 |
+
"epoch": 587.1,
|
| 192792 |
+
"learning_rate": 8.839695512820513e-06,
|
| 192793 |
+
"loss": 0.3072,
|
| 192794 |
+
"step": 73430
|
| 192795 |
+
},
|
| 192796 |
+
{
|
| 192797 |
+
"epoch": 587.14,
|
| 192798 |
+
"learning_rate": 8.839615384615385e-06,
|
| 192799 |
+
"loss": 0.3478,
|
| 192800 |
+
"step": 73435
|
| 192801 |
+
},
|
| 192802 |
+
{
|
| 192803 |
+
"epoch": 587.18,
|
| 192804 |
+
"learning_rate": 8.839535256410258e-06,
|
| 192805 |
+
"loss": 0.5605,
|
| 192806 |
+
"step": 73440
|
| 192807 |
+
},
|
| 192808 |
+
{
|
| 192809 |
+
"epoch": 587.22,
|
| 192810 |
+
"learning_rate": 8.839455128205129e-06,
|
| 192811 |
+
"loss": 1.1978,
|
| 192812 |
+
"step": 73445
|
| 192813 |
+
},
|
| 192814 |
+
{
|
| 192815 |
+
"epoch": 587.26,
|
| 192816 |
+
"learning_rate": 8.839375e-06,
|
| 192817 |
+
"loss": 0.308,
|
| 192818 |
+
"step": 73450
|
| 192819 |
+
},
|
| 192820 |
+
{
|
| 192821 |
+
"epoch": 587.3,
|
| 192822 |
+
"learning_rate": 8.839294871794872e-06,
|
| 192823 |
+
"loss": 0.2844,
|
| 192824 |
+
"step": 73455
|
| 192825 |
+
},
|
| 192826 |
+
{
|
| 192827 |
+
"epoch": 587.34,
|
| 192828 |
+
"learning_rate": 8.839214743589745e-06,
|
| 192829 |
+
"loss": 0.3765,
|
| 192830 |
+
"step": 73460
|
| 192831 |
+
},
|
| 192832 |
+
{
|
| 192833 |
+
"epoch": 587.38,
|
| 192834 |
+
"learning_rate": 8.839134615384616e-06,
|
| 192835 |
+
"loss": 0.6947,
|
| 192836 |
+
"step": 73465
|
| 192837 |
+
},
|
| 192838 |
+
{
|
| 192839 |
+
"epoch": 587.42,
|
| 192840 |
+
"learning_rate": 8.839054487179488e-06,
|
| 192841 |
+
"loss": 0.9649,
|
| 192842 |
+
"step": 73470
|
| 192843 |
+
},
|
| 192844 |
+
{
|
| 192845 |
+
"epoch": 587.46,
|
| 192846 |
+
"learning_rate": 8.838974358974359e-06,
|
| 192847 |
+
"loss": 0.2923,
|
| 192848 |
+
"step": 73475
|
| 192849 |
+
},
|
| 192850 |
+
{
|
| 192851 |
+
"epoch": 587.5,
|
| 192852 |
+
"learning_rate": 8.838894230769232e-06,
|
| 192853 |
+
"loss": 0.3179,
|
| 192854 |
+
"step": 73480
|
| 192855 |
+
},
|
| 192856 |
+
{
|
| 192857 |
+
"epoch": 587.54,
|
| 192858 |
+
"learning_rate": 8.838814102564103e-06,
|
| 192859 |
+
"loss": 0.3462,
|
| 192860 |
+
"step": 73485
|
| 192861 |
+
},
|
| 192862 |
+
{
|
| 192863 |
+
"epoch": 587.58,
|
| 192864 |
+
"learning_rate": 8.838733974358975e-06,
|
| 192865 |
+
"loss": 0.7362,
|
| 192866 |
+
"step": 73490
|
| 192867 |
+
},
|
| 192868 |
+
{
|
| 192869 |
+
"epoch": 587.62,
|
| 192870 |
+
"learning_rate": 8.838653846153848e-06,
|
| 192871 |
+
"loss": 1.1776,
|
| 192872 |
+
"step": 73495
|
| 192873 |
+
},
|
| 192874 |
+
{
|
| 192875 |
+
"epoch": 587.66,
|
| 192876 |
+
"learning_rate": 8.838573717948719e-06,
|
| 192877 |
+
"loss": 0.3046,
|
| 192878 |
+
"step": 73500
|
| 192879 |
+
},
|
| 192880 |
+
{
|
| 192881 |
+
"epoch": 587.7,
|
| 192882 |
+
"learning_rate": 8.83849358974359e-06,
|
| 192883 |
+
"loss": 0.3213,
|
| 192884 |
+
"step": 73505
|
| 192885 |
+
},
|
| 192886 |
+
{
|
| 192887 |
+
"epoch": 587.74,
|
| 192888 |
+
"learning_rate": 8.838413461538462e-06,
|
| 192889 |
+
"loss": 0.3917,
|
| 192890 |
+
"step": 73510
|
| 192891 |
+
},
|
| 192892 |
+
{
|
| 192893 |
+
"epoch": 587.78,
|
| 192894 |
+
"learning_rate": 8.838333333333335e-06,
|
| 192895 |
+
"loss": 0.8034,
|
| 192896 |
+
"step": 73515
|
| 192897 |
+
},
|
| 192898 |
+
{
|
| 192899 |
+
"epoch": 587.82,
|
| 192900 |
+
"learning_rate": 8.838253205128206e-06,
|
| 192901 |
+
"loss": 0.8914,
|
| 192902 |
+
"step": 73520
|
| 192903 |
+
},
|
| 192904 |
+
{
|
| 192905 |
+
"epoch": 587.86,
|
| 192906 |
+
"learning_rate": 8.838173076923078e-06,
|
| 192907 |
+
"loss": 0.2794,
|
| 192908 |
+
"step": 73525
|
| 192909 |
+
},
|
| 192910 |
+
{
|
| 192911 |
+
"epoch": 587.9,
|
| 192912 |
+
"learning_rate": 8.838092948717949e-06,
|
| 192913 |
+
"loss": 0.5335,
|
| 192914 |
+
"step": 73530
|
| 192915 |
+
},
|
| 192916 |
+
{
|
| 192917 |
+
"epoch": 587.94,
|
| 192918 |
+
"learning_rate": 8.838012820512822e-06,
|
| 192919 |
+
"loss": 0.3818,
|
| 192920 |
+
"step": 73535
|
| 192921 |
+
},
|
| 192922 |
+
{
|
| 192923 |
+
"epoch": 587.98,
|
| 192924 |
+
"learning_rate": 8.837932692307693e-06,
|
| 192925 |
+
"loss": 0.6372,
|
| 192926 |
+
"step": 73540
|
| 192927 |
+
},
|
| 192928 |
+
{
|
| 192929 |
+
"epoch": 588.0,
|
| 192930 |
+
"eval_loss": 0.40962567925453186,
|
| 192931 |
+
"eval_runtime": 39.6457,
|
| 192932 |
+
"eval_samples_per_second": 21.188,
|
| 192933 |
+
"eval_steps_per_second": 0.681,
|
| 192934 |
+
"eval_wer": 0.1885777450257922,
|
| 192935 |
+
"step": 73543
|
| 192936 |
+
},
|
| 192937 |
+
{
|
| 192938 |
+
"epoch": 593.02,
|
| 192939 |
+
"learning_rate": 8.83786858974359e-06,
|
| 192940 |
+
"loss": 0.4499,
|
| 192941 |
+
"step": 73545
|
| 192942 |
+
},
|
| 192943 |
+
{
|
| 192944 |
+
"epoch": 593.06,
|
| 192945 |
+
"learning_rate": 8.837788461538462e-06,
|
| 192946 |
+
"loss": 0.3547,
|
| 192947 |
+
"step": 73550
|
| 192948 |
+
},
|
| 192949 |
+
{
|
| 192950 |
+
"epoch": 593.1,
|
| 192951 |
+
"learning_rate": 8.837708333333333e-06,
|
| 192952 |
+
"loss": 0.3296,
|
| 192953 |
+
"step": 73555
|
| 192954 |
+
},
|
| 192955 |
+
{
|
| 192956 |
+
"epoch": 593.14,
|
| 192957 |
+
"learning_rate": 8.837628205128207e-06,
|
| 192958 |
+
"loss": 0.3436,
|
| 192959 |
+
"step": 73560
|
| 192960 |
+
},
|
| 192961 |
+
{
|
| 192962 |
+
"epoch": 593.18,
|
| 192963 |
+
"learning_rate": 8.837548076923078e-06,
|
| 192964 |
+
"loss": 0.6371,
|
| 192965 |
+
"step": 73565
|
| 192966 |
+
},
|
| 192967 |
+
{
|
| 192968 |
+
"epoch": 593.22,
|
| 192969 |
+
"learning_rate": 8.83746794871795e-06,
|
| 192970 |
+
"loss": 1.066,
|
| 192971 |
+
"step": 73570
|
| 192972 |
+
},
|
| 192973 |
+
{
|
| 192974 |
+
"epoch": 593.26,
|
| 192975 |
+
"learning_rate": 8.837387820512822e-06,
|
| 192976 |
+
"loss": 0.3452,
|
| 192977 |
+
"step": 73575
|
| 192978 |
+
},
|
| 192979 |
+
{
|
| 192980 |
+
"epoch": 593.3,
|
| 192981 |
+
"learning_rate": 8.837307692307694e-06,
|
| 192982 |
+
"loss": 0.3008,
|
| 192983 |
+
"step": 73580
|
| 192984 |
+
},
|
| 192985 |
+
{
|
| 192986 |
+
"epoch": 593.34,
|
| 192987 |
+
"learning_rate": 8.837227564102565e-06,
|
| 192988 |
+
"loss": 0.3361,
|
| 192989 |
+
"step": 73585
|
| 192990 |
+
},
|
| 192991 |
+
{
|
| 192992 |
+
"epoch": 593.38,
|
| 192993 |
+
"learning_rate": 8.837147435897436e-06,
|
| 192994 |
+
"loss": 0.6165,
|
| 192995 |
+
"step": 73590
|
| 192996 |
+
},
|
| 192997 |
+
{
|
| 192998 |
+
"epoch": 593.42,
|
| 192999 |
+
"learning_rate": 8.83706730769231e-06,
|
| 193000 |
+
"loss": 0.9722,
|
| 193001 |
+
"step": 73595
|
| 193002 |
+
},
|
| 193003 |
+
{
|
| 193004 |
+
"epoch": 593.46,
|
| 193005 |
+
"learning_rate": 8.836987179487179e-06,
|
| 193006 |
+
"loss": 0.3145,
|
| 193007 |
+
"step": 73600
|
| 193008 |
+
},
|
| 193009 |
+
{
|
| 193010 |
+
"epoch": 593.5,
|
| 193011 |
+
"learning_rate": 8.836907051282052e-06,
|
| 193012 |
+
"loss": 0.2744,
|
| 193013 |
+
"step": 73605
|
| 193014 |
+
},
|
| 193015 |
+
{
|
| 193016 |
+
"epoch": 593.54,
|
| 193017 |
+
"learning_rate": 8.836826923076923e-06,
|
| 193018 |
+
"loss": 0.3529,
|
| 193019 |
+
"step": 73610
|
| 193020 |
+
},
|
| 193021 |
+
{
|
| 193022 |
+
"epoch": 593.58,
|
| 193023 |
+
"learning_rate": 8.836746794871795e-06,
|
| 193024 |
+
"loss": 0.588,
|
| 193025 |
+
"step": 73615
|
| 193026 |
+
},
|
| 193027 |
+
{
|
| 193028 |
+
"epoch": 593.62,
|
| 193029 |
+
"learning_rate": 8.836666666666668e-06,
|
| 193030 |
+
"loss": 1.2817,
|
| 193031 |
+
"step": 73620
|
| 193032 |
+
},
|
| 193033 |
+
{
|
| 193034 |
+
"epoch": 593.66,
|
| 193035 |
+
"learning_rate": 8.83658653846154e-06,
|
| 193036 |
+
"loss": 0.2927,
|
| 193037 |
+
"step": 73625
|
| 193038 |
+
},
|
| 193039 |
+
{
|
| 193040 |
+
"epoch": 593.7,
|
| 193041 |
+
"learning_rate": 8.83650641025641e-06,
|
| 193042 |
+
"loss": 0.3421,
|
| 193043 |
+
"step": 73630
|
| 193044 |
+
},
|
| 193045 |
+
{
|
| 193046 |
+
"epoch": 593.74,
|
| 193047 |
+
"learning_rate": 8.836426282051282e-06,
|
| 193048 |
+
"loss": 0.3321,
|
| 193049 |
+
"step": 73635
|
| 193050 |
+
},
|
| 193051 |
+
{
|
| 193052 |
+
"epoch": 593.78,
|
| 193053 |
+
"learning_rate": 8.836346153846155e-06,
|
| 193054 |
+
"loss": 0.7771,
|
| 193055 |
+
"step": 73640
|
| 193056 |
+
},
|
| 193057 |
+
{
|
| 193058 |
+
"epoch": 593.82,
|
| 193059 |
+
"learning_rate": 8.836282051282052e-06,
|
| 193060 |
+
"loss": Infinity,
|
| 193061 |
+
"step": 73645
|
| 193062 |
+
},
|
| 193063 |
+
{
|
| 193064 |
+
"epoch": 593.86,
|
| 193065 |
+
"learning_rate": 8.836201923076924e-06,
|
| 193066 |
+
"loss": NaN,
|
| 193067 |
+
"step": 73650
|
| 193068 |
+
},
|
| 193069 |
+
{
|
| 193070 |
+
"epoch": 593.9,
|
| 193071 |
+
"learning_rate": 8.836121794871795e-06,
|
| 193072 |
+
"loss": NaN,
|
| 193073 |
+
"step": 73655
|
| 193074 |
+
},
|
| 193075 |
+
{
|
| 193076 |
+
"epoch": 593.94,
|
| 193077 |
+
"learning_rate": 8.836041666666667e-06,
|
| 193078 |
+
"loss": NaN,
|
| 193079 |
+
"step": 73660
|
| 193080 |
+
},
|
| 193081 |
+
{
|
| 193082 |
+
"epoch": 593.98,
|
| 193083 |
+
"learning_rate": 8.83596153846154e-06,
|
| 193084 |
+
"loss": NaN,
|
| 193085 |
+
"step": 73665
|
| 193086 |
+
},
|
| 193087 |
+
{
|
| 193088 |
+
"epoch": 594.0,
|
| 193089 |
+
"eval_loss": 0.3711722195148468,
|
| 193090 |
+
"eval_runtime": 38.93,
|
| 193091 |
+
"eval_samples_per_second": 21.577,
|
| 193092 |
+
"eval_steps_per_second": 0.694,
|
| 193093 |
+
"eval_wer": 0.19046610169491526,
|
| 193094 |
+
"step": 73667
|
| 193095 |
+
},
|
| 193096 |
+
{
|
| 193097 |
+
"epoch": 589.02,
|
| 193098 |
+
"learning_rate": 8.835881410256411e-06,
|
| 193099 |
+
"loss": 0.4863,
|
| 193100 |
+
"step": 73670
|
| 193101 |
+
},
|
| 193102 |
+
{
|
| 193103 |
+
"epoch": 589.06,
|
| 193104 |
+
"learning_rate": 8.835801282051282e-06,
|
| 193105 |
+
"loss": 0.2866,
|
| 193106 |
+
"step": 73675
|
| 193107 |
+
},
|
| 193108 |
+
{
|
| 193109 |
+
"epoch": 589.1,
|
| 193110 |
+
"learning_rate": 8.835721153846154e-06,
|
| 193111 |
+
"loss": 0.3515,
|
| 193112 |
+
"step": 73680
|
| 193113 |
+
},
|
| 193114 |
+
{
|
| 193115 |
+
"epoch": 589.14,
|
| 193116 |
+
"learning_rate": 8.835641025641027e-06,
|
| 193117 |
+
"loss": 0.4241,
|
| 193118 |
+
"step": 73685
|
| 193119 |
+
},
|
| 193120 |
+
{
|
| 193121 |
+
"epoch": 589.18,
|
| 193122 |
+
"learning_rate": 8.835560897435898e-06,
|
| 193123 |
+
"loss": 0.7713,
|
| 193124 |
+
"step": 73690
|
| 193125 |
+
},
|
| 193126 |
+
{
|
| 193127 |
+
"epoch": 589.22,
|
| 193128 |
+
"learning_rate": 8.83548076923077e-06,
|
| 193129 |
+
"loss": 0.9977,
|
| 193130 |
+
"step": 73695
|
| 193131 |
+
},
|
| 193132 |
+
{
|
| 193133 |
+
"epoch": 589.26,
|
| 193134 |
+
"learning_rate": 8.835400641025642e-06,
|
| 193135 |
+
"loss": 0.3414,
|
| 193136 |
+
"step": 73700
|
| 193137 |
+
},
|
| 193138 |
+
{
|
| 193139 |
+
"epoch": 589.3,
|
| 193140 |
+
"learning_rate": 8.835320512820514e-06,
|
| 193141 |
+
"loss": 0.319,
|
| 193142 |
+
"step": 73705
|
| 193143 |
+
},
|
| 193144 |
+
{
|
| 193145 |
+
"epoch": 589.34,
|
| 193146 |
+
"learning_rate": 8.835240384615385e-06,
|
| 193147 |
+
"loss": 0.4791,
|
| 193148 |
+
"step": 73710
|
| 193149 |
+
},
|
| 193150 |
+
{
|
| 193151 |
+
"epoch": 589.38,
|
| 193152 |
+
"learning_rate": 8.835160256410257e-06,
|
| 193153 |
+
"loss": 0.8164,
|
| 193154 |
+
"step": 73715
|
| 193155 |
+
},
|
| 193156 |
+
{
|
| 193157 |
+
"epoch": 589.42,
|
| 193158 |
+
"learning_rate": 8.83508012820513e-06,
|
| 193159 |
+
"loss": 0.8898,
|
| 193160 |
+
"step": 73720
|
| 193161 |
+
},
|
| 193162 |
+
{
|
| 193163 |
+
"epoch": 589.46,
|
| 193164 |
+
"learning_rate": 8.835000000000001e-06,
|
| 193165 |
+
"loss": 0.3485,
|
| 193166 |
+
"step": 73725
|
| 193167 |
+
},
|
| 193168 |
+
{
|
| 193169 |
+
"epoch": 589.5,
|
| 193170 |
+
"learning_rate": 8.834919871794872e-06,
|
| 193171 |
+
"loss": 0.3385,
|
| 193172 |
+
"step": 73730
|
| 193173 |
+
},
|
| 193174 |
+
{
|
| 193175 |
+
"epoch": 589.54,
|
| 193176 |
+
"learning_rate": 8.834839743589745e-06,
|
| 193177 |
+
"loss": 0.4159,
|
| 193178 |
+
"step": 73735
|
| 193179 |
+
},
|
| 193180 |
+
{
|
| 193181 |
+
"epoch": 589.58,
|
| 193182 |
+
"learning_rate": 8.834759615384617e-06,
|
| 193183 |
+
"loss": 0.8097,
|
| 193184 |
+
"step": 73740
|
| 193185 |
+
},
|
| 193186 |
+
{
|
| 193187 |
+
"epoch": 589.62,
|
| 193188 |
+
"learning_rate": 8.834679487179488e-06,
|
| 193189 |
+
"loss": 0.9052,
|
| 193190 |
+
"step": 73745
|
| 193191 |
+
},
|
| 193192 |
+
{
|
| 193193 |
+
"epoch": 589.66,
|
| 193194 |
+
"learning_rate": 8.83459935897436e-06,
|
| 193195 |
+
"loss": 0.2916,
|
| 193196 |
+
"step": 73750
|
| 193197 |
+
},
|
| 193198 |
+
{
|
| 193199 |
+
"epoch": 589.7,
|
| 193200 |
+
"learning_rate": 8.834519230769233e-06,
|
| 193201 |
+
"loss": 0.3518,
|
| 193202 |
+
"step": 73755
|
| 193203 |
+
},
|
| 193204 |
+
{
|
| 193205 |
+
"epoch": 589.74,
|
| 193206 |
+
"learning_rate": 8.834439102564102e-06,
|
| 193207 |
+
"loss": 0.4072,
|
| 193208 |
+
"step": 73760
|
| 193209 |
+
},
|
| 193210 |
+
{
|
| 193211 |
+
"epoch": 589.78,
|
| 193212 |
+
"learning_rate": 8.834358974358975e-06,
|
| 193213 |
+
"loss": 0.7854,
|
| 193214 |
+
"step": 73765
|
| 193215 |
+
},
|
| 193216 |
+
{
|
| 193217 |
+
"epoch": 589.82,
|
| 193218 |
+
"learning_rate": 8.834278846153847e-06,
|
| 193219 |
+
"loss": 0.8298,
|
| 193220 |
+
"step": 73770
|
| 193221 |
+
},
|
| 193222 |
+
{
|
| 193223 |
+
"epoch": 589.86,
|
| 193224 |
+
"learning_rate": 8.834198717948718e-06,
|
| 193225 |
+
"loss": 0.3155,
|
| 193226 |
+
"step": 73775
|
| 193227 |
+
},
|
| 193228 |
+
{
|
| 193229 |
+
"epoch": 589.9,
|
| 193230 |
+
"learning_rate": 8.83411858974359e-06,
|
| 193231 |
+
"loss": 0.342,
|
| 193232 |
+
"step": 73780
|
| 193233 |
+
},
|
| 193234 |
+
{
|
| 193235 |
+
"epoch": 589.94,
|
| 193236 |
+
"learning_rate": 8.834038461538462e-06,
|
| 193237 |
+
"loss": 0.3902,
|
| 193238 |
+
"step": 73785
|
| 193239 |
+
},
|
| 193240 |
+
{
|
| 193241 |
+
"epoch": 589.98,
|
| 193242 |
+
"learning_rate": 8.833958333333334e-06,
|
| 193243 |
+
"loss": 0.7749,
|
| 193244 |
+
"step": 73790
|
| 193245 |
+
},
|
| 193246 |
+
{
|
| 193247 |
+
"epoch": 590.0,
|
| 193248 |
+
"eval_loss": 0.4138658046722412,
|
| 193249 |
+
"eval_runtime": 40.8862,
|
| 193250 |
+
"eval_samples_per_second": 20.545,
|
| 193251 |
+
"eval_steps_per_second": 0.66,
|
| 193252 |
+
"eval_wer": 0.19385140905209222,
|
| 193253 |
+
"step": 73792
|
| 193254 |
+
},
|
| 193255 |
+
{
|
| 193256 |
+
"epoch": 595.02,
|
| 193257 |
+
"learning_rate": 8.833878205128205e-06,
|
| 193258 |
+
"loss": 0.3728,
|
| 193259 |
+
"step": 73795
|
| 193260 |
+
},
|
| 193261 |
+
{
|
| 193262 |
+
"epoch": 595.06,
|
| 193263 |
+
"learning_rate": 8.833798076923078e-06,
|
| 193264 |
+
"loss": 0.2469,
|
| 193265 |
+
"step": 73800
|
| 193266 |
+
},
|
| 193267 |
+
{
|
| 193268 |
+
"epoch": 595.1,
|
| 193269 |
+
"learning_rate": 8.83371794871795e-06,
|
| 193270 |
+
"loss": 0.3617,
|
| 193271 |
+
"step": 73805
|
| 193272 |
+
},
|
| 193273 |
+
{
|
| 193274 |
+
"epoch": 595.14,
|
| 193275 |
+
"learning_rate": 8.83363782051282e-06,
|
| 193276 |
+
"loss": 0.5017,
|
| 193277 |
+
"step": 73810
|
| 193278 |
+
},
|
| 193279 |
+
{
|
| 193280 |
+
"epoch": 595.18,
|
| 193281 |
+
"learning_rate": 8.833557692307692e-06,
|
| 193282 |
+
"loss": 0.7071,
|
| 193283 |
+
"step": 73815
|
| 193284 |
+
},
|
| 193285 |
+
{
|
| 193286 |
+
"epoch": 595.22,
|
| 193287 |
+
"learning_rate": 8.833477564102565e-06,
|
| 193288 |
+
"loss": 0.8614,
|
| 193289 |
+
"step": 73820
|
| 193290 |
+
},
|
| 193291 |
+
{
|
| 193292 |
+
"epoch": 595.27,
|
| 193293 |
+
"learning_rate": 8.833397435897437e-06,
|
| 193294 |
+
"loss": 0.3067,
|
| 193295 |
+
"step": 73825
|
| 193296 |
+
},
|
| 193297 |
+
{
|
| 193298 |
+
"epoch": 595.31,
|
| 193299 |
+
"learning_rate": 8.833317307692308e-06,
|
| 193300 |
+
"loss": 0.3224,
|
| 193301 |
+
"step": 73830
|
| 193302 |
+
},
|
| 193303 |
+
{
|
| 193304 |
+
"epoch": 595.35,
|
| 193305 |
+
"learning_rate": 8.833237179487181e-06,
|
| 193306 |
+
"loss": 0.4127,
|
| 193307 |
+
"step": 73835
|
| 193308 |
+
},
|
| 193309 |
+
{
|
| 193310 |
+
"epoch": 595.39,
|
| 193311 |
+
"learning_rate": 8.833157051282052e-06,
|
| 193312 |
+
"loss": 0.8873,
|
| 193313 |
+
"step": 73840
|
| 193314 |
+
},
|
| 193315 |
+
{
|
| 193316 |
+
"epoch": 595.43,
|
| 193317 |
+
"learning_rate": 8.833076923076924e-06,
|
| 193318 |
+
"loss": 0.9318,
|
| 193319 |
+
"step": 73845
|
| 193320 |
+
},
|
| 193321 |
+
{
|
| 193322 |
+
"epoch": 595.47,
|
| 193323 |
+
"learning_rate": 8.832996794871795e-06,
|
| 193324 |
+
"loss": 0.3216,
|
| 193325 |
+
"step": 73850
|
| 193326 |
+
},
|
| 193327 |
+
{
|
| 193328 |
+
"epoch": 595.51,
|
| 193329 |
+
"learning_rate": 8.832916666666668e-06,
|
| 193330 |
+
"loss": 0.3496,
|
| 193331 |
+
"step": 73855
|
| 193332 |
+
},
|
| 193333 |
+
{
|
| 193334 |
+
"epoch": 595.55,
|
| 193335 |
+
"learning_rate": 8.83283653846154e-06,
|
| 193336 |
+
"loss": 0.3752,
|
| 193337 |
+
"step": 73860
|
| 193338 |
+
},
|
| 193339 |
+
{
|
| 193340 |
+
"epoch": 595.59,
|
| 193341 |
+
"learning_rate": 8.832756410256411e-06,
|
| 193342 |
+
"loss": 0.7592,
|
| 193343 |
+
"step": 73865
|
| 193344 |
+
},
|
| 193345 |
+
{
|
| 193346 |
+
"epoch": 595.63,
|
| 193347 |
+
"learning_rate": 8.832676282051282e-06,
|
| 193348 |
+
"loss": 0.927,
|
| 193349 |
+
"step": 73870
|
| 193350 |
+
},
|
| 193351 |
+
{
|
| 193352 |
+
"epoch": 595.67,
|
| 193353 |
+
"learning_rate": 8.832596153846155e-06,
|
| 193354 |
+
"loss": 0.3161,
|
| 193355 |
+
"step": 73875
|
| 193356 |
+
},
|
| 193357 |
+
{
|
| 193358 |
+
"epoch": 595.71,
|
| 193359 |
+
"learning_rate": 8.832516025641027e-06,
|
| 193360 |
+
"loss": 0.361,
|
| 193361 |
+
"step": 73880
|
| 193362 |
+
},
|
| 193363 |
+
{
|
| 193364 |
+
"epoch": 595.75,
|
| 193365 |
+
"learning_rate": 8.832435897435898e-06,
|
| 193366 |
+
"loss": 0.3999,
|
| 193367 |
+
"step": 73885
|
| 193368 |
+
},
|
| 193369 |
+
{
|
| 193370 |
+
"epoch": 595.79,
|
| 193371 |
+
"learning_rate": 8.832355769230771e-06,
|
| 193372 |
+
"loss": 0.7727,
|
| 193373 |
+
"step": 73890
|
| 193374 |
+
},
|
| 193375 |
+
{
|
| 193376 |
+
"epoch": 595.83,
|
| 193377 |
+
"learning_rate": 8.832275641025642e-06,
|
| 193378 |
+
"loss": 0.9198,
|
| 193379 |
+
"step": 73895
|
| 193380 |
+
},
|
| 193381 |
+
{
|
| 193382 |
+
"epoch": 595.87,
|
| 193383 |
+
"learning_rate": 8.832195512820514e-06,
|
| 193384 |
+
"loss": 0.4088,
|
| 193385 |
+
"step": 73900
|
| 193386 |
+
},
|
| 193387 |
+
{
|
| 193388 |
+
"epoch": 595.91,
|
| 193389 |
+
"learning_rate": 8.832115384615385e-06,
|
| 193390 |
+
"loss": 0.3442,
|
| 193391 |
+
"step": 73905
|
| 193392 |
+
},
|
| 193393 |
+
{
|
| 193394 |
+
"epoch": 595.95,
|
| 193395 |
+
"learning_rate": 8.832035256410258e-06,
|
| 193396 |
+
"loss": 0.3927,
|
| 193397 |
+
"step": 73910
|
| 193398 |
+
},
|
| 193399 |
+
{
|
| 193400 |
+
"epoch": 595.99,
|
| 193401 |
+
"learning_rate": 8.831955128205128e-06,
|
| 193402 |
+
"loss": 0.8166,
|
| 193403 |
+
"step": 73915
|
| 193404 |
+
},
|
| 193405 |
+
{
|
| 193406 |
+
"epoch": 596.0,
|
| 193407 |
+
"eval_loss": 0.4341099262237549,
|
| 193408 |
+
"eval_runtime": 39.1108,
|
| 193409 |
+
"eval_samples_per_second": 21.477,
|
| 193410 |
+
"eval_steps_per_second": 0.69,
|
| 193411 |
+
"eval_wer": 0.1899564928840056,
|
| 193412 |
+
"step": 73916
|
| 193413 |
}
|
| 193414 |
],
|
| 193415 |
+
"max_steps": 620000,
|
| 193416 |
"num_train_epochs": 5000,
|
| 193417 |
+
"total_flos": 2.08014258346129e+20,
|
| 193418 |
"trial_name": null,
|
| 193419 |
"trial_params": null
|
| 193420 |
}
|
model-bin/finetune/base/{checkpoint-73294 → checkpoint-73916}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629872387.216304/events.out.tfevents.1629872387.7e498afd5545.905.63
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fab1213ee83b8ab20ed67b0a41b3a7119718916a3e5fc6cccf397fe0a158bb8a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629873040.9577622/events.out.tfevents.1629873040.7e498afd5545.905.65
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cfe40c82647fbf57e2fd91377769c78c9405cf05d96cd1b805e9a0f8ce9062bd
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629873815.6117036/events.out.tfevents.1629873815.7e498afd5545.905.67
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c5b7bdee9da9550fb64c64cdf7bdf3102bda3f4693fd0fe1c270d805a39e816f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629874460.1476715/events.out.tfevents.1629874460.7e498afd5545.905.69
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f7d36d0c7d2cfa270e17d1103291cafe23e90e17e5fca22ebae077f8d20a4d8f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629875104.9909084/events.out.tfevents.1629875104.7e498afd5545.905.71
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c7483bed69bd1f30ee7f80da8fb30aa906d7d000597764347fbd2dae5f0d7529
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629872387.7e498afd5545.905.62
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bc690dba20d9dc7f9438804ccd3bf313426dfd646917e757957a7e1c86f46562
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629873040.7e498afd5545.905.64
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49ba294d0a143840803fe76ce3b2051519f8a957a43ea2fe6fde9a03154bc358
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629873815.7e498afd5545.905.66
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5098ab750bc4e7c2bde49ac1cd790815cded43f504e1ddb933d7da48d49af9f3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629874460.7e498afd5545.905.68
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2aee550599220ed9f542bac07d9127929e530ee706fa895d29696cd84afd3e4c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629875104.7e498afd5545.905.70
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9cfb270bf0972cef4eface5ce1c36791ac9766aff700c59d6b0455da12c2b290
|
| 3 |
+
size 8622
|