"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629989066.542981/events.out.tfevents.1629989066.8e89bd551565.924.221 +3 -0
- model-bin/finetune/base/log/1629989517.413555/events.out.tfevents.1629989517.8e89bd551565.924.223 +3 -0
- model-bin/finetune/base/log/1629989953.1397705/events.out.tfevents.1629989955.8e89bd551565.924.225 +3 -0
- model-bin/finetune/base/log/1629990389.2203898/events.out.tfevents.1629990389.8e89bd551565.924.227 +3 -0
- model-bin/finetune/base/log/1629990808.4787564/events.out.tfevents.1629990808.8e89bd551565.924.229 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629989066.8e89bd551565.924.220 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629989515.8e89bd551565.924.222 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629989953.8e89bd551565.924.224 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629990387.8e89bd551565.924.226 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629990808.8e89bd551565.924.228 +3 -0
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/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:af9a4e20d30a3e47d0abded938fb1582578c9d5ee770f7527ba53e206e18f6d8
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/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:c47ac21d1538c72e68b61b0a2654c31bb911aff70da729d3d2bb4d25a67783a9
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/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:f7c00aebbf192aaff34a5eef9868352d3c5c2370e9cfe027619f32134f7a85d5
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/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 sha256:703b2c951df4efdfabf698ea4ebf8efc6d144661a2d4c8aee86aa621f308f4ea
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/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 sha256:2fb0a4131c45a097c22d7fefe521e60e1cd099dc67af2792275d48b90f13c0c5
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/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,
|
|
@@ -229200,11 +229200,806 @@
|
|
| 229200 |
"eval_steps_per_second": 0.651,
|
| 229201 |
"eval_wer": 0.17922283901665345,
|
| 229202 |
"step": 102049
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 229203 |
}
|
| 229204 |
],
|
| 229205 |
"max_steps": 625000,
|
| 229206 |
"num_train_epochs": 5000,
|
| 229207 |
-
"total_flos": 2.
|
| 229208 |
"trial_name": null,
|
| 229209 |
"trial_params": null
|
| 229210 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 821.0,
|
| 5 |
+
"global_step": 102673,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 229200 |
"eval_steps_per_second": 0.651,
|
| 229201 |
"eval_wer": 0.17922283901665345,
|
| 229202 |
"step": 102049
|
| 229203 |
+
},
|
| 229204 |
+
{
|
| 229205 |
+
"epoch": 816.01,
|
| 229206 |
+
"learning_rate": 8.38125e-06,
|
| 229207 |
+
"loss": 0.3945,
|
| 229208 |
+
"step": 102050
|
| 229209 |
+
},
|
| 229210 |
+
{
|
| 229211 |
+
"epoch": 816.05,
|
| 229212 |
+
"learning_rate": 8.381169871794872e-06,
|
| 229213 |
+
"loss": 0.3024,
|
| 229214 |
+
"step": 102055
|
| 229215 |
+
},
|
| 229216 |
+
{
|
| 229217 |
+
"epoch": 816.09,
|
| 229218 |
+
"learning_rate": 8.381089743589745e-06,
|
| 229219 |
+
"loss": 0.3532,
|
| 229220 |
+
"step": 102060
|
| 229221 |
+
},
|
| 229222 |
+
{
|
| 229223 |
+
"epoch": 816.13,
|
| 229224 |
+
"learning_rate": 8.381009615384616e-06,
|
| 229225 |
+
"loss": 0.308,
|
| 229226 |
+
"step": 102065
|
| 229227 |
+
},
|
| 229228 |
+
{
|
| 229229 |
+
"epoch": 816.17,
|
| 229230 |
+
"learning_rate": 8.380929487179488e-06,
|
| 229231 |
+
"loss": 0.6114,
|
| 229232 |
+
"step": 102070
|
| 229233 |
+
},
|
| 229234 |
+
{
|
| 229235 |
+
"epoch": 816.21,
|
| 229236 |
+
"learning_rate": 8.38084935897436e-06,
|
| 229237 |
+
"loss": 1.3084,
|
| 229238 |
+
"step": 102075
|
| 229239 |
+
},
|
| 229240 |
+
{
|
| 229241 |
+
"epoch": 816.25,
|
| 229242 |
+
"learning_rate": 8.380769230769232e-06,
|
| 229243 |
+
"loss": 0.4907,
|
| 229244 |
+
"step": 102080
|
| 229245 |
+
},
|
| 229246 |
+
{
|
| 229247 |
+
"epoch": 816.29,
|
| 229248 |
+
"learning_rate": 8.380689102564103e-06,
|
| 229249 |
+
"loss": 0.4826,
|
| 229250 |
+
"step": 102085
|
| 229251 |
+
},
|
| 229252 |
+
{
|
| 229253 |
+
"epoch": 816.33,
|
| 229254 |
+
"learning_rate": 8.380608974358975e-06,
|
| 229255 |
+
"loss": 0.3124,
|
| 229256 |
+
"step": 102090
|
| 229257 |
+
},
|
| 229258 |
+
{
|
| 229259 |
+
"epoch": 816.37,
|
| 229260 |
+
"learning_rate": 8.380528846153848e-06,
|
| 229261 |
+
"loss": 0.5426,
|
| 229262 |
+
"step": 102095
|
| 229263 |
+
},
|
| 229264 |
+
{
|
| 229265 |
+
"epoch": 816.41,
|
| 229266 |
+
"learning_rate": 8.380448717948717e-06,
|
| 229267 |
+
"loss": 1.1867,
|
| 229268 |
+
"step": 102100
|
| 229269 |
+
},
|
| 229270 |
+
{
|
| 229271 |
+
"epoch": 816.45,
|
| 229272 |
+
"learning_rate": 8.38036858974359e-06,
|
| 229273 |
+
"loss": 0.3174,
|
| 229274 |
+
"step": 102105
|
| 229275 |
+
},
|
| 229276 |
+
{
|
| 229277 |
+
"epoch": 816.49,
|
| 229278 |
+
"learning_rate": 8.380288461538464e-06,
|
| 229279 |
+
"loss": 0.4763,
|
| 229280 |
+
"step": 102110
|
| 229281 |
+
},
|
| 229282 |
+
{
|
| 229283 |
+
"epoch": 816.53,
|
| 229284 |
+
"learning_rate": 8.380208333333333e-06,
|
| 229285 |
+
"loss": 0.2736,
|
| 229286 |
+
"step": 102115
|
| 229287 |
+
},
|
| 229288 |
+
{
|
| 229289 |
+
"epoch": 816.57,
|
| 229290 |
+
"learning_rate": 8.380128205128206e-06,
|
| 229291 |
+
"loss": 0.4283,
|
| 229292 |
+
"step": 102120
|
| 229293 |
+
},
|
| 229294 |
+
{
|
| 229295 |
+
"epoch": 816.61,
|
| 229296 |
+
"learning_rate": 8.380048076923078e-06,
|
| 229297 |
+
"loss": 1.0865,
|
| 229298 |
+
"step": 102125
|
| 229299 |
+
},
|
| 229300 |
+
{
|
| 229301 |
+
"epoch": 816.65,
|
| 229302 |
+
"learning_rate": 8.379967948717949e-06,
|
| 229303 |
+
"loss": 0.3076,
|
| 229304 |
+
"step": 102130
|
| 229305 |
+
},
|
| 229306 |
+
{
|
| 229307 |
+
"epoch": 816.69,
|
| 229308 |
+
"learning_rate": 8.37988782051282e-06,
|
| 229309 |
+
"loss": 0.2716,
|
| 229310 |
+
"step": 102135
|
| 229311 |
+
},
|
| 229312 |
+
{
|
| 229313 |
+
"epoch": 816.73,
|
| 229314 |
+
"learning_rate": 8.379807692307693e-06,
|
| 229315 |
+
"loss": 0.3148,
|
| 229316 |
+
"step": 102140
|
| 229317 |
+
},
|
| 229318 |
+
{
|
| 229319 |
+
"epoch": 816.76,
|
| 229320 |
+
"learning_rate": 8.379727564102565e-06,
|
| 229321 |
+
"loss": 0.4542,
|
| 229322 |
+
"step": 102145
|
| 229323 |
+
},
|
| 229324 |
+
{
|
| 229325 |
+
"epoch": 816.8,
|
| 229326 |
+
"learning_rate": 8.379647435897436e-06,
|
| 229327 |
+
"loss": 1.1537,
|
| 229328 |
+
"step": 102150
|
| 229329 |
+
},
|
| 229330 |
+
{
|
| 229331 |
+
"epoch": 816.84,
|
| 229332 |
+
"learning_rate": 8.379567307692307e-06,
|
| 229333 |
+
"loss": 0.345,
|
| 229334 |
+
"step": 102155
|
| 229335 |
+
},
|
| 229336 |
+
{
|
| 229337 |
+
"epoch": 816.88,
|
| 229338 |
+
"learning_rate": 8.37948717948718e-06,
|
| 229339 |
+
"loss": 0.3189,
|
| 229340 |
+
"step": 102160
|
| 229341 |
+
},
|
| 229342 |
+
{
|
| 229343 |
+
"epoch": 816.92,
|
| 229344 |
+
"learning_rate": 8.379407051282052e-06,
|
| 229345 |
+
"loss": 0.3174,
|
| 229346 |
+
"step": 102165
|
| 229347 |
+
},
|
| 229348 |
+
{
|
| 229349 |
+
"epoch": 816.96,
|
| 229350 |
+
"learning_rate": 8.379326923076923e-06,
|
| 229351 |
+
"loss": 0.5369,
|
| 229352 |
+
"step": 102170
|
| 229353 |
+
},
|
| 229354 |
+
{
|
| 229355 |
+
"epoch": 817.0,
|
| 229356 |
+
"eval_loss": 0.43944230675697327,
|
| 229357 |
+
"eval_runtime": 40.1764,
|
| 229358 |
+
"eval_samples_per_second": 20.908,
|
| 229359 |
+
"eval_steps_per_second": 0.672,
|
| 229360 |
+
"eval_wer": 0.18577132486388384,
|
| 229361 |
+
"step": 102174
|
| 229362 |
+
},
|
| 229363 |
+
{
|
| 229364 |
+
"epoch": 823.01,
|
| 229365 |
+
"learning_rate": 8.37926282051282e-06,
|
| 229366 |
+
"loss": 0.2977,
|
| 229367 |
+
"step": 102175
|
| 229368 |
+
},
|
| 229369 |
+
{
|
| 229370 |
+
"epoch": 823.05,
|
| 229371 |
+
"learning_rate": 8.379182692307694e-06,
|
| 229372 |
+
"loss": 0.2942,
|
| 229373 |
+
"step": 102180
|
| 229374 |
+
},
|
| 229375 |
+
{
|
| 229376 |
+
"epoch": 823.09,
|
| 229377 |
+
"learning_rate": 8.379102564102565e-06,
|
| 229378 |
+
"loss": 0.2977,
|
| 229379 |
+
"step": 102185
|
| 229380 |
+
},
|
| 229381 |
+
{
|
| 229382 |
+
"epoch": 823.13,
|
| 229383 |
+
"learning_rate": 8.379022435897436e-06,
|
| 229384 |
+
"loss": 0.3607,
|
| 229385 |
+
"step": 102190
|
| 229386 |
+
},
|
| 229387 |
+
{
|
| 229388 |
+
"epoch": 823.17,
|
| 229389 |
+
"learning_rate": 8.378942307692308e-06,
|
| 229390 |
+
"loss": 0.4314,
|
| 229391 |
+
"step": 102195
|
| 229392 |
+
},
|
| 229393 |
+
{
|
| 229394 |
+
"epoch": 823.21,
|
| 229395 |
+
"learning_rate": 8.37886217948718e-06,
|
| 229396 |
+
"loss": 1.193,
|
| 229397 |
+
"step": 102200
|
| 229398 |
+
},
|
| 229399 |
+
{
|
| 229400 |
+
"epoch": 823.25,
|
| 229401 |
+
"learning_rate": 8.378782051282052e-06,
|
| 229402 |
+
"loss": 0.3182,
|
| 229403 |
+
"step": 102205
|
| 229404 |
+
},
|
| 229405 |
+
{
|
| 229406 |
+
"epoch": 823.29,
|
| 229407 |
+
"learning_rate": 8.378701923076924e-06,
|
| 229408 |
+
"loss": 0.2171,
|
| 229409 |
+
"step": 102210
|
| 229410 |
+
},
|
| 229411 |
+
{
|
| 229412 |
+
"epoch": 823.33,
|
| 229413 |
+
"learning_rate": 8.378621794871795e-06,
|
| 229414 |
+
"loss": 0.3123,
|
| 229415 |
+
"step": 102215
|
| 229416 |
+
},
|
| 229417 |
+
{
|
| 229418 |
+
"epoch": 823.37,
|
| 229419 |
+
"learning_rate": 8.378541666666668e-06,
|
| 229420 |
+
"loss": 0.4648,
|
| 229421 |
+
"step": 102220
|
| 229422 |
+
},
|
| 229423 |
+
{
|
| 229424 |
+
"epoch": 823.41,
|
| 229425 |
+
"learning_rate": 8.37846153846154e-06,
|
| 229426 |
+
"loss": 1.104,
|
| 229427 |
+
"step": 102225
|
| 229428 |
+
},
|
| 229429 |
+
{
|
| 229430 |
+
"epoch": 823.45,
|
| 229431 |
+
"learning_rate": 8.37838141025641e-06,
|
| 229432 |
+
"loss": 0.3179,
|
| 229433 |
+
"step": 102230
|
| 229434 |
+
},
|
| 229435 |
+
{
|
| 229436 |
+
"epoch": 823.49,
|
| 229437 |
+
"learning_rate": 8.378301282051284e-06,
|
| 229438 |
+
"loss": 0.3032,
|
| 229439 |
+
"step": 102235
|
| 229440 |
+
},
|
| 229441 |
+
{
|
| 229442 |
+
"epoch": 823.53,
|
| 229443 |
+
"learning_rate": 8.378221153846155e-06,
|
| 229444 |
+
"loss": 0.3243,
|
| 229445 |
+
"step": 102240
|
| 229446 |
+
},
|
| 229447 |
+
{
|
| 229448 |
+
"epoch": 823.57,
|
| 229449 |
+
"learning_rate": 8.378141025641026e-06,
|
| 229450 |
+
"loss": 0.5687,
|
| 229451 |
+
"step": 102245
|
| 229452 |
+
},
|
| 229453 |
+
{
|
| 229454 |
+
"epoch": 823.61,
|
| 229455 |
+
"learning_rate": 8.378060897435898e-06,
|
| 229456 |
+
"loss": 1.1412,
|
| 229457 |
+
"step": 102250
|
| 229458 |
+
},
|
| 229459 |
+
{
|
| 229460 |
+
"epoch": 823.65,
|
| 229461 |
+
"learning_rate": 8.37798076923077e-06,
|
| 229462 |
+
"loss": 0.3057,
|
| 229463 |
+
"step": 102255
|
| 229464 |
+
},
|
| 229465 |
+
{
|
| 229466 |
+
"epoch": 823.69,
|
| 229467 |
+
"learning_rate": 8.377900641025642e-06,
|
| 229468 |
+
"loss": 0.2491,
|
| 229469 |
+
"step": 102260
|
| 229470 |
+
},
|
| 229471 |
+
{
|
| 229472 |
+
"epoch": 823.73,
|
| 229473 |
+
"learning_rate": 8.377820512820514e-06,
|
| 229474 |
+
"loss": 0.3076,
|
| 229475 |
+
"step": 102265
|
| 229476 |
+
},
|
| 229477 |
+
{
|
| 229478 |
+
"epoch": 823.77,
|
| 229479 |
+
"learning_rate": 8.377740384615385e-06,
|
| 229480 |
+
"loss": 0.4681,
|
| 229481 |
+
"step": 102270
|
| 229482 |
+
},
|
| 229483 |
+
{
|
| 229484 |
+
"epoch": 823.81,
|
| 229485 |
+
"learning_rate": 8.377660256410258e-06,
|
| 229486 |
+
"loss": 1.0723,
|
| 229487 |
+
"step": 102275
|
| 229488 |
+
},
|
| 229489 |
+
{
|
| 229490 |
+
"epoch": 823.85,
|
| 229491 |
+
"learning_rate": 8.377580128205128e-06,
|
| 229492 |
+
"loss": 0.3152,
|
| 229493 |
+
"step": 102280
|
| 229494 |
+
},
|
| 229495 |
+
{
|
| 229496 |
+
"epoch": 823.9,
|
| 229497 |
+
"learning_rate": 8.3775e-06,
|
| 229498 |
+
"loss": 0.303,
|
| 229499 |
+
"step": 102285
|
| 229500 |
+
},
|
| 229501 |
+
{
|
| 229502 |
+
"epoch": 823.94,
|
| 229503 |
+
"learning_rate": 8.377419871794874e-06,
|
| 229504 |
+
"loss": 0.3279,
|
| 229505 |
+
"step": 102290
|
| 229506 |
+
},
|
| 229507 |
+
{
|
| 229508 |
+
"epoch": 823.98,
|
| 229509 |
+
"learning_rate": 8.377339743589743e-06,
|
| 229510 |
+
"loss": 0.5272,
|
| 229511 |
+
"step": 102295
|
| 229512 |
+
},
|
| 229513 |
+
{
|
| 229514 |
+
"epoch": 824.0,
|
| 229515 |
+
"eval_loss": 0.40923911333084106,
|
| 229516 |
+
"eval_runtime": 38.3052,
|
| 229517 |
+
"eval_samples_per_second": 21.929,
|
| 229518 |
+
"eval_steps_per_second": 0.705,
|
| 229519 |
+
"eval_wer": 0.18392390357246588,
|
| 229520 |
+
"step": 102298
|
| 229521 |
+
},
|
| 229522 |
+
{
|
| 229523 |
+
"epoch": 818.02,
|
| 229524 |
+
"learning_rate": 8.377259615384616e-06,
|
| 229525 |
+
"loss": 0.4523,
|
| 229526 |
+
"step": 102300
|
| 229527 |
+
},
|
| 229528 |
+
{
|
| 229529 |
+
"epoch": 818.06,
|
| 229530 |
+
"learning_rate": 8.377179487179488e-06,
|
| 229531 |
+
"loss": 0.296,
|
| 229532 |
+
"step": 102305
|
| 229533 |
+
},
|
| 229534 |
+
{
|
| 229535 |
+
"epoch": 818.1,
|
| 229536 |
+
"learning_rate": 8.37709935897436e-06,
|
| 229537 |
+
"loss": 0.26,
|
| 229538 |
+
"step": 102310
|
| 229539 |
+
},
|
| 229540 |
+
{
|
| 229541 |
+
"epoch": 818.14,
|
| 229542 |
+
"learning_rate": 8.37701923076923e-06,
|
| 229543 |
+
"loss": 0.3082,
|
| 229544 |
+
"step": 102315
|
| 229545 |
+
},
|
| 229546 |
+
{
|
| 229547 |
+
"epoch": 818.18,
|
| 229548 |
+
"learning_rate": 8.376939102564104e-06,
|
| 229549 |
+
"loss": 0.6886,
|
| 229550 |
+
"step": 102320
|
| 229551 |
+
},
|
| 229552 |
+
{
|
| 229553 |
+
"epoch": 818.22,
|
| 229554 |
+
"learning_rate": 8.376858974358975e-06,
|
| 229555 |
+
"loss": 0.9145,
|
| 229556 |
+
"step": 102325
|
| 229557 |
+
},
|
| 229558 |
+
{
|
| 229559 |
+
"epoch": 818.26,
|
| 229560 |
+
"learning_rate": 8.376778846153846e-06,
|
| 229561 |
+
"loss": 0.3013,
|
| 229562 |
+
"step": 102330
|
| 229563 |
+
},
|
| 229564 |
+
{
|
| 229565 |
+
"epoch": 818.3,
|
| 229566 |
+
"learning_rate": 8.37669871794872e-06,
|
| 229567 |
+
"loss": 0.3044,
|
| 229568 |
+
"step": 102335
|
| 229569 |
+
},
|
| 229570 |
+
{
|
| 229571 |
+
"epoch": 818.34,
|
| 229572 |
+
"learning_rate": 8.37661858974359e-06,
|
| 229573 |
+
"loss": 0.4015,
|
| 229574 |
+
"step": 102340
|
| 229575 |
+
},
|
| 229576 |
+
{
|
| 229577 |
+
"epoch": 818.38,
|
| 229578 |
+
"learning_rate": 8.376538461538462e-06,
|
| 229579 |
+
"loss": 0.617,
|
| 229580 |
+
"step": 102345
|
| 229581 |
+
},
|
| 229582 |
+
{
|
| 229583 |
+
"epoch": 818.42,
|
| 229584 |
+
"learning_rate": 8.376458333333333e-06,
|
| 229585 |
+
"loss": 1.2272,
|
| 229586 |
+
"step": 102350
|
| 229587 |
+
},
|
| 229588 |
+
{
|
| 229589 |
+
"epoch": 818.46,
|
| 229590 |
+
"learning_rate": 8.376378205128206e-06,
|
| 229591 |
+
"loss": 0.2681,
|
| 229592 |
+
"step": 102355
|
| 229593 |
+
},
|
| 229594 |
+
{
|
| 229595 |
+
"epoch": 818.5,
|
| 229596 |
+
"learning_rate": 8.376298076923078e-06,
|
| 229597 |
+
"loss": 0.259,
|
| 229598 |
+
"step": 102360
|
| 229599 |
+
},
|
| 229600 |
+
{
|
| 229601 |
+
"epoch": 818.54,
|
| 229602 |
+
"learning_rate": 8.37621794871795e-06,
|
| 229603 |
+
"loss": 0.3254,
|
| 229604 |
+
"step": 102365
|
| 229605 |
+
},
|
| 229606 |
+
{
|
| 229607 |
+
"epoch": 818.58,
|
| 229608 |
+
"learning_rate": 8.37613782051282e-06,
|
| 229609 |
+
"loss": 0.5868,
|
| 229610 |
+
"step": 102370
|
| 229611 |
+
},
|
| 229612 |
+
{
|
| 229613 |
+
"epoch": 818.62,
|
| 229614 |
+
"learning_rate": 8.376057692307694e-06,
|
| 229615 |
+
"loss": 0.9538,
|
| 229616 |
+
"step": 102375
|
| 229617 |
+
},
|
| 229618 |
+
{
|
| 229619 |
+
"epoch": 818.66,
|
| 229620 |
+
"learning_rate": 8.375977564102565e-06,
|
| 229621 |
+
"loss": 0.2679,
|
| 229622 |
+
"step": 102380
|
| 229623 |
+
},
|
| 229624 |
+
{
|
| 229625 |
+
"epoch": 818.7,
|
| 229626 |
+
"learning_rate": 8.375897435897436e-06,
|
| 229627 |
+
"loss": 0.3279,
|
| 229628 |
+
"step": 102385
|
| 229629 |
+
},
|
| 229630 |
+
{
|
| 229631 |
+
"epoch": 818.74,
|
| 229632 |
+
"learning_rate": 8.37581730769231e-06,
|
| 229633 |
+
"loss": 0.3469,
|
| 229634 |
+
"step": 102390
|
| 229635 |
+
},
|
| 229636 |
+
{
|
| 229637 |
+
"epoch": 818.78,
|
| 229638 |
+
"learning_rate": 8.37573717948718e-06,
|
| 229639 |
+
"loss": 0.6476,
|
| 229640 |
+
"step": 102395
|
| 229641 |
+
},
|
| 229642 |
+
{
|
| 229643 |
+
"epoch": 818.82,
|
| 229644 |
+
"learning_rate": 8.375657051282052e-06,
|
| 229645 |
+
"loss": 0.8834,
|
| 229646 |
+
"step": 102400
|
| 229647 |
+
},
|
| 229648 |
+
{
|
| 229649 |
+
"epoch": 818.86,
|
| 229650 |
+
"learning_rate": 8.375576923076923e-06,
|
| 229651 |
+
"loss": 0.3755,
|
| 229652 |
+
"step": 102405
|
| 229653 |
+
},
|
| 229654 |
+
{
|
| 229655 |
+
"epoch": 818.9,
|
| 229656 |
+
"learning_rate": 8.375496794871797e-06,
|
| 229657 |
+
"loss": 0.3062,
|
| 229658 |
+
"step": 102410
|
| 229659 |
+
},
|
| 229660 |
+
{
|
| 229661 |
+
"epoch": 818.94,
|
| 229662 |
+
"learning_rate": 8.375416666666666e-06,
|
| 229663 |
+
"loss": 0.2884,
|
| 229664 |
+
"step": 102415
|
| 229665 |
+
},
|
| 229666 |
+
{
|
| 229667 |
+
"epoch": 818.98,
|
| 229668 |
+
"learning_rate": 8.37533653846154e-06,
|
| 229669 |
+
"loss": 0.6774,
|
| 229670 |
+
"step": 102420
|
| 229671 |
+
},
|
| 229672 |
+
{
|
| 229673 |
+
"epoch": 819.0,
|
| 229674 |
+
"eval_loss": 0.3606967628002167,
|
| 229675 |
+
"eval_runtime": 38.349,
|
| 229676 |
+
"eval_samples_per_second": 21.904,
|
| 229677 |
+
"eval_steps_per_second": 0.704,
|
| 229678 |
+
"eval_wer": 0.18543682150505872,
|
| 229679 |
+
"step": 102423
|
| 229680 |
+
},
|
| 229681 |
+
{
|
| 229682 |
+
"epoch": 819.02,
|
| 229683 |
+
"learning_rate": 8.37525641025641e-06,
|
| 229684 |
+
"loss": 0.4049,
|
| 229685 |
+
"step": 102425
|
| 229686 |
+
},
|
| 229687 |
+
{
|
| 229688 |
+
"epoch": 819.06,
|
| 229689 |
+
"learning_rate": 8.375176282051282e-06,
|
| 229690 |
+
"loss": 0.2873,
|
| 229691 |
+
"step": 102430
|
| 229692 |
+
},
|
| 229693 |
+
{
|
| 229694 |
+
"epoch": 819.1,
|
| 229695 |
+
"learning_rate": 8.375096153846155e-06,
|
| 229696 |
+
"loss": 0.3085,
|
| 229697 |
+
"step": 102435
|
| 229698 |
+
},
|
| 229699 |
+
{
|
| 229700 |
+
"epoch": 819.14,
|
| 229701 |
+
"learning_rate": 8.375016025641026e-06,
|
| 229702 |
+
"loss": 0.3548,
|
| 229703 |
+
"step": 102440
|
| 229704 |
+
},
|
| 229705 |
+
{
|
| 229706 |
+
"epoch": 819.18,
|
| 229707 |
+
"learning_rate": 8.374935897435898e-06,
|
| 229708 |
+
"loss": 0.579,
|
| 229709 |
+
"step": 102445
|
| 229710 |
+
},
|
| 229711 |
+
{
|
| 229712 |
+
"epoch": 819.22,
|
| 229713 |
+
"learning_rate": 8.374855769230769e-06,
|
| 229714 |
+
"loss": 0.9847,
|
| 229715 |
+
"step": 102450
|
| 229716 |
+
},
|
| 229717 |
+
{
|
| 229718 |
+
"epoch": 819.26,
|
| 229719 |
+
"learning_rate": 8.374775641025642e-06,
|
| 229720 |
+
"loss": 0.3314,
|
| 229721 |
+
"step": 102455
|
| 229722 |
+
},
|
| 229723 |
+
{
|
| 229724 |
+
"epoch": 819.3,
|
| 229725 |
+
"learning_rate": 8.374695512820513e-06,
|
| 229726 |
+
"loss": 0.292,
|
| 229727 |
+
"step": 102460
|
| 229728 |
+
},
|
| 229729 |
+
{
|
| 229730 |
+
"epoch": 819.34,
|
| 229731 |
+
"learning_rate": 8.374615384615385e-06,
|
| 229732 |
+
"loss": 0.3688,
|
| 229733 |
+
"step": 102465
|
| 229734 |
+
},
|
| 229735 |
+
{
|
| 229736 |
+
"epoch": 819.38,
|
| 229737 |
+
"learning_rate": 8.374535256410256e-06,
|
| 229738 |
+
"loss": 0.6313,
|
| 229739 |
+
"step": 102470
|
| 229740 |
+
},
|
| 229741 |
+
{
|
| 229742 |
+
"epoch": 819.42,
|
| 229743 |
+
"learning_rate": 8.37445512820513e-06,
|
| 229744 |
+
"loss": 0.9745,
|
| 229745 |
+
"step": 102475
|
| 229746 |
+
},
|
| 229747 |
+
{
|
| 229748 |
+
"epoch": 819.46,
|
| 229749 |
+
"learning_rate": 8.374375e-06,
|
| 229750 |
+
"loss": 0.3424,
|
| 229751 |
+
"step": 102480
|
| 229752 |
+
},
|
| 229753 |
+
{
|
| 229754 |
+
"epoch": 819.5,
|
| 229755 |
+
"learning_rate": 8.374294871794872e-06,
|
| 229756 |
+
"loss": 0.3116,
|
| 229757 |
+
"step": 102485
|
| 229758 |
+
},
|
| 229759 |
+
{
|
| 229760 |
+
"epoch": 819.54,
|
| 229761 |
+
"learning_rate": 8.374214743589745e-06,
|
| 229762 |
+
"loss": 0.3602,
|
| 229763 |
+
"step": 102490
|
| 229764 |
+
},
|
| 229765 |
+
{
|
| 229766 |
+
"epoch": 819.58,
|
| 229767 |
+
"learning_rate": 8.374134615384616e-06,
|
| 229768 |
+
"loss": 0.6469,
|
| 229769 |
+
"step": 102495
|
| 229770 |
+
},
|
| 229771 |
+
{
|
| 229772 |
+
"epoch": 819.62,
|
| 229773 |
+
"learning_rate": 8.374054487179488e-06,
|
| 229774 |
+
"loss": 1.1507,
|
| 229775 |
+
"step": 102500
|
| 229776 |
+
},
|
| 229777 |
+
{
|
| 229778 |
+
"epoch": 819.66,
|
| 229779 |
+
"learning_rate": 8.373974358974359e-06,
|
| 229780 |
+
"loss": 0.2987,
|
| 229781 |
+
"step": 102505
|
| 229782 |
+
},
|
| 229783 |
+
{
|
| 229784 |
+
"epoch": 819.7,
|
| 229785 |
+
"learning_rate": 8.373894230769232e-06,
|
| 229786 |
+
"loss": 0.4105,
|
| 229787 |
+
"step": 102510
|
| 229788 |
+
},
|
| 229789 |
+
{
|
| 229790 |
+
"epoch": 819.74,
|
| 229791 |
+
"learning_rate": 8.373814102564104e-06,
|
| 229792 |
+
"loss": 0.3374,
|
| 229793 |
+
"step": 102515
|
| 229794 |
+
},
|
| 229795 |
+
{
|
| 229796 |
+
"epoch": 819.78,
|
| 229797 |
+
"learning_rate": 8.373733974358975e-06,
|
| 229798 |
+
"loss": 0.6295,
|
| 229799 |
+
"step": 102520
|
| 229800 |
+
},
|
| 229801 |
+
{
|
| 229802 |
+
"epoch": 819.82,
|
| 229803 |
+
"learning_rate": 8.373653846153846e-06,
|
| 229804 |
+
"loss": 1.0377,
|
| 229805 |
+
"step": 102525
|
| 229806 |
+
},
|
| 229807 |
+
{
|
| 229808 |
+
"epoch": 819.86,
|
| 229809 |
+
"learning_rate": 8.37357371794872e-06,
|
| 229810 |
+
"loss": 0.5091,
|
| 229811 |
+
"step": 102530
|
| 229812 |
+
},
|
| 229813 |
+
{
|
| 229814 |
+
"epoch": 819.9,
|
| 229815 |
+
"learning_rate": 8.37349358974359e-06,
|
| 229816 |
+
"loss": 0.2433,
|
| 229817 |
+
"step": 102535
|
| 229818 |
+
},
|
| 229819 |
+
{
|
| 229820 |
+
"epoch": 819.94,
|
| 229821 |
+
"learning_rate": 8.373413461538462e-06,
|
| 229822 |
+
"loss": 0.3735,
|
| 229823 |
+
"step": 102540
|
| 229824 |
+
},
|
| 229825 |
+
{
|
| 229826 |
+
"epoch": 819.98,
|
| 229827 |
+
"learning_rate": 8.373333333333335e-06,
|
| 229828 |
+
"loss": 0.6356,
|
| 229829 |
+
"step": 102545
|
| 229830 |
+
},
|
| 229831 |
+
{
|
| 229832 |
+
"epoch": 820.0,
|
| 229833 |
+
"eval_loss": 0.36974748969078064,
|
| 229834 |
+
"eval_runtime": 37.6433,
|
| 229835 |
+
"eval_samples_per_second": 22.341,
|
| 229836 |
+
"eval_steps_per_second": 0.717,
|
| 229837 |
+
"eval_wer": 0.17474260205918352,
|
| 229838 |
+
"step": 102548
|
| 229839 |
+
},
|
| 229840 |
+
{
|
| 229841 |
+
"epoch": 820.02,
|
| 229842 |
+
"learning_rate": 8.373253205128206e-06,
|
| 229843 |
+
"loss": 0.3477,
|
| 229844 |
+
"step": 102550
|
| 229845 |
+
},
|
| 229846 |
+
{
|
| 229847 |
+
"epoch": 820.06,
|
| 229848 |
+
"learning_rate": 8.373173076923078e-06,
|
| 229849 |
+
"loss": 0.261,
|
| 229850 |
+
"step": 102555
|
| 229851 |
+
},
|
| 229852 |
+
{
|
| 229853 |
+
"epoch": 820.1,
|
| 229854 |
+
"learning_rate": 8.373092948717949e-06,
|
| 229855 |
+
"loss": 0.353,
|
| 229856 |
+
"step": 102560
|
| 229857 |
+
},
|
| 229858 |
+
{
|
| 229859 |
+
"epoch": 820.14,
|
| 229860 |
+
"learning_rate": 8.373012820512822e-06,
|
| 229861 |
+
"loss": 0.3394,
|
| 229862 |
+
"step": 102565
|
| 229863 |
+
},
|
| 229864 |
+
{
|
| 229865 |
+
"epoch": 820.18,
|
| 229866 |
+
"learning_rate": 8.372932692307692e-06,
|
| 229867 |
+
"loss": 0.6584,
|
| 229868 |
+
"step": 102570
|
| 229869 |
+
},
|
| 229870 |
+
{
|
| 229871 |
+
"epoch": 820.22,
|
| 229872 |
+
"learning_rate": 8.372852564102565e-06,
|
| 229873 |
+
"loss": 1.0039,
|
| 229874 |
+
"step": 102575
|
| 229875 |
+
},
|
| 229876 |
+
{
|
| 229877 |
+
"epoch": 820.26,
|
| 229878 |
+
"learning_rate": 8.372772435897436e-06,
|
| 229879 |
+
"loss": 0.338,
|
| 229880 |
+
"step": 102580
|
| 229881 |
+
},
|
| 229882 |
+
{
|
| 229883 |
+
"epoch": 820.3,
|
| 229884 |
+
"learning_rate": 8.372692307692308e-06,
|
| 229885 |
+
"loss": 0.2874,
|
| 229886 |
+
"step": 102585
|
| 229887 |
+
},
|
| 229888 |
+
{
|
| 229889 |
+
"epoch": 820.34,
|
| 229890 |
+
"learning_rate": 8.37261217948718e-06,
|
| 229891 |
+
"loss": 0.3527,
|
| 229892 |
+
"step": 102590
|
| 229893 |
+
},
|
| 229894 |
+
{
|
| 229895 |
+
"epoch": 820.38,
|
| 229896 |
+
"learning_rate": 8.372532051282052e-06,
|
| 229897 |
+
"loss": 0.5877,
|
| 229898 |
+
"step": 102595
|
| 229899 |
+
},
|
| 229900 |
+
{
|
| 229901 |
+
"epoch": 820.42,
|
| 229902 |
+
"learning_rate": 8.372451923076923e-06,
|
| 229903 |
+
"loss": 1.0461,
|
| 229904 |
+
"step": 102600
|
| 229905 |
+
},
|
| 229906 |
+
{
|
| 229907 |
+
"epoch": 820.46,
|
| 229908 |
+
"learning_rate": 8.372371794871795e-06,
|
| 229909 |
+
"loss": 0.3091,
|
| 229910 |
+
"step": 102605
|
| 229911 |
+
},
|
| 229912 |
+
{
|
| 229913 |
+
"epoch": 820.5,
|
| 229914 |
+
"learning_rate": 8.372291666666668e-06,
|
| 229915 |
+
"loss": 0.2585,
|
| 229916 |
+
"step": 102610
|
| 229917 |
+
},
|
| 229918 |
+
{
|
| 229919 |
+
"epoch": 820.54,
|
| 229920 |
+
"learning_rate": 8.372211538461539e-06,
|
| 229921 |
+
"loss": 0.336,
|
| 229922 |
+
"step": 102615
|
| 229923 |
+
},
|
| 229924 |
+
{
|
| 229925 |
+
"epoch": 820.58,
|
| 229926 |
+
"learning_rate": 8.37213141025641e-06,
|
| 229927 |
+
"loss": 1.2123,
|
| 229928 |
+
"step": 102620
|
| 229929 |
+
},
|
| 229930 |
+
{
|
| 229931 |
+
"epoch": 820.62,
|
| 229932 |
+
"learning_rate": 8.372051282051282e-06,
|
| 229933 |
+
"loss": 1.0318,
|
| 229934 |
+
"step": 102625
|
| 229935 |
+
},
|
| 229936 |
+
{
|
| 229937 |
+
"epoch": 820.66,
|
| 229938 |
+
"learning_rate": 8.371971153846155e-06,
|
| 229939 |
+
"loss": 0.3275,
|
| 229940 |
+
"step": 102630
|
| 229941 |
+
},
|
| 229942 |
+
{
|
| 229943 |
+
"epoch": 820.7,
|
| 229944 |
+
"learning_rate": 8.371891025641026e-06,
|
| 229945 |
+
"loss": 0.2638,
|
| 229946 |
+
"step": 102635
|
| 229947 |
+
},
|
| 229948 |
+
{
|
| 229949 |
+
"epoch": 820.74,
|
| 229950 |
+
"learning_rate": 8.371810897435898e-06,
|
| 229951 |
+
"loss": 0.3076,
|
| 229952 |
+
"step": 102640
|
| 229953 |
+
},
|
| 229954 |
+
{
|
| 229955 |
+
"epoch": 820.78,
|
| 229956 |
+
"learning_rate": 8.37173076923077e-06,
|
| 229957 |
+
"loss": 0.6243,
|
| 229958 |
+
"step": 102645
|
| 229959 |
+
},
|
| 229960 |
+
{
|
| 229961 |
+
"epoch": 820.82,
|
| 229962 |
+
"learning_rate": 8.371650641025642e-06,
|
| 229963 |
+
"loss": 1.2342,
|
| 229964 |
+
"step": 102650
|
| 229965 |
+
},
|
| 229966 |
+
{
|
| 229967 |
+
"epoch": 820.86,
|
| 229968 |
+
"learning_rate": 8.371570512820513e-06,
|
| 229969 |
+
"loss": 0.2736,
|
| 229970 |
+
"step": 102655
|
| 229971 |
+
},
|
| 229972 |
+
{
|
| 229973 |
+
"epoch": 820.9,
|
| 229974 |
+
"learning_rate": 8.371490384615385e-06,
|
| 229975 |
+
"loss": 0.2979,
|
| 229976 |
+
"step": 102660
|
| 229977 |
+
},
|
| 229978 |
+
{
|
| 229979 |
+
"epoch": 820.94,
|
| 229980 |
+
"learning_rate": 8.371410256410258e-06,
|
| 229981 |
+
"loss": 0.3252,
|
| 229982 |
+
"step": 102665
|
| 229983 |
+
},
|
| 229984 |
+
{
|
| 229985 |
+
"epoch": 820.98,
|
| 229986 |
+
"learning_rate": 8.37133012820513e-06,
|
| 229987 |
+
"loss": 0.6398,
|
| 229988 |
+
"step": 102670
|
| 229989 |
+
},
|
| 229990 |
+
{
|
| 229991 |
+
"epoch": 821.0,
|
| 229992 |
+
"eval_loss": 0.4211651682853699,
|
| 229993 |
+
"eval_runtime": 38.6049,
|
| 229994 |
+
"eval_samples_per_second": 21.759,
|
| 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": 2.8894841678586194e+20,
|
| 230003 |
"trial_name": null,
|
| 230004 |
"trial_params": null
|
| 230005 |
}
|
model-bin/finetune/base/{checkpoint-102049 → checkpoint-102673}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629989066.542981/events.out.tfevents.1629989066.8e89bd551565.924.221
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa27cfc53a8fa8217a44c734e8b906e34324baf019631832174b0fa66976826f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629989517.413555/events.out.tfevents.1629989517.8e89bd551565.924.223
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:48ac41c0d6d5b07c90036f1da40290099727e5aaade9f8d5900619281316cd8b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629989953.1397705/events.out.tfevents.1629989955.8e89bd551565.924.225
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:965b38a1bc2195bfe2823367d0f1369b3816dbad47da6843237bda7fc3dd032d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629990389.2203898/events.out.tfevents.1629990389.8e89bd551565.924.227
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:622e55cfa4a600b2263fca555278b2f32086791fc0ad71a0a4fcbdf8900a17bc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629990808.4787564/events.out.tfevents.1629990808.8e89bd551565.924.229
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e1ce4c99b3e757bca429749b880f5c0efde1bc83fa50a1a46abaf5aa833569a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629989066.8e89bd551565.924.220
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d7426f66202278e393a44d93cf47bd60b253857ece761b153596f22018528e48
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629989515.8e89bd551565.924.222
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77decc7f56b71d1f68ee84a26aea5fa44edf06292ee9b90028be2fe4f84549c9
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629989953.8e89bd551565.924.224
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b6c3e53915781b1e58d059337219628b2d3deae7a10d23ce2a953141704dedef
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629990387.8e89bd551565.924.226
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4e2e7bbfe6b42c3b5dc1a25367d69fd4d48dddf0aa9cfec7438616cba5b7d6de
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629990808.8e89bd551565.924.228
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e769f7c1edcbe397d82925f186c9daab84515e83d29d1f8b9c19be818e93bb64
|
| 3 |
+
size 8622
|