"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629941989.4511316/events.out.tfevents.1629941989.8e89bd551565.924.11 +3 -0
- model-bin/finetune/base/log/1629942431.5525227/events.out.tfevents.1629942431.8e89bd551565.924.13 +3 -0
- model-bin/finetune/base/log/1629942858.3612747/events.out.tfevents.1629942858.8e89bd551565.924.15 +3 -0
- model-bin/finetune/base/log/1629943284.2321923/events.out.tfevents.1629943284.8e89bd551565.924.17 +3 -0
- model-bin/finetune/base/log/1629943702.5924923/events.out.tfevents.1629943702.8e89bd551565.924.19 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629941989.8e89bd551565.924.10 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629942431.8e89bd551565.924.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629942858.8e89bd551565.924.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629943284.8e89bd551565.924.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629943702.8e89bd551565.924.18 +3 -0
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/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 sha256:2443b8ea1d53aac7f389153148ffabe5ce7712b63b8d05b786b70b8a85200a25
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/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:d42f5b531eb03f0e314693855366c0745885141031e2c6fb93cab7ad1340bbb1
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14567
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:91ed9200715cb38ca621fab9af9f5470bcc041ea0ae387481d03723f806e19db
|
| 3 |
size 14567
|
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/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 sha256:0f396b66cec689ed47334ef06d8b2f62cc0b8112d2b9e35e8e685b8ace420b29
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/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:ee56de8be16189110ea7ac83a11716df996819bb79cfacd43f9aafb201925200
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -212577,11 +212577,800 @@
|
|
| 212577 |
"eval_steps_per_second": 0.703,
|
| 212578 |
"eval_wer": 0.1821153566758681,
|
| 212579 |
"step": 88981
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 212580 |
}
|
| 212581 |
],
|
| 212582 |
-
"max_steps":
|
| 212583 |
"num_train_epochs": 5000,
|
| 212584 |
-
"total_flos": 2.
|
| 212585 |
"trial_name": null,
|
| 212586 |
"trial_params": null
|
| 212587 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
+
"epoch": 716.0,
|
| 5 |
+
"global_step": 89603,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 212577 |
"eval_steps_per_second": 0.703,
|
| 212578 |
"eval_wer": 0.1821153566758681,
|
| 212579 |
"step": 88981
|
| 212580 |
+
},
|
| 212581 |
+
{
|
| 212582 |
+
"epoch": 711.03,
|
| 212583 |
+
"learning_rate": 8.590560897435898e-06,
|
| 212584 |
+
"loss": 0.3179,
|
| 212585 |
+
"step": 88985
|
| 212586 |
+
},
|
| 212587 |
+
{
|
| 212588 |
+
"epoch": 711.07,
|
| 212589 |
+
"learning_rate": 8.590480769230771e-06,
|
| 212590 |
+
"loss": 0.2987,
|
| 212591 |
+
"step": 88990
|
| 212592 |
+
},
|
| 212593 |
+
{
|
| 212594 |
+
"epoch": 711.11,
|
| 212595 |
+
"learning_rate": 8.59040064102564e-06,
|
| 212596 |
+
"loss": 0.3108,
|
| 212597 |
+
"step": 88995
|
| 212598 |
+
},
|
| 212599 |
+
{
|
| 212600 |
+
"epoch": 711.15,
|
| 212601 |
+
"learning_rate": 8.590320512820514e-06,
|
| 212602 |
+
"loss": 0.392,
|
| 212603 |
+
"step": 89000
|
| 212604 |
+
},
|
| 212605 |
+
{
|
| 212606 |
+
"epoch": 711.19,
|
| 212607 |
+
"learning_rate": 8.590240384615385e-06,
|
| 212608 |
+
"loss": 0.8451,
|
| 212609 |
+
"step": 89005
|
| 212610 |
+
},
|
| 212611 |
+
{
|
| 212612 |
+
"epoch": 711.23,
|
| 212613 |
+
"learning_rate": 8.590160256410256e-06,
|
| 212614 |
+
"loss": 0.613,
|
| 212615 |
+
"step": 89010
|
| 212616 |
+
},
|
| 212617 |
+
{
|
| 212618 |
+
"epoch": 711.27,
|
| 212619 |
+
"learning_rate": 8.59008012820513e-06,
|
| 212620 |
+
"loss": 0.3105,
|
| 212621 |
+
"step": 89015
|
| 212622 |
+
},
|
| 212623 |
+
{
|
| 212624 |
+
"epoch": 711.31,
|
| 212625 |
+
"learning_rate": 8.59e-06,
|
| 212626 |
+
"loss": 0.2981,
|
| 212627 |
+
"step": 89020
|
| 212628 |
+
},
|
| 212629 |
+
{
|
| 212630 |
+
"epoch": 711.35,
|
| 212631 |
+
"learning_rate": 8.589919871794872e-06,
|
| 212632 |
+
"loss": 0.4424,
|
| 212633 |
+
"step": 89025
|
| 212634 |
+
},
|
| 212635 |
+
{
|
| 212636 |
+
"epoch": 711.39,
|
| 212637 |
+
"learning_rate": 8.589839743589743e-06,
|
| 212638 |
+
"loss": 0.9266,
|
| 212639 |
+
"step": 89030
|
| 212640 |
+
},
|
| 212641 |
+
{
|
| 212642 |
+
"epoch": 711.43,
|
| 212643 |
+
"learning_rate": 8.589759615384617e-06,
|
| 212644 |
+
"loss": 0.6477,
|
| 212645 |
+
"step": 89035
|
| 212646 |
+
},
|
| 212647 |
+
{
|
| 212648 |
+
"epoch": 711.47,
|
| 212649 |
+
"learning_rate": 8.589679487179488e-06,
|
| 212650 |
+
"loss": 0.3203,
|
| 212651 |
+
"step": 89040
|
| 212652 |
+
},
|
| 212653 |
+
{
|
| 212654 |
+
"epoch": 711.51,
|
| 212655 |
+
"learning_rate": 8.58959935897436e-06,
|
| 212656 |
+
"loss": 0.361,
|
| 212657 |
+
"step": 89045
|
| 212658 |
+
},
|
| 212659 |
+
{
|
| 212660 |
+
"epoch": 711.55,
|
| 212661 |
+
"learning_rate": 8.58951923076923e-06,
|
| 212662 |
+
"loss": 0.4658,
|
| 212663 |
+
"step": 89050
|
| 212664 |
+
},
|
| 212665 |
+
{
|
| 212666 |
+
"epoch": 711.59,
|
| 212667 |
+
"learning_rate": 8.589439102564104e-06,
|
| 212668 |
+
"loss": 0.9329,
|
| 212669 |
+
"step": 89055
|
| 212670 |
+
},
|
| 212671 |
+
{
|
| 212672 |
+
"epoch": 711.63,
|
| 212673 |
+
"learning_rate": 8.589358974358975e-06,
|
| 212674 |
+
"loss": 0.6051,
|
| 212675 |
+
"step": 89060
|
| 212676 |
+
},
|
| 212677 |
+
{
|
| 212678 |
+
"epoch": 711.67,
|
| 212679 |
+
"learning_rate": 8.589278846153846e-06,
|
| 212680 |
+
"loss": 0.2512,
|
| 212681 |
+
"step": 89065
|
| 212682 |
+
},
|
| 212683 |
+
{
|
| 212684 |
+
"epoch": 711.71,
|
| 212685 |
+
"learning_rate": 8.58919871794872e-06,
|
| 212686 |
+
"loss": 0.2798,
|
| 212687 |
+
"step": 89070
|
| 212688 |
+
},
|
| 212689 |
+
{
|
| 212690 |
+
"epoch": 711.75,
|
| 212691 |
+
"learning_rate": 8.58911858974359e-06,
|
| 212692 |
+
"loss": 0.4504,
|
| 212693 |
+
"step": 89075
|
| 212694 |
+
},
|
| 212695 |
+
{
|
| 212696 |
+
"epoch": 711.79,
|
| 212697 |
+
"learning_rate": 8.589038461538462e-06,
|
| 212698 |
+
"loss": 0.9256,
|
| 212699 |
+
"step": 89080
|
| 212700 |
+
},
|
| 212701 |
+
{
|
| 212702 |
+
"epoch": 711.83,
|
| 212703 |
+
"learning_rate": 8.588958333333334e-06,
|
| 212704 |
+
"loss": 0.7481,
|
| 212705 |
+
"step": 89085
|
| 212706 |
+
},
|
| 212707 |
+
{
|
| 212708 |
+
"epoch": 711.87,
|
| 212709 |
+
"learning_rate": 8.588878205128207e-06,
|
| 212710 |
+
"loss": 0.2726,
|
| 212711 |
+
"step": 89090
|
| 212712 |
+
},
|
| 212713 |
+
{
|
| 212714 |
+
"epoch": 711.91,
|
| 212715 |
+
"learning_rate": 8.588798076923078e-06,
|
| 212716 |
+
"loss": 0.313,
|
| 212717 |
+
"step": 89095
|
| 212718 |
+
},
|
| 212719 |
+
{
|
| 212720 |
+
"epoch": 711.95,
|
| 212721 |
+
"learning_rate": 8.58871794871795e-06,
|
| 212722 |
+
"loss": 0.3964,
|
| 212723 |
+
"step": 89100
|
| 212724 |
+
},
|
| 212725 |
+
{
|
| 212726 |
+
"epoch": 711.99,
|
| 212727 |
+
"learning_rate": 8.58863782051282e-06,
|
| 212728 |
+
"loss": 0.9138,
|
| 212729 |
+
"step": 89105
|
| 212730 |
+
},
|
| 212731 |
+
{
|
| 212732 |
+
"epoch": 712.0,
|
| 212733 |
+
"eval_loss": 0.3886844217777252,
|
| 212734 |
+
"eval_runtime": 39.4344,
|
| 212735 |
+
"eval_samples_per_second": 21.327,
|
| 212736 |
+
"eval_steps_per_second": 0.685,
|
| 212737 |
+
"eval_wer": 0.19554527046572173,
|
| 212738 |
+
"step": 89106
|
| 212739 |
+
},
|
| 212740 |
+
{
|
| 212741 |
+
"epoch": 718.03,
|
| 212742 |
+
"learning_rate": 8.588557692307694e-06,
|
| 212743 |
+
"loss": 0.3705,
|
| 212744 |
+
"step": 89110
|
| 212745 |
+
},
|
| 212746 |
+
{
|
| 212747 |
+
"epoch": 718.07,
|
| 212748 |
+
"learning_rate": 8.588477564102565e-06,
|
| 212749 |
+
"loss": 0.2904,
|
| 212750 |
+
"step": 89115
|
| 212751 |
+
},
|
| 212752 |
+
{
|
| 212753 |
+
"epoch": 718.11,
|
| 212754 |
+
"learning_rate": 8.588397435897436e-06,
|
| 212755 |
+
"loss": 0.3454,
|
| 212756 |
+
"step": 89120
|
| 212757 |
+
},
|
| 212758 |
+
{
|
| 212759 |
+
"epoch": 718.15,
|
| 212760 |
+
"learning_rate": 8.58831730769231e-06,
|
| 212761 |
+
"loss": 0.3967,
|
| 212762 |
+
"step": 89125
|
| 212763 |
+
},
|
| 212764 |
+
{
|
| 212765 |
+
"epoch": 718.19,
|
| 212766 |
+
"learning_rate": 8.588237179487179e-06,
|
| 212767 |
+
"loss": 0.915,
|
| 212768 |
+
"step": 89130
|
| 212769 |
+
},
|
| 212770 |
+
{
|
| 212771 |
+
"epoch": 718.23,
|
| 212772 |
+
"learning_rate": 8.588157051282052e-06,
|
| 212773 |
+
"loss": 0.7099,
|
| 212774 |
+
"step": 89135
|
| 212775 |
+
},
|
| 212776 |
+
{
|
| 212777 |
+
"epoch": 718.27,
|
| 212778 |
+
"learning_rate": 8.588076923076924e-06,
|
| 212779 |
+
"loss": 0.2788,
|
| 212780 |
+
"step": 89140
|
| 212781 |
+
},
|
| 212782 |
+
{
|
| 212783 |
+
"epoch": 718.31,
|
| 212784 |
+
"learning_rate": 8.587996794871795e-06,
|
| 212785 |
+
"loss": 0.2902,
|
| 212786 |
+
"step": 89145
|
| 212787 |
+
},
|
| 212788 |
+
{
|
| 212789 |
+
"epoch": 718.35,
|
| 212790 |
+
"learning_rate": 8.587916666666666e-06,
|
| 212791 |
+
"loss": 0.3752,
|
| 212792 |
+
"step": 89150
|
| 212793 |
+
},
|
| 212794 |
+
{
|
| 212795 |
+
"epoch": 718.39,
|
| 212796 |
+
"learning_rate": 8.58783653846154e-06,
|
| 212797 |
+
"loss": 0.8795,
|
| 212798 |
+
"step": 89155
|
| 212799 |
+
},
|
| 212800 |
+
{
|
| 212801 |
+
"epoch": 718.43,
|
| 212802 |
+
"learning_rate": 8.58775641025641e-06,
|
| 212803 |
+
"loss": 0.824,
|
| 212804 |
+
"step": 89160
|
| 212805 |
+
},
|
| 212806 |
+
{
|
| 212807 |
+
"epoch": 718.47,
|
| 212808 |
+
"learning_rate": 8.587676282051282e-06,
|
| 212809 |
+
"loss": 0.3274,
|
| 212810 |
+
"step": 89165
|
| 212811 |
+
},
|
| 212812 |
+
{
|
| 212813 |
+
"epoch": 718.51,
|
| 212814 |
+
"learning_rate": 8.587596153846155e-06,
|
| 212815 |
+
"loss": 0.3495,
|
| 212816 |
+
"step": 89170
|
| 212817 |
+
},
|
| 212818 |
+
{
|
| 212819 |
+
"epoch": 718.55,
|
| 212820 |
+
"learning_rate": 8.587516025641026e-06,
|
| 212821 |
+
"loss": 0.41,
|
| 212822 |
+
"step": 89175
|
| 212823 |
+
},
|
| 212824 |
+
{
|
| 212825 |
+
"epoch": 718.59,
|
| 212826 |
+
"learning_rate": 8.587435897435898e-06,
|
| 212827 |
+
"loss": 1.0472,
|
| 212828 |
+
"step": 89180
|
| 212829 |
+
},
|
| 212830 |
+
{
|
| 212831 |
+
"epoch": 718.63,
|
| 212832 |
+
"learning_rate": 8.587355769230769e-06,
|
| 212833 |
+
"loss": 0.6203,
|
| 212834 |
+
"step": 89185
|
| 212835 |
+
},
|
| 212836 |
+
{
|
| 212837 |
+
"epoch": 718.67,
|
| 212838 |
+
"learning_rate": 8.587275641025642e-06,
|
| 212839 |
+
"loss": 0.3644,
|
| 212840 |
+
"step": 89190
|
| 212841 |
+
},
|
| 212842 |
+
{
|
| 212843 |
+
"epoch": 718.71,
|
| 212844 |
+
"learning_rate": 8.587195512820514e-06,
|
| 212845 |
+
"loss": 0.2647,
|
| 212846 |
+
"step": 89195
|
| 212847 |
+
},
|
| 212848 |
+
{
|
| 212849 |
+
"epoch": 718.76,
|
| 212850 |
+
"learning_rate": 8.587115384615385e-06,
|
| 212851 |
+
"loss": 0.4371,
|
| 212852 |
+
"step": 89200
|
| 212853 |
+
},
|
| 212854 |
+
{
|
| 212855 |
+
"epoch": 718.8,
|
| 212856 |
+
"learning_rate": 8.587035256410256e-06,
|
| 212857 |
+
"loss": 1.1203,
|
| 212858 |
+
"step": 89205
|
| 212859 |
+
},
|
| 212860 |
+
{
|
| 212861 |
+
"epoch": 718.84,
|
| 212862 |
+
"learning_rate": 8.58695512820513e-06,
|
| 212863 |
+
"loss": 0.652,
|
| 212864 |
+
"step": 89210
|
| 212865 |
+
},
|
| 212866 |
+
{
|
| 212867 |
+
"epoch": 718.88,
|
| 212868 |
+
"learning_rate": 8.586875e-06,
|
| 212869 |
+
"loss": 0.3579,
|
| 212870 |
+
"step": 89215
|
| 212871 |
+
},
|
| 212872 |
+
{
|
| 212873 |
+
"epoch": 718.92,
|
| 212874 |
+
"learning_rate": 8.586794871794872e-06,
|
| 212875 |
+
"loss": 0.3215,
|
| 212876 |
+
"step": 89220
|
| 212877 |
+
},
|
| 212878 |
+
{
|
| 212879 |
+
"epoch": 718.96,
|
| 212880 |
+
"learning_rate": 8.586714743589745e-06,
|
| 212881 |
+
"loss": 0.4431,
|
| 212882 |
+
"step": 89225
|
| 212883 |
+
},
|
| 212884 |
+
{
|
| 212885 |
+
"epoch": 719.0,
|
| 212886 |
+
"learning_rate": 8.586634615384616e-06,
|
| 212887 |
+
"loss": 1.1711,
|
| 212888 |
+
"step": 89230
|
| 212889 |
+
},
|
| 212890 |
+
{
|
| 212891 |
+
"epoch": 719.0,
|
| 212892 |
+
"eval_loss": 0.44025877118110657,
|
| 212893 |
+
"eval_runtime": 38.1224,
|
| 212894 |
+
"eval_samples_per_second": 22.061,
|
| 212895 |
+
"eval_steps_per_second": 0.708,
|
| 212896 |
+
"eval_wer": 0.19066512768720242,
|
| 212897 |
+
"step": 89230
|
| 212898 |
+
},
|
| 212899 |
+
{
|
| 212900 |
+
"epoch": 713.04,
|
| 212901 |
+
"learning_rate": 8.586554487179488e-06,
|
| 212902 |
+
"loss": 0.3305,
|
| 212903 |
+
"step": 89235
|
| 212904 |
+
},
|
| 212905 |
+
{
|
| 212906 |
+
"epoch": 713.08,
|
| 212907 |
+
"learning_rate": 8.58647435897436e-06,
|
| 212908 |
+
"loss": 0.2897,
|
| 212909 |
+
"step": 89240
|
| 212910 |
+
},
|
| 212911 |
+
{
|
| 212912 |
+
"epoch": 713.12,
|
| 212913 |
+
"learning_rate": 8.586394230769232e-06,
|
| 212914 |
+
"loss": 0.3075,
|
| 212915 |
+
"step": 89245
|
| 212916 |
+
},
|
| 212917 |
+
{
|
| 212918 |
+
"epoch": 713.16,
|
| 212919 |
+
"learning_rate": 8.586314102564104e-06,
|
| 212920 |
+
"loss": 0.4379,
|
| 212921 |
+
"step": 89250
|
| 212922 |
+
},
|
| 212923 |
+
{
|
| 212924 |
+
"epoch": 713.2,
|
| 212925 |
+
"learning_rate": 8.586233974358975e-06,
|
| 212926 |
+
"loss": 1.2201,
|
| 212927 |
+
"step": 89255
|
| 212928 |
+
},
|
| 212929 |
+
{
|
| 212930 |
+
"epoch": 713.24,
|
| 212931 |
+
"learning_rate": 8.586153846153848e-06,
|
| 212932 |
+
"loss": 0.3432,
|
| 212933 |
+
"step": 89260
|
| 212934 |
+
},
|
| 212935 |
+
{
|
| 212936 |
+
"epoch": 713.28,
|
| 212937 |
+
"learning_rate": 8.58607371794872e-06,
|
| 212938 |
+
"loss": 0.2984,
|
| 212939 |
+
"step": 89265
|
| 212940 |
+
},
|
| 212941 |
+
{
|
| 212942 |
+
"epoch": 713.32,
|
| 212943 |
+
"learning_rate": 8.58599358974359e-06,
|
| 212944 |
+
"loss": 0.362,
|
| 212945 |
+
"step": 89270
|
| 212946 |
+
},
|
| 212947 |
+
{
|
| 212948 |
+
"epoch": 713.36,
|
| 212949 |
+
"learning_rate": 8.585913461538462e-06,
|
| 212950 |
+
"loss": 0.4702,
|
| 212951 |
+
"step": 89275
|
| 212952 |
+
},
|
| 212953 |
+
{
|
| 212954 |
+
"epoch": 713.4,
|
| 212955 |
+
"learning_rate": 8.585833333333335e-06,
|
| 212956 |
+
"loss": 1.2665,
|
| 212957 |
+
"step": 89280
|
| 212958 |
+
},
|
| 212959 |
+
{
|
| 212960 |
+
"epoch": 713.44,
|
| 212961 |
+
"learning_rate": 8.585753205128205e-06,
|
| 212962 |
+
"loss": 0.2856,
|
| 212963 |
+
"step": 89285
|
| 212964 |
+
},
|
| 212965 |
+
{
|
| 212966 |
+
"epoch": 713.48,
|
| 212967 |
+
"learning_rate": 8.585673076923078e-06,
|
| 212968 |
+
"loss": 0.3187,
|
| 212969 |
+
"step": 89290
|
| 212970 |
+
},
|
| 212971 |
+
{
|
| 212972 |
+
"epoch": 713.52,
|
| 212973 |
+
"learning_rate": 8.58559294871795e-06,
|
| 212974 |
+
"loss": 0.2725,
|
| 212975 |
+
"step": 89295
|
| 212976 |
+
},
|
| 212977 |
+
{
|
| 212978 |
+
"epoch": 713.56,
|
| 212979 |
+
"learning_rate": 8.58551282051282e-06,
|
| 212980 |
+
"loss": 0.4075,
|
| 212981 |
+
"step": 89300
|
| 212982 |
+
},
|
| 212983 |
+
{
|
| 212984 |
+
"epoch": 713.6,
|
| 212985 |
+
"learning_rate": 8.585432692307692e-06,
|
| 212986 |
+
"loss": 1.0961,
|
| 212987 |
+
"step": 89305
|
| 212988 |
+
},
|
| 212989 |
+
{
|
| 212990 |
+
"epoch": 713.64,
|
| 212991 |
+
"learning_rate": 8.585352564102565e-06,
|
| 212992 |
+
"loss": 0.3646,
|
| 212993 |
+
"step": 89310
|
| 212994 |
+
},
|
| 212995 |
+
{
|
| 212996 |
+
"epoch": 713.68,
|
| 212997 |
+
"learning_rate": 8.585272435897436e-06,
|
| 212998 |
+
"loss": 0.2833,
|
| 212999 |
+
"step": 89315
|
| 213000 |
+
},
|
| 213001 |
+
{
|
| 213002 |
+
"epoch": 713.72,
|
| 213003 |
+
"learning_rate": 8.585192307692308e-06,
|
| 213004 |
+
"loss": 0.3468,
|
| 213005 |
+
"step": 89320
|
| 213006 |
+
},
|
| 213007 |
+
{
|
| 213008 |
+
"epoch": 713.76,
|
| 213009 |
+
"learning_rate": 8.58511217948718e-06,
|
| 213010 |
+
"loss": 0.4216,
|
| 213011 |
+
"step": 89325
|
| 213012 |
+
},
|
| 213013 |
+
{
|
| 213014 |
+
"epoch": 713.8,
|
| 213015 |
+
"learning_rate": 8.585032051282052e-06,
|
| 213016 |
+
"loss": 1.2307,
|
| 213017 |
+
"step": 89330
|
| 213018 |
+
},
|
| 213019 |
+
{
|
| 213020 |
+
"epoch": 713.84,
|
| 213021 |
+
"learning_rate": 8.584951923076923e-06,
|
| 213022 |
+
"loss": 0.3273,
|
| 213023 |
+
"step": 89335
|
| 213024 |
+
},
|
| 213025 |
+
{
|
| 213026 |
+
"epoch": 713.88,
|
| 213027 |
+
"learning_rate": 8.584871794871795e-06,
|
| 213028 |
+
"loss": 0.2878,
|
| 213029 |
+
"step": 89340
|
| 213030 |
+
},
|
| 213031 |
+
{
|
| 213032 |
+
"epoch": 713.92,
|
| 213033 |
+
"learning_rate": 8.584791666666668e-06,
|
| 213034 |
+
"loss": 0.2904,
|
| 213035 |
+
"step": 89345
|
| 213036 |
+
},
|
| 213037 |
+
{
|
| 213038 |
+
"epoch": 713.96,
|
| 213039 |
+
"learning_rate": 8.58471153846154e-06,
|
| 213040 |
+
"loss": 0.5359,
|
| 213041 |
+
"step": 89350
|
| 213042 |
+
},
|
| 213043 |
+
{
|
| 213044 |
+
"epoch": 714.0,
|
| 213045 |
+
"learning_rate": 8.58463141025641e-06,
|
| 213046 |
+
"loss": 1.3037,
|
| 213047 |
+
"step": 89355
|
| 213048 |
+
},
|
| 213049 |
+
{
|
| 213050 |
+
"epoch": 714.0,
|
| 213051 |
+
"eval_loss": 0.38318589329719543,
|
| 213052 |
+
"eval_runtime": 37.6593,
|
| 213053 |
+
"eval_samples_per_second": 22.332,
|
| 213054 |
+
"eval_steps_per_second": 0.717,
|
| 213055 |
+
"eval_wer": 0.18946828768126556,
|
| 213056 |
+
"step": 89355
|
| 213057 |
+
},
|
| 213058 |
+
{
|
| 213059 |
+
"epoch": 726.04,
|
| 213060 |
+
"learning_rate": 8.584551282051284e-06,
|
| 213061 |
+
"loss": 0.3571,
|
| 213062 |
+
"step": 89360
|
| 213063 |
+
},
|
| 213064 |
+
{
|
| 213065 |
+
"epoch": 726.08,
|
| 213066 |
+
"learning_rate": 8.584471153846155e-06,
|
| 213067 |
+
"loss": 0.3214,
|
| 213068 |
+
"step": 89365
|
| 213069 |
+
},
|
| 213070 |
+
{
|
| 213071 |
+
"epoch": 726.12,
|
| 213072 |
+
"learning_rate": 8.584391025641026e-06,
|
| 213073 |
+
"loss": 0.3781,
|
| 213074 |
+
"step": 89370
|
| 213075 |
+
},
|
| 213076 |
+
{
|
| 213077 |
+
"epoch": 726.16,
|
| 213078 |
+
"learning_rate": 8.584310897435898e-06,
|
| 213079 |
+
"loss": 0.4707,
|
| 213080 |
+
"step": 89375
|
| 213081 |
+
},
|
| 213082 |
+
{
|
| 213083 |
+
"epoch": 726.2,
|
| 213084 |
+
"learning_rate": 8.58423076923077e-06,
|
| 213085 |
+
"loss": 1.139,
|
| 213086 |
+
"step": 89380
|
| 213087 |
+
},
|
| 213088 |
+
{
|
| 213089 |
+
"epoch": 726.24,
|
| 213090 |
+
"learning_rate": 8.584150641025642e-06,
|
| 213091 |
+
"loss": 0.3475,
|
| 213092 |
+
"step": 89385
|
| 213093 |
+
},
|
| 213094 |
+
{
|
| 213095 |
+
"epoch": 726.28,
|
| 213096 |
+
"learning_rate": 8.584070512820513e-06,
|
| 213097 |
+
"loss": 0.3237,
|
| 213098 |
+
"step": 89390
|
| 213099 |
+
},
|
| 213100 |
+
{
|
| 213101 |
+
"epoch": 726.32,
|
| 213102 |
+
"learning_rate": 8.583990384615385e-06,
|
| 213103 |
+
"loss": 0.3205,
|
| 213104 |
+
"step": 89395
|
| 213105 |
+
},
|
| 213106 |
+
{
|
| 213107 |
+
"epoch": 726.36,
|
| 213108 |
+
"learning_rate": 8.583910256410258e-06,
|
| 213109 |
+
"loss": 0.4546,
|
| 213110 |
+
"step": 89400
|
| 213111 |
+
},
|
| 213112 |
+
{
|
| 213113 |
+
"epoch": 726.4,
|
| 213114 |
+
"learning_rate": 8.583830128205128e-06,
|
| 213115 |
+
"loss": 1.338,
|
| 213116 |
+
"step": 89405
|
| 213117 |
+
},
|
| 213118 |
+
{
|
| 213119 |
+
"epoch": 726.45,
|
| 213120 |
+
"learning_rate": 8.58375e-06,
|
| 213121 |
+
"loss": 0.3786,
|
| 213122 |
+
"step": 89410
|
| 213123 |
+
},
|
| 213124 |
+
{
|
| 213125 |
+
"epoch": 726.49,
|
| 213126 |
+
"learning_rate": 8.583669871794874e-06,
|
| 213127 |
+
"loss": 0.2822,
|
| 213128 |
+
"step": 89415
|
| 213129 |
+
},
|
| 213130 |
+
{
|
| 213131 |
+
"epoch": 726.53,
|
| 213132 |
+
"learning_rate": 8.583589743589743e-06,
|
| 213133 |
+
"loss": 0.3013,
|
| 213134 |
+
"step": 89420
|
| 213135 |
+
},
|
| 213136 |
+
{
|
| 213137 |
+
"epoch": 726.57,
|
| 213138 |
+
"learning_rate": 8.583509615384616e-06,
|
| 213139 |
+
"loss": 0.4148,
|
| 213140 |
+
"step": 89425
|
| 213141 |
+
},
|
| 213142 |
+
{
|
| 213143 |
+
"epoch": 726.61,
|
| 213144 |
+
"learning_rate": 8.583429487179488e-06,
|
| 213145 |
+
"loss": 1.1928,
|
| 213146 |
+
"step": 89430
|
| 213147 |
+
},
|
| 213148 |
+
{
|
| 213149 |
+
"epoch": 726.65,
|
| 213150 |
+
"learning_rate": 8.583349358974359e-06,
|
| 213151 |
+
"loss": 0.2964,
|
| 213152 |
+
"step": 89435
|
| 213153 |
+
},
|
| 213154 |
+
{
|
| 213155 |
+
"epoch": 726.69,
|
| 213156 |
+
"learning_rate": 8.58326923076923e-06,
|
| 213157 |
+
"loss": 0.3008,
|
| 213158 |
+
"step": 89440
|
| 213159 |
+
},
|
| 213160 |
+
{
|
| 213161 |
+
"epoch": 726.73,
|
| 213162 |
+
"learning_rate": 8.583189102564104e-06,
|
| 213163 |
+
"loss": 0.3086,
|
| 213164 |
+
"step": 89445
|
| 213165 |
+
},
|
| 213166 |
+
{
|
| 213167 |
+
"epoch": 726.77,
|
| 213168 |
+
"learning_rate": 8.583108974358975e-06,
|
| 213169 |
+
"loss": 0.4667,
|
| 213170 |
+
"step": 89450
|
| 213171 |
+
},
|
| 213172 |
+
{
|
| 213173 |
+
"epoch": 726.81,
|
| 213174 |
+
"learning_rate": 8.583028846153846e-06,
|
| 213175 |
+
"loss": 1.1449,
|
| 213176 |
+
"step": 89455
|
| 213177 |
+
},
|
| 213178 |
+
{
|
| 213179 |
+
"epoch": 726.85,
|
| 213180 |
+
"learning_rate": 8.58294871794872e-06,
|
| 213181 |
+
"loss": 0.3306,
|
| 213182 |
+
"step": 89460
|
| 213183 |
+
},
|
| 213184 |
+
{
|
| 213185 |
+
"epoch": 726.89,
|
| 213186 |
+
"learning_rate": 8.58286858974359e-06,
|
| 213187 |
+
"loss": 0.2821,
|
| 213188 |
+
"step": 89465
|
| 213189 |
+
},
|
| 213190 |
+
{
|
| 213191 |
+
"epoch": 726.93,
|
| 213192 |
+
"learning_rate": 8.582788461538462e-06,
|
| 213193 |
+
"loss": 0.4959,
|
| 213194 |
+
"step": 89470
|
| 213195 |
+
},
|
| 213196 |
+
{
|
| 213197 |
+
"epoch": 726.97,
|
| 213198 |
+
"learning_rate": 8.582708333333333e-06,
|
| 213199 |
+
"loss": 0.522,
|
| 213200 |
+
"step": 89475
|
| 213201 |
+
},
|
| 213202 |
+
{
|
| 213203 |
+
"epoch": 727.0,
|
| 213204 |
+
"eval_loss": 0.37962403893470764,
|
| 213205 |
+
"eval_runtime": 37.8556,
|
| 213206 |
+
"eval_samples_per_second": 22.216,
|
| 213207 |
+
"eval_steps_per_second": 0.713,
|
| 213208 |
+
"eval_wer": 0.18697681309726646,
|
| 213209 |
+
"step": 89478
|
| 213210 |
+
},
|
| 213211 |
+
{
|
| 213212 |
+
"epoch": 715.02,
|
| 213213 |
+
"learning_rate": 8.582628205128206e-06,
|
| 213214 |
+
"loss": 0.3868,
|
| 213215 |
+
"step": 89480
|
| 213216 |
+
},
|
| 213217 |
+
{
|
| 213218 |
+
"epoch": 715.06,
|
| 213219 |
+
"learning_rate": 8.582548076923078e-06,
|
| 213220 |
+
"loss": 0.3219,
|
| 213221 |
+
"step": 89485
|
| 213222 |
+
},
|
| 213223 |
+
{
|
| 213224 |
+
"epoch": 715.1,
|
| 213225 |
+
"learning_rate": 8.582467948717949e-06,
|
| 213226 |
+
"loss": 0.2923,
|
| 213227 |
+
"step": 89490
|
| 213228 |
+
},
|
| 213229 |
+
{
|
| 213230 |
+
"epoch": 715.14,
|
| 213231 |
+
"learning_rate": 8.58238782051282e-06,
|
| 213232 |
+
"loss": 0.3483,
|
| 213233 |
+
"step": 89495
|
| 213234 |
+
},
|
| 213235 |
+
{
|
| 213236 |
+
"epoch": 715.18,
|
| 213237 |
+
"learning_rate": 8.582307692307694e-06,
|
| 213238 |
+
"loss": 0.6396,
|
| 213239 |
+
"step": 89500
|
| 213240 |
+
},
|
| 213241 |
+
{
|
| 213242 |
+
"epoch": 715.22,
|
| 213243 |
+
"learning_rate": 8.582227564102565e-06,
|
| 213244 |
+
"loss": 0.9591,
|
| 213245 |
+
"step": 89505
|
| 213246 |
+
},
|
| 213247 |
+
{
|
| 213248 |
+
"epoch": 715.26,
|
| 213249 |
+
"learning_rate": 8.582147435897436e-06,
|
| 213250 |
+
"loss": 0.2791,
|
| 213251 |
+
"step": 89510
|
| 213252 |
+
},
|
| 213253 |
+
{
|
| 213254 |
+
"epoch": 715.3,
|
| 213255 |
+
"learning_rate": 8.58206730769231e-06,
|
| 213256 |
+
"loss": 0.3353,
|
| 213257 |
+
"step": 89515
|
| 213258 |
+
},
|
| 213259 |
+
{
|
| 213260 |
+
"epoch": 715.34,
|
| 213261 |
+
"learning_rate": 8.58198717948718e-06,
|
| 213262 |
+
"loss": 0.4082,
|
| 213263 |
+
"step": 89520
|
| 213264 |
+
},
|
| 213265 |
+
{
|
| 213266 |
+
"epoch": 715.38,
|
| 213267 |
+
"learning_rate": 8.581907051282052e-06,
|
| 213268 |
+
"loss": 0.6496,
|
| 213269 |
+
"step": 89525
|
| 213270 |
+
},
|
| 213271 |
+
{
|
| 213272 |
+
"epoch": 715.42,
|
| 213273 |
+
"learning_rate": 8.581826923076923e-06,
|
| 213274 |
+
"loss": 1.0216,
|
| 213275 |
+
"step": 89530
|
| 213276 |
+
},
|
| 213277 |
+
{
|
| 213278 |
+
"epoch": 715.46,
|
| 213279 |
+
"learning_rate": 8.581746794871796e-06,
|
| 213280 |
+
"loss": 0.2943,
|
| 213281 |
+
"step": 89535
|
| 213282 |
+
},
|
| 213283 |
+
{
|
| 213284 |
+
"epoch": 715.5,
|
| 213285 |
+
"learning_rate": 8.581666666666668e-06,
|
| 213286 |
+
"loss": 0.2755,
|
| 213287 |
+
"step": 89540
|
| 213288 |
+
},
|
| 213289 |
+
{
|
| 213290 |
+
"epoch": 715.54,
|
| 213291 |
+
"learning_rate": 8.581586538461539e-06,
|
| 213292 |
+
"loss": 0.3416,
|
| 213293 |
+
"step": 89545
|
| 213294 |
+
},
|
| 213295 |
+
{
|
| 213296 |
+
"epoch": 715.58,
|
| 213297 |
+
"learning_rate": 8.58150641025641e-06,
|
| 213298 |
+
"loss": 0.618,
|
| 213299 |
+
"step": 89550
|
| 213300 |
+
},
|
| 213301 |
+
{
|
| 213302 |
+
"epoch": 715.62,
|
| 213303 |
+
"learning_rate": 8.581426282051284e-06,
|
| 213304 |
+
"loss": 1.0209,
|
| 213305 |
+
"step": 89555
|
| 213306 |
+
},
|
| 213307 |
+
{
|
| 213308 |
+
"epoch": 715.66,
|
| 213309 |
+
"learning_rate": 8.581346153846155e-06,
|
| 213310 |
+
"loss": 0.2874,
|
| 213311 |
+
"step": 89560
|
| 213312 |
+
},
|
| 213313 |
+
{
|
| 213314 |
+
"epoch": 715.7,
|
| 213315 |
+
"learning_rate": 8.581266025641026e-06,
|
| 213316 |
+
"loss": 0.2984,
|
| 213317 |
+
"step": 89565
|
| 213318 |
+
},
|
| 213319 |
+
{
|
| 213320 |
+
"epoch": 715.74,
|
| 213321 |
+
"learning_rate": 8.5811858974359e-06,
|
| 213322 |
+
"loss": 0.3482,
|
| 213323 |
+
"step": 89570
|
| 213324 |
+
},
|
| 213325 |
+
{
|
| 213326 |
+
"epoch": 715.78,
|
| 213327 |
+
"learning_rate": 8.581105769230769e-06,
|
| 213328 |
+
"loss": 0.6483,
|
| 213329 |
+
"step": 89575
|
| 213330 |
+
},
|
| 213331 |
+
{
|
| 213332 |
+
"epoch": 715.82,
|
| 213333 |
+
"learning_rate": 8.581025641025642e-06,
|
| 213334 |
+
"loss": 1.0368,
|
| 213335 |
+
"step": 89580
|
| 213336 |
+
},
|
| 213337 |
+
{
|
| 213338 |
+
"epoch": 715.86,
|
| 213339 |
+
"learning_rate": 8.580945512820513e-06,
|
| 213340 |
+
"loss": 0.2969,
|
| 213341 |
+
"step": 89585
|
| 213342 |
+
},
|
| 213343 |
+
{
|
| 213344 |
+
"epoch": 715.9,
|
| 213345 |
+
"learning_rate": 8.580865384615385e-06,
|
| 213346 |
+
"loss": 0.3009,
|
| 213347 |
+
"step": 89590
|
| 213348 |
+
},
|
| 213349 |
+
{
|
| 213350 |
+
"epoch": 715.94,
|
| 213351 |
+
"learning_rate": 8.580785256410256e-06,
|
| 213352 |
+
"loss": 0.3862,
|
| 213353 |
+
"step": 89595
|
| 213354 |
+
},
|
| 213355 |
+
{
|
| 213356 |
+
"epoch": 715.98,
|
| 213357 |
+
"learning_rate": 8.58070512820513e-06,
|
| 213358 |
+
"loss": 0.6047,
|
| 213359 |
+
"step": 89600
|
| 213360 |
+
},
|
| 213361 |
+
{
|
| 213362 |
+
"epoch": 716.0,
|
| 213363 |
+
"eval_loss": 0.42192313075065613,
|
| 213364 |
+
"eval_runtime": 38.1277,
|
| 213365 |
+
"eval_samples_per_second": 22.084,
|
| 213366 |
+
"eval_steps_per_second": 0.708,
|
| 213367 |
+
"eval_wer": 0.18980540226546616,
|
| 213368 |
+
"step": 89603
|
| 213369 |
}
|
| 213370 |
],
|
| 213371 |
+
"max_steps": 625000,
|
| 213372 |
"num_train_epochs": 5000,
|
| 213373 |
+
"total_flos": 2.521538624421044e+20,
|
| 213374 |
"trial_name": null,
|
| 213375 |
"trial_params": null
|
| 213376 |
}
|
model-bin/finetune/base/{checkpoint-88981 → checkpoint-89603}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629941989.4511316/events.out.tfevents.1629941989.8e89bd551565.924.11
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7cfe52ce5123f221d5430ba4bcc21c0098e5e3ec4b6a2fca4e799fddda0c3fb8
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629942431.5525227/events.out.tfevents.1629942431.8e89bd551565.924.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ec92524f1117d47d6597750660ed02d2202182c1c8a7cecc2507723ce6f435f9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629942858.3612747/events.out.tfevents.1629942858.8e89bd551565.924.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a6e16c69a0b188bb907d6b765efb936fd01ad8b86c55782750896e7eb44b1f9e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629943284.2321923/events.out.tfevents.1629943284.8e89bd551565.924.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df7d94a62b0805c1896bb2704d09db0ab4a2e6177763c3befc9f0f59d1494743
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629943702.5924923/events.out.tfevents.1629943702.8e89bd551565.924.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a2f80f84df64b7af186cd8a8c7a1d7c4baa7bb66c57c8305efcb236d680881c4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629941989.8e89bd551565.924.10
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5da17eb9052a1bd3e736635e8a2e22936429b3faedc65f6050d33e34ec409525
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629942431.8e89bd551565.924.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d68afe14b9644be97f1bc2680ae0e318ef494c029bfa45de4bfd04d2fdcb55af
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629942858.8e89bd551565.924.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a95517ca0d559b2325de22f8b61ff900de7468f44bb80d0aa6f95a0c64f18bf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629943284.8e89bd551565.924.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2de3335406e8b08ee93085cff73d46c08ea4ff85a8a6d90a7a306600ff6f7810
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629943702.8e89bd551565.924.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:774bc1c6320128780f1d19b5ee725bb98c22e1270909a68ae4558c5a3839af2e
|
| 3 |
+
size 8622
|