"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629991265.2188263/events.out.tfevents.1629991265.8e89bd551565.924.231 +3 -0
- model-bin/finetune/base/log/1629991684.7530313/events.out.tfevents.1629991685.8e89bd551565.924.233 +3 -0
- model-bin/finetune/base/log/1629992115.6724274/events.out.tfevents.1629992115.8e89bd551565.924.235 +3 -0
- model-bin/finetune/base/log/1629992549.6115065/events.out.tfevents.1629992549.8e89bd551565.924.237 +3 -0
- model-bin/finetune/base/log/1629992990.702717/events.out.tfevents.1629992990.8e89bd551565.924.239 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629991265.8e89bd551565.924.230 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629991684.8e89bd551565.924.232 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629992115.8e89bd551565.924.234 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629992549.8e89bd551565.924.236 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629992990.8e89bd551565.924.238 +3 -0
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/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 sha256:9857b78c463eaa385a304848157348739a41eb8da315dfeeaa2d0a3ec71e75a8
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/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 sha256:14fb77230a7ee63d8bf5f9f184067800cd9ffc8bb19caf3281f975ac41a0fcb9
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
-
size
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:af11dafbe5807298fa685ae5cccc7951004d27d1d92f3f39b987c08874b4ccb8
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/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 sha256:69d87179115fbc6b93e2b095c017d1bc9093ab62fbae15eb403587041f7292fd
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/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:d5a33c6409adafb3ae851e601f2030d4c7390f928113bd1e1c69ada32803f1df
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/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,
|
|
@@ -229995,11 +229995,806 @@
|
|
| 229995 |
"eval_steps_per_second": 0.699,
|
| 229996 |
"eval_wer": 0.1812651176427472,
|
| 229997 |
"step": 102673
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 229998 |
}
|
| 229999 |
],
|
| 230000 |
"max_steps": 625000,
|
| 230001 |
"num_train_epochs": 5000,
|
| 230002 |
-
"total_flos": 2.
|
| 230003 |
"trial_name": null,
|
| 230004 |
"trial_params": null
|
| 230005 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 826.0,
|
| 5 |
+
"global_step": 103297,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 229995 |
"eval_steps_per_second": 0.699,
|
| 229996 |
"eval_wer": 0.1812651176427472,
|
| 229997 |
"step": 102673
|
| 229998 |
+
},
|
| 229999 |
+
{
|
| 230000 |
+
"epoch": 821.02,
|
| 230001 |
+
"learning_rate": 8.37125e-06,
|
| 230002 |
+
"loss": 0.446,
|
| 230003 |
+
"step": 102675
|
| 230004 |
+
},
|
| 230005 |
+
{
|
| 230006 |
+
"epoch": 821.06,
|
| 230007 |
+
"learning_rate": 8.371169871794872e-06,
|
| 230008 |
+
"loss": 0.2875,
|
| 230009 |
+
"step": 102680
|
| 230010 |
+
},
|
| 230011 |
+
{
|
| 230012 |
+
"epoch": 821.1,
|
| 230013 |
+
"learning_rate": 8.371089743589745e-06,
|
| 230014 |
+
"loss": 0.3178,
|
| 230015 |
+
"step": 102685
|
| 230016 |
+
},
|
| 230017 |
+
{
|
| 230018 |
+
"epoch": 821.14,
|
| 230019 |
+
"learning_rate": 8.371009615384616e-06,
|
| 230020 |
+
"loss": 0.3415,
|
| 230021 |
+
"step": 102690
|
| 230022 |
+
},
|
| 230023 |
+
{
|
| 230024 |
+
"epoch": 821.18,
|
| 230025 |
+
"learning_rate": 8.370929487179488e-06,
|
| 230026 |
+
"loss": 0.7006,
|
| 230027 |
+
"step": 102695
|
| 230028 |
+
},
|
| 230029 |
+
{
|
| 230030 |
+
"epoch": 821.22,
|
| 230031 |
+
"learning_rate": 8.37084935897436e-06,
|
| 230032 |
+
"loss": 1.073,
|
| 230033 |
+
"step": 102700
|
| 230034 |
+
},
|
| 230035 |
+
{
|
| 230036 |
+
"epoch": 821.26,
|
| 230037 |
+
"learning_rate": 8.37076923076923e-06,
|
| 230038 |
+
"loss": 0.3145,
|
| 230039 |
+
"step": 102705
|
| 230040 |
+
},
|
| 230041 |
+
{
|
| 230042 |
+
"epoch": 821.3,
|
| 230043 |
+
"learning_rate": 8.370689102564103e-06,
|
| 230044 |
+
"loss": 0.2885,
|
| 230045 |
+
"step": 102710
|
| 230046 |
+
},
|
| 230047 |
+
{
|
| 230048 |
+
"epoch": 821.34,
|
| 230049 |
+
"learning_rate": 8.370608974358975e-06,
|
| 230050 |
+
"loss": 0.3705,
|
| 230051 |
+
"step": 102715
|
| 230052 |
+
},
|
| 230053 |
+
{
|
| 230054 |
+
"epoch": 821.38,
|
| 230055 |
+
"learning_rate": 8.370528846153846e-06,
|
| 230056 |
+
"loss": 0.6628,
|
| 230057 |
+
"step": 102720
|
| 230058 |
+
},
|
| 230059 |
+
{
|
| 230060 |
+
"epoch": 821.42,
|
| 230061 |
+
"learning_rate": 8.370448717948718e-06,
|
| 230062 |
+
"loss": 1.0318,
|
| 230063 |
+
"step": 102725
|
| 230064 |
+
},
|
| 230065 |
+
{
|
| 230066 |
+
"epoch": 821.46,
|
| 230067 |
+
"learning_rate": 8.37036858974359e-06,
|
| 230068 |
+
"loss": 0.3119,
|
| 230069 |
+
"step": 102730
|
| 230070 |
+
},
|
| 230071 |
+
{
|
| 230072 |
+
"epoch": 821.5,
|
| 230073 |
+
"learning_rate": 8.370288461538462e-06,
|
| 230074 |
+
"loss": 0.3151,
|
| 230075 |
+
"step": 102735
|
| 230076 |
+
},
|
| 230077 |
+
{
|
| 230078 |
+
"epoch": 821.54,
|
| 230079 |
+
"learning_rate": 8.370208333333333e-06,
|
| 230080 |
+
"loss": 0.3474,
|
| 230081 |
+
"step": 102740
|
| 230082 |
+
},
|
| 230083 |
+
{
|
| 230084 |
+
"epoch": 821.58,
|
| 230085 |
+
"learning_rate": 8.370128205128206e-06,
|
| 230086 |
+
"loss": 0.6315,
|
| 230087 |
+
"step": 102745
|
| 230088 |
+
},
|
| 230089 |
+
{
|
| 230090 |
+
"epoch": 821.62,
|
| 230091 |
+
"learning_rate": 8.370048076923078e-06,
|
| 230092 |
+
"loss": 1.0638,
|
| 230093 |
+
"step": 102750
|
| 230094 |
+
},
|
| 230095 |
+
{
|
| 230096 |
+
"epoch": 821.66,
|
| 230097 |
+
"learning_rate": 8.369967948717949e-06,
|
| 230098 |
+
"loss": 0.3256,
|
| 230099 |
+
"step": 102755
|
| 230100 |
+
},
|
| 230101 |
+
{
|
| 230102 |
+
"epoch": 821.7,
|
| 230103 |
+
"learning_rate": 8.36988782051282e-06,
|
| 230104 |
+
"loss": 0.2922,
|
| 230105 |
+
"step": 102760
|
| 230106 |
+
},
|
| 230107 |
+
{
|
| 230108 |
+
"epoch": 821.74,
|
| 230109 |
+
"learning_rate": 8.369807692307693e-06,
|
| 230110 |
+
"loss": 0.5094,
|
| 230111 |
+
"step": 102765
|
| 230112 |
+
},
|
| 230113 |
+
{
|
| 230114 |
+
"epoch": 821.78,
|
| 230115 |
+
"learning_rate": 8.369727564102565e-06,
|
| 230116 |
+
"loss": 0.7097,
|
| 230117 |
+
"step": 102770
|
| 230118 |
+
},
|
| 230119 |
+
{
|
| 230120 |
+
"epoch": 821.82,
|
| 230121 |
+
"learning_rate": 8.369647435897436e-06,
|
| 230122 |
+
"loss": 0.9577,
|
| 230123 |
+
"step": 102775
|
| 230124 |
+
},
|
| 230125 |
+
{
|
| 230126 |
+
"epoch": 821.86,
|
| 230127 |
+
"learning_rate": 8.369567307692308e-06,
|
| 230128 |
+
"loss": 0.3296,
|
| 230129 |
+
"step": 102780
|
| 230130 |
+
},
|
| 230131 |
+
{
|
| 230132 |
+
"epoch": 821.9,
|
| 230133 |
+
"learning_rate": 8.36948717948718e-06,
|
| 230134 |
+
"loss": 0.2805,
|
| 230135 |
+
"step": 102785
|
| 230136 |
+
},
|
| 230137 |
+
{
|
| 230138 |
+
"epoch": 821.94,
|
| 230139 |
+
"learning_rate": 8.369407051282052e-06,
|
| 230140 |
+
"loss": 0.4308,
|
| 230141 |
+
"step": 102790
|
| 230142 |
+
},
|
| 230143 |
+
{
|
| 230144 |
+
"epoch": 821.98,
|
| 230145 |
+
"learning_rate": 8.369326923076923e-06,
|
| 230146 |
+
"loss": 0.6183,
|
| 230147 |
+
"step": 102795
|
| 230148 |
+
},
|
| 230149 |
+
{
|
| 230150 |
+
"epoch": 822.0,
|
| 230151 |
+
"eval_loss": 0.3813929855823517,
|
| 230152 |
+
"eval_runtime": 39.3722,
|
| 230153 |
+
"eval_samples_per_second": 21.335,
|
| 230154 |
+
"eval_steps_per_second": 0.686,
|
| 230155 |
+
"eval_wer": 0.1866103202846975,
|
| 230156 |
+
"step": 102798
|
| 230157 |
+
},
|
| 230158 |
+
{
|
| 230159 |
+
"epoch": 822.02,
|
| 230160 |
+
"learning_rate": 8.369246794871796e-06,
|
| 230161 |
+
"loss": 0.3868,
|
| 230162 |
+
"step": 102800
|
| 230163 |
+
},
|
| 230164 |
+
{
|
| 230165 |
+
"epoch": 822.06,
|
| 230166 |
+
"learning_rate": 8.369166666666668e-06,
|
| 230167 |
+
"loss": 0.4333,
|
| 230168 |
+
"step": 102805
|
| 230169 |
+
},
|
| 230170 |
+
{
|
| 230171 |
+
"epoch": 822.1,
|
| 230172 |
+
"learning_rate": 8.369086538461539e-06,
|
| 230173 |
+
"loss": 0.2607,
|
| 230174 |
+
"step": 102810
|
| 230175 |
+
},
|
| 230176 |
+
{
|
| 230177 |
+
"epoch": 822.14,
|
| 230178 |
+
"learning_rate": 8.36900641025641e-06,
|
| 230179 |
+
"loss": 0.3474,
|
| 230180 |
+
"step": 102815
|
| 230181 |
+
},
|
| 230182 |
+
{
|
| 230183 |
+
"epoch": 822.18,
|
| 230184 |
+
"learning_rate": 8.368926282051283e-06,
|
| 230185 |
+
"loss": 0.6296,
|
| 230186 |
+
"step": 102820
|
| 230187 |
+
},
|
| 230188 |
+
{
|
| 230189 |
+
"epoch": 822.22,
|
| 230190 |
+
"learning_rate": 8.368846153846155e-06,
|
| 230191 |
+
"loss": 1.1556,
|
| 230192 |
+
"step": 102825
|
| 230193 |
+
},
|
| 230194 |
+
{
|
| 230195 |
+
"epoch": 822.25,
|
| 230196 |
+
"learning_rate": 8.368766025641026e-06,
|
| 230197 |
+
"loss": 0.2646,
|
| 230198 |
+
"step": 102830
|
| 230199 |
+
},
|
| 230200 |
+
{
|
| 230201 |
+
"epoch": 822.29,
|
| 230202 |
+
"learning_rate": 8.3686858974359e-06,
|
| 230203 |
+
"loss": 0.2644,
|
| 230204 |
+
"step": 102835
|
| 230205 |
+
},
|
| 230206 |
+
{
|
| 230207 |
+
"epoch": 822.33,
|
| 230208 |
+
"learning_rate": 8.36860576923077e-06,
|
| 230209 |
+
"loss": 0.3483,
|
| 230210 |
+
"step": 102840
|
| 230211 |
+
},
|
| 230212 |
+
{
|
| 230213 |
+
"epoch": 822.37,
|
| 230214 |
+
"learning_rate": 8.368525641025642e-06,
|
| 230215 |
+
"loss": 0.7282,
|
| 230216 |
+
"step": 102845
|
| 230217 |
+
},
|
| 230218 |
+
{
|
| 230219 |
+
"epoch": 822.41,
|
| 230220 |
+
"learning_rate": 8.368445512820513e-06,
|
| 230221 |
+
"loss": 1.0165,
|
| 230222 |
+
"step": 102850
|
| 230223 |
+
},
|
| 230224 |
+
{
|
| 230225 |
+
"epoch": 822.45,
|
| 230226 |
+
"learning_rate": 8.368365384615386e-06,
|
| 230227 |
+
"loss": 0.27,
|
| 230228 |
+
"step": 102855
|
| 230229 |
+
},
|
| 230230 |
+
{
|
| 230231 |
+
"epoch": 822.49,
|
| 230232 |
+
"learning_rate": 8.368285256410256e-06,
|
| 230233 |
+
"loss": 0.3203,
|
| 230234 |
+
"step": 102860
|
| 230235 |
+
},
|
| 230236 |
+
{
|
| 230237 |
+
"epoch": 822.53,
|
| 230238 |
+
"learning_rate": 8.368205128205129e-06,
|
| 230239 |
+
"loss": 0.3605,
|
| 230240 |
+
"step": 102865
|
| 230241 |
+
},
|
| 230242 |
+
{
|
| 230243 |
+
"epoch": 822.57,
|
| 230244 |
+
"learning_rate": 8.368125e-06,
|
| 230245 |
+
"loss": 0.6079,
|
| 230246 |
+
"step": 102870
|
| 230247 |
+
},
|
| 230248 |
+
{
|
| 230249 |
+
"epoch": 822.61,
|
| 230250 |
+
"learning_rate": 8.368044871794872e-06,
|
| 230251 |
+
"loss": 0.9338,
|
| 230252 |
+
"step": 102875
|
| 230253 |
+
},
|
| 230254 |
+
{
|
| 230255 |
+
"epoch": 822.65,
|
| 230256 |
+
"learning_rate": 8.367964743589743e-06,
|
| 230257 |
+
"loss": 0.361,
|
| 230258 |
+
"step": 102880
|
| 230259 |
+
},
|
| 230260 |
+
{
|
| 230261 |
+
"epoch": 822.69,
|
| 230262 |
+
"learning_rate": 8.367884615384616e-06,
|
| 230263 |
+
"loss": 0.2758,
|
| 230264 |
+
"step": 102885
|
| 230265 |
+
},
|
| 230266 |
+
{
|
| 230267 |
+
"epoch": 822.73,
|
| 230268 |
+
"learning_rate": 8.367804487179488e-06,
|
| 230269 |
+
"loss": 0.3147,
|
| 230270 |
+
"step": 102890
|
| 230271 |
+
},
|
| 230272 |
+
{
|
| 230273 |
+
"epoch": 822.77,
|
| 230274 |
+
"learning_rate": 8.367724358974359e-06,
|
| 230275 |
+
"loss": 0.57,
|
| 230276 |
+
"step": 102895
|
| 230277 |
+
},
|
| 230278 |
+
{
|
| 230279 |
+
"epoch": 822.81,
|
| 230280 |
+
"learning_rate": 8.367644230769232e-06,
|
| 230281 |
+
"loss": 1.06,
|
| 230282 |
+
"step": 102900
|
| 230283 |
+
},
|
| 230284 |
+
{
|
| 230285 |
+
"epoch": 822.85,
|
| 230286 |
+
"learning_rate": 8.367564102564103e-06,
|
| 230287 |
+
"loss": 0.3122,
|
| 230288 |
+
"step": 102905
|
| 230289 |
+
},
|
| 230290 |
+
{
|
| 230291 |
+
"epoch": 822.89,
|
| 230292 |
+
"learning_rate": 8.367483974358975e-06,
|
| 230293 |
+
"loss": 0.3036,
|
| 230294 |
+
"step": 102910
|
| 230295 |
+
},
|
| 230296 |
+
{
|
| 230297 |
+
"epoch": 822.93,
|
| 230298 |
+
"learning_rate": 8.367403846153846e-06,
|
| 230299 |
+
"loss": 0.3218,
|
| 230300 |
+
"step": 102915
|
| 230301 |
+
},
|
| 230302 |
+
{
|
| 230303 |
+
"epoch": 822.97,
|
| 230304 |
+
"learning_rate": 8.367323717948719e-06,
|
| 230305 |
+
"loss": 0.6319,
|
| 230306 |
+
"step": 102920
|
| 230307 |
+
},
|
| 230308 |
+
{
|
| 230309 |
+
"epoch": 823.0,
|
| 230310 |
+
"eval_loss": 0.4378843903541565,
|
| 230311 |
+
"eval_runtime": 39.7689,
|
| 230312 |
+
"eval_samples_per_second": 21.122,
|
| 230313 |
+
"eval_steps_per_second": 0.679,
|
| 230314 |
+
"eval_wer": 0.18468633335773368,
|
| 230315 |
+
"step": 102923
|
| 230316 |
+
},
|
| 230317 |
+
{
|
| 230318 |
+
"epoch": 830.02,
|
| 230319 |
+
"learning_rate": 8.36724358974359e-06,
|
| 230320 |
+
"loss": 0.3786,
|
| 230321 |
+
"step": 102925
|
| 230322 |
+
},
|
| 230323 |
+
{
|
| 230324 |
+
"epoch": 830.06,
|
| 230325 |
+
"learning_rate": 8.367163461538462e-06,
|
| 230326 |
+
"loss": 0.2767,
|
| 230327 |
+
"step": 102930
|
| 230328 |
+
},
|
| 230329 |
+
{
|
| 230330 |
+
"epoch": 830.1,
|
| 230331 |
+
"learning_rate": 8.367083333333335e-06,
|
| 230332 |
+
"loss": 0.2652,
|
| 230333 |
+
"step": 102935
|
| 230334 |
+
},
|
| 230335 |
+
{
|
| 230336 |
+
"epoch": 830.14,
|
| 230337 |
+
"learning_rate": 8.367003205128206e-06,
|
| 230338 |
+
"loss": 0.3667,
|
| 230339 |
+
"step": 102940
|
| 230340 |
+
},
|
| 230341 |
+
{
|
| 230342 |
+
"epoch": 830.18,
|
| 230343 |
+
"learning_rate": 8.366923076923078e-06,
|
| 230344 |
+
"loss": 0.6871,
|
| 230345 |
+
"step": 102945
|
| 230346 |
+
},
|
| 230347 |
+
{
|
| 230348 |
+
"epoch": 830.22,
|
| 230349 |
+
"learning_rate": 8.366842948717949e-06,
|
| 230350 |
+
"loss": 1.112,
|
| 230351 |
+
"step": 102950
|
| 230352 |
+
},
|
| 230353 |
+
{
|
| 230354 |
+
"epoch": 830.26,
|
| 230355 |
+
"learning_rate": 8.366762820512822e-06,
|
| 230356 |
+
"loss": 0.2261,
|
| 230357 |
+
"step": 102955
|
| 230358 |
+
},
|
| 230359 |
+
{
|
| 230360 |
+
"epoch": 830.3,
|
| 230361 |
+
"learning_rate": 8.366682692307693e-06,
|
| 230362 |
+
"loss": 0.2711,
|
| 230363 |
+
"step": 102960
|
| 230364 |
+
},
|
| 230365 |
+
{
|
| 230366 |
+
"epoch": 830.34,
|
| 230367 |
+
"learning_rate": 8.366602564102565e-06,
|
| 230368 |
+
"loss": 0.3583,
|
| 230369 |
+
"step": 102965
|
| 230370 |
+
},
|
| 230371 |
+
{
|
| 230372 |
+
"epoch": 830.38,
|
| 230373 |
+
"learning_rate": 8.366522435897436e-06,
|
| 230374 |
+
"loss": 0.5891,
|
| 230375 |
+
"step": 102970
|
| 230376 |
+
},
|
| 230377 |
+
{
|
| 230378 |
+
"epoch": 830.42,
|
| 230379 |
+
"learning_rate": 8.366442307692309e-06,
|
| 230380 |
+
"loss": 0.952,
|
| 230381 |
+
"step": 102975
|
| 230382 |
+
},
|
| 230383 |
+
{
|
| 230384 |
+
"epoch": 830.46,
|
| 230385 |
+
"learning_rate": 8.366362179487179e-06,
|
| 230386 |
+
"loss": 0.2706,
|
| 230387 |
+
"step": 102980
|
| 230388 |
+
},
|
| 230389 |
+
{
|
| 230390 |
+
"epoch": 830.5,
|
| 230391 |
+
"learning_rate": 8.366282051282052e-06,
|
| 230392 |
+
"loss": 0.2699,
|
| 230393 |
+
"step": 102985
|
| 230394 |
+
},
|
| 230395 |
+
{
|
| 230396 |
+
"epoch": 830.54,
|
| 230397 |
+
"learning_rate": 8.366201923076925e-06,
|
| 230398 |
+
"loss": 0.33,
|
| 230399 |
+
"step": 102990
|
| 230400 |
+
},
|
| 230401 |
+
{
|
| 230402 |
+
"epoch": 830.58,
|
| 230403 |
+
"learning_rate": 8.366121794871796e-06,
|
| 230404 |
+
"loss": 0.6562,
|
| 230405 |
+
"step": 102995
|
| 230406 |
+
},
|
| 230407 |
+
{
|
| 230408 |
+
"epoch": 830.62,
|
| 230409 |
+
"learning_rate": 8.366041666666668e-06,
|
| 230410 |
+
"loss": 0.9395,
|
| 230411 |
+
"step": 103000
|
| 230412 |
+
},
|
| 230413 |
+
{
|
| 230414 |
+
"epoch": 830.66,
|
| 230415 |
+
"learning_rate": 8.365961538461539e-06,
|
| 230416 |
+
"loss": 0.2931,
|
| 230417 |
+
"step": 103005
|
| 230418 |
+
},
|
| 230419 |
+
{
|
| 230420 |
+
"epoch": 830.7,
|
| 230421 |
+
"learning_rate": 8.365881410256412e-06,
|
| 230422 |
+
"loss": 0.2713,
|
| 230423 |
+
"step": 103010
|
| 230424 |
+
},
|
| 230425 |
+
{
|
| 230426 |
+
"epoch": 830.74,
|
| 230427 |
+
"learning_rate": 8.365801282051282e-06,
|
| 230428 |
+
"loss": 0.3263,
|
| 230429 |
+
"step": 103015
|
| 230430 |
+
},
|
| 230431 |
+
{
|
| 230432 |
+
"epoch": 830.78,
|
| 230433 |
+
"learning_rate": 8.365721153846155e-06,
|
| 230434 |
+
"loss": 0.6797,
|
| 230435 |
+
"step": 103020
|
| 230436 |
+
},
|
| 230437 |
+
{
|
| 230438 |
+
"epoch": 830.82,
|
| 230439 |
+
"learning_rate": 8.365641025641026e-06,
|
| 230440 |
+
"loss": 0.9683,
|
| 230441 |
+
"step": 103025
|
| 230442 |
+
},
|
| 230443 |
+
{
|
| 230444 |
+
"epoch": 830.86,
|
| 230445 |
+
"learning_rate": 8.365560897435897e-06,
|
| 230446 |
+
"loss": 0.2865,
|
| 230447 |
+
"step": 103030
|
| 230448 |
+
},
|
| 230449 |
+
{
|
| 230450 |
+
"epoch": 830.9,
|
| 230451 |
+
"learning_rate": 8.36548076923077e-06,
|
| 230452 |
+
"loss": 0.2816,
|
| 230453 |
+
"step": 103035
|
| 230454 |
+
},
|
| 230455 |
+
{
|
| 230456 |
+
"epoch": 830.94,
|
| 230457 |
+
"learning_rate": 8.365400641025642e-06,
|
| 230458 |
+
"loss": 0.3849,
|
| 230459 |
+
"step": 103040
|
| 230460 |
+
},
|
| 230461 |
+
{
|
| 230462 |
+
"epoch": 830.98,
|
| 230463 |
+
"learning_rate": 8.365320512820513e-06,
|
| 230464 |
+
"loss": 0.8167,
|
| 230465 |
+
"step": 103045
|
| 230466 |
+
},
|
| 230467 |
+
{
|
| 230468 |
+
"epoch": 831.0,
|
| 230469 |
+
"eval_loss": 0.34371402859687805,
|
| 230470 |
+
"eval_runtime": 41.2146,
|
| 230471 |
+
"eval_samples_per_second": 20.381,
|
| 230472 |
+
"eval_steps_per_second": 0.655,
|
| 230473 |
+
"eval_wer": 0.18333213178574004,
|
| 230474 |
+
"step": 103047
|
| 230475 |
+
},
|
| 230476 |
+
{
|
| 230477 |
+
"epoch": 824.02,
|
| 230478 |
+
"learning_rate": 8.365240384615385e-06,
|
| 230479 |
+
"loss": 0.3935,
|
| 230480 |
+
"step": 103050
|
| 230481 |
+
},
|
| 230482 |
+
{
|
| 230483 |
+
"epoch": 824.06,
|
| 230484 |
+
"learning_rate": 8.365160256410258e-06,
|
| 230485 |
+
"loss": 0.2855,
|
| 230486 |
+
"step": 103055
|
| 230487 |
+
},
|
| 230488 |
+
{
|
| 230489 |
+
"epoch": 824.1,
|
| 230490 |
+
"learning_rate": 8.365080128205129e-06,
|
| 230491 |
+
"loss": 0.2965,
|
| 230492 |
+
"step": 103060
|
| 230493 |
+
},
|
| 230494 |
+
{
|
| 230495 |
+
"epoch": 824.14,
|
| 230496 |
+
"learning_rate": 8.365e-06,
|
| 230497 |
+
"loss": 0.4175,
|
| 230498 |
+
"step": 103065
|
| 230499 |
+
},
|
| 230500 |
+
{
|
| 230501 |
+
"epoch": 824.18,
|
| 230502 |
+
"learning_rate": 8.364919871794872e-06,
|
| 230503 |
+
"loss": 0.7754,
|
| 230504 |
+
"step": 103070
|
| 230505 |
+
},
|
| 230506 |
+
{
|
| 230507 |
+
"epoch": 824.22,
|
| 230508 |
+
"learning_rate": 8.364839743589745e-06,
|
| 230509 |
+
"loss": 0.8846,
|
| 230510 |
+
"step": 103075
|
| 230511 |
+
},
|
| 230512 |
+
{
|
| 230513 |
+
"epoch": 824.26,
|
| 230514 |
+
"learning_rate": 8.364759615384616e-06,
|
| 230515 |
+
"loss": 0.2775,
|
| 230516 |
+
"step": 103080
|
| 230517 |
+
},
|
| 230518 |
+
{
|
| 230519 |
+
"epoch": 824.3,
|
| 230520 |
+
"learning_rate": 8.364679487179488e-06,
|
| 230521 |
+
"loss": 0.3054,
|
| 230522 |
+
"step": 103085
|
| 230523 |
+
},
|
| 230524 |
+
{
|
| 230525 |
+
"epoch": 824.34,
|
| 230526 |
+
"learning_rate": 8.36459935897436e-06,
|
| 230527 |
+
"loss": 0.4014,
|
| 230528 |
+
"step": 103090
|
| 230529 |
+
},
|
| 230530 |
+
{
|
| 230531 |
+
"epoch": 824.38,
|
| 230532 |
+
"learning_rate": 8.364519230769232e-06,
|
| 230533 |
+
"loss": 0.7294,
|
| 230534 |
+
"step": 103095
|
| 230535 |
+
},
|
| 230536 |
+
{
|
| 230537 |
+
"epoch": 824.42,
|
| 230538 |
+
"learning_rate": 8.364439102564103e-06,
|
| 230539 |
+
"loss": 0.9003,
|
| 230540 |
+
"step": 103100
|
| 230541 |
+
},
|
| 230542 |
+
{
|
| 230543 |
+
"epoch": 824.46,
|
| 230544 |
+
"learning_rate": 8.364358974358975e-06,
|
| 230545 |
+
"loss": 0.2668,
|
| 230546 |
+
"step": 103105
|
| 230547 |
+
},
|
| 230548 |
+
{
|
| 230549 |
+
"epoch": 824.5,
|
| 230550 |
+
"learning_rate": 8.364278846153848e-06,
|
| 230551 |
+
"loss": 0.2774,
|
| 230552 |
+
"step": 103110
|
| 230553 |
+
},
|
| 230554 |
+
{
|
| 230555 |
+
"epoch": 824.54,
|
| 230556 |
+
"learning_rate": 8.364198717948719e-06,
|
| 230557 |
+
"loss": 0.3793,
|
| 230558 |
+
"step": 103115
|
| 230559 |
+
},
|
| 230560 |
+
{
|
| 230561 |
+
"epoch": 824.58,
|
| 230562 |
+
"learning_rate": 8.36411858974359e-06,
|
| 230563 |
+
"loss": 0.8735,
|
| 230564 |
+
"step": 103120
|
| 230565 |
+
},
|
| 230566 |
+
{
|
| 230567 |
+
"epoch": 824.62,
|
| 230568 |
+
"learning_rate": 8.364038461538462e-06,
|
| 230569 |
+
"loss": 0.799,
|
| 230570 |
+
"step": 103125
|
| 230571 |
+
},
|
| 230572 |
+
{
|
| 230573 |
+
"epoch": 824.66,
|
| 230574 |
+
"learning_rate": 8.363958333333335e-06,
|
| 230575 |
+
"loss": 0.2933,
|
| 230576 |
+
"step": 103130
|
| 230577 |
+
},
|
| 230578 |
+
{
|
| 230579 |
+
"epoch": 824.7,
|
| 230580 |
+
"learning_rate": 8.363878205128206e-06,
|
| 230581 |
+
"loss": 0.3628,
|
| 230582 |
+
"step": 103135
|
| 230583 |
+
},
|
| 230584 |
+
{
|
| 230585 |
+
"epoch": 824.74,
|
| 230586 |
+
"learning_rate": 8.363798076923078e-06,
|
| 230587 |
+
"loss": 0.3653,
|
| 230588 |
+
"step": 103140
|
| 230589 |
+
},
|
| 230590 |
+
{
|
| 230591 |
+
"epoch": 824.78,
|
| 230592 |
+
"learning_rate": 8.36371794871795e-06,
|
| 230593 |
+
"loss": 0.8661,
|
| 230594 |
+
"step": 103145
|
| 230595 |
+
},
|
| 230596 |
+
{
|
| 230597 |
+
"epoch": 824.82,
|
| 230598 |
+
"learning_rate": 8.36363782051282e-06,
|
| 230599 |
+
"loss": 0.8644,
|
| 230600 |
+
"step": 103150
|
| 230601 |
+
},
|
| 230602 |
+
{
|
| 230603 |
+
"epoch": 824.86,
|
| 230604 |
+
"learning_rate": 8.363557692307693e-06,
|
| 230605 |
+
"loss": 0.2742,
|
| 230606 |
+
"step": 103155
|
| 230607 |
+
},
|
| 230608 |
+
{
|
| 230609 |
+
"epoch": 824.9,
|
| 230610 |
+
"learning_rate": 8.363477564102565e-06,
|
| 230611 |
+
"loss": 0.29,
|
| 230612 |
+
"step": 103160
|
| 230613 |
+
},
|
| 230614 |
+
{
|
| 230615 |
+
"epoch": 824.94,
|
| 230616 |
+
"learning_rate": 8.363397435897436e-06,
|
| 230617 |
+
"loss": 0.345,
|
| 230618 |
+
"step": 103165
|
| 230619 |
+
},
|
| 230620 |
+
{
|
| 230621 |
+
"epoch": 824.98,
|
| 230622 |
+
"learning_rate": 8.363317307692307e-06,
|
| 230623 |
+
"loss": 0.7565,
|
| 230624 |
+
"step": 103170
|
| 230625 |
+
},
|
| 230626 |
+
{
|
| 230627 |
+
"epoch": 825.0,
|
| 230628 |
+
"eval_loss": 0.43166759610176086,
|
| 230629 |
+
"eval_runtime": 40.4182,
|
| 230630 |
+
"eval_samples_per_second": 20.783,
|
| 230631 |
+
"eval_steps_per_second": 0.668,
|
| 230632 |
+
"eval_wer": 0.18569872958257713,
|
| 230633 |
+
"step": 103172
|
| 230634 |
+
},
|
| 230635 |
+
{
|
| 230636 |
+
"epoch": 825.02,
|
| 230637 |
+
"learning_rate": 8.36323717948718e-06,
|
| 230638 |
+
"loss": 0.3739,
|
| 230639 |
+
"step": 103175
|
| 230640 |
+
},
|
| 230641 |
+
{
|
| 230642 |
+
"epoch": 825.06,
|
| 230643 |
+
"learning_rate": 8.363157051282052e-06,
|
| 230644 |
+
"loss": 0.3119,
|
| 230645 |
+
"step": 103180
|
| 230646 |
+
},
|
| 230647 |
+
{
|
| 230648 |
+
"epoch": 825.1,
|
| 230649 |
+
"learning_rate": 8.363076923076923e-06,
|
| 230650 |
+
"loss": 0.31,
|
| 230651 |
+
"step": 103185
|
| 230652 |
+
},
|
| 230653 |
+
{
|
| 230654 |
+
"epoch": 825.14,
|
| 230655 |
+
"learning_rate": 8.362996794871796e-06,
|
| 230656 |
+
"loss": 0.3199,
|
| 230657 |
+
"step": 103190
|
| 230658 |
+
},
|
| 230659 |
+
{
|
| 230660 |
+
"epoch": 825.18,
|
| 230661 |
+
"learning_rate": 8.362916666666668e-06,
|
| 230662 |
+
"loss": 0.7039,
|
| 230663 |
+
"step": 103195
|
| 230664 |
+
},
|
| 230665 |
+
{
|
| 230666 |
+
"epoch": 825.22,
|
| 230667 |
+
"learning_rate": 8.362836538461539e-06,
|
| 230668 |
+
"loss": 0.8929,
|
| 230669 |
+
"step": 103200
|
| 230670 |
+
},
|
| 230671 |
+
{
|
| 230672 |
+
"epoch": 825.26,
|
| 230673 |
+
"learning_rate": 8.36275641025641e-06,
|
| 230674 |
+
"loss": 0.315,
|
| 230675 |
+
"step": 103205
|
| 230676 |
+
},
|
| 230677 |
+
{
|
| 230678 |
+
"epoch": 825.3,
|
| 230679 |
+
"learning_rate": 8.362676282051283e-06,
|
| 230680 |
+
"loss": 0.3098,
|
| 230681 |
+
"step": 103210
|
| 230682 |
+
},
|
| 230683 |
+
{
|
| 230684 |
+
"epoch": 825.34,
|
| 230685 |
+
"learning_rate": 8.362596153846155e-06,
|
| 230686 |
+
"loss": 0.3446,
|
| 230687 |
+
"step": 103215
|
| 230688 |
+
},
|
| 230689 |
+
{
|
| 230690 |
+
"epoch": 825.38,
|
| 230691 |
+
"learning_rate": 8.362516025641026e-06,
|
| 230692 |
+
"loss": 0.8006,
|
| 230693 |
+
"step": 103220
|
| 230694 |
+
},
|
| 230695 |
+
{
|
| 230696 |
+
"epoch": 825.42,
|
| 230697 |
+
"learning_rate": 8.362435897435897e-06,
|
| 230698 |
+
"loss": 0.8486,
|
| 230699 |
+
"step": 103225
|
| 230700 |
+
},
|
| 230701 |
+
{
|
| 230702 |
+
"epoch": 825.46,
|
| 230703 |
+
"learning_rate": 8.36235576923077e-06,
|
| 230704 |
+
"loss": 0.2663,
|
| 230705 |
+
"step": 103230
|
| 230706 |
+
},
|
| 230707 |
+
{
|
| 230708 |
+
"epoch": 825.5,
|
| 230709 |
+
"learning_rate": 8.362275641025642e-06,
|
| 230710 |
+
"loss": 0.3005,
|
| 230711 |
+
"step": 103235
|
| 230712 |
+
},
|
| 230713 |
+
{
|
| 230714 |
+
"epoch": 825.54,
|
| 230715 |
+
"learning_rate": 8.362195512820513e-06,
|
| 230716 |
+
"loss": 0.3577,
|
| 230717 |
+
"step": 103240
|
| 230718 |
+
},
|
| 230719 |
+
{
|
| 230720 |
+
"epoch": 825.58,
|
| 230721 |
+
"learning_rate": 8.362115384615386e-06,
|
| 230722 |
+
"loss": 0.6979,
|
| 230723 |
+
"step": 103245
|
| 230724 |
+
},
|
| 230725 |
+
{
|
| 230726 |
+
"epoch": 825.62,
|
| 230727 |
+
"learning_rate": 8.362035256410258e-06,
|
| 230728 |
+
"loss": 0.9191,
|
| 230729 |
+
"step": 103250
|
| 230730 |
+
},
|
| 230731 |
+
{
|
| 230732 |
+
"epoch": 825.66,
|
| 230733 |
+
"learning_rate": 8.361955128205129e-06,
|
| 230734 |
+
"loss": 0.378,
|
| 230735 |
+
"step": 103255
|
| 230736 |
+
},
|
| 230737 |
+
{
|
| 230738 |
+
"epoch": 825.7,
|
| 230739 |
+
"learning_rate": 8.361875e-06,
|
| 230740 |
+
"loss": 0.4107,
|
| 230741 |
+
"step": 103260
|
| 230742 |
+
},
|
| 230743 |
+
{
|
| 230744 |
+
"epoch": 825.74,
|
| 230745 |
+
"learning_rate": 8.361794871794873e-06,
|
| 230746 |
+
"loss": 0.3645,
|
| 230747 |
+
"step": 103265
|
| 230748 |
+
},
|
| 230749 |
+
{
|
| 230750 |
+
"epoch": 825.78,
|
| 230751 |
+
"learning_rate": 8.361714743589745e-06,
|
| 230752 |
+
"loss": 0.7475,
|
| 230753 |
+
"step": 103270
|
| 230754 |
+
},
|
| 230755 |
+
{
|
| 230756 |
+
"epoch": 825.82,
|
| 230757 |
+
"learning_rate": 8.361634615384616e-06,
|
| 230758 |
+
"loss": 0.9194,
|
| 230759 |
+
"step": 103275
|
| 230760 |
+
},
|
| 230761 |
+
{
|
| 230762 |
+
"epoch": 825.86,
|
| 230763 |
+
"learning_rate": 8.361554487179487e-06,
|
| 230764 |
+
"loss": 0.2677,
|
| 230765 |
+
"step": 103280
|
| 230766 |
+
},
|
| 230767 |
+
{
|
| 230768 |
+
"epoch": 825.9,
|
| 230769 |
+
"learning_rate": 8.36147435897436e-06,
|
| 230770 |
+
"loss": 0.3263,
|
| 230771 |
+
"step": 103285
|
| 230772 |
+
},
|
| 230773 |
+
{
|
| 230774 |
+
"epoch": 825.94,
|
| 230775 |
+
"learning_rate": 8.361394230769232e-06,
|
| 230776 |
+
"loss": 0.3464,
|
| 230777 |
+
"step": 103290
|
| 230778 |
+
},
|
| 230779 |
+
{
|
| 230780 |
+
"epoch": 825.98,
|
| 230781 |
+
"learning_rate": 8.361314102564103e-06,
|
| 230782 |
+
"loss": 0.7888,
|
| 230783 |
+
"step": 103295
|
| 230784 |
+
},
|
| 230785 |
+
{
|
| 230786 |
+
"epoch": 826.0,
|
| 230787 |
+
"eval_loss": 0.3905450701713562,
|
| 230788 |
+
"eval_runtime": 39.4956,
|
| 230789 |
+
"eval_samples_per_second": 21.268,
|
| 230790 |
+
"eval_steps_per_second": 0.684,
|
| 230791 |
+
"eval_wer": 0.18283473714783618,
|
| 230792 |
+
"step": 103297
|
| 230793 |
}
|
| 230794 |
],
|
| 230795 |
"max_steps": 625000,
|
| 230796 |
"num_train_epochs": 5000,
|
| 230797 |
+
"total_flos": 2.907039073185294e+20,
|
| 230798 |
"trial_name": null,
|
| 230799 |
"trial_params": null
|
| 230800 |
}
|
model-bin/finetune/base/{checkpoint-102673 → checkpoint-103297}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629991265.2188263/events.out.tfevents.1629991265.8e89bd551565.924.231
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:289983e3423fb3b8ad381bac2d394b391b4ff7831b415e69f03954ac9197ac4c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629991684.7530313/events.out.tfevents.1629991685.8e89bd551565.924.233
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d057641351534481c77776d1f93e52b829833441a23af1e724dad0937f733e50
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629992115.6724274/events.out.tfevents.1629992115.8e89bd551565.924.235
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fd16089afa9ff59c71d70463bda29f56f257a32bf85ec5de97a3ac3c3db1dcaf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629992549.6115065/events.out.tfevents.1629992549.8e89bd551565.924.237
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e6bf2ea4e1465ea80023c02e1a380f353158bafe1f0347d53fc1bacf779095a9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629992990.702717/events.out.tfevents.1629992990.8e89bd551565.924.239
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:013702e11027c4dde07c2e805088b9256061267b8fcf184381aedf73eea6ada0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629991265.8e89bd551565.924.230
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:13fd9a91b77c508bf659a845dacd18e04ab36036b1b997f45240212b8a12058b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629991684.8e89bd551565.924.232
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:61c0a55089332d2ee76889257f844e55e072cbcc9e80cca9cd11bfef0e76341a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629992115.8e89bd551565.924.234
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:312a89c31238d2c77b4b106808aeecb7965a3a12286738a3fffa160ca15ea11c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629992549.8e89bd551565.924.236
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e72781dc5458da05e63fdcb14c86c88366de22295a7210893d7f412922fb6caf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629992990.8e89bd551565.924.238
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:246596c08706ad35c8268a592651ff86f33328aa5dc5acbc047212fe8d267d37
|
| 3 |
+
size 8622
|