"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629995693.1376626/events.out.tfevents.1629995693.8e89bd551565.924.251 +3 -0
- model-bin/finetune/base/log/1629996128.4661825/events.out.tfevents.1629996129.8e89bd551565.924.253 +3 -0
- model-bin/finetune/base/log/1629996648.4800694/events.out.tfevents.1629996648.8e89bd551565.924.255 +3 -0
- model-bin/finetune/base/log/1629997104.0112085/events.out.tfevents.1629997104.8e89bd551565.924.257 +3 -0
- model-bin/finetune/base/log/1629997541.0580804/events.out.tfevents.1629997541.8e89bd551565.924.259 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629995693.8e89bd551565.924.250 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629996128.8e89bd551565.924.252 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629996648.8e89bd551565.924.254 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629997104.8e89bd551565.924.256 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629997541.8e89bd551565.924.258 +3 -0
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/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 sha256:18d1d0e7b18117085c5c6980f6665a7ddd2e685a7c21bedd75352320eae612e1
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/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 sha256:491fbf5bec3ceab2b3029b1588b8a53cc7c6dc6f1a3b0f743dd919485a4809fb
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:863a206f5996b23822226dcc80d7f21f039bee94d1f6943796990274de49b558
|
| 3 |
+
size 14375
|
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/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 sha256:52590b6b41ae98301e51f804dfdc3089d29db3b95cf917c1c4047b381e792adb
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/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:da5ef322dcad50f2827ba3193e02d6f59c8fd27dd174d3cf7d38cbad2827c51d
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -231579,11 +231579,800 @@
|
|
| 231579 |
"eval_steps_per_second": 0.686,
|
| 231580 |
"eval_wer": 0.18322231169021302,
|
| 231581 |
"step": 103918
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 231582 |
}
|
| 231583 |
],
|
| 231584 |
"max_steps": 620000,
|
| 231585 |
"num_train_epochs": 5000,
|
| 231586 |
-
"total_flos": 2.
|
| 231587 |
"trial_name": null,
|
| 231588 |
"trial_params": null
|
| 231589 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 842.995983935743,
|
| 5 |
+
"global_step": 104539,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 231579 |
"eval_steps_per_second": 0.686,
|
| 231580 |
"eval_wer": 0.18322231169021302,
|
| 231581 |
"step": 103918
|
| 231582 |
+
},
|
| 231583 |
+
{
|
| 231584 |
+
"epoch": 838.02,
|
| 231585 |
+
"learning_rate": 8.351298076923077e-06,
|
| 231586 |
+
"loss": 0.2745,
|
| 231587 |
+
"step": 103920
|
| 231588 |
+
},
|
| 231589 |
+
{
|
| 231590 |
+
"epoch": 838.06,
|
| 231591 |
+
"learning_rate": 8.351217948717949e-06,
|
| 231592 |
+
"loss": 0.2572,
|
| 231593 |
+
"step": 103925
|
| 231594 |
+
},
|
| 231595 |
+
{
|
| 231596 |
+
"epoch": 838.1,
|
| 231597 |
+
"learning_rate": 8.351137820512822e-06,
|
| 231598 |
+
"loss": 0.3528,
|
| 231599 |
+
"step": 103930
|
| 231600 |
+
},
|
| 231601 |
+
{
|
| 231602 |
+
"epoch": 838.14,
|
| 231603 |
+
"learning_rate": 8.351057692307693e-06,
|
| 231604 |
+
"loss": 0.3604,
|
| 231605 |
+
"step": 103935
|
| 231606 |
+
},
|
| 231607 |
+
{
|
| 231608 |
+
"epoch": 838.18,
|
| 231609 |
+
"learning_rate": 8.350977564102564e-06,
|
| 231610 |
+
"loss": 0.7321,
|
| 231611 |
+
"step": 103940
|
| 231612 |
+
},
|
| 231613 |
+
{
|
| 231614 |
+
"epoch": 838.22,
|
| 231615 |
+
"learning_rate": 8.350897435897437e-06,
|
| 231616 |
+
"loss": 0.9929,
|
| 231617 |
+
"step": 103945
|
| 231618 |
+
},
|
| 231619 |
+
{
|
| 231620 |
+
"epoch": 838.26,
|
| 231621 |
+
"learning_rate": 8.350817307692309e-06,
|
| 231622 |
+
"loss": 0.2765,
|
| 231623 |
+
"step": 103950
|
| 231624 |
+
},
|
| 231625 |
+
{
|
| 231626 |
+
"epoch": 838.3,
|
| 231627 |
+
"learning_rate": 8.35073717948718e-06,
|
| 231628 |
+
"loss": 0.2527,
|
| 231629 |
+
"step": 103955
|
| 231630 |
+
},
|
| 231631 |
+
{
|
| 231632 |
+
"epoch": 838.34,
|
| 231633 |
+
"learning_rate": 8.350657051282051e-06,
|
| 231634 |
+
"loss": 0.3403,
|
| 231635 |
+
"step": 103960
|
| 231636 |
+
},
|
| 231637 |
+
{
|
| 231638 |
+
"epoch": 838.38,
|
| 231639 |
+
"learning_rate": 8.350576923076925e-06,
|
| 231640 |
+
"loss": 0.6165,
|
| 231641 |
+
"step": 103965
|
| 231642 |
+
},
|
| 231643 |
+
{
|
| 231644 |
+
"epoch": 838.42,
|
| 231645 |
+
"learning_rate": 8.350496794871796e-06,
|
| 231646 |
+
"loss": 0.9087,
|
| 231647 |
+
"step": 103970
|
| 231648 |
+
},
|
| 231649 |
+
{
|
| 231650 |
+
"epoch": 838.46,
|
| 231651 |
+
"learning_rate": 8.350416666666667e-06,
|
| 231652 |
+
"loss": 0.3288,
|
| 231653 |
+
"step": 103975
|
| 231654 |
+
},
|
| 231655 |
+
{
|
| 231656 |
+
"epoch": 838.5,
|
| 231657 |
+
"learning_rate": 8.350336538461539e-06,
|
| 231658 |
+
"loss": 0.3491,
|
| 231659 |
+
"step": 103980
|
| 231660 |
+
},
|
| 231661 |
+
{
|
| 231662 |
+
"epoch": 838.54,
|
| 231663 |
+
"learning_rate": 8.350256410256412e-06,
|
| 231664 |
+
"loss": 0.3221,
|
| 231665 |
+
"step": 103985
|
| 231666 |
+
},
|
| 231667 |
+
{
|
| 231668 |
+
"epoch": 838.58,
|
| 231669 |
+
"learning_rate": 8.350176282051283e-06,
|
| 231670 |
+
"loss": 0.6123,
|
| 231671 |
+
"step": 103990
|
| 231672 |
+
},
|
| 231673 |
+
{
|
| 231674 |
+
"epoch": 838.62,
|
| 231675 |
+
"learning_rate": 8.350096153846154e-06,
|
| 231676 |
+
"loss": 0.9183,
|
| 231677 |
+
"step": 103995
|
| 231678 |
+
},
|
| 231679 |
+
{
|
| 231680 |
+
"epoch": 838.66,
|
| 231681 |
+
"learning_rate": 8.350016025641027e-06,
|
| 231682 |
+
"loss": 0.3169,
|
| 231683 |
+
"step": 104000
|
| 231684 |
+
},
|
| 231685 |
+
{
|
| 231686 |
+
"epoch": 838.7,
|
| 231687 |
+
"learning_rate": 8.349935897435897e-06,
|
| 231688 |
+
"loss": 0.3932,
|
| 231689 |
+
"step": 104005
|
| 231690 |
+
},
|
| 231691 |
+
{
|
| 231692 |
+
"epoch": 838.74,
|
| 231693 |
+
"learning_rate": 8.34985576923077e-06,
|
| 231694 |
+
"loss": 0.3419,
|
| 231695 |
+
"step": 104010
|
| 231696 |
+
},
|
| 231697 |
+
{
|
| 231698 |
+
"epoch": 838.78,
|
| 231699 |
+
"learning_rate": 8.349775641025642e-06,
|
| 231700 |
+
"loss": 0.638,
|
| 231701 |
+
"step": 104015
|
| 231702 |
+
},
|
| 231703 |
+
{
|
| 231704 |
+
"epoch": 838.82,
|
| 231705 |
+
"learning_rate": 8.349695512820513e-06,
|
| 231706 |
+
"loss": 1.0426,
|
| 231707 |
+
"step": 104020
|
| 231708 |
+
},
|
| 231709 |
+
{
|
| 231710 |
+
"epoch": 838.86,
|
| 231711 |
+
"learning_rate": 8.349615384615384e-06,
|
| 231712 |
+
"loss": 0.2742,
|
| 231713 |
+
"step": 104025
|
| 231714 |
+
},
|
| 231715 |
+
{
|
| 231716 |
+
"epoch": 838.9,
|
| 231717 |
+
"learning_rate": 8.349535256410257e-06,
|
| 231718 |
+
"loss": 0.3108,
|
| 231719 |
+
"step": 104030
|
| 231720 |
+
},
|
| 231721 |
+
{
|
| 231722 |
+
"epoch": 838.94,
|
| 231723 |
+
"learning_rate": 8.349455128205129e-06,
|
| 231724 |
+
"loss": 0.3617,
|
| 231725 |
+
"step": 104035
|
| 231726 |
+
},
|
| 231727 |
+
{
|
| 231728 |
+
"epoch": 838.98,
|
| 231729 |
+
"learning_rate": 8.349375e-06,
|
| 231730 |
+
"loss": 0.6442,
|
| 231731 |
+
"step": 104040
|
| 231732 |
+
},
|
| 231733 |
+
{
|
| 231734 |
+
"epoch": 839.0,
|
| 231735 |
+
"eval_loss": 0.35273829102516174,
|
| 231736 |
+
"eval_runtime": 41.2938,
|
| 231737 |
+
"eval_samples_per_second": 20.342,
|
| 231738 |
+
"eval_steps_per_second": 0.654,
|
| 231739 |
+
"eval_wer": 0.17864609617186936,
|
| 231740 |
+
"step": 104042
|
| 231741 |
+
},
|
| 231742 |
+
{
|
| 231743 |
+
"epoch": 839.02,
|
| 231744 |
+
"learning_rate": 8.349294871794873e-06,
|
| 231745 |
+
"loss": 0.3248,
|
| 231746 |
+
"step": 104045
|
| 231747 |
+
},
|
| 231748 |
+
{
|
| 231749 |
+
"epoch": 839.06,
|
| 231750 |
+
"learning_rate": 8.349214743589744e-06,
|
| 231751 |
+
"loss": 0.2771,
|
| 231752 |
+
"step": 104050
|
| 231753 |
+
},
|
| 231754 |
+
{
|
| 231755 |
+
"epoch": 839.1,
|
| 231756 |
+
"learning_rate": 8.349134615384616e-06,
|
| 231757 |
+
"loss": 0.2696,
|
| 231758 |
+
"step": 104055
|
| 231759 |
+
},
|
| 231760 |
+
{
|
| 231761 |
+
"epoch": 839.14,
|
| 231762 |
+
"learning_rate": 8.349054487179487e-06,
|
| 231763 |
+
"loss": 0.3619,
|
| 231764 |
+
"step": 104060
|
| 231765 |
+
},
|
| 231766 |
+
{
|
| 231767 |
+
"epoch": 839.18,
|
| 231768 |
+
"learning_rate": 8.34897435897436e-06,
|
| 231769 |
+
"loss": 0.7768,
|
| 231770 |
+
"step": 104065
|
| 231771 |
+
},
|
| 231772 |
+
{
|
| 231773 |
+
"epoch": 839.22,
|
| 231774 |
+
"learning_rate": 8.348894230769232e-06,
|
| 231775 |
+
"loss": 0.8096,
|
| 231776 |
+
"step": 104070
|
| 231777 |
+
},
|
| 231778 |
+
{
|
| 231779 |
+
"epoch": 839.27,
|
| 231780 |
+
"learning_rate": 8.348814102564103e-06,
|
| 231781 |
+
"loss": 0.2898,
|
| 231782 |
+
"step": 104075
|
| 231783 |
+
},
|
| 231784 |
+
{
|
| 231785 |
+
"epoch": 839.31,
|
| 231786 |
+
"learning_rate": 8.348733974358974e-06,
|
| 231787 |
+
"loss": 0.2886,
|
| 231788 |
+
"step": 104080
|
| 231789 |
+
},
|
| 231790 |
+
{
|
| 231791 |
+
"epoch": 839.35,
|
| 231792 |
+
"learning_rate": 8.348653846153847e-06,
|
| 231793 |
+
"loss": 0.3345,
|
| 231794 |
+
"step": 104085
|
| 231795 |
+
},
|
| 231796 |
+
{
|
| 231797 |
+
"epoch": 839.39,
|
| 231798 |
+
"learning_rate": 8.348573717948719e-06,
|
| 231799 |
+
"loss": 0.7645,
|
| 231800 |
+
"step": 104090
|
| 231801 |
+
},
|
| 231802 |
+
{
|
| 231803 |
+
"epoch": 839.43,
|
| 231804 |
+
"learning_rate": 8.34849358974359e-06,
|
| 231805 |
+
"loss": 0.9751,
|
| 231806 |
+
"step": 104095
|
| 231807 |
+
},
|
| 231808 |
+
{
|
| 231809 |
+
"epoch": 839.47,
|
| 231810 |
+
"learning_rate": 8.348413461538463e-06,
|
| 231811 |
+
"loss": 0.2766,
|
| 231812 |
+
"step": 104100
|
| 231813 |
+
},
|
| 231814 |
+
{
|
| 231815 |
+
"epoch": 839.51,
|
| 231816 |
+
"learning_rate": 8.348333333333334e-06,
|
| 231817 |
+
"loss": 0.3699,
|
| 231818 |
+
"step": 104105
|
| 231819 |
+
},
|
| 231820 |
+
{
|
| 231821 |
+
"epoch": 839.55,
|
| 231822 |
+
"learning_rate": 8.348253205128206e-06,
|
| 231823 |
+
"loss": 0.4221,
|
| 231824 |
+
"step": 104110
|
| 231825 |
+
},
|
| 231826 |
+
{
|
| 231827 |
+
"epoch": 839.59,
|
| 231828 |
+
"learning_rate": 8.348173076923077e-06,
|
| 231829 |
+
"loss": 0.7732,
|
| 231830 |
+
"step": 104115
|
| 231831 |
+
},
|
| 231832 |
+
{
|
| 231833 |
+
"epoch": 839.63,
|
| 231834 |
+
"learning_rate": 8.34809294871795e-06,
|
| 231835 |
+
"loss": 0.8807,
|
| 231836 |
+
"step": 104120
|
| 231837 |
+
},
|
| 231838 |
+
{
|
| 231839 |
+
"epoch": 839.67,
|
| 231840 |
+
"learning_rate": 8.348012820512822e-06,
|
| 231841 |
+
"loss": 0.2793,
|
| 231842 |
+
"step": 104125
|
| 231843 |
+
},
|
| 231844 |
+
{
|
| 231845 |
+
"epoch": 839.71,
|
| 231846 |
+
"learning_rate": 8.347932692307693e-06,
|
| 231847 |
+
"loss": 0.2918,
|
| 231848 |
+
"step": 104130
|
| 231849 |
+
},
|
| 231850 |
+
{
|
| 231851 |
+
"epoch": 839.75,
|
| 231852 |
+
"learning_rate": 8.347852564102564e-06,
|
| 231853 |
+
"loss": 0.3833,
|
| 231854 |
+
"step": 104135
|
| 231855 |
+
},
|
| 231856 |
+
{
|
| 231857 |
+
"epoch": 839.79,
|
| 231858 |
+
"learning_rate": 8.347772435897437e-06,
|
| 231859 |
+
"loss": 0.7802,
|
| 231860 |
+
"step": 104140
|
| 231861 |
+
},
|
| 231862 |
+
{
|
| 231863 |
+
"epoch": 839.83,
|
| 231864 |
+
"learning_rate": 8.347692307692309e-06,
|
| 231865 |
+
"loss": 0.8053,
|
| 231866 |
+
"step": 104145
|
| 231867 |
+
},
|
| 231868 |
+
{
|
| 231869 |
+
"epoch": 839.87,
|
| 231870 |
+
"learning_rate": 8.34761217948718e-06,
|
| 231871 |
+
"loss": 0.362,
|
| 231872 |
+
"step": 104150
|
| 231873 |
+
},
|
| 231874 |
+
{
|
| 231875 |
+
"epoch": 839.91,
|
| 231876 |
+
"learning_rate": 8.347532051282053e-06,
|
| 231877 |
+
"loss": 0.3359,
|
| 231878 |
+
"step": 104155
|
| 231879 |
+
},
|
| 231880 |
+
{
|
| 231881 |
+
"epoch": 839.95,
|
| 231882 |
+
"learning_rate": 8.347451923076923e-06,
|
| 231883 |
+
"loss": 0.3976,
|
| 231884 |
+
"step": 104160
|
| 231885 |
+
},
|
| 231886 |
+
{
|
| 231887 |
+
"epoch": 839.99,
|
| 231888 |
+
"learning_rate": 8.347371794871796e-06,
|
| 231889 |
+
"loss": 0.8711,
|
| 231890 |
+
"step": 104165
|
| 231891 |
+
},
|
| 231892 |
+
{
|
| 231893 |
+
"epoch": 840.0,
|
| 231894 |
+
"eval_loss": 0.42207542061805725,
|
| 231895 |
+
"eval_runtime": 40.7022,
|
| 231896 |
+
"eval_samples_per_second": 20.638,
|
| 231897 |
+
"eval_steps_per_second": 0.663,
|
| 231898 |
+
"eval_wer": 0.18555353901996371,
|
| 231899 |
+
"step": 104166
|
| 231900 |
+
},
|
| 231901 |
+
{
|
| 231902 |
+
"epoch": 833.03,
|
| 231903 |
+
"learning_rate": 8.347291666666667e-06,
|
| 231904 |
+
"loss": 0.347,
|
| 231905 |
+
"step": 104170
|
| 231906 |
+
},
|
| 231907 |
+
{
|
| 231908 |
+
"epoch": 833.07,
|
| 231909 |
+
"learning_rate": 8.347211538461539e-06,
|
| 231910 |
+
"loss": 0.3221,
|
| 231911 |
+
"step": 104175
|
| 231912 |
+
},
|
| 231913 |
+
{
|
| 231914 |
+
"epoch": 833.11,
|
| 231915 |
+
"learning_rate": 8.34713141025641e-06,
|
| 231916 |
+
"loss": 0.2602,
|
| 231917 |
+
"step": 104180
|
| 231918 |
+
},
|
| 231919 |
+
{
|
| 231920 |
+
"epoch": 833.15,
|
| 231921 |
+
"learning_rate": 8.347051282051283e-06,
|
| 231922 |
+
"loss": 0.3825,
|
| 231923 |
+
"step": 104185
|
| 231924 |
+
},
|
| 231925 |
+
{
|
| 231926 |
+
"epoch": 833.19,
|
| 231927 |
+
"learning_rate": 8.346971153846154e-06,
|
| 231928 |
+
"loss": 0.8882,
|
| 231929 |
+
"step": 104190
|
| 231930 |
+
},
|
| 231931 |
+
{
|
| 231932 |
+
"epoch": 833.23,
|
| 231933 |
+
"learning_rate": 8.346891025641026e-06,
|
| 231934 |
+
"loss": 0.7291,
|
| 231935 |
+
"step": 104195
|
| 231936 |
+
},
|
| 231937 |
+
{
|
| 231938 |
+
"epoch": 833.27,
|
| 231939 |
+
"learning_rate": 8.346810897435899e-06,
|
| 231940 |
+
"loss": 0.312,
|
| 231941 |
+
"step": 104200
|
| 231942 |
+
},
|
| 231943 |
+
{
|
| 231944 |
+
"epoch": 833.31,
|
| 231945 |
+
"learning_rate": 8.34673076923077e-06,
|
| 231946 |
+
"loss": 0.2868,
|
| 231947 |
+
"step": 104205
|
| 231948 |
+
},
|
| 231949 |
+
{
|
| 231950 |
+
"epoch": 833.35,
|
| 231951 |
+
"learning_rate": 8.346650641025641e-06,
|
| 231952 |
+
"loss": 0.4478,
|
| 231953 |
+
"step": 104210
|
| 231954 |
+
},
|
| 231955 |
+
{
|
| 231956 |
+
"epoch": 833.39,
|
| 231957 |
+
"learning_rate": 8.346570512820513e-06,
|
| 231958 |
+
"loss": 0.8855,
|
| 231959 |
+
"step": 104215
|
| 231960 |
+
},
|
| 231961 |
+
{
|
| 231962 |
+
"epoch": 833.43,
|
| 231963 |
+
"learning_rate": 8.346490384615386e-06,
|
| 231964 |
+
"loss": 0.7151,
|
| 231965 |
+
"step": 104220
|
| 231966 |
+
},
|
| 231967 |
+
{
|
| 231968 |
+
"epoch": 833.47,
|
| 231969 |
+
"learning_rate": 8.346410256410257e-06,
|
| 231970 |
+
"loss": 0.3058,
|
| 231971 |
+
"step": 104225
|
| 231972 |
+
},
|
| 231973 |
+
{
|
| 231974 |
+
"epoch": 833.51,
|
| 231975 |
+
"learning_rate": 8.346330128205129e-06,
|
| 231976 |
+
"loss": 0.3211,
|
| 231977 |
+
"step": 104230
|
| 231978 |
+
},
|
| 231979 |
+
{
|
| 231980 |
+
"epoch": 833.55,
|
| 231981 |
+
"learning_rate": 8.34625e-06,
|
| 231982 |
+
"loss": 0.4314,
|
| 231983 |
+
"step": 104235
|
| 231984 |
+
},
|
| 231985 |
+
{
|
| 231986 |
+
"epoch": 833.59,
|
| 231987 |
+
"learning_rate": 8.346169871794873e-06,
|
| 231988 |
+
"loss": 0.9153,
|
| 231989 |
+
"step": 104240
|
| 231990 |
+
},
|
| 231991 |
+
{
|
| 231992 |
+
"epoch": 833.63,
|
| 231993 |
+
"learning_rate": 8.346089743589744e-06,
|
| 231994 |
+
"loss": 0.6461,
|
| 231995 |
+
"step": 104245
|
| 231996 |
+
},
|
| 231997 |
+
{
|
| 231998 |
+
"epoch": 833.67,
|
| 231999 |
+
"learning_rate": 8.346009615384616e-06,
|
| 232000 |
+
"loss": 0.3166,
|
| 232001 |
+
"step": 104250
|
| 232002 |
+
},
|
| 232003 |
+
{
|
| 232004 |
+
"epoch": 833.71,
|
| 232005 |
+
"learning_rate": 8.345929487179489e-06,
|
| 232006 |
+
"loss": 0.3198,
|
| 232007 |
+
"step": 104255
|
| 232008 |
+
},
|
| 232009 |
+
{
|
| 232010 |
+
"epoch": 833.75,
|
| 232011 |
+
"learning_rate": 8.34584935897436e-06,
|
| 232012 |
+
"loss": 0.4173,
|
| 232013 |
+
"step": 104260
|
| 232014 |
+
},
|
| 232015 |
+
{
|
| 232016 |
+
"epoch": 833.79,
|
| 232017 |
+
"learning_rate": 8.345769230769231e-06,
|
| 232018 |
+
"loss": 0.9187,
|
| 232019 |
+
"step": 104265
|
| 232020 |
+
},
|
| 232021 |
+
{
|
| 232022 |
+
"epoch": 833.83,
|
| 232023 |
+
"learning_rate": 8.345689102564103e-06,
|
| 232024 |
+
"loss": 0.6422,
|
| 232025 |
+
"step": 104270
|
| 232026 |
+
},
|
| 232027 |
+
{
|
| 232028 |
+
"epoch": 833.87,
|
| 232029 |
+
"learning_rate": 8.345608974358976e-06,
|
| 232030 |
+
"loss": 0.2888,
|
| 232031 |
+
"step": 104275
|
| 232032 |
+
},
|
| 232033 |
+
{
|
| 232034 |
+
"epoch": 833.91,
|
| 232035 |
+
"learning_rate": 8.345528846153847e-06,
|
| 232036 |
+
"loss": 0.3323,
|
| 232037 |
+
"step": 104280
|
| 232038 |
+
},
|
| 232039 |
+
{
|
| 232040 |
+
"epoch": 833.95,
|
| 232041 |
+
"learning_rate": 8.345448717948719e-06,
|
| 232042 |
+
"loss": 0.4291,
|
| 232043 |
+
"step": 104285
|
| 232044 |
+
},
|
| 232045 |
+
{
|
| 232046 |
+
"epoch": 833.99,
|
| 232047 |
+
"learning_rate": 8.34536858974359e-06,
|
| 232048 |
+
"loss": 1.0071,
|
| 232049 |
+
"step": 104290
|
| 232050 |
+
},
|
| 232051 |
+
{
|
| 232052 |
+
"epoch": 834.0,
|
| 232053 |
+
"eval_loss": 0.3745848536491394,
|
| 232054 |
+
"eval_runtime": 40.2099,
|
| 232055 |
+
"eval_samples_per_second": 20.841,
|
| 232056 |
+
"eval_steps_per_second": 0.671,
|
| 232057 |
+
"eval_wer": 0.18355542539171182,
|
| 232058 |
+
"step": 104291
|
| 232059 |
+
},
|
| 232060 |
+
{
|
| 232061 |
+
"epoch": 841.03,
|
| 232062 |
+
"learning_rate": 8.345288461538463e-06,
|
| 232063 |
+
"loss": 0.3266,
|
| 232064 |
+
"step": 104295
|
| 232065 |
+
},
|
| 232066 |
+
{
|
| 232067 |
+
"epoch": 841.07,
|
| 232068 |
+
"learning_rate": 8.345208333333334e-06,
|
| 232069 |
+
"loss": 0.3229,
|
| 232070 |
+
"step": 104300
|
| 232071 |
+
},
|
| 232072 |
+
{
|
| 232073 |
+
"epoch": 841.11,
|
| 232074 |
+
"learning_rate": 8.345128205128206e-06,
|
| 232075 |
+
"loss": 0.3198,
|
| 232076 |
+
"step": 104305
|
| 232077 |
+
},
|
| 232078 |
+
{
|
| 232079 |
+
"epoch": 841.15,
|
| 232080 |
+
"learning_rate": 8.345048076923079e-06,
|
| 232081 |
+
"loss": 0.3572,
|
| 232082 |
+
"step": 104310
|
| 232083 |
+
},
|
| 232084 |
+
{
|
| 232085 |
+
"epoch": 841.19,
|
| 232086 |
+
"learning_rate": 8.344967948717948e-06,
|
| 232087 |
+
"loss": 0.8764,
|
| 232088 |
+
"step": 104315
|
| 232089 |
+
},
|
| 232090 |
+
{
|
| 232091 |
+
"epoch": 841.23,
|
| 232092 |
+
"learning_rate": 8.344887820512821e-06,
|
| 232093 |
+
"loss": 0.6498,
|
| 232094 |
+
"step": 104320
|
| 232095 |
+
},
|
| 232096 |
+
{
|
| 232097 |
+
"epoch": 841.27,
|
| 232098 |
+
"learning_rate": 8.344807692307693e-06,
|
| 232099 |
+
"loss": 0.2689,
|
| 232100 |
+
"step": 104325
|
| 232101 |
+
},
|
| 232102 |
+
{
|
| 232103 |
+
"epoch": 841.31,
|
| 232104 |
+
"learning_rate": 8.344727564102564e-06,
|
| 232105 |
+
"loss": 0.2865,
|
| 232106 |
+
"step": 104330
|
| 232107 |
+
},
|
| 232108 |
+
{
|
| 232109 |
+
"epoch": 841.35,
|
| 232110 |
+
"learning_rate": 8.344647435897436e-06,
|
| 232111 |
+
"loss": 0.3679,
|
| 232112 |
+
"step": 104335
|
| 232113 |
+
},
|
| 232114 |
+
{
|
| 232115 |
+
"epoch": 841.39,
|
| 232116 |
+
"learning_rate": 8.344567307692309e-06,
|
| 232117 |
+
"loss": 0.9102,
|
| 232118 |
+
"step": 104340
|
| 232119 |
+
},
|
| 232120 |
+
{
|
| 232121 |
+
"epoch": 841.43,
|
| 232122 |
+
"learning_rate": 8.34448717948718e-06,
|
| 232123 |
+
"loss": 0.6518,
|
| 232124 |
+
"step": 104345
|
| 232125 |
+
},
|
| 232126 |
+
{
|
| 232127 |
+
"epoch": 841.47,
|
| 232128 |
+
"learning_rate": 8.344407051282051e-06,
|
| 232129 |
+
"loss": 0.3188,
|
| 232130 |
+
"step": 104350
|
| 232131 |
+
},
|
| 232132 |
+
{
|
| 232133 |
+
"epoch": 841.51,
|
| 232134 |
+
"learning_rate": 8.344326923076924e-06,
|
| 232135 |
+
"loss": 0.2554,
|
| 232136 |
+
"step": 104355
|
| 232137 |
+
},
|
| 232138 |
+
{
|
| 232139 |
+
"epoch": 841.55,
|
| 232140 |
+
"learning_rate": 8.344246794871796e-06,
|
| 232141 |
+
"loss": 0.3999,
|
| 232142 |
+
"step": 104360
|
| 232143 |
+
},
|
| 232144 |
+
{
|
| 232145 |
+
"epoch": 841.59,
|
| 232146 |
+
"learning_rate": 8.344166666666667e-06,
|
| 232147 |
+
"loss": 0.8557,
|
| 232148 |
+
"step": 104365
|
| 232149 |
+
},
|
| 232150 |
+
{
|
| 232151 |
+
"epoch": 841.63,
|
| 232152 |
+
"learning_rate": 8.344086538461538e-06,
|
| 232153 |
+
"loss": 0.6551,
|
| 232154 |
+
"step": 104370
|
| 232155 |
+
},
|
| 232156 |
+
{
|
| 232157 |
+
"epoch": 841.67,
|
| 232158 |
+
"learning_rate": 8.344006410256412e-06,
|
| 232159 |
+
"loss": 0.2773,
|
| 232160 |
+
"step": 104375
|
| 232161 |
+
},
|
| 232162 |
+
{
|
| 232163 |
+
"epoch": 841.71,
|
| 232164 |
+
"learning_rate": 8.343926282051283e-06,
|
| 232165 |
+
"loss": 0.258,
|
| 232166 |
+
"step": 104380
|
| 232167 |
+
},
|
| 232168 |
+
{
|
| 232169 |
+
"epoch": 841.76,
|
| 232170 |
+
"learning_rate": 8.343846153846154e-06,
|
| 232171 |
+
"loss": 0.3869,
|
| 232172 |
+
"step": 104385
|
| 232173 |
+
},
|
| 232174 |
+
{
|
| 232175 |
+
"epoch": 841.8,
|
| 232176 |
+
"learning_rate": 8.343766025641026e-06,
|
| 232177 |
+
"loss": 0.967,
|
| 232178 |
+
"step": 104390
|
| 232179 |
+
},
|
| 232180 |
+
{
|
| 232181 |
+
"epoch": 841.84,
|
| 232182 |
+
"learning_rate": 8.343685897435899e-06,
|
| 232183 |
+
"loss": 0.5489,
|
| 232184 |
+
"step": 104395
|
| 232185 |
+
},
|
| 232186 |
+
{
|
| 232187 |
+
"epoch": 841.88,
|
| 232188 |
+
"learning_rate": 8.34360576923077e-06,
|
| 232189 |
+
"loss": 0.2847,
|
| 232190 |
+
"step": 104400
|
| 232191 |
+
},
|
| 232192 |
+
{
|
| 232193 |
+
"epoch": 841.92,
|
| 232194 |
+
"learning_rate": 8.343525641025641e-06,
|
| 232195 |
+
"loss": 0.3259,
|
| 232196 |
+
"step": 104405
|
| 232197 |
+
},
|
| 232198 |
+
{
|
| 232199 |
+
"epoch": 841.96,
|
| 232200 |
+
"learning_rate": 8.343445512820514e-06,
|
| 232201 |
+
"loss": 0.4115,
|
| 232202 |
+
"step": 104410
|
| 232203 |
+
},
|
| 232204 |
+
{
|
| 232205 |
+
"epoch": 842.0,
|
| 232206 |
+
"learning_rate": 8.343365384615386e-06,
|
| 232207 |
+
"loss": 1.1216,
|
| 232208 |
+
"step": 104415
|
| 232209 |
+
},
|
| 232210 |
+
{
|
| 232211 |
+
"epoch": 842.0,
|
| 232212 |
+
"eval_loss": 0.3493131697177887,
|
| 232213 |
+
"eval_runtime": 38.6328,
|
| 232214 |
+
"eval_samples_per_second": 21.666,
|
| 232215 |
+
"eval_steps_per_second": 0.699,
|
| 232216 |
+
"eval_wer": 0.1871700296521299,
|
| 232217 |
+
"step": 104415
|
| 232218 |
+
},
|
| 232219 |
+
{
|
| 232220 |
+
"epoch": 842.04,
|
| 232221 |
+
"learning_rate": 8.343285256410257e-06,
|
| 232222 |
+
"loss": 0.3497,
|
| 232223 |
+
"step": 104420
|
| 232224 |
+
},
|
| 232225 |
+
{
|
| 232226 |
+
"epoch": 842.08,
|
| 232227 |
+
"learning_rate": 8.343205128205128e-06,
|
| 232228 |
+
"loss": 0.318,
|
| 232229 |
+
"step": 104425
|
| 232230 |
+
},
|
| 232231 |
+
{
|
| 232232 |
+
"epoch": 842.12,
|
| 232233 |
+
"learning_rate": 8.343125000000002e-06,
|
| 232234 |
+
"loss": 0.3279,
|
| 232235 |
+
"step": 104430
|
| 232236 |
+
},
|
| 232237 |
+
{
|
| 232238 |
+
"epoch": 842.16,
|
| 232239 |
+
"learning_rate": 8.343044871794871e-06,
|
| 232240 |
+
"loss": 0.4165,
|
| 232241 |
+
"step": 104435
|
| 232242 |
+
},
|
| 232243 |
+
{
|
| 232244 |
+
"epoch": 842.2,
|
| 232245 |
+
"learning_rate": 8.342964743589744e-06,
|
| 232246 |
+
"loss": 1.2737,
|
| 232247 |
+
"step": 104440
|
| 232248 |
+
},
|
| 232249 |
+
{
|
| 232250 |
+
"epoch": 842.24,
|
| 232251 |
+
"learning_rate": 8.342884615384617e-06,
|
| 232252 |
+
"loss": 0.3312,
|
| 232253 |
+
"step": 104445
|
| 232254 |
+
},
|
| 232255 |
+
{
|
| 232256 |
+
"epoch": 842.28,
|
| 232257 |
+
"learning_rate": 8.342804487179487e-06,
|
| 232258 |
+
"loss": 0.2809,
|
| 232259 |
+
"step": 104450
|
| 232260 |
+
},
|
| 232261 |
+
{
|
| 232262 |
+
"epoch": 842.32,
|
| 232263 |
+
"learning_rate": 8.34272435897436e-06,
|
| 232264 |
+
"loss": 0.2927,
|
| 232265 |
+
"step": 104455
|
| 232266 |
+
},
|
| 232267 |
+
{
|
| 232268 |
+
"epoch": 842.36,
|
| 232269 |
+
"learning_rate": 8.342644230769231e-06,
|
| 232270 |
+
"loss": 0.4408,
|
| 232271 |
+
"step": 104460
|
| 232272 |
+
},
|
| 232273 |
+
{
|
| 232274 |
+
"epoch": 842.4,
|
| 232275 |
+
"learning_rate": 8.342564102564103e-06,
|
| 232276 |
+
"loss": 1.1952,
|
| 232277 |
+
"step": 104465
|
| 232278 |
+
},
|
| 232279 |
+
{
|
| 232280 |
+
"epoch": 842.44,
|
| 232281 |
+
"learning_rate": 8.342483974358974e-06,
|
| 232282 |
+
"loss": 0.5169,
|
| 232283 |
+
"step": 104470
|
| 232284 |
+
},
|
| 232285 |
+
{
|
| 232286 |
+
"epoch": 842.48,
|
| 232287 |
+
"learning_rate": 8.342403846153847e-06,
|
| 232288 |
+
"loss": 0.2839,
|
| 232289 |
+
"step": 104475
|
| 232290 |
+
},
|
| 232291 |
+
{
|
| 232292 |
+
"epoch": 842.52,
|
| 232293 |
+
"learning_rate": 8.342323717948719e-06,
|
| 232294 |
+
"loss": 0.3011,
|
| 232295 |
+
"step": 104480
|
| 232296 |
+
},
|
| 232297 |
+
{
|
| 232298 |
+
"epoch": 842.56,
|
| 232299 |
+
"learning_rate": 8.34224358974359e-06,
|
| 232300 |
+
"loss": 0.4053,
|
| 232301 |
+
"step": 104485
|
| 232302 |
+
},
|
| 232303 |
+
{
|
| 232304 |
+
"epoch": 842.6,
|
| 232305 |
+
"learning_rate": 8.342163461538461e-06,
|
| 232306 |
+
"loss": 1.216,
|
| 232307 |
+
"step": 104490
|
| 232308 |
+
},
|
| 232309 |
+
{
|
| 232310 |
+
"epoch": 842.64,
|
| 232311 |
+
"learning_rate": 8.342083333333334e-06,
|
| 232312 |
+
"loss": 0.3982,
|
| 232313 |
+
"step": 104495
|
| 232314 |
+
},
|
| 232315 |
+
{
|
| 232316 |
+
"epoch": 842.68,
|
| 232317 |
+
"learning_rate": 8.342003205128206e-06,
|
| 232318 |
+
"loss": 0.3052,
|
| 232319 |
+
"step": 104500
|
| 232320 |
+
},
|
| 232321 |
+
{
|
| 232322 |
+
"epoch": 842.72,
|
| 232323 |
+
"learning_rate": 8.341923076923077e-06,
|
| 232324 |
+
"loss": 0.3062,
|
| 232325 |
+
"step": 104505
|
| 232326 |
+
},
|
| 232327 |
+
{
|
| 232328 |
+
"epoch": 842.76,
|
| 232329 |
+
"learning_rate": 8.34184294871795e-06,
|
| 232330 |
+
"loss": 0.4384,
|
| 232331 |
+
"step": 104510
|
| 232332 |
+
},
|
| 232333 |
+
{
|
| 232334 |
+
"epoch": 842.8,
|
| 232335 |
+
"learning_rate": 8.341762820512821e-06,
|
| 232336 |
+
"loss": 1.0651,
|
| 232337 |
+
"step": 104515
|
| 232338 |
+
},
|
| 232339 |
+
{
|
| 232340 |
+
"epoch": 842.84,
|
| 232341 |
+
"learning_rate": 8.341682692307693e-06,
|
| 232342 |
+
"loss": 0.3653,
|
| 232343 |
+
"step": 104520
|
| 232344 |
+
},
|
| 232345 |
+
{
|
| 232346 |
+
"epoch": 842.88,
|
| 232347 |
+
"learning_rate": 8.341602564102564e-06,
|
| 232348 |
+
"loss": 0.2773,
|
| 232349 |
+
"step": 104525
|
| 232350 |
+
},
|
| 232351 |
+
{
|
| 232352 |
+
"epoch": 842.92,
|
| 232353 |
+
"learning_rate": 8.341522435897437e-06,
|
| 232354 |
+
"loss": 0.3293,
|
| 232355 |
+
"step": 104530
|
| 232356 |
+
},
|
| 232357 |
+
{
|
| 232358 |
+
"epoch": 842.96,
|
| 232359 |
+
"learning_rate": 8.341442307692309e-06,
|
| 232360 |
+
"loss": 0.5005,
|
| 232361 |
+
"step": 104535
|
| 232362 |
+
},
|
| 232363 |
+
{
|
| 232364 |
+
"epoch": 843.0,
|
| 232365 |
+
"eval_loss": 0.38071903586387634,
|
| 232366 |
+
"eval_runtime": 39.7203,
|
| 232367 |
+
"eval_samples_per_second": 21.098,
|
| 232368 |
+
"eval_steps_per_second": 0.68,
|
| 232369 |
+
"eval_wer": 0.18273625239569513,
|
| 232370 |
+
"step": 104539
|
| 232371 |
}
|
| 232372 |
],
|
| 232373 |
"max_steps": 620000,
|
| 232374 |
"num_train_epochs": 5000,
|
| 232375 |
+
"total_flos": 2.9420435882603774e+20,
|
| 232376 |
"trial_name": null,
|
| 232377 |
"trial_params": null
|
| 232378 |
}
|
model-bin/finetune/base/{checkpoint-103918 → checkpoint-104539}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629995693.1376626/events.out.tfevents.1629995693.8e89bd551565.924.251
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0e4bc3acb124c42d484dbabf88d21e231d725f982e77f48bdbd37d1203e02155
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629996128.4661825/events.out.tfevents.1629996129.8e89bd551565.924.253
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1a613f4d8efe43ba38dcc6ecede800e02f85b5daaab321050c023bcbb2b0a55f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629996648.4800694/events.out.tfevents.1629996648.8e89bd551565.924.255
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:32fbff97f311ec84719302a8a9eae8ccb82d2e34a2816795fefbc1ac2b6bc286
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629997104.0112085/events.out.tfevents.1629997104.8e89bd551565.924.257
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:36b3da19efc5e1df9126ba79f7c6bd5d788eededd35a568c7b72e39438dc5994
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629997541.0580804/events.out.tfevents.1629997541.8e89bd551565.924.259
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7e7ae8d8e64747f877d97e0ecbbffab661c74155f45d449c94f550f913c79e70
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629995693.8e89bd551565.924.250
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0242544dd702433852891d50f4cc660aa6a8fac0f3ac1abb63c5ddf84bcc6a91
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629996128.8e89bd551565.924.252
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f0968a8a9eb32f58a83fd93bea19f510bdaee87a50fd7dcb19e74c4259f13f44
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629996648.8e89bd551565.924.254
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8cbf4ade00deacc89491ea814480b14fe1704e8919bda47a88fedcd3651890da
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629997104.8e89bd551565.924.256
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3fe8cc646a2439c9d920c71bf82c83b02b1c5b40ad2fbe0525af7ee250246b7d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629997541.8e89bd551565.924.258
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:94dfcbb96d61aec4c4ad0d0460a30fe18b7f938b5f40c7320d730ebb5ca9ef84
|
| 3 |
+
size 8462
|