"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629719560.684198/events.out.tfevents.1629719560.74272264b15c.932.123 +3 -0
- model-bin/finetune/base/log/1629720325.471121/events.out.tfevents.1629720325.74272264b15c.932.125 +3 -0
- model-bin/finetune/base/log/1629720980.0787914/events.out.tfevents.1629720980.74272264b15c.932.127 +3 -0
- model-bin/finetune/base/log/1629721622.799041/events.out.tfevents.1629721622.74272264b15c.932.129 +3 -0
- model-bin/finetune/base/log/1629722285.2883925/events.out.tfevents.1629722285.74272264b15c.932.131 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629719560.74272264b15c.932.122 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629720325.74272264b15c.932.124 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629720980.74272264b15c.932.126 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629721622.74272264b15c.932.128 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629722285.74272264b15c.932.130 +3 -0
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8841af70c3b5a37c5f0208a2993ba7e106422854c421c49a6ea08acdbb1bc9f6
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/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 sha256:8e5a540e970bfd0918840a392ab6a5b9c69f61260253aa6e87e3ca76a87752ab
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/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 sha256:050a4bd100c9e7d99a9c4c7cbf65aebe9747344205c1f7a1e46682ba93e397e3
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/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:dc92ae4af93d2420b4a5cc8088763dea6bc27db9efdd8427b6879f8c565a4cbc
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/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 sha256:9c8d1fe594378115fc3c20cbdbad9b8d73134ec6543bce3ba9fc67e6a4447f18
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -157632,11 +157632,800 @@
|
|
| 157632 |
"eval_steps_per_second": 0.707,
|
| 157633 |
"eval_wer": 0.21347227391142537,
|
| 157634 |
"step": 45796
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 157635 |
}
|
| 157636 |
],
|
| 157637 |
"max_steps": 625000,
|
| 157638 |
"num_train_epochs": 5000,
|
| 157639 |
-
"total_flos": 1.
|
| 157640 |
"trial_name": null,
|
| 157641 |
"trial_params": null
|
| 157642 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.2003221555132523,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-42807",
|
| 4 |
+
"epoch": 371.0,
|
| 5 |
+
"global_step": 46418,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 157632 |
"eval_steps_per_second": 0.707,
|
| 157633 |
"eval_wer": 0.21347227391142537,
|
| 157634 |
"step": 45796
|
| 157635 |
+
},
|
| 157636 |
+
{
|
| 157637 |
+
"epoch": 366.03,
|
| 157638 |
+
"learning_rate": 9.28232371794872e-06,
|
| 157639 |
+
"loss": 0.4708,
|
| 157640 |
+
"step": 45800
|
| 157641 |
+
},
|
| 157642 |
+
{
|
| 157643 |
+
"epoch": 366.07,
|
| 157644 |
+
"learning_rate": 9.282243589743591e-06,
|
| 157645 |
+
"loss": 0.346,
|
| 157646 |
+
"step": 45805
|
| 157647 |
+
},
|
| 157648 |
+
{
|
| 157649 |
+
"epoch": 366.11,
|
| 157650 |
+
"learning_rate": 9.282163461538462e-06,
|
| 157651 |
+
"loss": 0.4004,
|
| 157652 |
+
"step": 45810
|
| 157653 |
+
},
|
| 157654 |
+
{
|
| 157655 |
+
"epoch": 366.15,
|
| 157656 |
+
"learning_rate": 9.282083333333334e-06,
|
| 157657 |
+
"loss": 0.4102,
|
| 157658 |
+
"step": 45815
|
| 157659 |
+
},
|
| 157660 |
+
{
|
| 157661 |
+
"epoch": 366.19,
|
| 157662 |
+
"learning_rate": 9.282003205128207e-06,
|
| 157663 |
+
"loss": 1.0689,
|
| 157664 |
+
"step": 45820
|
| 157665 |
+
},
|
| 157666 |
+
{
|
| 157667 |
+
"epoch": 366.23,
|
| 157668 |
+
"learning_rate": 9.281923076923076e-06,
|
| 157669 |
+
"loss": 0.7857,
|
| 157670 |
+
"step": 45825
|
| 157671 |
+
},
|
| 157672 |
+
{
|
| 157673 |
+
"epoch": 366.27,
|
| 157674 |
+
"learning_rate": 9.28184294871795e-06,
|
| 157675 |
+
"loss": 0.3702,
|
| 157676 |
+
"step": 45830
|
| 157677 |
+
},
|
| 157678 |
+
{
|
| 157679 |
+
"epoch": 366.31,
|
| 157680 |
+
"learning_rate": 9.281762820512821e-06,
|
| 157681 |
+
"loss": 0.36,
|
| 157682 |
+
"step": 45835
|
| 157683 |
+
},
|
| 157684 |
+
{
|
| 157685 |
+
"epoch": 366.35,
|
| 157686 |
+
"learning_rate": 9.281682692307692e-06,
|
| 157687 |
+
"loss": 0.4491,
|
| 157688 |
+
"step": 45840
|
| 157689 |
+
},
|
| 157690 |
+
{
|
| 157691 |
+
"epoch": 366.39,
|
| 157692 |
+
"learning_rate": 9.281602564102565e-06,
|
| 157693 |
+
"loss": 1.0663,
|
| 157694 |
+
"step": 45845
|
| 157695 |
+
},
|
| 157696 |
+
{
|
| 157697 |
+
"epoch": 366.43,
|
| 157698 |
+
"learning_rate": 9.281522435897437e-06,
|
| 157699 |
+
"loss": 0.7977,
|
| 157700 |
+
"step": 45850
|
| 157701 |
+
},
|
| 157702 |
+
{
|
| 157703 |
+
"epoch": 366.47,
|
| 157704 |
+
"learning_rate": 9.281442307692308e-06,
|
| 157705 |
+
"loss": 0.3708,
|
| 157706 |
+
"step": 45855
|
| 157707 |
+
},
|
| 157708 |
+
{
|
| 157709 |
+
"epoch": 366.51,
|
| 157710 |
+
"learning_rate": 9.28136217948718e-06,
|
| 157711 |
+
"loss": 0.3979,
|
| 157712 |
+
"step": 45860
|
| 157713 |
+
},
|
| 157714 |
+
{
|
| 157715 |
+
"epoch": 366.55,
|
| 157716 |
+
"learning_rate": 9.281282051282052e-06,
|
| 157717 |
+
"loss": 0.4949,
|
| 157718 |
+
"step": 45865
|
| 157719 |
+
},
|
| 157720 |
+
{
|
| 157721 |
+
"epoch": 366.59,
|
| 157722 |
+
"learning_rate": 9.281201923076924e-06,
|
| 157723 |
+
"loss": 0.9667,
|
| 157724 |
+
"step": 45870
|
| 157725 |
+
},
|
| 157726 |
+
{
|
| 157727 |
+
"epoch": 366.63,
|
| 157728 |
+
"learning_rate": 9.281121794871795e-06,
|
| 157729 |
+
"loss": 0.8203,
|
| 157730 |
+
"step": 45875
|
| 157731 |
+
},
|
| 157732 |
+
{
|
| 157733 |
+
"epoch": 366.67,
|
| 157734 |
+
"learning_rate": 9.281041666666666e-06,
|
| 157735 |
+
"loss": 0.3286,
|
| 157736 |
+
"step": 45880
|
| 157737 |
+
},
|
| 157738 |
+
{
|
| 157739 |
+
"epoch": 366.71,
|
| 157740 |
+
"learning_rate": 9.28096153846154e-06,
|
| 157741 |
+
"loss": 0.4312,
|
| 157742 |
+
"step": 45885
|
| 157743 |
+
},
|
| 157744 |
+
{
|
| 157745 |
+
"epoch": 366.75,
|
| 157746 |
+
"learning_rate": 9.280881410256411e-06,
|
| 157747 |
+
"loss": 0.483,
|
| 157748 |
+
"step": 45890
|
| 157749 |
+
},
|
| 157750 |
+
{
|
| 157751 |
+
"epoch": 366.79,
|
| 157752 |
+
"learning_rate": 9.280801282051282e-06,
|
| 157753 |
+
"loss": 1.1711,
|
| 157754 |
+
"step": 45895
|
| 157755 |
+
},
|
| 157756 |
+
{
|
| 157757 |
+
"epoch": 366.83,
|
| 157758 |
+
"learning_rate": 9.280721153846155e-06,
|
| 157759 |
+
"loss": 0.8138,
|
| 157760 |
+
"step": 45900
|
| 157761 |
+
},
|
| 157762 |
+
{
|
| 157763 |
+
"epoch": 366.87,
|
| 157764 |
+
"learning_rate": 9.280641025641027e-06,
|
| 157765 |
+
"loss": 0.3247,
|
| 157766 |
+
"step": 45905
|
| 157767 |
+
},
|
| 157768 |
+
{
|
| 157769 |
+
"epoch": 366.91,
|
| 157770 |
+
"learning_rate": 9.280560897435898e-06,
|
| 157771 |
+
"loss": 0.3805,
|
| 157772 |
+
"step": 45910
|
| 157773 |
+
},
|
| 157774 |
+
{
|
| 157775 |
+
"epoch": 366.95,
|
| 157776 |
+
"learning_rate": 9.28048076923077e-06,
|
| 157777 |
+
"loss": 0.4875,
|
| 157778 |
+
"step": 45915
|
| 157779 |
+
},
|
| 157780 |
+
{
|
| 157781 |
+
"epoch": 366.99,
|
| 157782 |
+
"learning_rate": 9.280400641025642e-06,
|
| 157783 |
+
"loss": 1.3315,
|
| 157784 |
+
"step": 45920
|
| 157785 |
+
},
|
| 157786 |
+
{
|
| 157787 |
+
"epoch": 367.0,
|
| 157788 |
+
"eval_loss": 0.42447608709335327,
|
| 157789 |
+
"eval_runtime": 40.8992,
|
| 157790 |
+
"eval_samples_per_second": 20.636,
|
| 157791 |
+
"eval_steps_per_second": 0.66,
|
| 157792 |
+
"eval_wer": 0.2110683064157653,
|
| 157793 |
+
"step": 45921
|
| 157794 |
+
},
|
| 157795 |
+
{
|
| 157796 |
+
"epoch": 370.03,
|
| 157797 |
+
"learning_rate": 9.280320512820514e-06,
|
| 157798 |
+
"loss": 0.5274,
|
| 157799 |
+
"step": 45925
|
| 157800 |
+
},
|
| 157801 |
+
{
|
| 157802 |
+
"epoch": 370.07,
|
| 157803 |
+
"learning_rate": 9.280240384615385e-06,
|
| 157804 |
+
"loss": 0.3571,
|
| 157805 |
+
"step": 45930
|
| 157806 |
+
},
|
| 157807 |
+
{
|
| 157808 |
+
"epoch": 370.11,
|
| 157809 |
+
"learning_rate": 9.280160256410257e-06,
|
| 157810 |
+
"loss": 0.3882,
|
| 157811 |
+
"step": 45935
|
| 157812 |
+
},
|
| 157813 |
+
{
|
| 157814 |
+
"epoch": 370.15,
|
| 157815 |
+
"learning_rate": 9.28008012820513e-06,
|
| 157816 |
+
"loss": 0.5463,
|
| 157817 |
+
"step": 45940
|
| 157818 |
+
},
|
| 157819 |
+
{
|
| 157820 |
+
"epoch": 370.19,
|
| 157821 |
+
"learning_rate": 9.280000000000001e-06,
|
| 157822 |
+
"loss": 1.0035,
|
| 157823 |
+
"step": 45945
|
| 157824 |
+
},
|
| 157825 |
+
{
|
| 157826 |
+
"epoch": 370.23,
|
| 157827 |
+
"learning_rate": 9.279919871794872e-06,
|
| 157828 |
+
"loss": 0.7083,
|
| 157829 |
+
"step": 45950
|
| 157830 |
+
},
|
| 157831 |
+
{
|
| 157832 |
+
"epoch": 370.27,
|
| 157833 |
+
"learning_rate": 9.279839743589745e-06,
|
| 157834 |
+
"loss": 0.315,
|
| 157835 |
+
"step": 45955
|
| 157836 |
+
},
|
| 157837 |
+
{
|
| 157838 |
+
"epoch": 370.31,
|
| 157839 |
+
"learning_rate": 9.279759615384617e-06,
|
| 157840 |
+
"loss": 0.4206,
|
| 157841 |
+
"step": 45960
|
| 157842 |
+
},
|
| 157843 |
+
{
|
| 157844 |
+
"epoch": 370.35,
|
| 157845 |
+
"learning_rate": 9.279679487179488e-06,
|
| 157846 |
+
"loss": 0.4858,
|
| 157847 |
+
"step": 45965
|
| 157848 |
+
},
|
| 157849 |
+
{
|
| 157850 |
+
"epoch": 370.4,
|
| 157851 |
+
"learning_rate": 9.27959935897436e-06,
|
| 157852 |
+
"loss": 0.9684,
|
| 157853 |
+
"step": 45970
|
| 157854 |
+
},
|
| 157855 |
+
{
|
| 157856 |
+
"epoch": 370.44,
|
| 157857 |
+
"learning_rate": 9.279519230769232e-06,
|
| 157858 |
+
"loss": 0.6774,
|
| 157859 |
+
"step": 45975
|
| 157860 |
+
},
|
| 157861 |
+
{
|
| 157862 |
+
"epoch": 370.48,
|
| 157863 |
+
"learning_rate": 9.279439102564102e-06,
|
| 157864 |
+
"loss": 0.3357,
|
| 157865 |
+
"step": 45980
|
| 157866 |
+
},
|
| 157867 |
+
{
|
| 157868 |
+
"epoch": 370.52,
|
| 157869 |
+
"learning_rate": 9.279358974358975e-06,
|
| 157870 |
+
"loss": 0.3506,
|
| 157871 |
+
"step": 45985
|
| 157872 |
+
},
|
| 157873 |
+
{
|
| 157874 |
+
"epoch": 370.56,
|
| 157875 |
+
"learning_rate": 9.279278846153848e-06,
|
| 157876 |
+
"loss": 0.4169,
|
| 157877 |
+
"step": 45990
|
| 157878 |
+
},
|
| 157879 |
+
{
|
| 157880 |
+
"epoch": 370.6,
|
| 157881 |
+
"learning_rate": 9.279198717948718e-06,
|
| 157882 |
+
"loss": 0.9841,
|
| 157883 |
+
"step": 45995
|
| 157884 |
+
},
|
| 157885 |
+
{
|
| 157886 |
+
"epoch": 370.64,
|
| 157887 |
+
"learning_rate": 9.279118589743591e-06,
|
| 157888 |
+
"loss": 0.7478,
|
| 157889 |
+
"step": 46000
|
| 157890 |
+
},
|
| 157891 |
+
{
|
| 157892 |
+
"epoch": 370.68,
|
| 157893 |
+
"learning_rate": 9.279038461538462e-06,
|
| 157894 |
+
"loss": 0.3052,
|
| 157895 |
+
"step": 46005
|
| 157896 |
+
},
|
| 157897 |
+
{
|
| 157898 |
+
"epoch": 370.72,
|
| 157899 |
+
"learning_rate": 9.278958333333334e-06,
|
| 157900 |
+
"loss": 0.4429,
|
| 157901 |
+
"step": 46010
|
| 157902 |
+
},
|
| 157903 |
+
{
|
| 157904 |
+
"epoch": 370.76,
|
| 157905 |
+
"learning_rate": 9.278878205128205e-06,
|
| 157906 |
+
"loss": 0.5115,
|
| 157907 |
+
"step": 46015
|
| 157908 |
+
},
|
| 157909 |
+
{
|
| 157910 |
+
"epoch": 370.8,
|
| 157911 |
+
"learning_rate": 9.278798076923078e-06,
|
| 157912 |
+
"loss": 1.0209,
|
| 157913 |
+
"step": 46020
|
| 157914 |
+
},
|
| 157915 |
+
{
|
| 157916 |
+
"epoch": 370.84,
|
| 157917 |
+
"learning_rate": 9.27871794871795e-06,
|
| 157918 |
+
"loss": 0.7911,
|
| 157919 |
+
"step": 46025
|
| 157920 |
+
},
|
| 157921 |
+
{
|
| 157922 |
+
"epoch": 370.88,
|
| 157923 |
+
"learning_rate": 9.27863782051282e-06,
|
| 157924 |
+
"loss": 0.3474,
|
| 157925 |
+
"step": 46030
|
| 157926 |
+
},
|
| 157927 |
+
{
|
| 157928 |
+
"epoch": 370.92,
|
| 157929 |
+
"learning_rate": 9.278557692307692e-06,
|
| 157930 |
+
"loss": 0.41,
|
| 157931 |
+
"step": 46035
|
| 157932 |
+
},
|
| 157933 |
+
{
|
| 157934 |
+
"epoch": 370.96,
|
| 157935 |
+
"learning_rate": 9.278477564102565e-06,
|
| 157936 |
+
"loss": 0.5987,
|
| 157937 |
+
"step": 46040
|
| 157938 |
+
},
|
| 157939 |
+
{
|
| 157940 |
+
"epoch": 371.0,
|
| 157941 |
+
"learning_rate": 9.278397435897437e-06,
|
| 157942 |
+
"loss": 1.3948,
|
| 157943 |
+
"step": 46045
|
| 157944 |
+
},
|
| 157945 |
+
{
|
| 157946 |
+
"epoch": 371.0,
|
| 157947 |
+
"eval_loss": 0.48161518573760986,
|
| 157948 |
+
"eval_runtime": 38.0151,
|
| 157949 |
+
"eval_samples_per_second": 21.965,
|
| 157950 |
+
"eval_steps_per_second": 0.71,
|
| 157951 |
+
"eval_wer": 0.2056692242114237,
|
| 157952 |
+
"step": 46045
|
| 157953 |
+
},
|
| 157954 |
+
{
|
| 157955 |
+
"epoch": 371.04,
|
| 157956 |
+
"learning_rate": 9.278317307692308e-06,
|
| 157957 |
+
"loss": 0.3475,
|
| 157958 |
+
"step": 46050
|
| 157959 |
+
},
|
| 157960 |
+
{
|
| 157961 |
+
"epoch": 371.08,
|
| 157962 |
+
"learning_rate": 9.278237179487181e-06,
|
| 157963 |
+
"loss": 0.3976,
|
| 157964 |
+
"step": 46055
|
| 157965 |
+
},
|
| 157966 |
+
{
|
| 157967 |
+
"epoch": 371.12,
|
| 157968 |
+
"learning_rate": 9.278157051282052e-06,
|
| 157969 |
+
"loss": 0.3443,
|
| 157970 |
+
"step": 46060
|
| 157971 |
+
},
|
| 157972 |
+
{
|
| 157973 |
+
"epoch": 371.16,
|
| 157974 |
+
"learning_rate": 9.278076923076924e-06,
|
| 157975 |
+
"loss": 0.5226,
|
| 157976 |
+
"step": 46065
|
| 157977 |
+
},
|
| 157978 |
+
{
|
| 157979 |
+
"epoch": 371.2,
|
| 157980 |
+
"learning_rate": 9.277996794871795e-06,
|
| 157981 |
+
"loss": 1.4301,
|
| 157982 |
+
"step": 46070
|
| 157983 |
+
},
|
| 157984 |
+
{
|
| 157985 |
+
"epoch": 371.24,
|
| 157986 |
+
"learning_rate": 9.277916666666668e-06,
|
| 157987 |
+
"loss": 0.3512,
|
| 157988 |
+
"step": 46075
|
| 157989 |
+
},
|
| 157990 |
+
{
|
| 157991 |
+
"epoch": 371.28,
|
| 157992 |
+
"learning_rate": 9.27783653846154e-06,
|
| 157993 |
+
"loss": 0.3918,
|
| 157994 |
+
"step": 46080
|
| 157995 |
+
},
|
| 157996 |
+
{
|
| 157997 |
+
"epoch": 371.32,
|
| 157998 |
+
"learning_rate": 9.27775641025641e-06,
|
| 157999 |
+
"loss": 0.3949,
|
| 158000 |
+
"step": 46085
|
| 158001 |
+
},
|
| 158002 |
+
{
|
| 158003 |
+
"epoch": 371.36,
|
| 158004 |
+
"learning_rate": 9.277676282051284e-06,
|
| 158005 |
+
"loss": 0.5323,
|
| 158006 |
+
"step": 46090
|
| 158007 |
+
},
|
| 158008 |
+
{
|
| 158009 |
+
"epoch": 371.4,
|
| 158010 |
+
"learning_rate": 9.277596153846155e-06,
|
| 158011 |
+
"loss": 1.2795,
|
| 158012 |
+
"step": 46095
|
| 158013 |
+
},
|
| 158014 |
+
{
|
| 158015 |
+
"epoch": 371.44,
|
| 158016 |
+
"learning_rate": 9.277516025641027e-06,
|
| 158017 |
+
"loss": 0.5222,
|
| 158018 |
+
"step": 46100
|
| 158019 |
+
},
|
| 158020 |
+
{
|
| 158021 |
+
"epoch": 371.48,
|
| 158022 |
+
"learning_rate": 9.277435897435898e-06,
|
| 158023 |
+
"loss": 0.3962,
|
| 158024 |
+
"step": 46105
|
| 158025 |
+
},
|
| 158026 |
+
{
|
| 158027 |
+
"epoch": 371.52,
|
| 158028 |
+
"learning_rate": 9.277355769230771e-06,
|
| 158029 |
+
"loss": 0.3715,
|
| 158030 |
+
"step": 46110
|
| 158031 |
+
},
|
| 158032 |
+
{
|
| 158033 |
+
"epoch": 371.56,
|
| 158034 |
+
"learning_rate": 9.27727564102564e-06,
|
| 158035 |
+
"loss": 0.5905,
|
| 158036 |
+
"step": 46115
|
| 158037 |
+
},
|
| 158038 |
+
{
|
| 158039 |
+
"epoch": 371.6,
|
| 158040 |
+
"learning_rate": 9.277195512820514e-06,
|
| 158041 |
+
"loss": 1.2249,
|
| 158042 |
+
"step": 46120
|
| 158043 |
+
},
|
| 158044 |
+
{
|
| 158045 |
+
"epoch": 371.65,
|
| 158046 |
+
"learning_rate": 9.277115384615385e-06,
|
| 158047 |
+
"loss": 0.3703,
|
| 158048 |
+
"step": 46125
|
| 158049 |
+
},
|
| 158050 |
+
{
|
| 158051 |
+
"epoch": 371.69,
|
| 158052 |
+
"learning_rate": 9.277035256410256e-06,
|
| 158053 |
+
"loss": 0.4298,
|
| 158054 |
+
"step": 46130
|
| 158055 |
+
},
|
| 158056 |
+
{
|
| 158057 |
+
"epoch": 371.73,
|
| 158058 |
+
"learning_rate": 9.276955128205128e-06,
|
| 158059 |
+
"loss": 0.4247,
|
| 158060 |
+
"step": 46135
|
| 158061 |
+
},
|
| 158062 |
+
{
|
| 158063 |
+
"epoch": 371.77,
|
| 158064 |
+
"learning_rate": 9.276875000000001e-06,
|
| 158065 |
+
"loss": 0.4587,
|
| 158066 |
+
"step": 46140
|
| 158067 |
+
},
|
| 158068 |
+
{
|
| 158069 |
+
"epoch": 371.81,
|
| 158070 |
+
"learning_rate": 9.276794871794872e-06,
|
| 158071 |
+
"loss": 1.3956,
|
| 158072 |
+
"step": 46145
|
| 158073 |
+
},
|
| 158074 |
+
{
|
| 158075 |
+
"epoch": 371.85,
|
| 158076 |
+
"learning_rate": 9.276714743589744e-06,
|
| 158077 |
+
"loss": 0.3604,
|
| 158078 |
+
"step": 46150
|
| 158079 |
+
},
|
| 158080 |
+
{
|
| 158081 |
+
"epoch": 371.89,
|
| 158082 |
+
"learning_rate": 9.276634615384617e-06,
|
| 158083 |
+
"loss": 0.4535,
|
| 158084 |
+
"step": 46155
|
| 158085 |
+
},
|
| 158086 |
+
{
|
| 158087 |
+
"epoch": 371.93,
|
| 158088 |
+
"learning_rate": 9.276554487179488e-06,
|
| 158089 |
+
"loss": 0.4185,
|
| 158090 |
+
"step": 46160
|
| 158091 |
+
},
|
| 158092 |
+
{
|
| 158093 |
+
"epoch": 371.97,
|
| 158094 |
+
"learning_rate": 9.27647435897436e-06,
|
| 158095 |
+
"loss": 0.585,
|
| 158096 |
+
"step": 46165
|
| 158097 |
+
},
|
| 158098 |
+
{
|
| 158099 |
+
"epoch": 372.0,
|
| 158100 |
+
"eval_loss": 0.4676620364189148,
|
| 158101 |
+
"eval_runtime": 39.9801,
|
| 158102 |
+
"eval_samples_per_second": 20.91,
|
| 158103 |
+
"eval_steps_per_second": 0.675,
|
| 158104 |
+
"eval_wer": 0.20456064770110674,
|
| 158105 |
+
"step": 46169
|
| 158106 |
+
},
|
| 158107 |
+
{
|
| 158108 |
+
"epoch": 372.01,
|
| 158109 |
+
"learning_rate": 9.27639423076923e-06,
|
| 158110 |
+
"loss": 0.4551,
|
| 158111 |
+
"step": 46170
|
| 158112 |
+
},
|
| 158113 |
+
{
|
| 158114 |
+
"epoch": 372.05,
|
| 158115 |
+
"learning_rate": 9.276314102564104e-06,
|
| 158116 |
+
"loss": 0.4301,
|
| 158117 |
+
"step": 46175
|
| 158118 |
+
},
|
| 158119 |
+
{
|
| 158120 |
+
"epoch": 372.09,
|
| 158121 |
+
"learning_rate": 9.276233974358975e-06,
|
| 158122 |
+
"loss": 0.3309,
|
| 158123 |
+
"step": 46180
|
| 158124 |
+
},
|
| 158125 |
+
{
|
| 158126 |
+
"epoch": 372.13,
|
| 158127 |
+
"learning_rate": 9.276153846153846e-06,
|
| 158128 |
+
"loss": 0.41,
|
| 158129 |
+
"step": 46185
|
| 158130 |
+
},
|
| 158131 |
+
{
|
| 158132 |
+
"epoch": 372.17,
|
| 158133 |
+
"learning_rate": 9.27607371794872e-06,
|
| 158134 |
+
"loss": 0.651,
|
| 158135 |
+
"step": 46190
|
| 158136 |
+
},
|
| 158137 |
+
{
|
| 158138 |
+
"epoch": 372.21,
|
| 158139 |
+
"learning_rate": 9.275993589743591e-06,
|
| 158140 |
+
"loss": 1.2111,
|
| 158141 |
+
"step": 46195
|
| 158142 |
+
},
|
| 158143 |
+
{
|
| 158144 |
+
"epoch": 372.25,
|
| 158145 |
+
"learning_rate": 9.275913461538462e-06,
|
| 158146 |
+
"loss": 0.345,
|
| 158147 |
+
"step": 46200
|
| 158148 |
+
},
|
| 158149 |
+
{
|
| 158150 |
+
"epoch": 372.29,
|
| 158151 |
+
"learning_rate": 9.275833333333334e-06,
|
| 158152 |
+
"loss": 0.3448,
|
| 158153 |
+
"step": 46205
|
| 158154 |
+
},
|
| 158155 |
+
{
|
| 158156 |
+
"epoch": 372.33,
|
| 158157 |
+
"learning_rate": 9.275753205128207e-06,
|
| 158158 |
+
"loss": 0.3657,
|
| 158159 |
+
"step": 46210
|
| 158160 |
+
},
|
| 158161 |
+
{
|
| 158162 |
+
"epoch": 372.37,
|
| 158163 |
+
"learning_rate": 9.275673076923078e-06,
|
| 158164 |
+
"loss": 0.7764,
|
| 158165 |
+
"step": 46215
|
| 158166 |
+
},
|
| 158167 |
+
{
|
| 158168 |
+
"epoch": 372.41,
|
| 158169 |
+
"learning_rate": 9.27559294871795e-06,
|
| 158170 |
+
"loss": 1.1147,
|
| 158171 |
+
"step": 46220
|
| 158172 |
+
},
|
| 158173 |
+
{
|
| 158174 |
+
"epoch": 372.45,
|
| 158175 |
+
"learning_rate": 9.27551282051282e-06,
|
| 158176 |
+
"loss": 0.3512,
|
| 158177 |
+
"step": 46225
|
| 158178 |
+
},
|
| 158179 |
+
{
|
| 158180 |
+
"epoch": 372.49,
|
| 158181 |
+
"learning_rate": 9.275432692307694e-06,
|
| 158182 |
+
"loss": 0.3299,
|
| 158183 |
+
"step": 46230
|
| 158184 |
+
},
|
| 158185 |
+
{
|
| 158186 |
+
"epoch": 372.53,
|
| 158187 |
+
"learning_rate": 9.275352564102565e-06,
|
| 158188 |
+
"loss": 0.3989,
|
| 158189 |
+
"step": 46235
|
| 158190 |
+
},
|
| 158191 |
+
{
|
| 158192 |
+
"epoch": 372.57,
|
| 158193 |
+
"learning_rate": 9.275272435897437e-06,
|
| 158194 |
+
"loss": 0.6187,
|
| 158195 |
+
"step": 46240
|
| 158196 |
+
},
|
| 158197 |
+
{
|
| 158198 |
+
"epoch": 372.61,
|
| 158199 |
+
"learning_rate": 9.27519230769231e-06,
|
| 158200 |
+
"loss": 1.3198,
|
| 158201 |
+
"step": 46245
|
| 158202 |
+
},
|
| 158203 |
+
{
|
| 158204 |
+
"epoch": 372.65,
|
| 158205 |
+
"learning_rate": 9.275112179487181e-06,
|
| 158206 |
+
"loss": 0.3802,
|
| 158207 |
+
"step": 46250
|
| 158208 |
+
},
|
| 158209 |
+
{
|
| 158210 |
+
"epoch": 372.69,
|
| 158211 |
+
"learning_rate": 9.275032051282052e-06,
|
| 158212 |
+
"loss": 0.3415,
|
| 158213 |
+
"step": 46255
|
| 158214 |
+
},
|
| 158215 |
+
{
|
| 158216 |
+
"epoch": 372.73,
|
| 158217 |
+
"learning_rate": 9.274951923076924e-06,
|
| 158218 |
+
"loss": 0.3431,
|
| 158219 |
+
"step": 46260
|
| 158220 |
+
},
|
| 158221 |
+
{
|
| 158222 |
+
"epoch": 372.77,
|
| 158223 |
+
"learning_rate": 9.274871794871797e-06,
|
| 158224 |
+
"loss": 0.6262,
|
| 158225 |
+
"step": 46265
|
| 158226 |
+
},
|
| 158227 |
+
{
|
| 158228 |
+
"epoch": 372.81,
|
| 158229 |
+
"learning_rate": 9.274791666666666e-06,
|
| 158230 |
+
"loss": 1.2644,
|
| 158231 |
+
"step": 46270
|
| 158232 |
+
},
|
| 158233 |
+
{
|
| 158234 |
+
"epoch": 372.85,
|
| 158235 |
+
"learning_rate": 9.27471153846154e-06,
|
| 158236 |
+
"loss": 0.3902,
|
| 158237 |
+
"step": 46275
|
| 158238 |
+
},
|
| 158239 |
+
{
|
| 158240 |
+
"epoch": 372.89,
|
| 158241 |
+
"learning_rate": 9.27463141025641e-06,
|
| 158242 |
+
"loss": 0.2856,
|
| 158243 |
+
"step": 46280
|
| 158244 |
+
},
|
| 158245 |
+
{
|
| 158246 |
+
"epoch": 372.93,
|
| 158247 |
+
"learning_rate": 9.274551282051282e-06,
|
| 158248 |
+
"loss": 0.4002,
|
| 158249 |
+
"step": 46285
|
| 158250 |
+
},
|
| 158251 |
+
{
|
| 158252 |
+
"epoch": 372.97,
|
| 158253 |
+
"learning_rate": 9.274471153846155e-06,
|
| 158254 |
+
"loss": 0.6261,
|
| 158255 |
+
"step": 46290
|
| 158256 |
+
},
|
| 158257 |
+
{
|
| 158258 |
+
"epoch": 373.0,
|
| 158259 |
+
"eval_loss": 0.4447263479232788,
|
| 158260 |
+
"eval_runtime": 41.0181,
|
| 158261 |
+
"eval_samples_per_second": 20.381,
|
| 158262 |
+
"eval_steps_per_second": 0.658,
|
| 158263 |
+
"eval_wer": 0.216405516643801,
|
| 158264 |
+
"step": 46293
|
| 158265 |
+
},
|
| 158266 |
+
{
|
| 158267 |
+
"epoch": 370.02,
|
| 158268 |
+
"learning_rate": 9.274391025641027e-06,
|
| 158269 |
+
"loss": 0.3643,
|
| 158270 |
+
"step": 46295
|
| 158271 |
+
},
|
| 158272 |
+
{
|
| 158273 |
+
"epoch": 370.06,
|
| 158274 |
+
"learning_rate": 9.274310897435898e-06,
|
| 158275 |
+
"loss": 0.3806,
|
| 158276 |
+
"step": 46300
|
| 158277 |
+
},
|
| 158278 |
+
{
|
| 158279 |
+
"epoch": 370.1,
|
| 158280 |
+
"learning_rate": 9.27423076923077e-06,
|
| 158281 |
+
"loss": 0.3132,
|
| 158282 |
+
"step": 46305
|
| 158283 |
+
},
|
| 158284 |
+
{
|
| 158285 |
+
"epoch": 370.14,
|
| 158286 |
+
"learning_rate": 9.274150641025642e-06,
|
| 158287 |
+
"loss": 0.3786,
|
| 158288 |
+
"step": 46310
|
| 158289 |
+
},
|
| 158290 |
+
{
|
| 158291 |
+
"epoch": 370.18,
|
| 158292 |
+
"learning_rate": 9.274070512820514e-06,
|
| 158293 |
+
"loss": 0.7365,
|
| 158294 |
+
"step": 46315
|
| 158295 |
+
},
|
| 158296 |
+
{
|
| 158297 |
+
"epoch": 370.22,
|
| 158298 |
+
"learning_rate": 9.273990384615385e-06,
|
| 158299 |
+
"loss": 1.1359,
|
| 158300 |
+
"step": 46320
|
| 158301 |
+
},
|
| 158302 |
+
{
|
| 158303 |
+
"epoch": 370.26,
|
| 158304 |
+
"learning_rate": 9.273910256410256e-06,
|
| 158305 |
+
"loss": 0.3751,
|
| 158306 |
+
"step": 46325
|
| 158307 |
+
},
|
| 158308 |
+
{
|
| 158309 |
+
"epoch": 370.3,
|
| 158310 |
+
"learning_rate": 9.27383012820513e-06,
|
| 158311 |
+
"loss": 0.6026,
|
| 158312 |
+
"step": 46330
|
| 158313 |
+
},
|
| 158314 |
+
{
|
| 158315 |
+
"epoch": 370.34,
|
| 158316 |
+
"learning_rate": 9.27375e-06,
|
| 158317 |
+
"loss": 0.4709,
|
| 158318 |
+
"step": 46335
|
| 158319 |
+
},
|
| 158320 |
+
{
|
| 158321 |
+
"epoch": 370.38,
|
| 158322 |
+
"learning_rate": 9.273669871794872e-06,
|
| 158323 |
+
"loss": 0.8101,
|
| 158324 |
+
"step": 46340
|
| 158325 |
+
},
|
| 158326 |
+
{
|
| 158327 |
+
"epoch": 370.42,
|
| 158328 |
+
"learning_rate": 9.273589743589745e-06,
|
| 158329 |
+
"loss": 0.9702,
|
| 158330 |
+
"step": 46345
|
| 158331 |
+
},
|
| 158332 |
+
{
|
| 158333 |
+
"epoch": 370.46,
|
| 158334 |
+
"learning_rate": 9.273509615384617e-06,
|
| 158335 |
+
"loss": 0.3566,
|
| 158336 |
+
"step": 46350
|
| 158337 |
+
},
|
| 158338 |
+
{
|
| 158339 |
+
"epoch": 370.5,
|
| 158340 |
+
"learning_rate": 9.273429487179488e-06,
|
| 158341 |
+
"loss": 0.3402,
|
| 158342 |
+
"step": 46355
|
| 158343 |
+
},
|
| 158344 |
+
{
|
| 158345 |
+
"epoch": 370.54,
|
| 158346 |
+
"learning_rate": 9.27334935897436e-06,
|
| 158347 |
+
"loss": 0.4004,
|
| 158348 |
+
"step": 46360
|
| 158349 |
+
},
|
| 158350 |
+
{
|
| 158351 |
+
"epoch": 370.58,
|
| 158352 |
+
"learning_rate": 9.273269230769232e-06,
|
| 158353 |
+
"loss": 0.7466,
|
| 158354 |
+
"step": 46365
|
| 158355 |
+
},
|
| 158356 |
+
{
|
| 158357 |
+
"epoch": 370.62,
|
| 158358 |
+
"learning_rate": 9.273189102564104e-06,
|
| 158359 |
+
"loss": 1.1429,
|
| 158360 |
+
"step": 46370
|
| 158361 |
+
},
|
| 158362 |
+
{
|
| 158363 |
+
"epoch": 370.66,
|
| 158364 |
+
"learning_rate": 9.273108974358975e-06,
|
| 158365 |
+
"loss": 0.3775,
|
| 158366 |
+
"step": 46375
|
| 158367 |
+
},
|
| 158368 |
+
{
|
| 158369 |
+
"epoch": 370.7,
|
| 158370 |
+
"learning_rate": 9.273028846153846e-06,
|
| 158371 |
+
"loss": 0.3599,
|
| 158372 |
+
"step": 46380
|
| 158373 |
+
},
|
| 158374 |
+
{
|
| 158375 |
+
"epoch": 370.74,
|
| 158376 |
+
"learning_rate": 9.27294871794872e-06,
|
| 158377 |
+
"loss": 0.4465,
|
| 158378 |
+
"step": 46385
|
| 158379 |
+
},
|
| 158380 |
+
{
|
| 158381 |
+
"epoch": 370.78,
|
| 158382 |
+
"learning_rate": 9.27286858974359e-06,
|
| 158383 |
+
"loss": 0.7581,
|
| 158384 |
+
"step": 46390
|
| 158385 |
+
},
|
| 158386 |
+
{
|
| 158387 |
+
"epoch": 370.82,
|
| 158388 |
+
"learning_rate": 9.272788461538462e-06,
|
| 158389 |
+
"loss": 1.1816,
|
| 158390 |
+
"step": 46395
|
| 158391 |
+
},
|
| 158392 |
+
{
|
| 158393 |
+
"epoch": 370.86,
|
| 158394 |
+
"learning_rate": 9.272708333333335e-06,
|
| 158395 |
+
"loss": 0.3272,
|
| 158396 |
+
"step": 46400
|
| 158397 |
+
},
|
| 158398 |
+
{
|
| 158399 |
+
"epoch": 370.9,
|
| 158400 |
+
"learning_rate": 9.272628205128205e-06,
|
| 158401 |
+
"loss": 0.3828,
|
| 158402 |
+
"step": 46405
|
| 158403 |
+
},
|
| 158404 |
+
{
|
| 158405 |
+
"epoch": 370.94,
|
| 158406 |
+
"learning_rate": 9.272548076923078e-06,
|
| 158407 |
+
"loss": 0.4614,
|
| 158408 |
+
"step": 46410
|
| 158409 |
+
},
|
| 158410 |
+
{
|
| 158411 |
+
"epoch": 370.98,
|
| 158412 |
+
"learning_rate": 9.27246794871795e-06,
|
| 158413 |
+
"loss": 0.8639,
|
| 158414 |
+
"step": 46415
|
| 158415 |
+
},
|
| 158416 |
+
{
|
| 158417 |
+
"epoch": 371.0,
|
| 158418 |
+
"eval_loss": 0.41137728095054626,
|
| 158419 |
+
"eval_runtime": 40.1796,
|
| 158420 |
+
"eval_samples_per_second": 20.807,
|
| 158421 |
+
"eval_steps_per_second": 0.672,
|
| 158422 |
+
"eval_wer": 0.20991401112797167,
|
| 158423 |
+
"step": 46418
|
| 158424 |
}
|
| 158425 |
],
|
| 158426 |
"max_steps": 625000,
|
| 158427 |
"num_train_epochs": 5000,
|
| 158428 |
+
"total_flos": 1.306004945083258e+20,
|
| 158429 |
"trial_name": null,
|
| 158430 |
"trial_params": null
|
| 158431 |
}
|
model-bin/finetune/base/{checkpoint-45796 → checkpoint-46418}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629719560.684198/events.out.tfevents.1629719560.74272264b15c.932.123
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d348f06fd11e01b8bf667b85e3ce969dcea9ce7e01e8944cf7feba8553647dd3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629720325.471121/events.out.tfevents.1629720325.74272264b15c.932.125
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:54c94f35c089f19418b813ee6ed14559db5c8a43a26451f9f62be3751a988528
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629720980.0787914/events.out.tfevents.1629720980.74272264b15c.932.127
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:44402ecf76f41ba3780387d837f02f3ef0d8c2a9ce716f37f15132223e89d89a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629721622.799041/events.out.tfevents.1629721622.74272264b15c.932.129
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:71ec7add12b5d93f30713fa1b81bbe3a43b422ea4d369ec65c8cd69849a1dca1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629722285.2883925/events.out.tfevents.1629722285.74272264b15c.932.131
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b43b35a993db39482cf61f80777872309400375cacebadedb62baab4a47b4479
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629719560.74272264b15c.932.122
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:06959a40ff266c2e361a236db224ced71922dc3200a0c213f4ec876e6c4db875
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629720325.74272264b15c.932.124
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ad9ad8672e41f4a0e6f0103e87185fd0cfe8d26d0515ca3e5e7ad9394558f73f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629720980.74272264b15c.932.126
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3dda258a2aa94580ba2a130dd32b5163ca22cdae5c24a54d0125e99b128cfbf9
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629721622.74272264b15c.932.128
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dafdbcd97c3b16a224b53d91f7b7ac607f2ea9c9f4aa9b93ab25d926c9bb0732
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629722285.74272264b15c.932.130
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:49c5fdeb1748a457de913b1df724b50b0f4095f50e1b821fc5527982855069b7
|
| 3 |
+
size 8622
|