"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629858970.8441613/events.out.tfevents.1629858970.7e498afd5545.905.23 +3 -0
- model-bin/finetune/base/log/1629859624.9463222/events.out.tfevents.1629859624.7e498afd5545.905.25 +3 -0
- model-bin/finetune/base/log/1629860278.0745044/events.out.tfevents.1629860278.7e498afd5545.905.27 +3 -0
- model-bin/finetune/base/log/1629860932.5132222/events.out.tfevents.1629860932.7e498afd5545.905.29 +3 -0
- model-bin/finetune/base/log/1629861583.8800406/events.out.tfevents.1629861583.7e498afd5545.905.31 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629858970.7e498afd5545.905.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629859624.7e498afd5545.905.24 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629860278.7e498afd5545.905.26 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629860932.7e498afd5545.905.28 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629861583.7e498afd5545.905.30 +3 -0
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/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 sha256:1c2cf74949ac6fbe57789fc93a1ba947fc6fea6315730c11480f522422c2cb32
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/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:b94fe687400b28878fb9b1535714b9e18e88910b1486e085968c125fe165b608
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/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:a44b4a00afd33832285479c87b3d0330ccd6122276d09600764d244ca2e0d676
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/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:fa5ffd5ce4a042224436871219db3b89c25b62d29b4a4605aafd1e3b9552f733
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/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 sha256:34abd4dd1bca145f62df90529f947d27c07749481bac26350eaca06e62acd0fe
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/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,
|
|
@@ -189453,11 +189453,800 @@
|
|
| 189453 |
"eval_steps_per_second": 0.673,
|
| 189454 |
"eval_wer": 0.18833597008485545,
|
| 189455 |
"step": 70809
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 189456 |
}
|
| 189457 |
],
|
| 189458 |
"max_steps": 620000,
|
| 189459 |
"num_train_epochs": 5000,
|
| 189460 |
-
"total_flos":
|
| 189461 |
"trial_name": null,
|
| 189462 |
"trial_params": null
|
| 189463 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 575.995983935743,
|
| 5 |
+
"global_step": 71429,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 189453 |
"eval_steps_per_second": 0.673,
|
| 189454 |
"eval_wer": 0.18833597008485545,
|
| 189455 |
"step": 70809
|
| 189456 |
+
},
|
| 189457 |
+
{
|
| 189458 |
+
"epoch": 571.01,
|
| 189459 |
+
"learning_rate": 8.881666666666667e-06,
|
| 189460 |
+
"loss": 0.3513,
|
| 189461 |
+
"step": 70810
|
| 189462 |
+
},
|
| 189463 |
+
{
|
| 189464 |
+
"epoch": 571.05,
|
| 189465 |
+
"learning_rate": 8.88158653846154e-06,
|
| 189466 |
+
"loss": 0.2739,
|
| 189467 |
+
"step": 70815
|
| 189468 |
+
},
|
| 189469 |
+
{
|
| 189470 |
+
"epoch": 571.09,
|
| 189471 |
+
"learning_rate": 8.881506410256411e-06,
|
| 189472 |
+
"loss": 0.3412,
|
| 189473 |
+
"step": 70820
|
| 189474 |
+
},
|
| 189475 |
+
{
|
| 189476 |
+
"epoch": 571.13,
|
| 189477 |
+
"learning_rate": 8.881426282051282e-06,
|
| 189478 |
+
"loss": 0.3873,
|
| 189479 |
+
"step": 70825
|
| 189480 |
+
},
|
| 189481 |
+
{
|
| 189482 |
+
"epoch": 571.17,
|
| 189483 |
+
"learning_rate": 8.881346153846154e-06,
|
| 189484 |
+
"loss": 0.5661,
|
| 189485 |
+
"step": 70830
|
| 189486 |
+
},
|
| 189487 |
+
{
|
| 189488 |
+
"epoch": 571.21,
|
| 189489 |
+
"learning_rate": 8.881266025641027e-06,
|
| 189490 |
+
"loss": 1.1271,
|
| 189491 |
+
"step": 70835
|
| 189492 |
+
},
|
| 189493 |
+
{
|
| 189494 |
+
"epoch": 571.25,
|
| 189495 |
+
"learning_rate": 8.881185897435898e-06,
|
| 189496 |
+
"loss": 0.3545,
|
| 189497 |
+
"step": 70840
|
| 189498 |
+
},
|
| 189499 |
+
{
|
| 189500 |
+
"epoch": 571.29,
|
| 189501 |
+
"learning_rate": 8.88110576923077e-06,
|
| 189502 |
+
"loss": 0.2882,
|
| 189503 |
+
"step": 70845
|
| 189504 |
+
},
|
| 189505 |
+
{
|
| 189506 |
+
"epoch": 571.33,
|
| 189507 |
+
"learning_rate": 8.881025641025641e-06,
|
| 189508 |
+
"loss": 0.352,
|
| 189509 |
+
"step": 70850
|
| 189510 |
+
},
|
| 189511 |
+
{
|
| 189512 |
+
"epoch": 571.37,
|
| 189513 |
+
"learning_rate": 8.880945512820514e-06,
|
| 189514 |
+
"loss": 0.5543,
|
| 189515 |
+
"step": 70855
|
| 189516 |
+
},
|
| 189517 |
+
{
|
| 189518 |
+
"epoch": 571.41,
|
| 189519 |
+
"learning_rate": 8.880865384615385e-06,
|
| 189520 |
+
"loss": 1.203,
|
| 189521 |
+
"step": 70860
|
| 189522 |
+
},
|
| 189523 |
+
{
|
| 189524 |
+
"epoch": 571.45,
|
| 189525 |
+
"learning_rate": 8.880785256410257e-06,
|
| 189526 |
+
"loss": 0.3559,
|
| 189527 |
+
"step": 70865
|
| 189528 |
+
},
|
| 189529 |
+
{
|
| 189530 |
+
"epoch": 571.49,
|
| 189531 |
+
"learning_rate": 8.88070512820513e-06,
|
| 189532 |
+
"loss": 0.3581,
|
| 189533 |
+
"step": 70870
|
| 189534 |
+
},
|
| 189535 |
+
{
|
| 189536 |
+
"epoch": 571.53,
|
| 189537 |
+
"learning_rate": 8.880625000000001e-06,
|
| 189538 |
+
"loss": 0.3475,
|
| 189539 |
+
"step": 70875
|
| 189540 |
+
},
|
| 189541 |
+
{
|
| 189542 |
+
"epoch": 571.57,
|
| 189543 |
+
"learning_rate": 8.880544871794872e-06,
|
| 189544 |
+
"loss": 0.6108,
|
| 189545 |
+
"step": 70880
|
| 189546 |
+
},
|
| 189547 |
+
{
|
| 189548 |
+
"epoch": 571.61,
|
| 189549 |
+
"learning_rate": 8.880464743589744e-06,
|
| 189550 |
+
"loss": 1.0645,
|
| 189551 |
+
"step": 70885
|
| 189552 |
+
},
|
| 189553 |
+
{
|
| 189554 |
+
"epoch": 571.65,
|
| 189555 |
+
"learning_rate": 8.880384615384617e-06,
|
| 189556 |
+
"loss": 0.4175,
|
| 189557 |
+
"step": 70890
|
| 189558 |
+
},
|
| 189559 |
+
{
|
| 189560 |
+
"epoch": 571.69,
|
| 189561 |
+
"learning_rate": 8.880304487179488e-06,
|
| 189562 |
+
"loss": 0.3068,
|
| 189563 |
+
"step": 70895
|
| 189564 |
+
},
|
| 189565 |
+
{
|
| 189566 |
+
"epoch": 571.73,
|
| 189567 |
+
"learning_rate": 8.88022435897436e-06,
|
| 189568 |
+
"loss": 0.3434,
|
| 189569 |
+
"step": 70900
|
| 189570 |
+
},
|
| 189571 |
+
{
|
| 189572 |
+
"epoch": 571.77,
|
| 189573 |
+
"learning_rate": 8.880144230769233e-06,
|
| 189574 |
+
"loss": 0.5246,
|
| 189575 |
+
"step": 70905
|
| 189576 |
+
},
|
| 189577 |
+
{
|
| 189578 |
+
"epoch": 571.81,
|
| 189579 |
+
"learning_rate": 8.880064102564104e-06,
|
| 189580 |
+
"loss": 1.2047,
|
| 189581 |
+
"step": 70910
|
| 189582 |
+
},
|
| 189583 |
+
{
|
| 189584 |
+
"epoch": 571.85,
|
| 189585 |
+
"learning_rate": 8.879983974358975e-06,
|
| 189586 |
+
"loss": 0.3039,
|
| 189587 |
+
"step": 70915
|
| 189588 |
+
},
|
| 189589 |
+
{
|
| 189590 |
+
"epoch": 571.9,
|
| 189591 |
+
"learning_rate": 8.879903846153847e-06,
|
| 189592 |
+
"loss": 0.3366,
|
| 189593 |
+
"step": 70920
|
| 189594 |
+
},
|
| 189595 |
+
{
|
| 189596 |
+
"epoch": 571.94,
|
| 189597 |
+
"learning_rate": 8.87982371794872e-06,
|
| 189598 |
+
"loss": 0.412,
|
| 189599 |
+
"step": 70925
|
| 189600 |
+
},
|
| 189601 |
+
{
|
| 189602 |
+
"epoch": 571.98,
|
| 189603 |
+
"learning_rate": 8.87974358974359e-06,
|
| 189604 |
+
"loss": 0.6262,
|
| 189605 |
+
"step": 70930
|
| 189606 |
+
},
|
| 189607 |
+
{
|
| 189608 |
+
"epoch": 572.0,
|
| 189609 |
+
"eval_loss": 0.37808266282081604,
|
| 189610 |
+
"eval_runtime": 39.9883,
|
| 189611 |
+
"eval_samples_per_second": 21.006,
|
| 189612 |
+
"eval_steps_per_second": 0.675,
|
| 189613 |
+
"eval_wer": 0.1896881435716387,
|
| 189614 |
+
"step": 70933
|
| 189615 |
+
},
|
| 189616 |
+
{
|
| 189617 |
+
"epoch": 572.02,
|
| 189618 |
+
"learning_rate": 8.879663461538462e-06,
|
| 189619 |
+
"loss": 0.4067,
|
| 189620 |
+
"step": 70935
|
| 189621 |
+
},
|
| 189622 |
+
{
|
| 189623 |
+
"epoch": 572.06,
|
| 189624 |
+
"learning_rate": 8.879583333333334e-06,
|
| 189625 |
+
"loss": 0.3122,
|
| 189626 |
+
"step": 70940
|
| 189627 |
+
},
|
| 189628 |
+
{
|
| 189629 |
+
"epoch": 572.1,
|
| 189630 |
+
"learning_rate": 8.879503205128205e-06,
|
| 189631 |
+
"loss": 0.3144,
|
| 189632 |
+
"step": 70945
|
| 189633 |
+
},
|
| 189634 |
+
{
|
| 189635 |
+
"epoch": 572.14,
|
| 189636 |
+
"learning_rate": 8.879423076923077e-06,
|
| 189637 |
+
"loss": 0.3966,
|
| 189638 |
+
"step": 70950
|
| 189639 |
+
},
|
| 189640 |
+
{
|
| 189641 |
+
"epoch": 572.18,
|
| 189642 |
+
"learning_rate": 8.87934294871795e-06,
|
| 189643 |
+
"loss": 0.7134,
|
| 189644 |
+
"step": 70955
|
| 189645 |
+
},
|
| 189646 |
+
{
|
| 189647 |
+
"epoch": 572.22,
|
| 189648 |
+
"learning_rate": 8.879262820512821e-06,
|
| 189649 |
+
"loss": 1.0855,
|
| 189650 |
+
"step": 70960
|
| 189651 |
+
},
|
| 189652 |
+
{
|
| 189653 |
+
"epoch": 572.26,
|
| 189654 |
+
"learning_rate": 8.879182692307692e-06,
|
| 189655 |
+
"loss": 0.2845,
|
| 189656 |
+
"step": 70965
|
| 189657 |
+
},
|
| 189658 |
+
{
|
| 189659 |
+
"epoch": 572.3,
|
| 189660 |
+
"learning_rate": 8.879102564102565e-06,
|
| 189661 |
+
"loss": 0.3495,
|
| 189662 |
+
"step": 70970
|
| 189663 |
+
},
|
| 189664 |
+
{
|
| 189665 |
+
"epoch": 572.34,
|
| 189666 |
+
"learning_rate": 8.879022435897437e-06,
|
| 189667 |
+
"loss": 0.325,
|
| 189668 |
+
"step": 70975
|
| 189669 |
+
},
|
| 189670 |
+
{
|
| 189671 |
+
"epoch": 572.38,
|
| 189672 |
+
"learning_rate": 8.878942307692308e-06,
|
| 189673 |
+
"loss": 0.6465,
|
| 189674 |
+
"step": 70980
|
| 189675 |
+
},
|
| 189676 |
+
{
|
| 189677 |
+
"epoch": 572.42,
|
| 189678 |
+
"learning_rate": 8.87886217948718e-06,
|
| 189679 |
+
"loss": 1.0453,
|
| 189680 |
+
"step": 70985
|
| 189681 |
+
},
|
| 189682 |
+
{
|
| 189683 |
+
"epoch": 572.46,
|
| 189684 |
+
"learning_rate": 8.878782051282052e-06,
|
| 189685 |
+
"loss": 0.3642,
|
| 189686 |
+
"step": 70990
|
| 189687 |
+
},
|
| 189688 |
+
{
|
| 189689 |
+
"epoch": 572.5,
|
| 189690 |
+
"learning_rate": 8.878701923076924e-06,
|
| 189691 |
+
"loss": 0.3005,
|
| 189692 |
+
"step": 70995
|
| 189693 |
+
},
|
| 189694 |
+
{
|
| 189695 |
+
"epoch": 572.54,
|
| 189696 |
+
"learning_rate": 8.878621794871795e-06,
|
| 189697 |
+
"loss": 0.3992,
|
| 189698 |
+
"step": 71000
|
| 189699 |
+
},
|
| 189700 |
+
{
|
| 189701 |
+
"epoch": 572.58,
|
| 189702 |
+
"learning_rate": 8.878541666666668e-06,
|
| 189703 |
+
"loss": 0.6974,
|
| 189704 |
+
"step": 71005
|
| 189705 |
+
},
|
| 189706 |
+
{
|
| 189707 |
+
"epoch": 572.62,
|
| 189708 |
+
"learning_rate": 8.87846153846154e-06,
|
| 189709 |
+
"loss": 0.9939,
|
| 189710 |
+
"step": 71010
|
| 189711 |
+
},
|
| 189712 |
+
{
|
| 189713 |
+
"epoch": 572.66,
|
| 189714 |
+
"learning_rate": 8.878381410256411e-06,
|
| 189715 |
+
"loss": 0.3028,
|
| 189716 |
+
"step": 71015
|
| 189717 |
+
},
|
| 189718 |
+
{
|
| 189719 |
+
"epoch": 572.7,
|
| 189720 |
+
"learning_rate": 8.878301282051282e-06,
|
| 189721 |
+
"loss": 0.3952,
|
| 189722 |
+
"step": 71020
|
| 189723 |
+
},
|
| 189724 |
+
{
|
| 189725 |
+
"epoch": 572.74,
|
| 189726 |
+
"learning_rate": 8.878221153846155e-06,
|
| 189727 |
+
"loss": 0.3883,
|
| 189728 |
+
"step": 71025
|
| 189729 |
+
},
|
| 189730 |
+
{
|
| 189731 |
+
"epoch": 572.78,
|
| 189732 |
+
"learning_rate": 8.878141025641027e-06,
|
| 189733 |
+
"loss": 0.6341,
|
| 189734 |
+
"step": 71030
|
| 189735 |
+
},
|
| 189736 |
+
{
|
| 189737 |
+
"epoch": 572.82,
|
| 189738 |
+
"learning_rate": 8.878060897435898e-06,
|
| 189739 |
+
"loss": 1.0706,
|
| 189740 |
+
"step": 71035
|
| 189741 |
+
},
|
| 189742 |
+
{
|
| 189743 |
+
"epoch": 572.86,
|
| 189744 |
+
"learning_rate": 8.87798076923077e-06,
|
| 189745 |
+
"loss": 0.2861,
|
| 189746 |
+
"step": 71040
|
| 189747 |
+
},
|
| 189748 |
+
{
|
| 189749 |
+
"epoch": 572.9,
|
| 189750 |
+
"learning_rate": 8.877900641025642e-06,
|
| 189751 |
+
"loss": 0.4219,
|
| 189752 |
+
"step": 71045
|
| 189753 |
+
},
|
| 189754 |
+
{
|
| 189755 |
+
"epoch": 572.94,
|
| 189756 |
+
"learning_rate": 8.877820512820514e-06,
|
| 189757 |
+
"loss": 0.3828,
|
| 189758 |
+
"step": 71050
|
| 189759 |
+
},
|
| 189760 |
+
{
|
| 189761 |
+
"epoch": 572.98,
|
| 189762 |
+
"learning_rate": 8.877740384615385e-06,
|
| 189763 |
+
"loss": 0.6617,
|
| 189764 |
+
"step": 71055
|
| 189765 |
+
},
|
| 189766 |
+
{
|
| 189767 |
+
"epoch": 573.0,
|
| 189768 |
+
"eval_loss": 0.37939906120300293,
|
| 189769 |
+
"eval_runtime": 39.3467,
|
| 189770 |
+
"eval_samples_per_second": 21.349,
|
| 189771 |
+
"eval_steps_per_second": 0.686,
|
| 189772 |
+
"eval_wer": 0.19251534649804009,
|
| 189773 |
+
"step": 71057
|
| 189774 |
+
},
|
| 189775 |
+
{
|
| 189776 |
+
"epoch": 573.02,
|
| 189777 |
+
"learning_rate": 8.877660256410258e-06,
|
| 189778 |
+
"loss": 0.3238,
|
| 189779 |
+
"step": 71060
|
| 189780 |
+
},
|
| 189781 |
+
{
|
| 189782 |
+
"epoch": 573.06,
|
| 189783 |
+
"learning_rate": 8.87758012820513e-06,
|
| 189784 |
+
"loss": 0.3552,
|
| 189785 |
+
"step": 71065
|
| 189786 |
+
},
|
| 189787 |
+
{
|
| 189788 |
+
"epoch": 573.1,
|
| 189789 |
+
"learning_rate": 8.877500000000001e-06,
|
| 189790 |
+
"loss": 0.3396,
|
| 189791 |
+
"step": 71070
|
| 189792 |
+
},
|
| 189793 |
+
{
|
| 189794 |
+
"epoch": 573.14,
|
| 189795 |
+
"learning_rate": 8.877419871794872e-06,
|
| 189796 |
+
"loss": 0.3658,
|
| 189797 |
+
"step": 71075
|
| 189798 |
+
},
|
| 189799 |
+
{
|
| 189800 |
+
"epoch": 573.18,
|
| 189801 |
+
"learning_rate": 8.877339743589745e-06,
|
| 189802 |
+
"loss": 0.8553,
|
| 189803 |
+
"step": 71080
|
| 189804 |
+
},
|
| 189805 |
+
{
|
| 189806 |
+
"epoch": 573.22,
|
| 189807 |
+
"learning_rate": 8.877259615384615e-06,
|
| 189808 |
+
"loss": 0.8966,
|
| 189809 |
+
"step": 71085
|
| 189810 |
+
},
|
| 189811 |
+
{
|
| 189812 |
+
"epoch": 573.27,
|
| 189813 |
+
"learning_rate": 8.877179487179488e-06,
|
| 189814 |
+
"loss": 0.4122,
|
| 189815 |
+
"step": 71090
|
| 189816 |
+
},
|
| 189817 |
+
{
|
| 189818 |
+
"epoch": 573.31,
|
| 189819 |
+
"learning_rate": 8.87709935897436e-06,
|
| 189820 |
+
"loss": 0.3168,
|
| 189821 |
+
"step": 71095
|
| 189822 |
+
},
|
| 189823 |
+
{
|
| 189824 |
+
"epoch": 573.35,
|
| 189825 |
+
"learning_rate": 8.87701923076923e-06,
|
| 189826 |
+
"loss": 0.3781,
|
| 189827 |
+
"step": 71100
|
| 189828 |
+
},
|
| 189829 |
+
{
|
| 189830 |
+
"epoch": 573.39,
|
| 189831 |
+
"learning_rate": 8.876939102564102e-06,
|
| 189832 |
+
"loss": 0.7079,
|
| 189833 |
+
"step": 71105
|
| 189834 |
+
},
|
| 189835 |
+
{
|
| 189836 |
+
"epoch": 573.43,
|
| 189837 |
+
"learning_rate": 8.876858974358975e-06,
|
| 189838 |
+
"loss": 0.9013,
|
| 189839 |
+
"step": 71110
|
| 189840 |
+
},
|
| 189841 |
+
{
|
| 189842 |
+
"epoch": 573.47,
|
| 189843 |
+
"learning_rate": 8.876778846153847e-06,
|
| 189844 |
+
"loss": 0.3044,
|
| 189845 |
+
"step": 71115
|
| 189846 |
+
},
|
| 189847 |
+
{
|
| 189848 |
+
"epoch": 573.51,
|
| 189849 |
+
"learning_rate": 8.876698717948718e-06,
|
| 189850 |
+
"loss": 0.3062,
|
| 189851 |
+
"step": 71120
|
| 189852 |
+
},
|
| 189853 |
+
{
|
| 189854 |
+
"epoch": 573.55,
|
| 189855 |
+
"learning_rate": 8.876618589743591e-06,
|
| 189856 |
+
"loss": 0.4105,
|
| 189857 |
+
"step": 71125
|
| 189858 |
+
},
|
| 189859 |
+
{
|
| 189860 |
+
"epoch": 573.59,
|
| 189861 |
+
"learning_rate": 8.876538461538462e-06,
|
| 189862 |
+
"loss": 0.8214,
|
| 189863 |
+
"step": 71130
|
| 189864 |
+
},
|
| 189865 |
+
{
|
| 189866 |
+
"epoch": 573.63,
|
| 189867 |
+
"learning_rate": 8.876458333333334e-06,
|
| 189868 |
+
"loss": 0.739,
|
| 189869 |
+
"step": 71135
|
| 189870 |
+
},
|
| 189871 |
+
{
|
| 189872 |
+
"epoch": 573.67,
|
| 189873 |
+
"learning_rate": 8.876378205128205e-06,
|
| 189874 |
+
"loss": 0.2696,
|
| 189875 |
+
"step": 71140
|
| 189876 |
+
},
|
| 189877 |
+
{
|
| 189878 |
+
"epoch": 573.71,
|
| 189879 |
+
"learning_rate": 8.876298076923078e-06,
|
| 189880 |
+
"loss": 0.2784,
|
| 189881 |
+
"step": 71145
|
| 189882 |
+
},
|
| 189883 |
+
{
|
| 189884 |
+
"epoch": 573.75,
|
| 189885 |
+
"learning_rate": 8.87621794871795e-06,
|
| 189886 |
+
"loss": 0.3571,
|
| 189887 |
+
"step": 71150
|
| 189888 |
+
},
|
| 189889 |
+
{
|
| 189890 |
+
"epoch": 573.79,
|
| 189891 |
+
"learning_rate": 8.876137820512821e-06,
|
| 189892 |
+
"loss": 0.9113,
|
| 189893 |
+
"step": 71155
|
| 189894 |
+
},
|
| 189895 |
+
{
|
| 189896 |
+
"epoch": 573.83,
|
| 189897 |
+
"learning_rate": 8.876057692307694e-06,
|
| 189898 |
+
"loss": 0.9159,
|
| 189899 |
+
"step": 71160
|
| 189900 |
+
},
|
| 189901 |
+
{
|
| 189902 |
+
"epoch": 573.87,
|
| 189903 |
+
"learning_rate": 8.875977564102565e-06,
|
| 189904 |
+
"loss": 0.2806,
|
| 189905 |
+
"step": 71165
|
| 189906 |
+
},
|
| 189907 |
+
{
|
| 189908 |
+
"epoch": 573.91,
|
| 189909 |
+
"learning_rate": 8.875897435897437e-06,
|
| 189910 |
+
"loss": 0.3812,
|
| 189911 |
+
"step": 71170
|
| 189912 |
+
},
|
| 189913 |
+
{
|
| 189914 |
+
"epoch": 573.95,
|
| 189915 |
+
"learning_rate": 8.875817307692308e-06,
|
| 189916 |
+
"loss": 0.4332,
|
| 189917 |
+
"step": 71175
|
| 189918 |
+
},
|
| 189919 |
+
{
|
| 189920 |
+
"epoch": 573.99,
|
| 189921 |
+
"learning_rate": 8.875737179487181e-06,
|
| 189922 |
+
"loss": 0.8548,
|
| 189923 |
+
"step": 71180
|
| 189924 |
+
},
|
| 189925 |
+
{
|
| 189926 |
+
"epoch": 574.0,
|
| 189927 |
+
"eval_loss": 0.3877381682395935,
|
| 189928 |
+
"eval_runtime": 39.9108,
|
| 189929 |
+
"eval_samples_per_second": 21.047,
|
| 189930 |
+
"eval_steps_per_second": 0.677,
|
| 189931 |
+
"eval_wer": 0.19365056511410267,
|
| 189932 |
+
"step": 71181
|
| 189933 |
+
},
|
| 189934 |
+
{
|
| 189935 |
+
"epoch": 574.03,
|
| 189936 |
+
"learning_rate": 8.875657051282052e-06,
|
| 189937 |
+
"loss": 0.3677,
|
| 189938 |
+
"step": 71185
|
| 189939 |
+
},
|
| 189940 |
+
{
|
| 189941 |
+
"epoch": 574.07,
|
| 189942 |
+
"learning_rate": 8.875576923076924e-06,
|
| 189943 |
+
"loss": 0.3396,
|
| 189944 |
+
"step": 71190
|
| 189945 |
+
},
|
| 189946 |
+
{
|
| 189947 |
+
"epoch": 574.11,
|
| 189948 |
+
"learning_rate": 8.875496794871795e-06,
|
| 189949 |
+
"loss": 0.3444,
|
| 189950 |
+
"step": 71195
|
| 189951 |
+
},
|
| 189952 |
+
{
|
| 189953 |
+
"epoch": 574.15,
|
| 189954 |
+
"learning_rate": 8.875416666666668e-06,
|
| 189955 |
+
"loss": 0.3975,
|
| 189956 |
+
"step": 71200
|
| 189957 |
+
},
|
| 189958 |
+
{
|
| 189959 |
+
"epoch": 574.19,
|
| 189960 |
+
"learning_rate": 8.875336538461538e-06,
|
| 189961 |
+
"loss": 0.9141,
|
| 189962 |
+
"step": 71205
|
| 189963 |
+
},
|
| 189964 |
+
{
|
| 189965 |
+
"epoch": 574.23,
|
| 189966 |
+
"learning_rate": 8.875256410256411e-06,
|
| 189967 |
+
"loss": 0.7714,
|
| 189968 |
+
"step": 71210
|
| 189969 |
+
},
|
| 189970 |
+
{
|
| 189971 |
+
"epoch": 574.27,
|
| 189972 |
+
"learning_rate": 8.875176282051284e-06,
|
| 189973 |
+
"loss": 0.3316,
|
| 189974 |
+
"step": 71215
|
| 189975 |
+
},
|
| 189976 |
+
{
|
| 189977 |
+
"epoch": 574.31,
|
| 189978 |
+
"learning_rate": 8.875096153846154e-06,
|
| 189979 |
+
"loss": 0.3371,
|
| 189980 |
+
"step": 71220
|
| 189981 |
+
},
|
| 189982 |
+
{
|
| 189983 |
+
"epoch": 574.35,
|
| 189984 |
+
"learning_rate": 8.875016025641027e-06,
|
| 189985 |
+
"loss": 0.4133,
|
| 189986 |
+
"step": 71225
|
| 189987 |
+
},
|
| 189988 |
+
{
|
| 189989 |
+
"epoch": 574.39,
|
| 189990 |
+
"learning_rate": 8.874935897435898e-06,
|
| 189991 |
+
"loss": 1.0848,
|
| 189992 |
+
"step": 71230
|
| 189993 |
+
},
|
| 189994 |
+
{
|
| 189995 |
+
"epoch": 574.43,
|
| 189996 |
+
"learning_rate": 8.87485576923077e-06,
|
| 189997 |
+
"loss": 0.6723,
|
| 189998 |
+
"step": 71235
|
| 189999 |
+
},
|
| 190000 |
+
{
|
| 190001 |
+
"epoch": 574.47,
|
| 190002 |
+
"learning_rate": 8.87477564102564e-06,
|
| 190003 |
+
"loss": 0.2706,
|
| 190004 |
+
"step": 71240
|
| 190005 |
+
},
|
| 190006 |
+
{
|
| 190007 |
+
"epoch": 574.51,
|
| 190008 |
+
"learning_rate": 8.874695512820514e-06,
|
| 190009 |
+
"loss": 0.3278,
|
| 190010 |
+
"step": 71245
|
| 190011 |
+
},
|
| 190012 |
+
{
|
| 190013 |
+
"epoch": 574.55,
|
| 190014 |
+
"learning_rate": 8.874615384615385e-06,
|
| 190015 |
+
"loss": 0.3897,
|
| 190016 |
+
"step": 71250
|
| 190017 |
+
},
|
| 190018 |
+
{
|
| 190019 |
+
"epoch": 574.59,
|
| 190020 |
+
"learning_rate": 8.874535256410256e-06,
|
| 190021 |
+
"loss": 1.1266,
|
| 190022 |
+
"step": 71255
|
| 190023 |
+
},
|
| 190024 |
+
{
|
| 190025 |
+
"epoch": 574.63,
|
| 190026 |
+
"learning_rate": 8.87445512820513e-06,
|
| 190027 |
+
"loss": 0.8757,
|
| 190028 |
+
"step": 71260
|
| 190029 |
+
},
|
| 190030 |
+
{
|
| 190031 |
+
"epoch": 574.67,
|
| 190032 |
+
"learning_rate": 8.874375000000001e-06,
|
| 190033 |
+
"loss": 0.3173,
|
| 190034 |
+
"step": 71265
|
| 190035 |
+
},
|
| 190036 |
+
{
|
| 190037 |
+
"epoch": 574.71,
|
| 190038 |
+
"learning_rate": 8.874294871794872e-06,
|
| 190039 |
+
"loss": 0.2944,
|
| 190040 |
+
"step": 71270
|
| 190041 |
+
},
|
| 190042 |
+
{
|
| 190043 |
+
"epoch": 574.76,
|
| 190044 |
+
"learning_rate": 8.874214743589744e-06,
|
| 190045 |
+
"loss": 0.4019,
|
| 190046 |
+
"step": 71275
|
| 190047 |
+
},
|
| 190048 |
+
{
|
| 190049 |
+
"epoch": 574.8,
|
| 190050 |
+
"learning_rate": 8.874134615384617e-06,
|
| 190051 |
+
"loss": 1.0625,
|
| 190052 |
+
"step": 71280
|
| 190053 |
+
},
|
| 190054 |
+
{
|
| 190055 |
+
"epoch": 574.84,
|
| 190056 |
+
"learning_rate": 8.874054487179488e-06,
|
| 190057 |
+
"loss": 0.7614,
|
| 190058 |
+
"step": 71285
|
| 190059 |
+
},
|
| 190060 |
+
{
|
| 190061 |
+
"epoch": 574.88,
|
| 190062 |
+
"learning_rate": 8.87397435897436e-06,
|
| 190063 |
+
"loss": 0.2756,
|
| 190064 |
+
"step": 71290
|
| 190065 |
+
},
|
| 190066 |
+
{
|
| 190067 |
+
"epoch": 574.92,
|
| 190068 |
+
"learning_rate": 8.87389423076923e-06,
|
| 190069 |
+
"loss": 0.3798,
|
| 190070 |
+
"step": 71295
|
| 190071 |
+
},
|
| 190072 |
+
{
|
| 190073 |
+
"epoch": 574.96,
|
| 190074 |
+
"learning_rate": 8.873814102564104e-06,
|
| 190075 |
+
"loss": 0.4301,
|
| 190076 |
+
"step": 71300
|
| 190077 |
+
},
|
| 190078 |
+
{
|
| 190079 |
+
"epoch": 575.0,
|
| 190080 |
+
"learning_rate": 8.873733974358975e-06,
|
| 190081 |
+
"loss": 1.2421,
|
| 190082 |
+
"step": 71305
|
| 190083 |
+
},
|
| 190084 |
+
{
|
| 190085 |
+
"epoch": 575.0,
|
| 190086 |
+
"eval_loss": 0.4396674633026123,
|
| 190087 |
+
"eval_runtime": 40.1793,
|
| 190088 |
+
"eval_samples_per_second": 20.906,
|
| 190089 |
+
"eval_steps_per_second": 0.672,
|
| 190090 |
+
"eval_wer": 0.19375594063025517,
|
| 190091 |
+
"step": 71305
|
| 190092 |
+
},
|
| 190093 |
+
{
|
| 190094 |
+
"epoch": 575.04,
|
| 190095 |
+
"learning_rate": 8.873653846153847e-06,
|
| 190096 |
+
"loss": 0.3501,
|
| 190097 |
+
"step": 71310
|
| 190098 |
+
},
|
| 190099 |
+
{
|
| 190100 |
+
"epoch": 575.08,
|
| 190101 |
+
"learning_rate": 8.87357371794872e-06,
|
| 190102 |
+
"loss": 0.3542,
|
| 190103 |
+
"step": 71315
|
| 190104 |
+
},
|
| 190105 |
+
{
|
| 190106 |
+
"epoch": 575.12,
|
| 190107 |
+
"learning_rate": 8.873493589743591e-06,
|
| 190108 |
+
"loss": 0.3445,
|
| 190109 |
+
"step": 71320
|
| 190110 |
+
},
|
| 190111 |
+
{
|
| 190112 |
+
"epoch": 575.16,
|
| 190113 |
+
"learning_rate": 8.873413461538462e-06,
|
| 190114 |
+
"loss": 0.5054,
|
| 190115 |
+
"step": 71325
|
| 190116 |
+
},
|
| 190117 |
+
{
|
| 190118 |
+
"epoch": 575.2,
|
| 190119 |
+
"learning_rate": 8.873333333333334e-06,
|
| 190120 |
+
"loss": 1.209,
|
| 190121 |
+
"step": 71330
|
| 190122 |
+
},
|
| 190123 |
+
{
|
| 190124 |
+
"epoch": 575.24,
|
| 190125 |
+
"learning_rate": 8.873253205128207e-06,
|
| 190126 |
+
"loss": 0.3995,
|
| 190127 |
+
"step": 71335
|
| 190128 |
+
},
|
| 190129 |
+
{
|
| 190130 |
+
"epoch": 575.28,
|
| 190131 |
+
"learning_rate": 8.873173076923078e-06,
|
| 190132 |
+
"loss": 0.3047,
|
| 190133 |
+
"step": 71340
|
| 190134 |
+
},
|
| 190135 |
+
{
|
| 190136 |
+
"epoch": 575.32,
|
| 190137 |
+
"learning_rate": 8.87309294871795e-06,
|
| 190138 |
+
"loss": 0.3208,
|
| 190139 |
+
"step": 71345
|
| 190140 |
+
},
|
| 190141 |
+
{
|
| 190142 |
+
"epoch": 575.36,
|
| 190143 |
+
"learning_rate": 8.87301282051282e-06,
|
| 190144 |
+
"loss": 0.4865,
|
| 190145 |
+
"step": 71350
|
| 190146 |
+
},
|
| 190147 |
+
{
|
| 190148 |
+
"epoch": 575.4,
|
| 190149 |
+
"learning_rate": 8.872932692307694e-06,
|
| 190150 |
+
"loss": 1.2983,
|
| 190151 |
+
"step": 71355
|
| 190152 |
+
},
|
| 190153 |
+
{
|
| 190154 |
+
"epoch": 575.44,
|
| 190155 |
+
"learning_rate": 8.872852564102565e-06,
|
| 190156 |
+
"loss": 0.3453,
|
| 190157 |
+
"step": 71360
|
| 190158 |
+
},
|
| 190159 |
+
{
|
| 190160 |
+
"epoch": 575.48,
|
| 190161 |
+
"learning_rate": 8.872772435897437e-06,
|
| 190162 |
+
"loss": 0.3389,
|
| 190163 |
+
"step": 71365
|
| 190164 |
+
},
|
| 190165 |
+
{
|
| 190166 |
+
"epoch": 575.52,
|
| 190167 |
+
"learning_rate": 8.87269230769231e-06,
|
| 190168 |
+
"loss": 0.3255,
|
| 190169 |
+
"step": 71370
|
| 190170 |
+
},
|
| 190171 |
+
{
|
| 190172 |
+
"epoch": 575.56,
|
| 190173 |
+
"learning_rate": 8.87261217948718e-06,
|
| 190174 |
+
"loss": 0.4804,
|
| 190175 |
+
"step": 71375
|
| 190176 |
+
},
|
| 190177 |
+
{
|
| 190178 |
+
"epoch": 575.6,
|
| 190179 |
+
"learning_rate": 8.872532051282052e-06,
|
| 190180 |
+
"loss": 1.1065,
|
| 190181 |
+
"step": 71380
|
| 190182 |
+
},
|
| 190183 |
+
{
|
| 190184 |
+
"epoch": 575.64,
|
| 190185 |
+
"learning_rate": 8.872451923076924e-06,
|
| 190186 |
+
"loss": 0.3769,
|
| 190187 |
+
"step": 71385
|
| 190188 |
+
},
|
| 190189 |
+
{
|
| 190190 |
+
"epoch": 575.68,
|
| 190191 |
+
"learning_rate": 8.872371794871795e-06,
|
| 190192 |
+
"loss": 0.2841,
|
| 190193 |
+
"step": 71390
|
| 190194 |
+
},
|
| 190195 |
+
{
|
| 190196 |
+
"epoch": 575.72,
|
| 190197 |
+
"learning_rate": 8.872291666666666e-06,
|
| 190198 |
+
"loss": 0.3687,
|
| 190199 |
+
"step": 71395
|
| 190200 |
+
},
|
| 190201 |
+
{
|
| 190202 |
+
"epoch": 575.76,
|
| 190203 |
+
"learning_rate": 8.87221153846154e-06,
|
| 190204 |
+
"loss": 0.4895,
|
| 190205 |
+
"step": 71400
|
| 190206 |
+
},
|
| 190207 |
+
{
|
| 190208 |
+
"epoch": 575.8,
|
| 190209 |
+
"learning_rate": 8.87213141025641e-06,
|
| 190210 |
+
"loss": 1.1831,
|
| 190211 |
+
"step": 71405
|
| 190212 |
+
},
|
| 190213 |
+
{
|
| 190214 |
+
"epoch": 575.84,
|
| 190215 |
+
"learning_rate": 8.872051282051282e-06,
|
| 190216 |
+
"loss": 0.3517,
|
| 190217 |
+
"step": 71410
|
| 190218 |
+
},
|
| 190219 |
+
{
|
| 190220 |
+
"epoch": 575.88,
|
| 190221 |
+
"learning_rate": 8.871971153846155e-06,
|
| 190222 |
+
"loss": 0.3023,
|
| 190223 |
+
"step": 71415
|
| 190224 |
+
},
|
| 190225 |
+
{
|
| 190226 |
+
"epoch": 575.92,
|
| 190227 |
+
"learning_rate": 8.871891025641027e-06,
|
| 190228 |
+
"loss": 0.3109,
|
| 190229 |
+
"step": 71420
|
| 190230 |
+
},
|
| 190231 |
+
{
|
| 190232 |
+
"epoch": 575.96,
|
| 190233 |
+
"learning_rate": 8.871810897435898e-06,
|
| 190234 |
+
"loss": 0.5157,
|
| 190235 |
+
"step": 71425
|
| 190236 |
+
},
|
| 190237 |
+
{
|
| 190238 |
+
"epoch": 576.0,
|
| 190239 |
+
"eval_loss": 0.370493620634079,
|
| 190240 |
+
"eval_runtime": 40.7235,
|
| 190241 |
+
"eval_samples_per_second": 20.627,
|
| 190242 |
+
"eval_steps_per_second": 0.663,
|
| 190243 |
+
"eval_wer": 0.18538826318909307,
|
| 190244 |
+
"step": 71429
|
| 190245 |
}
|
| 190246 |
],
|
| 190247 |
"max_steps": 620000,
|
| 190248 |
"num_train_epochs": 5000,
|
| 190249 |
+
"total_flos": 2.0101646863500657e+20,
|
| 190250 |
"trial_name": null,
|
| 190251 |
"trial_params": null
|
| 190252 |
}
|
model-bin/finetune/base/{checkpoint-70809 → checkpoint-71429}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629858970.8441613/events.out.tfevents.1629858970.7e498afd5545.905.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bdda49e59d41b85957a8cccc5577fcf0f644c5611ecd244a897f22720490ca05
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629859624.9463222/events.out.tfevents.1629859624.7e498afd5545.905.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f0de7580033efd4f6523ea6ea4d985866eb98165f7ce6cef08b659fd4f77df42
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629860278.0745044/events.out.tfevents.1629860278.7e498afd5545.905.27
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f3764d5f0e85be9f251324fca32644721d02680b360ebd90ac23c0742e0a6f7
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629860932.5132222/events.out.tfevents.1629860932.7e498afd5545.905.29
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5e0087eb4fb9ddf06503c1a7b2305e2eb97a4422095bef32e827069366286069
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629861583.8800406/events.out.tfevents.1629861583.7e498afd5545.905.31
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:13c2e93be9000158ca59580c8a3592bd64fa86a07859832ca99e6402955c4981
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629858970.7e498afd5545.905.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:95341d235f3561e1cc248e9f54713ad5d5c383fc0af5c13f55763c13de9e32d3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629859624.7e498afd5545.905.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2c66a50c891088080220870c0899a1d489514ffd01056aad55cd7b982b2316b1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629860278.7e498afd5545.905.26
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49a2cad8d2fc8657e15e9a3c1596b3a90b0411c1665554e0fd6049efa13743ec
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629860932.7e498afd5545.905.28
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9e23d6fb0b7e0651fee49bbb66a50c47a62bbea9c783fb1b5f187de94c1bccca
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629861583.7e498afd5545.905.30
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:013cda62abdeb0bba2da058ab6dbf1e2e2fb143163b9e461ac2f41c2a39b07f1
|
| 3 |
+
size 8462
|