"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630136516.5433424/events.out.tfevents.1630136516.86bb0ddabf9b.4092.71 +3 -0
- model-bin/finetune/base/log/1630136910.4690464/events.out.tfevents.1630136910.86bb0ddabf9b.4092.73 +3 -0
- model-bin/finetune/base/log/1630137306.2544515/events.out.tfevents.1630137306.86bb0ddabf9b.4092.75 +3 -0
- model-bin/finetune/base/log/1630137697.289981/events.out.tfevents.1630137697.86bb0ddabf9b.4092.77 +3 -0
- model-bin/finetune/base/log/1630138093.0549505/events.out.tfevents.1630138093.86bb0ddabf9b.4092.79 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630136516.86bb0ddabf9b.4092.70 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630136910.86bb0ddabf9b.4092.72 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630137306.86bb0ddabf9b.4092.74 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630137697.86bb0ddabf9b.4092.76 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630138093.86bb0ddabf9b.4092.78 +3 -0
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/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 sha256:0ec9b4ca716593f4c0045c041f86101934ebc89734754023031ed5465bdabb1c
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/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:cec3ce895e777b0b559398c7eedd6d3c5a2ee4f564f98d9b959fa449c27bfc00
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/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:da69cf5f555ca72838098676ce3c29a68b9bbdaeff35f98b7e6a2d1f21631a42
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/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:ca89118762f5fcee3e0fd4bd629fd68bc9636cb4b91f886dcda93a3a4e9aaae4
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/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 sha256:3c117c3e6f981903b3cdb3170ac9334c2facfbc6eadbf698128fa29363fffa85
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -255003,11 +255003,806 @@
|
|
| 255003 |
"eval_steps_per_second": 0.752,
|
| 255004 |
"eval_wer": 0.18008964719491036,
|
| 255005 |
"step": 122329
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 255006 |
}
|
| 255007 |
],
|
| 255008 |
"max_steps": 625000,
|
| 255009 |
"num_train_epochs": 5000,
|
| 255010 |
-
"total_flos": 3.
|
| 255011 |
"trial_name": null,
|
| 255012 |
"trial_params": null
|
| 255013 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 982.9960159362549,
|
| 5 |
+
"global_step": 122953,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 255003 |
"eval_steps_per_second": 0.752,
|
| 255004 |
"eval_wer": 0.18008964719491036,
|
| 255005 |
"step": 122329
|
| 255006 |
+
},
|
| 255007 |
+
{
|
| 255008 |
+
"epoch": 978.01,
|
| 255009 |
+
"learning_rate": 8.056378205128206e-06,
|
| 255010 |
+
"loss": 0.3131,
|
| 255011 |
+
"step": 122330
|
| 255012 |
+
},
|
| 255013 |
+
{
|
| 255014 |
+
"epoch": 978.05,
|
| 255015 |
+
"learning_rate": 8.056298076923078e-06,
|
| 255016 |
+
"loss": 0.2839,
|
| 255017 |
+
"step": 122335
|
| 255018 |
+
},
|
| 255019 |
+
{
|
| 255020 |
+
"epoch": 978.09,
|
| 255021 |
+
"learning_rate": 8.056217948717949e-06,
|
| 255022 |
+
"loss": 0.3023,
|
| 255023 |
+
"step": 122340
|
| 255024 |
+
},
|
| 255025 |
+
{
|
| 255026 |
+
"epoch": 978.13,
|
| 255027 |
+
"learning_rate": 8.056137820512822e-06,
|
| 255028 |
+
"loss": 0.417,
|
| 255029 |
+
"step": 122345
|
| 255030 |
+
},
|
| 255031 |
+
{
|
| 255032 |
+
"epoch": 978.17,
|
| 255033 |
+
"learning_rate": 8.056057692307693e-06,
|
| 255034 |
+
"loss": 0.5113,
|
| 255035 |
+
"step": 122350
|
| 255036 |
+
},
|
| 255037 |
+
{
|
| 255038 |
+
"epoch": 978.21,
|
| 255039 |
+
"learning_rate": 8.055977564102565e-06,
|
| 255040 |
+
"loss": 1.0222,
|
| 255041 |
+
"step": 122355
|
| 255042 |
+
},
|
| 255043 |
+
{
|
| 255044 |
+
"epoch": 978.25,
|
| 255045 |
+
"learning_rate": 8.055897435897436e-06,
|
| 255046 |
+
"loss": 0.3167,
|
| 255047 |
+
"step": 122360
|
| 255048 |
+
},
|
| 255049 |
+
{
|
| 255050 |
+
"epoch": 978.29,
|
| 255051 |
+
"learning_rate": 8.055817307692309e-06,
|
| 255052 |
+
"loss": 0.3206,
|
| 255053 |
+
"step": 122365
|
| 255054 |
+
},
|
| 255055 |
+
{
|
| 255056 |
+
"epoch": 978.33,
|
| 255057 |
+
"learning_rate": 8.05573717948718e-06,
|
| 255058 |
+
"loss": 0.2958,
|
| 255059 |
+
"step": 122370
|
| 255060 |
+
},
|
| 255061 |
+
{
|
| 255062 |
+
"epoch": 978.37,
|
| 255063 |
+
"learning_rate": 8.055657051282052e-06,
|
| 255064 |
+
"loss": 0.5858,
|
| 255065 |
+
"step": 122375
|
| 255066 |
+
},
|
| 255067 |
+
{
|
| 255068 |
+
"epoch": 978.41,
|
| 255069 |
+
"learning_rate": 8.055576923076923e-06,
|
| 255070 |
+
"loss": 1.0429,
|
| 255071 |
+
"step": 122380
|
| 255072 |
+
},
|
| 255073 |
+
{
|
| 255074 |
+
"epoch": 978.45,
|
| 255075 |
+
"learning_rate": 8.055496794871796e-06,
|
| 255076 |
+
"loss": 0.3465,
|
| 255077 |
+
"step": 122385
|
| 255078 |
+
},
|
| 255079 |
+
{
|
| 255080 |
+
"epoch": 978.49,
|
| 255081 |
+
"learning_rate": 8.055416666666668e-06,
|
| 255082 |
+
"loss": 0.2864,
|
| 255083 |
+
"step": 122390
|
| 255084 |
+
},
|
| 255085 |
+
{
|
| 255086 |
+
"epoch": 978.53,
|
| 255087 |
+
"learning_rate": 8.055336538461539e-06,
|
| 255088 |
+
"loss": 0.2985,
|
| 255089 |
+
"step": 122395
|
| 255090 |
+
},
|
| 255091 |
+
{
|
| 255092 |
+
"epoch": 978.57,
|
| 255093 |
+
"learning_rate": 8.055256410256412e-06,
|
| 255094 |
+
"loss": 0.5527,
|
| 255095 |
+
"step": 122400
|
| 255096 |
+
},
|
| 255097 |
+
{
|
| 255098 |
+
"epoch": 978.61,
|
| 255099 |
+
"learning_rate": 8.055176282051282e-06,
|
| 255100 |
+
"loss": 1.1217,
|
| 255101 |
+
"step": 122405
|
| 255102 |
+
},
|
| 255103 |
+
{
|
| 255104 |
+
"epoch": 978.65,
|
| 255105 |
+
"learning_rate": 8.055096153846155e-06,
|
| 255106 |
+
"loss": 0.304,
|
| 255107 |
+
"step": 122410
|
| 255108 |
+
},
|
| 255109 |
+
{
|
| 255110 |
+
"epoch": 978.69,
|
| 255111 |
+
"learning_rate": 8.055016025641026e-06,
|
| 255112 |
+
"loss": 0.3157,
|
| 255113 |
+
"step": 122415
|
| 255114 |
+
},
|
| 255115 |
+
{
|
| 255116 |
+
"epoch": 978.73,
|
| 255117 |
+
"learning_rate": 8.054935897435897e-06,
|
| 255118 |
+
"loss": 0.3796,
|
| 255119 |
+
"step": 122420
|
| 255120 |
+
},
|
| 255121 |
+
{
|
| 255122 |
+
"epoch": 978.77,
|
| 255123 |
+
"learning_rate": 8.054855769230769e-06,
|
| 255124 |
+
"loss": 0.5745,
|
| 255125 |
+
"step": 122425
|
| 255126 |
+
},
|
| 255127 |
+
{
|
| 255128 |
+
"epoch": 978.81,
|
| 255129 |
+
"learning_rate": 8.054775641025642e-06,
|
| 255130 |
+
"loss": 1.2091,
|
| 255131 |
+
"step": 122430
|
| 255132 |
+
},
|
| 255133 |
+
{
|
| 255134 |
+
"epoch": 978.85,
|
| 255135 |
+
"learning_rate": 8.054695512820513e-06,
|
| 255136 |
+
"loss": 0.3528,
|
| 255137 |
+
"step": 122435
|
| 255138 |
+
},
|
| 255139 |
+
{
|
| 255140 |
+
"epoch": 978.89,
|
| 255141 |
+
"learning_rate": 8.054615384615385e-06,
|
| 255142 |
+
"loss": 0.3314,
|
| 255143 |
+
"step": 122440
|
| 255144 |
+
},
|
| 255145 |
+
{
|
| 255146 |
+
"epoch": 978.93,
|
| 255147 |
+
"learning_rate": 8.054535256410258e-06,
|
| 255148 |
+
"loss": 0.295,
|
| 255149 |
+
"step": 122445
|
| 255150 |
+
},
|
| 255151 |
+
{
|
| 255152 |
+
"epoch": 978.97,
|
| 255153 |
+
"learning_rate": 8.054455128205129e-06,
|
| 255154 |
+
"loss": 0.516,
|
| 255155 |
+
"step": 122450
|
| 255156 |
+
},
|
| 255157 |
+
{
|
| 255158 |
+
"epoch": 979.0,
|
| 255159 |
+
"eval_loss": 0.4413236975669861,
|
| 255160 |
+
"eval_runtime": 35.5298,
|
| 255161 |
+
"eval_samples_per_second": 23.698,
|
| 255162 |
+
"eval_steps_per_second": 0.76,
|
| 255163 |
+
"eval_wer": 0.18869865530163862,
|
| 255164 |
+
"step": 122454
|
| 255165 |
+
},
|
| 255166 |
+
{
|
| 255167 |
+
"epoch": 979.01,
|
| 255168 |
+
"learning_rate": 8.054375e-06,
|
| 255169 |
+
"loss": 0.5158,
|
| 255170 |
+
"step": 122455
|
| 255171 |
+
},
|
| 255172 |
+
{
|
| 255173 |
+
"epoch": 979.05,
|
| 255174 |
+
"learning_rate": 8.054294871794872e-06,
|
| 255175 |
+
"loss": 0.3207,
|
| 255176 |
+
"step": 122460
|
| 255177 |
+
},
|
| 255178 |
+
{
|
| 255179 |
+
"epoch": 979.09,
|
| 255180 |
+
"learning_rate": 8.054214743589745e-06,
|
| 255181 |
+
"loss": 0.3022,
|
| 255182 |
+
"step": 122465
|
| 255183 |
+
},
|
| 255184 |
+
{
|
| 255185 |
+
"epoch": 979.13,
|
| 255186 |
+
"learning_rate": 8.054134615384616e-06,
|
| 255187 |
+
"loss": 0.3093,
|
| 255188 |
+
"step": 122470
|
| 255189 |
+
},
|
| 255190 |
+
{
|
| 255191 |
+
"epoch": 979.17,
|
| 255192 |
+
"learning_rate": 8.054054487179487e-06,
|
| 255193 |
+
"loss": 0.4786,
|
| 255194 |
+
"step": 122475
|
| 255195 |
+
},
|
| 255196 |
+
{
|
| 255197 |
+
"epoch": 979.21,
|
| 255198 |
+
"learning_rate": 8.053974358974359e-06,
|
| 255199 |
+
"loss": 1.1211,
|
| 255200 |
+
"step": 122480
|
| 255201 |
+
},
|
| 255202 |
+
{
|
| 255203 |
+
"epoch": 979.25,
|
| 255204 |
+
"learning_rate": 8.053894230769232e-06,
|
| 255205 |
+
"loss": 0.33,
|
| 255206 |
+
"step": 122485
|
| 255207 |
+
},
|
| 255208 |
+
{
|
| 255209 |
+
"epoch": 979.29,
|
| 255210 |
+
"learning_rate": 8.053814102564103e-06,
|
| 255211 |
+
"loss": 0.2976,
|
| 255212 |
+
"step": 122490
|
| 255213 |
+
},
|
| 255214 |
+
{
|
| 255215 |
+
"epoch": 979.33,
|
| 255216 |
+
"learning_rate": 8.053733974358975e-06,
|
| 255217 |
+
"loss": 0.332,
|
| 255218 |
+
"step": 122495
|
| 255219 |
+
},
|
| 255220 |
+
{
|
| 255221 |
+
"epoch": 979.37,
|
| 255222 |
+
"learning_rate": 8.053653846153848e-06,
|
| 255223 |
+
"loss": 0.5332,
|
| 255224 |
+
"step": 122500
|
| 255225 |
+
},
|
| 255226 |
+
{
|
| 255227 |
+
"epoch": 979.41,
|
| 255228 |
+
"learning_rate": 8.053573717948719e-06,
|
| 255229 |
+
"loss": 1.1869,
|
| 255230 |
+
"step": 122505
|
| 255231 |
+
},
|
| 255232 |
+
{
|
| 255233 |
+
"epoch": 979.45,
|
| 255234 |
+
"learning_rate": 8.05349358974359e-06,
|
| 255235 |
+
"loss": 0.2747,
|
| 255236 |
+
"step": 122510
|
| 255237 |
+
},
|
| 255238 |
+
{
|
| 255239 |
+
"epoch": 979.49,
|
| 255240 |
+
"learning_rate": 8.053413461538462e-06,
|
| 255241 |
+
"loss": 0.2658,
|
| 255242 |
+
"step": 122515
|
| 255243 |
+
},
|
| 255244 |
+
{
|
| 255245 |
+
"epoch": 979.53,
|
| 255246 |
+
"learning_rate": 8.053333333333335e-06,
|
| 255247 |
+
"loss": 0.3075,
|
| 255248 |
+
"step": 122520
|
| 255249 |
+
},
|
| 255250 |
+
{
|
| 255251 |
+
"epoch": 979.57,
|
| 255252 |
+
"learning_rate": 8.053253205128206e-06,
|
| 255253 |
+
"loss": 0.462,
|
| 255254 |
+
"step": 122525
|
| 255255 |
+
},
|
| 255256 |
+
{
|
| 255257 |
+
"epoch": 979.61,
|
| 255258 |
+
"learning_rate": 8.053173076923077e-06,
|
| 255259 |
+
"loss": 1.0422,
|
| 255260 |
+
"step": 122530
|
| 255261 |
+
},
|
| 255262 |
+
{
|
| 255263 |
+
"epoch": 979.65,
|
| 255264 |
+
"learning_rate": 8.05309294871795e-06,
|
| 255265 |
+
"loss": 0.2877,
|
| 255266 |
+
"step": 122535
|
| 255267 |
+
},
|
| 255268 |
+
{
|
| 255269 |
+
"epoch": 979.69,
|
| 255270 |
+
"learning_rate": 8.053012820512822e-06,
|
| 255271 |
+
"loss": 0.3205,
|
| 255272 |
+
"step": 122540
|
| 255273 |
+
},
|
| 255274 |
+
{
|
| 255275 |
+
"epoch": 979.73,
|
| 255276 |
+
"learning_rate": 8.052932692307693e-06,
|
| 255277 |
+
"loss": 0.3306,
|
| 255278 |
+
"step": 122545
|
| 255279 |
+
},
|
| 255280 |
+
{
|
| 255281 |
+
"epoch": 979.77,
|
| 255282 |
+
"learning_rate": 8.052852564102565e-06,
|
| 255283 |
+
"loss": 0.5342,
|
| 255284 |
+
"step": 122550
|
| 255285 |
+
},
|
| 255286 |
+
{
|
| 255287 |
+
"epoch": 979.81,
|
| 255288 |
+
"learning_rate": 8.052772435897438e-06,
|
| 255289 |
+
"loss": 1.3037,
|
| 255290 |
+
"step": 122555
|
| 255291 |
+
},
|
| 255292 |
+
{
|
| 255293 |
+
"epoch": 979.85,
|
| 255294 |
+
"learning_rate": 8.052692307692307e-06,
|
| 255295 |
+
"loss": 0.3079,
|
| 255296 |
+
"step": 122560
|
| 255297 |
+
},
|
| 255298 |
+
{
|
| 255299 |
+
"epoch": 979.89,
|
| 255300 |
+
"learning_rate": 8.05261217948718e-06,
|
| 255301 |
+
"loss": 0.2873,
|
| 255302 |
+
"step": 122565
|
| 255303 |
+
},
|
| 255304 |
+
{
|
| 255305 |
+
"epoch": 979.93,
|
| 255306 |
+
"learning_rate": 8.052532051282052e-06,
|
| 255307 |
+
"loss": 0.3823,
|
| 255308 |
+
"step": 122570
|
| 255309 |
+
},
|
| 255310 |
+
{
|
| 255311 |
+
"epoch": 979.97,
|
| 255312 |
+
"learning_rate": 8.052451923076923e-06,
|
| 255313 |
+
"loss": 0.5631,
|
| 255314 |
+
"step": 122575
|
| 255315 |
+
},
|
| 255316 |
+
{
|
| 255317 |
+
"epoch": 980.0,
|
| 255318 |
+
"eval_loss": 0.3583020269870758,
|
| 255319 |
+
"eval_runtime": 36.4633,
|
| 255320 |
+
"eval_samples_per_second": 23.092,
|
| 255321 |
+
"eval_steps_per_second": 0.74,
|
| 255322 |
+
"eval_wer": 0.18102372034956304,
|
| 255323 |
+
"step": 122579
|
| 255324 |
+
},
|
| 255325 |
+
{
|
| 255326 |
+
"epoch": 988.01,
|
| 255327 |
+
"learning_rate": 8.052387820512822e-06,
|
| 255328 |
+
"loss": 0.3373,
|
| 255329 |
+
"step": 122580
|
| 255330 |
+
},
|
| 255331 |
+
{
|
| 255332 |
+
"epoch": 988.05,
|
| 255333 |
+
"learning_rate": 8.052307692307692e-06,
|
| 255334 |
+
"loss": 0.3258,
|
| 255335 |
+
"step": 122585
|
| 255336 |
+
},
|
| 255337 |
+
{
|
| 255338 |
+
"epoch": 988.09,
|
| 255339 |
+
"learning_rate": 8.052227564102565e-06,
|
| 255340 |
+
"loss": 0.3036,
|
| 255341 |
+
"step": 122590
|
| 255342 |
+
},
|
| 255343 |
+
{
|
| 255344 |
+
"epoch": 988.13,
|
| 255345 |
+
"learning_rate": 8.052147435897436e-06,
|
| 255346 |
+
"loss": 0.3302,
|
| 255347 |
+
"step": 122595
|
| 255348 |
+
},
|
| 255349 |
+
{
|
| 255350 |
+
"epoch": 988.17,
|
| 255351 |
+
"learning_rate": 8.052067307692308e-06,
|
| 255352 |
+
"loss": 0.4832,
|
| 255353 |
+
"step": 122600
|
| 255354 |
+
},
|
| 255355 |
+
{
|
| 255356 |
+
"epoch": 988.21,
|
| 255357 |
+
"learning_rate": 8.05198717948718e-06,
|
| 255358 |
+
"loss": 1.1441,
|
| 255359 |
+
"step": 122605
|
| 255360 |
+
},
|
| 255361 |
+
{
|
| 255362 |
+
"epoch": 988.25,
|
| 255363 |
+
"learning_rate": 8.051907051282052e-06,
|
| 255364 |
+
"loss": 0.2842,
|
| 255365 |
+
"step": 122610
|
| 255366 |
+
},
|
| 255367 |
+
{
|
| 255368 |
+
"epoch": 988.29,
|
| 255369 |
+
"learning_rate": 8.051826923076923e-06,
|
| 255370 |
+
"loss": 0.2901,
|
| 255371 |
+
"step": 122615
|
| 255372 |
+
},
|
| 255373 |
+
{
|
| 255374 |
+
"epoch": 988.33,
|
| 255375 |
+
"learning_rate": 8.051746794871795e-06,
|
| 255376 |
+
"loss": 0.342,
|
| 255377 |
+
"step": 122620
|
| 255378 |
+
},
|
| 255379 |
+
{
|
| 255380 |
+
"epoch": 988.37,
|
| 255381 |
+
"learning_rate": 8.051666666666668e-06,
|
| 255382 |
+
"loss": 0.5233,
|
| 255383 |
+
"step": 122625
|
| 255384 |
+
},
|
| 255385 |
+
{
|
| 255386 |
+
"epoch": 988.41,
|
| 255387 |
+
"learning_rate": 8.051586538461539e-06,
|
| 255388 |
+
"loss": 1.0718,
|
| 255389 |
+
"step": 122630
|
| 255390 |
+
},
|
| 255391 |
+
{
|
| 255392 |
+
"epoch": 988.45,
|
| 255393 |
+
"learning_rate": 8.05150641025641e-06,
|
| 255394 |
+
"loss": 0.2838,
|
| 255395 |
+
"step": 122635
|
| 255396 |
+
},
|
| 255397 |
+
{
|
| 255398 |
+
"epoch": 988.49,
|
| 255399 |
+
"learning_rate": 8.051426282051282e-06,
|
| 255400 |
+
"loss": 0.2908,
|
| 255401 |
+
"step": 122640
|
| 255402 |
+
},
|
| 255403 |
+
{
|
| 255404 |
+
"epoch": 988.53,
|
| 255405 |
+
"learning_rate": 8.051346153846155e-06,
|
| 255406 |
+
"loss": 0.3083,
|
| 255407 |
+
"step": 122645
|
| 255408 |
+
},
|
| 255409 |
+
{
|
| 255410 |
+
"epoch": 988.57,
|
| 255411 |
+
"learning_rate": 8.051266025641026e-06,
|
| 255412 |
+
"loss": 0.5269,
|
| 255413 |
+
"step": 122650
|
| 255414 |
+
},
|
| 255415 |
+
{
|
| 255416 |
+
"epoch": 988.61,
|
| 255417 |
+
"learning_rate": 8.051185897435898e-06,
|
| 255418 |
+
"loss": 1.1258,
|
| 255419 |
+
"step": 122655
|
| 255420 |
+
},
|
| 255421 |
+
{
|
| 255422 |
+
"epoch": 988.65,
|
| 255423 |
+
"learning_rate": 8.05110576923077e-06,
|
| 255424 |
+
"loss": 0.2821,
|
| 255425 |
+
"step": 122660
|
| 255426 |
+
},
|
| 255427 |
+
{
|
| 255428 |
+
"epoch": 988.69,
|
| 255429 |
+
"learning_rate": 8.051025641025642e-06,
|
| 255430 |
+
"loss": 0.2829,
|
| 255431 |
+
"step": 122665
|
| 255432 |
+
},
|
| 255433 |
+
{
|
| 255434 |
+
"epoch": 988.73,
|
| 255435 |
+
"learning_rate": 8.050945512820513e-06,
|
| 255436 |
+
"loss": 0.3298,
|
| 255437 |
+
"step": 122670
|
| 255438 |
+
},
|
| 255439 |
+
{
|
| 255440 |
+
"epoch": 988.77,
|
| 255441 |
+
"learning_rate": 8.050865384615385e-06,
|
| 255442 |
+
"loss": 0.4948,
|
| 255443 |
+
"step": 122675
|
| 255444 |
+
},
|
| 255445 |
+
{
|
| 255446 |
+
"epoch": 988.81,
|
| 255447 |
+
"learning_rate": 8.050785256410258e-06,
|
| 255448 |
+
"loss": 1.1248,
|
| 255449 |
+
"step": 122680
|
| 255450 |
+
},
|
| 255451 |
+
{
|
| 255452 |
+
"epoch": 988.85,
|
| 255453 |
+
"learning_rate": 8.05070512820513e-06,
|
| 255454 |
+
"loss": 0.2898,
|
| 255455 |
+
"step": 122685
|
| 255456 |
+
},
|
| 255457 |
+
{
|
| 255458 |
+
"epoch": 988.9,
|
| 255459 |
+
"learning_rate": 8.050625e-06,
|
| 255460 |
+
"loss": 0.3413,
|
| 255461 |
+
"step": 122690
|
| 255462 |
+
},
|
| 255463 |
+
{
|
| 255464 |
+
"epoch": 988.94,
|
| 255465 |
+
"learning_rate": 8.050544871794872e-06,
|
| 255466 |
+
"loss": 0.2792,
|
| 255467 |
+
"step": 122695
|
| 255468 |
+
},
|
| 255469 |
+
{
|
| 255470 |
+
"epoch": 988.98,
|
| 255471 |
+
"learning_rate": 8.050464743589745e-06,
|
| 255472 |
+
"loss": 0.5949,
|
| 255473 |
+
"step": 122700
|
| 255474 |
+
},
|
| 255475 |
+
{
|
| 255476 |
+
"epoch": 989.0,
|
| 255477 |
+
"eval_loss": 0.35863104462623596,
|
| 255478 |
+
"eval_runtime": 37.7799,
|
| 255479 |
+
"eval_samples_per_second": 22.287,
|
| 255480 |
+
"eval_steps_per_second": 0.715,
|
| 255481 |
+
"eval_wer": 0.18502449265220433,
|
| 255482 |
+
"step": 122703
|
| 255483 |
+
},
|
| 255484 |
+
{
|
| 255485 |
+
"epoch": 981.02,
|
| 255486 |
+
"learning_rate": 8.050384615384616e-06,
|
| 255487 |
+
"loss": 0.3343,
|
| 255488 |
+
"step": 122705
|
| 255489 |
+
},
|
| 255490 |
+
{
|
| 255491 |
+
"epoch": 981.06,
|
| 255492 |
+
"learning_rate": 8.050304487179488e-06,
|
| 255493 |
+
"loss": 0.2797,
|
| 255494 |
+
"step": 122710
|
| 255495 |
+
},
|
| 255496 |
+
{
|
| 255497 |
+
"epoch": 981.1,
|
| 255498 |
+
"learning_rate": 8.05022435897436e-06,
|
| 255499 |
+
"loss": 0.3152,
|
| 255500 |
+
"step": 122715
|
| 255501 |
+
},
|
| 255502 |
+
{
|
| 255503 |
+
"epoch": 981.14,
|
| 255504 |
+
"learning_rate": 8.050144230769232e-06,
|
| 255505 |
+
"loss": 0.3534,
|
| 255506 |
+
"step": 122720
|
| 255507 |
+
},
|
| 255508 |
+
{
|
| 255509 |
+
"epoch": 981.18,
|
| 255510 |
+
"learning_rate": 8.050064102564103e-06,
|
| 255511 |
+
"loss": 0.6074,
|
| 255512 |
+
"step": 122725
|
| 255513 |
+
},
|
| 255514 |
+
{
|
| 255515 |
+
"epoch": 981.22,
|
| 255516 |
+
"learning_rate": 8.049983974358975e-06,
|
| 255517 |
+
"loss": 1.0557,
|
| 255518 |
+
"step": 122730
|
| 255519 |
+
},
|
| 255520 |
+
{
|
| 255521 |
+
"epoch": 981.26,
|
| 255522 |
+
"learning_rate": 8.049903846153848e-06,
|
| 255523 |
+
"loss": 0.2889,
|
| 255524 |
+
"step": 122735
|
| 255525 |
+
},
|
| 255526 |
+
{
|
| 255527 |
+
"epoch": 981.3,
|
| 255528 |
+
"learning_rate": 8.049823717948718e-06,
|
| 255529 |
+
"loss": 0.3485,
|
| 255530 |
+
"step": 122740
|
| 255531 |
+
},
|
| 255532 |
+
{
|
| 255533 |
+
"epoch": 981.34,
|
| 255534 |
+
"learning_rate": 8.04974358974359e-06,
|
| 255535 |
+
"loss": 0.3523,
|
| 255536 |
+
"step": 122745
|
| 255537 |
+
},
|
| 255538 |
+
{
|
| 255539 |
+
"epoch": 981.38,
|
| 255540 |
+
"learning_rate": 8.049663461538462e-06,
|
| 255541 |
+
"loss": 0.6186,
|
| 255542 |
+
"step": 122750
|
| 255543 |
+
},
|
| 255544 |
+
{
|
| 255545 |
+
"epoch": 981.42,
|
| 255546 |
+
"learning_rate": 8.049583333333333e-06,
|
| 255547 |
+
"loss": 1.1316,
|
| 255548 |
+
"step": 122755
|
| 255549 |
+
},
|
| 255550 |
+
{
|
| 255551 |
+
"epoch": 981.46,
|
| 255552 |
+
"learning_rate": 8.049503205128206e-06,
|
| 255553 |
+
"loss": 0.2671,
|
| 255554 |
+
"step": 122760
|
| 255555 |
+
},
|
| 255556 |
+
{
|
| 255557 |
+
"epoch": 981.5,
|
| 255558 |
+
"learning_rate": 8.049423076923078e-06,
|
| 255559 |
+
"loss": 0.3189,
|
| 255560 |
+
"step": 122765
|
| 255561 |
+
},
|
| 255562 |
+
{
|
| 255563 |
+
"epoch": 981.54,
|
| 255564 |
+
"learning_rate": 8.049342948717949e-06,
|
| 255565 |
+
"loss": 0.4012,
|
| 255566 |
+
"step": 122770
|
| 255567 |
+
},
|
| 255568 |
+
{
|
| 255569 |
+
"epoch": 981.58,
|
| 255570 |
+
"learning_rate": 8.04926282051282e-06,
|
| 255571 |
+
"loss": 0.7237,
|
| 255572 |
+
"step": 122775
|
| 255573 |
+
},
|
| 255574 |
+
{
|
| 255575 |
+
"epoch": 981.62,
|
| 255576 |
+
"learning_rate": 8.049182692307693e-06,
|
| 255577 |
+
"loss": 0.9156,
|
| 255578 |
+
"step": 122780
|
| 255579 |
+
},
|
| 255580 |
+
{
|
| 255581 |
+
"epoch": 981.66,
|
| 255582 |
+
"learning_rate": 8.049102564102565e-06,
|
| 255583 |
+
"loss": 0.2938,
|
| 255584 |
+
"step": 122785
|
| 255585 |
+
},
|
| 255586 |
+
{
|
| 255587 |
+
"epoch": 981.7,
|
| 255588 |
+
"learning_rate": 8.049022435897436e-06,
|
| 255589 |
+
"loss": 0.3334,
|
| 255590 |
+
"step": 122790
|
| 255591 |
+
},
|
| 255592 |
+
{
|
| 255593 |
+
"epoch": 981.74,
|
| 255594 |
+
"learning_rate": 8.048942307692308e-06,
|
| 255595 |
+
"loss": 0.3146,
|
| 255596 |
+
"step": 122795
|
| 255597 |
+
},
|
| 255598 |
+
{
|
| 255599 |
+
"epoch": 981.78,
|
| 255600 |
+
"learning_rate": 8.04886217948718e-06,
|
| 255601 |
+
"loss": 0.6774,
|
| 255602 |
+
"step": 122800
|
| 255603 |
+
},
|
| 255604 |
+
{
|
| 255605 |
+
"epoch": 981.82,
|
| 255606 |
+
"learning_rate": 8.048782051282052e-06,
|
| 255607 |
+
"loss": 1.0139,
|
| 255608 |
+
"step": 122805
|
| 255609 |
+
},
|
| 255610 |
+
{
|
| 255611 |
+
"epoch": 981.86,
|
| 255612 |
+
"learning_rate": 8.048701923076923e-06,
|
| 255613 |
+
"loss": 0.2503,
|
| 255614 |
+
"step": 122810
|
| 255615 |
+
},
|
| 255616 |
+
{
|
| 255617 |
+
"epoch": 981.9,
|
| 255618 |
+
"learning_rate": 8.048621794871796e-06,
|
| 255619 |
+
"loss": 0.2785,
|
| 255620 |
+
"step": 122815
|
| 255621 |
+
},
|
| 255622 |
+
{
|
| 255623 |
+
"epoch": 981.94,
|
| 255624 |
+
"learning_rate": 8.048541666666668e-06,
|
| 255625 |
+
"loss": 0.3116,
|
| 255626 |
+
"step": 122820
|
| 255627 |
+
},
|
| 255628 |
+
{
|
| 255629 |
+
"epoch": 981.98,
|
| 255630 |
+
"learning_rate": 8.048461538461539e-06,
|
| 255631 |
+
"loss": 0.6077,
|
| 255632 |
+
"step": 122825
|
| 255633 |
+
},
|
| 255634 |
+
{
|
| 255635 |
+
"epoch": 982.0,
|
| 255636 |
+
"eval_loss": 0.3889777660369873,
|
| 255637 |
+
"eval_runtime": 36.1843,
|
| 255638 |
+
"eval_samples_per_second": 23.27,
|
| 255639 |
+
"eval_steps_per_second": 0.746,
|
| 255640 |
+
"eval_wer": 0.1833968347010551,
|
| 255641 |
+
"step": 122828
|
| 255642 |
+
},
|
| 255643 |
+
{
|
| 255644 |
+
"epoch": 982.02,
|
| 255645 |
+
"learning_rate": 8.04838141025641e-06,
|
| 255646 |
+
"loss": 0.3437,
|
| 255647 |
+
"step": 122830
|
| 255648 |
+
},
|
| 255649 |
+
{
|
| 255650 |
+
"epoch": 982.06,
|
| 255651 |
+
"learning_rate": 8.048301282051283e-06,
|
| 255652 |
+
"loss": 0.2759,
|
| 255653 |
+
"step": 122835
|
| 255654 |
+
},
|
| 255655 |
+
{
|
| 255656 |
+
"epoch": 982.1,
|
| 255657 |
+
"learning_rate": 8.048221153846155e-06,
|
| 255658 |
+
"loss": 0.3252,
|
| 255659 |
+
"step": 122840
|
| 255660 |
+
},
|
| 255661 |
+
{
|
| 255662 |
+
"epoch": 982.14,
|
| 255663 |
+
"learning_rate": 8.048141025641026e-06,
|
| 255664 |
+
"loss": 0.339,
|
| 255665 |
+
"step": 122845
|
| 255666 |
+
},
|
| 255667 |
+
{
|
| 255668 |
+
"epoch": 982.18,
|
| 255669 |
+
"learning_rate": 8.048060897435898e-06,
|
| 255670 |
+
"loss": 0.6332,
|
| 255671 |
+
"step": 122850
|
| 255672 |
+
},
|
| 255673 |
+
{
|
| 255674 |
+
"epoch": 982.22,
|
| 255675 |
+
"learning_rate": 8.04798076923077e-06,
|
| 255676 |
+
"loss": 0.9975,
|
| 255677 |
+
"step": 122855
|
| 255678 |
+
},
|
| 255679 |
+
{
|
| 255680 |
+
"epoch": 982.25,
|
| 255681 |
+
"learning_rate": 8.047900641025642e-06,
|
| 255682 |
+
"loss": 0.2914,
|
| 255683 |
+
"step": 122860
|
| 255684 |
+
},
|
| 255685 |
+
{
|
| 255686 |
+
"epoch": 982.29,
|
| 255687 |
+
"learning_rate": 8.047820512820513e-06,
|
| 255688 |
+
"loss": 0.244,
|
| 255689 |
+
"step": 122865
|
| 255690 |
+
},
|
| 255691 |
+
{
|
| 255692 |
+
"epoch": 982.33,
|
| 255693 |
+
"learning_rate": 8.047740384615386e-06,
|
| 255694 |
+
"loss": 0.3577,
|
| 255695 |
+
"step": 122870
|
| 255696 |
+
},
|
| 255697 |
+
{
|
| 255698 |
+
"epoch": 982.37,
|
| 255699 |
+
"learning_rate": 8.047660256410256e-06,
|
| 255700 |
+
"loss": 0.6809,
|
| 255701 |
+
"step": 122875
|
| 255702 |
+
},
|
| 255703 |
+
{
|
| 255704 |
+
"epoch": 982.41,
|
| 255705 |
+
"learning_rate": 8.047580128205129e-06,
|
| 255706 |
+
"loss": 1.0051,
|
| 255707 |
+
"step": 122880
|
| 255708 |
+
},
|
| 255709 |
+
{
|
| 255710 |
+
"epoch": 982.45,
|
| 255711 |
+
"learning_rate": 8.0475e-06,
|
| 255712 |
+
"loss": 0.2567,
|
| 255713 |
+
"step": 122885
|
| 255714 |
+
},
|
| 255715 |
+
{
|
| 255716 |
+
"epoch": 982.49,
|
| 255717 |
+
"learning_rate": 8.047419871794872e-06,
|
| 255718 |
+
"loss": 0.2913,
|
| 255719 |
+
"step": 122890
|
| 255720 |
+
},
|
| 255721 |
+
{
|
| 255722 |
+
"epoch": 982.53,
|
| 255723 |
+
"learning_rate": 8.047339743589743e-06,
|
| 255724 |
+
"loss": 0.3565,
|
| 255725 |
+
"step": 122895
|
| 255726 |
+
},
|
| 255727 |
+
{
|
| 255728 |
+
"epoch": 982.57,
|
| 255729 |
+
"learning_rate": 8.047259615384616e-06,
|
| 255730 |
+
"loss": 0.5235,
|
| 255731 |
+
"step": 122900
|
| 255732 |
+
},
|
| 255733 |
+
{
|
| 255734 |
+
"epoch": 982.61,
|
| 255735 |
+
"learning_rate": 8.047179487179488e-06,
|
| 255736 |
+
"loss": 1.0516,
|
| 255737 |
+
"step": 122905
|
| 255738 |
+
},
|
| 255739 |
+
{
|
| 255740 |
+
"epoch": 982.65,
|
| 255741 |
+
"learning_rate": 8.047099358974359e-06,
|
| 255742 |
+
"loss": 0.3071,
|
| 255743 |
+
"step": 122910
|
| 255744 |
+
},
|
| 255745 |
+
{
|
| 255746 |
+
"epoch": 982.69,
|
| 255747 |
+
"learning_rate": 8.047019230769232e-06,
|
| 255748 |
+
"loss": 0.3203,
|
| 255749 |
+
"step": 122915
|
| 255750 |
+
},
|
| 255751 |
+
{
|
| 255752 |
+
"epoch": 982.73,
|
| 255753 |
+
"learning_rate": 8.046939102564103e-06,
|
| 255754 |
+
"loss": 0.3126,
|
| 255755 |
+
"step": 122920
|
| 255756 |
+
},
|
| 255757 |
+
{
|
| 255758 |
+
"epoch": 982.77,
|
| 255759 |
+
"learning_rate": 8.046858974358975e-06,
|
| 255760 |
+
"loss": 0.6228,
|
| 255761 |
+
"step": 122925
|
| 255762 |
+
},
|
| 255763 |
+
{
|
| 255764 |
+
"epoch": 982.81,
|
| 255765 |
+
"learning_rate": 8.046778846153846e-06,
|
| 255766 |
+
"loss": 1.1762,
|
| 255767 |
+
"step": 122930
|
| 255768 |
+
},
|
| 255769 |
+
{
|
| 255770 |
+
"epoch": 982.85,
|
| 255771 |
+
"learning_rate": 8.046698717948719e-06,
|
| 255772 |
+
"loss": 0.3309,
|
| 255773 |
+
"step": 122935
|
| 255774 |
+
},
|
| 255775 |
+
{
|
| 255776 |
+
"epoch": 982.89,
|
| 255777 |
+
"learning_rate": 8.04661858974359e-06,
|
| 255778 |
+
"loss": 0.3249,
|
| 255779 |
+
"step": 122940
|
| 255780 |
+
},
|
| 255781 |
+
{
|
| 255782 |
+
"epoch": 982.93,
|
| 255783 |
+
"learning_rate": 8.046538461538462e-06,
|
| 255784 |
+
"loss": 0.2933,
|
| 255785 |
+
"step": 122945
|
| 255786 |
+
},
|
| 255787 |
+
{
|
| 255788 |
+
"epoch": 982.97,
|
| 255789 |
+
"learning_rate": 8.046458333333333e-06,
|
| 255790 |
+
"loss": 0.5956,
|
| 255791 |
+
"step": 122950
|
| 255792 |
+
},
|
| 255793 |
+
{
|
| 255794 |
+
"epoch": 983.0,
|
| 255795 |
+
"eval_loss": 0.3855600655078888,
|
| 255796 |
+
"eval_runtime": 36.3167,
|
| 255797 |
+
"eval_samples_per_second": 23.157,
|
| 255798 |
+
"eval_steps_per_second": 0.743,
|
| 255799 |
+
"eval_wer": 0.17894194687660953,
|
| 255800 |
+
"step": 122953
|
| 255801 |
}
|
| 255802 |
],
|
| 255803 |
"max_steps": 625000,
|
| 255804 |
"num_train_epochs": 5000,
|
| 255805 |
+
"total_flos": 3.460155832498785e+20,
|
| 255806 |
"trial_name": null,
|
| 255807 |
"trial_params": null
|
| 255808 |
}
|
model-bin/finetune/base/{checkpoint-122329 → checkpoint-122953}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630136516.5433424/events.out.tfevents.1630136516.86bb0ddabf9b.4092.71
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c720ae8f3f663c46e47c64a41968f635ada06f70a44cbfaee684ff42371633ef
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630136910.4690464/events.out.tfevents.1630136910.86bb0ddabf9b.4092.73
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5645f603465af0185a8e14f27f4ef3e5948f7aa243b3cd05568f03090848ee38
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630137306.2544515/events.out.tfevents.1630137306.86bb0ddabf9b.4092.75
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4c0aa3e74249d8582fbde1a041c8f6728a470ffdeca6c3f3959ad14107f5ea88
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630137697.289981/events.out.tfevents.1630137697.86bb0ddabf9b.4092.77
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dfecf2e9f77790fbf3c7dc24e7d42f5436fc9586a1f2e62d6db6708263590445
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630138093.0549505/events.out.tfevents.1630138093.86bb0ddabf9b.4092.79
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1f102867fa3ccc11746db2c598edb7ab2d0fc0bdb099d879949e3912f2d580fe
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630136516.86bb0ddabf9b.4092.70
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:114a61b5b24fadd2b36f53d779d5547556f36153d5b84212d2274ca5c9669e03
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630136910.86bb0ddabf9b.4092.72
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c0cb3f940a09c1a5f95f8b11a986bb5820b55b6765ec53d554ba9ca37ce439a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630137306.86bb0ddabf9b.4092.74
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:13cdcd91a2a1e4d1ac1eb4ac4b34c6b1149f73afb59ae70d870de249a1c01668
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630137697.86bb0ddabf9b.4092.76
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:19168fff7a3c657201cd2f549b666b03e6a4c5abd02c21978c783c252d6703ea
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630138093.86bb0ddabf9b.4092.78
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0c207437829dbb9b293116bc9578f1132cef6faefc33070d168eea5e7406375
|
| 3 |
+
size 8622
|