"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630134518.566483/events.out.tfevents.1630134518.86bb0ddabf9b.4092.61 +3 -0
- model-bin/finetune/base/log/1630134911.0423136/events.out.tfevents.1630134911.86bb0ddabf9b.4092.63 +3 -0
- model-bin/finetune/base/log/1630135301.2668922/events.out.tfevents.1630135301.86bb0ddabf9b.4092.65 +3 -0
- model-bin/finetune/base/log/1630135690.2425787/events.out.tfevents.1630135690.86bb0ddabf9b.4092.67 +3 -0
- model-bin/finetune/base/log/1630136081.7873354/events.out.tfevents.1630136081.86bb0ddabf9b.4092.69 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630134518.86bb0ddabf9b.4092.60 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630134911.86bb0ddabf9b.4092.62 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630135301.86bb0ddabf9b.4092.64 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630135690.86bb0ddabf9b.4092.66 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630136081.86bb0ddabf9b.4092.68 +3 -0
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/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 sha256:636ee02e97153db492c1e25610a1815b762f603251f71ecf979797a912931ee1
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/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 sha256:6276c6dd9dc2aaa4f75d517ca8508a8d43751e6445d190d706be9592e9b68088
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/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 sha256:32115bf128a0979d07acf3ff9faa7174ce3f4d18ad456f6b27cb05a856a9246b
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/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 sha256:145db2a6bd6fad33acea1d837e65e5d9dc1ffb472af1f46c6cced784aa941fb5
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/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 sha256:8fee945b453a68e33846c1e2b7e09fdb2acfb604bb05d5631e98a77b7387ea20
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/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,
|
|
@@ -254214,11 +254214,800 @@
|
|
| 254214 |
"eval_steps_per_second": 0.728,
|
| 254215 |
"eval_wer": 0.18579426172148356,
|
| 254216 |
"step": 121707
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 254217 |
}
|
| 254218 |
],
|
| 254219 |
"max_steps": 625000,
|
| 254220 |
"num_train_epochs": 5000,
|
| 254221 |
-
"total_flos": 3.
|
| 254222 |
"trial_name": null,
|
| 254223 |
"trial_params": null
|
| 254224 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 978.0,
|
| 5 |
+
"global_step": 122329,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 254214 |
"eval_steps_per_second": 0.728,
|
| 254215 |
"eval_wer": 0.18579426172148356,
|
| 254216 |
"step": 121707
|
| 254217 |
+
},
|
| 254218 |
+
{
|
| 254219 |
+
"epoch": 981.02,
|
| 254220 |
+
"learning_rate": 8.066314102564103e-06,
|
| 254221 |
+
"loss": 0.3619,
|
| 254222 |
+
"step": 121710
|
| 254223 |
+
},
|
| 254224 |
+
{
|
| 254225 |
+
"epoch": 981.06,
|
| 254226 |
+
"learning_rate": 8.066233974358975e-06,
|
| 254227 |
+
"loss": 0.2627,
|
| 254228 |
+
"step": 121715
|
| 254229 |
+
},
|
| 254230 |
+
{
|
| 254231 |
+
"epoch": 981.1,
|
| 254232 |
+
"learning_rate": 8.066153846153846e-06,
|
| 254233 |
+
"loss": 0.3249,
|
| 254234 |
+
"step": 121720
|
| 254235 |
+
},
|
| 254236 |
+
{
|
| 254237 |
+
"epoch": 981.14,
|
| 254238 |
+
"learning_rate": 8.066073717948718e-06,
|
| 254239 |
+
"loss": 0.3249,
|
| 254240 |
+
"step": 121725
|
| 254241 |
+
},
|
| 254242 |
+
{
|
| 254243 |
+
"epoch": 981.18,
|
| 254244 |
+
"learning_rate": 8.06599358974359e-06,
|
| 254245 |
+
"loss": 0.7026,
|
| 254246 |
+
"step": 121730
|
| 254247 |
+
},
|
| 254248 |
+
{
|
| 254249 |
+
"epoch": 981.22,
|
| 254250 |
+
"learning_rate": 8.065913461538462e-06,
|
| 254251 |
+
"loss": 0.8091,
|
| 254252 |
+
"step": 121735
|
| 254253 |
+
},
|
| 254254 |
+
{
|
| 254255 |
+
"epoch": 981.27,
|
| 254256 |
+
"learning_rate": 8.065833333333333e-06,
|
| 254257 |
+
"loss": 0.2785,
|
| 254258 |
+
"step": 121740
|
| 254259 |
+
},
|
| 254260 |
+
{
|
| 254261 |
+
"epoch": 981.31,
|
| 254262 |
+
"learning_rate": 8.065753205128206e-06,
|
| 254263 |
+
"loss": 0.2591,
|
| 254264 |
+
"step": 121745
|
| 254265 |
+
},
|
| 254266 |
+
{
|
| 254267 |
+
"epoch": 981.35,
|
| 254268 |
+
"learning_rate": 8.065673076923078e-06,
|
| 254269 |
+
"loss": 0.2905,
|
| 254270 |
+
"step": 121750
|
| 254271 |
+
},
|
| 254272 |
+
{
|
| 254273 |
+
"epoch": 981.39,
|
| 254274 |
+
"learning_rate": 8.065592948717949e-06,
|
| 254275 |
+
"loss": 0.709,
|
| 254276 |
+
"step": 121755
|
| 254277 |
+
},
|
| 254278 |
+
{
|
| 254279 |
+
"epoch": 981.43,
|
| 254280 |
+
"learning_rate": 8.06551282051282e-06,
|
| 254281 |
+
"loss": 0.7867,
|
| 254282 |
+
"step": 121760
|
| 254283 |
+
},
|
| 254284 |
+
{
|
| 254285 |
+
"epoch": 981.47,
|
| 254286 |
+
"learning_rate": 8.065432692307694e-06,
|
| 254287 |
+
"loss": 0.247,
|
| 254288 |
+
"step": 121765
|
| 254289 |
+
},
|
| 254290 |
+
{
|
| 254291 |
+
"epoch": 981.51,
|
| 254292 |
+
"learning_rate": 8.065352564102565e-06,
|
| 254293 |
+
"loss": 0.3309,
|
| 254294 |
+
"step": 121770
|
| 254295 |
+
},
|
| 254296 |
+
{
|
| 254297 |
+
"epoch": 981.55,
|
| 254298 |
+
"learning_rate": 8.065272435897436e-06,
|
| 254299 |
+
"loss": 0.3943,
|
| 254300 |
+
"step": 121775
|
| 254301 |
+
},
|
| 254302 |
+
{
|
| 254303 |
+
"epoch": 981.59,
|
| 254304 |
+
"learning_rate": 8.06519230769231e-06,
|
| 254305 |
+
"loss": 0.6668,
|
| 254306 |
+
"step": 121780
|
| 254307 |
+
},
|
| 254308 |
+
{
|
| 254309 |
+
"epoch": 981.63,
|
| 254310 |
+
"learning_rate": 8.06511217948718e-06,
|
| 254311 |
+
"loss": 0.8405,
|
| 254312 |
+
"step": 121785
|
| 254313 |
+
},
|
| 254314 |
+
{
|
| 254315 |
+
"epoch": 981.67,
|
| 254316 |
+
"learning_rate": 8.065032051282052e-06,
|
| 254317 |
+
"loss": 0.3276,
|
| 254318 |
+
"step": 121790
|
| 254319 |
+
},
|
| 254320 |
+
{
|
| 254321 |
+
"epoch": 981.71,
|
| 254322 |
+
"learning_rate": 8.064951923076923e-06,
|
| 254323 |
+
"loss": 0.3355,
|
| 254324 |
+
"step": 121795
|
| 254325 |
+
},
|
| 254326 |
+
{
|
| 254327 |
+
"epoch": 981.75,
|
| 254328 |
+
"learning_rate": 8.064871794871796e-06,
|
| 254329 |
+
"loss": 0.3806,
|
| 254330 |
+
"step": 121800
|
| 254331 |
+
},
|
| 254332 |
+
{
|
| 254333 |
+
"epoch": 981.79,
|
| 254334 |
+
"learning_rate": 8.064791666666668e-06,
|
| 254335 |
+
"loss": 0.9231,
|
| 254336 |
+
"step": 121805
|
| 254337 |
+
},
|
| 254338 |
+
{
|
| 254339 |
+
"epoch": 981.83,
|
| 254340 |
+
"learning_rate": 8.064711538461539e-06,
|
| 254341 |
+
"loss": 0.8035,
|
| 254342 |
+
"step": 121810
|
| 254343 |
+
},
|
| 254344 |
+
{
|
| 254345 |
+
"epoch": 981.87,
|
| 254346 |
+
"learning_rate": 8.06463141025641e-06,
|
| 254347 |
+
"loss": 0.286,
|
| 254348 |
+
"step": 121815
|
| 254349 |
+
},
|
| 254350 |
+
{
|
| 254351 |
+
"epoch": 981.91,
|
| 254352 |
+
"learning_rate": 8.064551282051284e-06,
|
| 254353 |
+
"loss": 0.2576,
|
| 254354 |
+
"step": 121820
|
| 254355 |
+
},
|
| 254356 |
+
{
|
| 254357 |
+
"epoch": 981.95,
|
| 254358 |
+
"learning_rate": 8.064471153846153e-06,
|
| 254359 |
+
"loss": 0.3875,
|
| 254360 |
+
"step": 121825
|
| 254361 |
+
},
|
| 254362 |
+
{
|
| 254363 |
+
"epoch": 981.99,
|
| 254364 |
+
"learning_rate": 8.064391025641026e-06,
|
| 254365 |
+
"loss": 0.9652,
|
| 254366 |
+
"step": 121830
|
| 254367 |
+
},
|
| 254368 |
+
{
|
| 254369 |
+
"epoch": 982.0,
|
| 254370 |
+
"eval_loss": 0.38985419273376465,
|
| 254371 |
+
"eval_runtime": 36.1301,
|
| 254372 |
+
"eval_samples_per_second": 23.305,
|
| 254373 |
+
"eval_steps_per_second": 0.747,
|
| 254374 |
+
"eval_wer": 0.183689917936694,
|
| 254375 |
+
"step": 121831
|
| 254376 |
+
},
|
| 254377 |
+
{
|
| 254378 |
+
"epoch": 974.03,
|
| 254379 |
+
"learning_rate": 8.0643108974359e-06,
|
| 254380 |
+
"loss": 0.3079,
|
| 254381 |
+
"step": 121835
|
| 254382 |
+
},
|
| 254383 |
+
{
|
| 254384 |
+
"epoch": 974.07,
|
| 254385 |
+
"learning_rate": 8.064230769230769e-06,
|
| 254386 |
+
"loss": 0.2465,
|
| 254387 |
+
"step": 121840
|
| 254388 |
+
},
|
| 254389 |
+
{
|
| 254390 |
+
"epoch": 974.11,
|
| 254391 |
+
"learning_rate": 8.064150641025642e-06,
|
| 254392 |
+
"loss": 0.3261,
|
| 254393 |
+
"step": 121845
|
| 254394 |
+
},
|
| 254395 |
+
{
|
| 254396 |
+
"epoch": 974.15,
|
| 254397 |
+
"learning_rate": 8.064070512820513e-06,
|
| 254398 |
+
"loss": 0.4024,
|
| 254399 |
+
"step": 121850
|
| 254400 |
+
},
|
| 254401 |
+
{
|
| 254402 |
+
"epoch": 974.19,
|
| 254403 |
+
"learning_rate": 8.063990384615385e-06,
|
| 254404 |
+
"loss": 0.931,
|
| 254405 |
+
"step": 121855
|
| 254406 |
+
},
|
| 254407 |
+
{
|
| 254408 |
+
"epoch": 974.23,
|
| 254409 |
+
"learning_rate": 8.063910256410256e-06,
|
| 254410 |
+
"loss": 0.7664,
|
| 254411 |
+
"step": 121860
|
| 254412 |
+
},
|
| 254413 |
+
{
|
| 254414 |
+
"epoch": 974.27,
|
| 254415 |
+
"learning_rate": 8.063830128205129e-06,
|
| 254416 |
+
"loss": 0.268,
|
| 254417 |
+
"step": 121865
|
| 254418 |
+
},
|
| 254419 |
+
{
|
| 254420 |
+
"epoch": 974.31,
|
| 254421 |
+
"learning_rate": 8.06375e-06,
|
| 254422 |
+
"loss": 0.3031,
|
| 254423 |
+
"step": 121870
|
| 254424 |
+
},
|
| 254425 |
+
{
|
| 254426 |
+
"epoch": 974.35,
|
| 254427 |
+
"learning_rate": 8.063669871794872e-06,
|
| 254428 |
+
"loss": 0.3893,
|
| 254429 |
+
"step": 121875
|
| 254430 |
+
},
|
| 254431 |
+
{
|
| 254432 |
+
"epoch": 974.39,
|
| 254433 |
+
"learning_rate": 8.063589743589745e-06,
|
| 254434 |
+
"loss": 0.9955,
|
| 254435 |
+
"step": 121880
|
| 254436 |
+
},
|
| 254437 |
+
{
|
| 254438 |
+
"epoch": 974.43,
|
| 254439 |
+
"learning_rate": 8.063509615384616e-06,
|
| 254440 |
+
"loss": 0.6156,
|
| 254441 |
+
"step": 121885
|
| 254442 |
+
},
|
| 254443 |
+
{
|
| 254444 |
+
"epoch": 974.47,
|
| 254445 |
+
"learning_rate": 8.063429487179488e-06,
|
| 254446 |
+
"loss": 0.2728,
|
| 254447 |
+
"step": 121890
|
| 254448 |
+
},
|
| 254449 |
+
{
|
| 254450 |
+
"epoch": 974.51,
|
| 254451 |
+
"learning_rate": 8.063349358974359e-06,
|
| 254452 |
+
"loss": 0.2574,
|
| 254453 |
+
"step": 121895
|
| 254454 |
+
},
|
| 254455 |
+
{
|
| 254456 |
+
"epoch": 974.55,
|
| 254457 |
+
"learning_rate": 8.063269230769232e-06,
|
| 254458 |
+
"loss": 0.3403,
|
| 254459 |
+
"step": 121900
|
| 254460 |
+
},
|
| 254461 |
+
{
|
| 254462 |
+
"epoch": 974.59,
|
| 254463 |
+
"learning_rate": 8.063189102564103e-06,
|
| 254464 |
+
"loss": 0.9278,
|
| 254465 |
+
"step": 121905
|
| 254466 |
+
},
|
| 254467 |
+
{
|
| 254468 |
+
"epoch": 974.63,
|
| 254469 |
+
"learning_rate": 8.063108974358975e-06,
|
| 254470 |
+
"loss": 0.6179,
|
| 254471 |
+
"step": 121910
|
| 254472 |
+
},
|
| 254473 |
+
{
|
| 254474 |
+
"epoch": 974.67,
|
| 254475 |
+
"learning_rate": 8.063028846153846e-06,
|
| 254476 |
+
"loss": 0.2698,
|
| 254477 |
+
"step": 121915
|
| 254478 |
+
},
|
| 254479 |
+
{
|
| 254480 |
+
"epoch": 974.71,
|
| 254481 |
+
"learning_rate": 8.06294871794872e-06,
|
| 254482 |
+
"loss": 0.2849,
|
| 254483 |
+
"step": 121920
|
| 254484 |
+
},
|
| 254485 |
+
{
|
| 254486 |
+
"epoch": 974.75,
|
| 254487 |
+
"learning_rate": 8.06286858974359e-06,
|
| 254488 |
+
"loss": 0.382,
|
| 254489 |
+
"step": 121925
|
| 254490 |
+
},
|
| 254491 |
+
{
|
| 254492 |
+
"epoch": 974.79,
|
| 254493 |
+
"learning_rate": 8.062788461538462e-06,
|
| 254494 |
+
"loss": 0.8746,
|
| 254495 |
+
"step": 121930
|
| 254496 |
+
},
|
| 254497 |
+
{
|
| 254498 |
+
"epoch": 974.83,
|
| 254499 |
+
"learning_rate": 8.062708333333335e-06,
|
| 254500 |
+
"loss": 0.6217,
|
| 254501 |
+
"step": 121935
|
| 254502 |
+
},
|
| 254503 |
+
{
|
| 254504 |
+
"epoch": 974.87,
|
| 254505 |
+
"learning_rate": 8.062628205128206e-06,
|
| 254506 |
+
"loss": 0.2796,
|
| 254507 |
+
"step": 121940
|
| 254508 |
+
},
|
| 254509 |
+
{
|
| 254510 |
+
"epoch": 974.91,
|
| 254511 |
+
"learning_rate": 8.062548076923078e-06,
|
| 254512 |
+
"loss": 0.3042,
|
| 254513 |
+
"step": 121945
|
| 254514 |
+
},
|
| 254515 |
+
{
|
| 254516 |
+
"epoch": 974.95,
|
| 254517 |
+
"learning_rate": 8.062467948717949e-06,
|
| 254518 |
+
"loss": 0.4009,
|
| 254519 |
+
"step": 121950
|
| 254520 |
+
},
|
| 254521 |
+
{
|
| 254522 |
+
"epoch": 974.99,
|
| 254523 |
+
"learning_rate": 8.062387820512822e-06,
|
| 254524 |
+
"loss": 1.0557,
|
| 254525 |
+
"step": 121955
|
| 254526 |
+
},
|
| 254527 |
+
{
|
| 254528 |
+
"epoch": 975.0,
|
| 254529 |
+
"eval_loss": 0.4002327620983124,
|
| 254530 |
+
"eval_runtime": 36.1482,
|
| 254531 |
+
"eval_samples_per_second": 23.265,
|
| 254532 |
+
"eval_steps_per_second": 0.747,
|
| 254533 |
+
"eval_wer": 0.18004561842395703,
|
| 254534 |
+
"step": 121956
|
| 254535 |
+
},
|
| 254536 |
+
{
|
| 254537 |
+
"epoch": 983.03,
|
| 254538 |
+
"learning_rate": 8.062307692307693e-06,
|
| 254539 |
+
"loss": 0.3859,
|
| 254540 |
+
"step": 121960
|
| 254541 |
+
},
|
| 254542 |
+
{
|
| 254543 |
+
"epoch": 983.07,
|
| 254544 |
+
"learning_rate": 8.062227564102565e-06,
|
| 254545 |
+
"loss": 0.2278,
|
| 254546 |
+
"step": 121965
|
| 254547 |
+
},
|
| 254548 |
+
{
|
| 254549 |
+
"epoch": 983.11,
|
| 254550 |
+
"learning_rate": 8.062147435897436e-06,
|
| 254551 |
+
"loss": 0.2952,
|
| 254552 |
+
"step": 121970
|
| 254553 |
+
},
|
| 254554 |
+
{
|
| 254555 |
+
"epoch": 983.15,
|
| 254556 |
+
"learning_rate": 8.06206730769231e-06,
|
| 254557 |
+
"loss": 0.337,
|
| 254558 |
+
"step": 121975
|
| 254559 |
+
},
|
| 254560 |
+
{
|
| 254561 |
+
"epoch": 983.19,
|
| 254562 |
+
"learning_rate": 8.06198717948718e-06,
|
| 254563 |
+
"loss": 1.0278,
|
| 254564 |
+
"step": 121980
|
| 254565 |
+
},
|
| 254566 |
+
{
|
| 254567 |
+
"epoch": 983.23,
|
| 254568 |
+
"learning_rate": 8.061907051282052e-06,
|
| 254569 |
+
"loss": 0.6666,
|
| 254570 |
+
"step": 121985
|
| 254571 |
+
},
|
| 254572 |
+
{
|
| 254573 |
+
"epoch": 983.27,
|
| 254574 |
+
"learning_rate": 8.061826923076925e-06,
|
| 254575 |
+
"loss": 0.2537,
|
| 254576 |
+
"step": 121990
|
| 254577 |
+
},
|
| 254578 |
+
{
|
| 254579 |
+
"epoch": 983.31,
|
| 254580 |
+
"learning_rate": 8.061746794871795e-06,
|
| 254581 |
+
"loss": 0.336,
|
| 254582 |
+
"step": 121995
|
| 254583 |
+
},
|
| 254584 |
+
{
|
| 254585 |
+
"epoch": 983.35,
|
| 254586 |
+
"learning_rate": 8.061666666666668e-06,
|
| 254587 |
+
"loss": 0.4248,
|
| 254588 |
+
"step": 122000
|
| 254589 |
+
},
|
| 254590 |
+
{
|
| 254591 |
+
"epoch": 983.4,
|
| 254592 |
+
"learning_rate": 8.061586538461539e-06,
|
| 254593 |
+
"loss": 0.8619,
|
| 254594 |
+
"step": 122005
|
| 254595 |
+
},
|
| 254596 |
+
{
|
| 254597 |
+
"epoch": 983.44,
|
| 254598 |
+
"learning_rate": 8.06150641025641e-06,
|
| 254599 |
+
"loss": 0.578,
|
| 254600 |
+
"step": 122010
|
| 254601 |
+
},
|
| 254602 |
+
{
|
| 254603 |
+
"epoch": 983.48,
|
| 254604 |
+
"learning_rate": 8.061426282051282e-06,
|
| 254605 |
+
"loss": 0.2749,
|
| 254606 |
+
"step": 122015
|
| 254607 |
+
},
|
| 254608 |
+
{
|
| 254609 |
+
"epoch": 983.52,
|
| 254610 |
+
"learning_rate": 8.061346153846155e-06,
|
| 254611 |
+
"loss": 0.2588,
|
| 254612 |
+
"step": 122020
|
| 254613 |
+
},
|
| 254614 |
+
{
|
| 254615 |
+
"epoch": 983.56,
|
| 254616 |
+
"learning_rate": 8.061266025641026e-06,
|
| 254617 |
+
"loss": 0.4149,
|
| 254618 |
+
"step": 122025
|
| 254619 |
+
},
|
| 254620 |
+
{
|
| 254621 |
+
"epoch": 983.6,
|
| 254622 |
+
"learning_rate": 8.061185897435898e-06,
|
| 254623 |
+
"loss": 0.9949,
|
| 254624 |
+
"step": 122030
|
| 254625 |
+
},
|
| 254626 |
+
{
|
| 254627 |
+
"epoch": 983.64,
|
| 254628 |
+
"learning_rate": 8.06110576923077e-06,
|
| 254629 |
+
"loss": 0.7007,
|
| 254630 |
+
"step": 122035
|
| 254631 |
+
},
|
| 254632 |
+
{
|
| 254633 |
+
"epoch": 983.68,
|
| 254634 |
+
"learning_rate": 8.061025641025642e-06,
|
| 254635 |
+
"loss": 0.2401,
|
| 254636 |
+
"step": 122040
|
| 254637 |
+
},
|
| 254638 |
+
{
|
| 254639 |
+
"epoch": 983.72,
|
| 254640 |
+
"learning_rate": 8.060945512820513e-06,
|
| 254641 |
+
"loss": 0.3475,
|
| 254642 |
+
"step": 122045
|
| 254643 |
+
},
|
| 254644 |
+
{
|
| 254645 |
+
"epoch": 983.76,
|
| 254646 |
+
"learning_rate": 8.060865384615385e-06,
|
| 254647 |
+
"loss": 0.409,
|
| 254648 |
+
"step": 122050
|
| 254649 |
+
},
|
| 254650 |
+
{
|
| 254651 |
+
"epoch": 983.8,
|
| 254652 |
+
"learning_rate": 8.060785256410258e-06,
|
| 254653 |
+
"loss": 0.8656,
|
| 254654 |
+
"step": 122055
|
| 254655 |
+
},
|
| 254656 |
+
{
|
| 254657 |
+
"epoch": 983.84,
|
| 254658 |
+
"learning_rate": 8.060705128205129e-06,
|
| 254659 |
+
"loss": 0.7616,
|
| 254660 |
+
"step": 122060
|
| 254661 |
+
},
|
| 254662 |
+
{
|
| 254663 |
+
"epoch": 983.88,
|
| 254664 |
+
"learning_rate": 8.060625e-06,
|
| 254665 |
+
"loss": 0.2378,
|
| 254666 |
+
"step": 122065
|
| 254667 |
+
},
|
| 254668 |
+
{
|
| 254669 |
+
"epoch": 983.92,
|
| 254670 |
+
"learning_rate": 8.060544871794872e-06,
|
| 254671 |
+
"loss": 0.341,
|
| 254672 |
+
"step": 122070
|
| 254673 |
+
},
|
| 254674 |
+
{
|
| 254675 |
+
"epoch": 983.96,
|
| 254676 |
+
"learning_rate": 8.060464743589745e-06,
|
| 254677 |
+
"loss": 0.4632,
|
| 254678 |
+
"step": 122075
|
| 254679 |
+
},
|
| 254680 |
+
{
|
| 254681 |
+
"epoch": 984.0,
|
| 254682 |
+
"learning_rate": 8.060384615384616e-06,
|
| 254683 |
+
"loss": 1.1266,
|
| 254684 |
+
"step": 122080
|
| 254685 |
+
},
|
| 254686 |
+
{
|
| 254687 |
+
"epoch": 984.0,
|
| 254688 |
+
"eval_loss": 0.42599642276763916,
|
| 254689 |
+
"eval_runtime": 36.0836,
|
| 254690 |
+
"eval_samples_per_second": 23.307,
|
| 254691 |
+
"eval_steps_per_second": 0.748,
|
| 254692 |
+
"eval_wer": 0.1875638592906145,
|
| 254693 |
+
"step": 122080
|
| 254694 |
+
},
|
| 254695 |
+
{
|
| 254696 |
+
"epoch": 984.04,
|
| 254697 |
+
"learning_rate": 8.060304487179488e-06,
|
| 254698 |
+
"loss": 0.3444,
|
| 254699 |
+
"step": 122085
|
| 254700 |
+
},
|
| 254701 |
+
{
|
| 254702 |
+
"epoch": 984.08,
|
| 254703 |
+
"learning_rate": 8.06022435897436e-06,
|
| 254704 |
+
"loss": 0.32,
|
| 254705 |
+
"step": 122090
|
| 254706 |
+
},
|
| 254707 |
+
{
|
| 254708 |
+
"epoch": 984.12,
|
| 254709 |
+
"learning_rate": 8.060144230769232e-06,
|
| 254710 |
+
"loss": 0.2597,
|
| 254711 |
+
"step": 122095
|
| 254712 |
+
},
|
| 254713 |
+
{
|
| 254714 |
+
"epoch": 984.16,
|
| 254715 |
+
"learning_rate": 8.060064102564103e-06,
|
| 254716 |
+
"loss": 0.4258,
|
| 254717 |
+
"step": 122100
|
| 254718 |
+
},
|
| 254719 |
+
{
|
| 254720 |
+
"epoch": 984.2,
|
| 254721 |
+
"learning_rate": 8.059983974358975e-06,
|
| 254722 |
+
"loss": 1.2592,
|
| 254723 |
+
"step": 122105
|
| 254724 |
+
},
|
| 254725 |
+
{
|
| 254726 |
+
"epoch": 984.24,
|
| 254727 |
+
"learning_rate": 8.059903846153848e-06,
|
| 254728 |
+
"loss": 0.319,
|
| 254729 |
+
"step": 122110
|
| 254730 |
+
},
|
| 254731 |
+
{
|
| 254732 |
+
"epoch": 984.28,
|
| 254733 |
+
"learning_rate": 8.059823717948717e-06,
|
| 254734 |
+
"loss": 0.2761,
|
| 254735 |
+
"step": 122115
|
| 254736 |
+
},
|
| 254737 |
+
{
|
| 254738 |
+
"epoch": 984.32,
|
| 254739 |
+
"learning_rate": 8.05974358974359e-06,
|
| 254740 |
+
"loss": 0.2952,
|
| 254741 |
+
"step": 122120
|
| 254742 |
+
},
|
| 254743 |
+
{
|
| 254744 |
+
"epoch": 984.36,
|
| 254745 |
+
"learning_rate": 8.059663461538462e-06,
|
| 254746 |
+
"loss": 0.4896,
|
| 254747 |
+
"step": 122125
|
| 254748 |
+
},
|
| 254749 |
+
{
|
| 254750 |
+
"epoch": 984.4,
|
| 254751 |
+
"learning_rate": 8.059583333333333e-06,
|
| 254752 |
+
"loss": 1.2025,
|
| 254753 |
+
"step": 122130
|
| 254754 |
+
},
|
| 254755 |
+
{
|
| 254756 |
+
"epoch": 984.44,
|
| 254757 |
+
"learning_rate": 8.059503205128206e-06,
|
| 254758 |
+
"loss": 0.3081,
|
| 254759 |
+
"step": 122135
|
| 254760 |
+
},
|
| 254761 |
+
{
|
| 254762 |
+
"epoch": 984.48,
|
| 254763 |
+
"learning_rate": 8.059423076923078e-06,
|
| 254764 |
+
"loss": 0.2566,
|
| 254765 |
+
"step": 122140
|
| 254766 |
+
},
|
| 254767 |
+
{
|
| 254768 |
+
"epoch": 984.52,
|
| 254769 |
+
"learning_rate": 8.059342948717949e-06,
|
| 254770 |
+
"loss": 0.2943,
|
| 254771 |
+
"step": 122145
|
| 254772 |
+
},
|
| 254773 |
+
{
|
| 254774 |
+
"epoch": 984.56,
|
| 254775 |
+
"learning_rate": 8.05926282051282e-06,
|
| 254776 |
+
"loss": 0.432,
|
| 254777 |
+
"step": 122150
|
| 254778 |
+
},
|
| 254779 |
+
{
|
| 254780 |
+
"epoch": 984.6,
|
| 254781 |
+
"learning_rate": 8.059182692307693e-06,
|
| 254782 |
+
"loss": 1.2402,
|
| 254783 |
+
"step": 122155
|
| 254784 |
+
},
|
| 254785 |
+
{
|
| 254786 |
+
"epoch": 984.65,
|
| 254787 |
+
"learning_rate": 8.059102564102565e-06,
|
| 254788 |
+
"loss": 0.3292,
|
| 254789 |
+
"step": 122160
|
| 254790 |
+
},
|
| 254791 |
+
{
|
| 254792 |
+
"epoch": 984.69,
|
| 254793 |
+
"learning_rate": 8.059022435897436e-06,
|
| 254794 |
+
"loss": 0.3257,
|
| 254795 |
+
"step": 122165
|
| 254796 |
+
},
|
| 254797 |
+
{
|
| 254798 |
+
"epoch": 984.73,
|
| 254799 |
+
"learning_rate": 8.058942307692307e-06,
|
| 254800 |
+
"loss": 0.2744,
|
| 254801 |
+
"step": 122170
|
| 254802 |
+
},
|
| 254803 |
+
{
|
| 254804 |
+
"epoch": 984.77,
|
| 254805 |
+
"learning_rate": 8.05886217948718e-06,
|
| 254806 |
+
"loss": 0.5238,
|
| 254807 |
+
"step": 122175
|
| 254808 |
+
},
|
| 254809 |
+
{
|
| 254810 |
+
"epoch": 984.81,
|
| 254811 |
+
"learning_rate": 8.058782051282052e-06,
|
| 254812 |
+
"loss": 1.13,
|
| 254813 |
+
"step": 122180
|
| 254814 |
+
},
|
| 254815 |
+
{
|
| 254816 |
+
"epoch": 984.85,
|
| 254817 |
+
"learning_rate": 8.058701923076923e-06,
|
| 254818 |
+
"loss": 0.3126,
|
| 254819 |
+
"step": 122185
|
| 254820 |
+
},
|
| 254821 |
+
{
|
| 254822 |
+
"epoch": 984.89,
|
| 254823 |
+
"learning_rate": 8.058621794871796e-06,
|
| 254824 |
+
"loss": 0.2472,
|
| 254825 |
+
"step": 122190
|
| 254826 |
+
},
|
| 254827 |
+
{
|
| 254828 |
+
"epoch": 984.93,
|
| 254829 |
+
"learning_rate": 8.058541666666668e-06,
|
| 254830 |
+
"loss": 0.2963,
|
| 254831 |
+
"step": 122195
|
| 254832 |
+
},
|
| 254833 |
+
{
|
| 254834 |
+
"epoch": 984.97,
|
| 254835 |
+
"learning_rate": 8.058461538461539e-06,
|
| 254836 |
+
"loss": 0.5364,
|
| 254837 |
+
"step": 122200
|
| 254838 |
+
},
|
| 254839 |
+
{
|
| 254840 |
+
"epoch": 985.0,
|
| 254841 |
+
"eval_loss": 0.44340452551841736,
|
| 254842 |
+
"eval_runtime": 36.0415,
|
| 254843 |
+
"eval_samples_per_second": 23.362,
|
| 254844 |
+
"eval_steps_per_second": 0.749,
|
| 254845 |
+
"eval_wer": 0.1748058921703795,
|
| 254846 |
+
"step": 122204
|
| 254847 |
+
},
|
| 254848 |
+
{
|
| 254849 |
+
"epoch": 977.01,
|
| 254850 |
+
"learning_rate": 8.05838141025641e-06,
|
| 254851 |
+
"loss": 0.3177,
|
| 254852 |
+
"step": 122205
|
| 254853 |
+
},
|
| 254854 |
+
{
|
| 254855 |
+
"epoch": 977.05,
|
| 254856 |
+
"learning_rate": 8.058301282051283e-06,
|
| 254857 |
+
"loss": 0.2794,
|
| 254858 |
+
"step": 122210
|
| 254859 |
+
},
|
| 254860 |
+
{
|
| 254861 |
+
"epoch": 977.09,
|
| 254862 |
+
"learning_rate": 8.058221153846155e-06,
|
| 254863 |
+
"loss": 0.2729,
|
| 254864 |
+
"step": 122215
|
| 254865 |
+
},
|
| 254866 |
+
{
|
| 254867 |
+
"epoch": 977.13,
|
| 254868 |
+
"learning_rate": 8.058141025641026e-06,
|
| 254869 |
+
"loss": 0.3982,
|
| 254870 |
+
"step": 122220
|
| 254871 |
+
},
|
| 254872 |
+
{
|
| 254873 |
+
"epoch": 977.17,
|
| 254874 |
+
"learning_rate": 8.058060897435897e-06,
|
| 254875 |
+
"loss": 0.5194,
|
| 254876 |
+
"step": 122225
|
| 254877 |
+
},
|
| 254878 |
+
{
|
| 254879 |
+
"epoch": 977.21,
|
| 254880 |
+
"learning_rate": 8.05798076923077e-06,
|
| 254881 |
+
"loss": 1.0576,
|
| 254882 |
+
"step": 122230
|
| 254883 |
+
},
|
| 254884 |
+
{
|
| 254885 |
+
"epoch": 977.25,
|
| 254886 |
+
"learning_rate": 8.057900641025642e-06,
|
| 254887 |
+
"loss": 0.3235,
|
| 254888 |
+
"step": 122235
|
| 254889 |
+
},
|
| 254890 |
+
{
|
| 254891 |
+
"epoch": 977.29,
|
| 254892 |
+
"learning_rate": 8.057820512820513e-06,
|
| 254893 |
+
"loss": 0.3022,
|
| 254894 |
+
"step": 122240
|
| 254895 |
+
},
|
| 254896 |
+
{
|
| 254897 |
+
"epoch": 977.33,
|
| 254898 |
+
"learning_rate": 8.057740384615386e-06,
|
| 254899 |
+
"loss": 0.4027,
|
| 254900 |
+
"step": 122245
|
| 254901 |
+
},
|
| 254902 |
+
{
|
| 254903 |
+
"epoch": 977.37,
|
| 254904 |
+
"learning_rate": 8.057660256410258e-06,
|
| 254905 |
+
"loss": 0.5891,
|
| 254906 |
+
"step": 122250
|
| 254907 |
+
},
|
| 254908 |
+
{
|
| 254909 |
+
"epoch": 977.41,
|
| 254910 |
+
"learning_rate": 8.057580128205129e-06,
|
| 254911 |
+
"loss": 0.9342,
|
| 254912 |
+
"step": 122255
|
| 254913 |
+
},
|
| 254914 |
+
{
|
| 254915 |
+
"epoch": 977.45,
|
| 254916 |
+
"learning_rate": 8.0575e-06,
|
| 254917 |
+
"loss": 0.2948,
|
| 254918 |
+
"step": 122260
|
| 254919 |
+
},
|
| 254920 |
+
{
|
| 254921 |
+
"epoch": 977.49,
|
| 254922 |
+
"learning_rate": 8.057419871794873e-06,
|
| 254923 |
+
"loss": 0.3486,
|
| 254924 |
+
"step": 122265
|
| 254925 |
+
},
|
| 254926 |
+
{
|
| 254927 |
+
"epoch": 977.53,
|
| 254928 |
+
"learning_rate": 8.057339743589743e-06,
|
| 254929 |
+
"loss": 0.3783,
|
| 254930 |
+
"step": 122270
|
| 254931 |
+
},
|
| 254932 |
+
{
|
| 254933 |
+
"epoch": 977.57,
|
| 254934 |
+
"learning_rate": 8.057259615384616e-06,
|
| 254935 |
+
"loss": 0.5175,
|
| 254936 |
+
"step": 122275
|
| 254937 |
+
},
|
| 254938 |
+
{
|
| 254939 |
+
"epoch": 977.61,
|
| 254940 |
+
"learning_rate": 8.05717948717949e-06,
|
| 254941 |
+
"loss": 1.0765,
|
| 254942 |
+
"step": 122280
|
| 254943 |
+
},
|
| 254944 |
+
{
|
| 254945 |
+
"epoch": 977.65,
|
| 254946 |
+
"learning_rate": 8.057099358974359e-06,
|
| 254947 |
+
"loss": 0.303,
|
| 254948 |
+
"step": 122285
|
| 254949 |
+
},
|
| 254950 |
+
{
|
| 254951 |
+
"epoch": 977.69,
|
| 254952 |
+
"learning_rate": 8.057019230769232e-06,
|
| 254953 |
+
"loss": 0.2748,
|
| 254954 |
+
"step": 122290
|
| 254955 |
+
},
|
| 254956 |
+
{
|
| 254957 |
+
"epoch": 977.73,
|
| 254958 |
+
"learning_rate": 8.056939102564103e-06,
|
| 254959 |
+
"loss": 0.3423,
|
| 254960 |
+
"step": 122295
|
| 254961 |
+
},
|
| 254962 |
+
{
|
| 254963 |
+
"epoch": 977.77,
|
| 254964 |
+
"learning_rate": 8.056858974358975e-06,
|
| 254965 |
+
"loss": 0.6243,
|
| 254966 |
+
"step": 122300
|
| 254967 |
+
},
|
| 254968 |
+
{
|
| 254969 |
+
"epoch": 977.81,
|
| 254970 |
+
"learning_rate": 8.056778846153846e-06,
|
| 254971 |
+
"loss": 1.0672,
|
| 254972 |
+
"step": 122305
|
| 254973 |
+
},
|
| 254974 |
+
{
|
| 254975 |
+
"epoch": 977.85,
|
| 254976 |
+
"learning_rate": 8.056698717948719e-06,
|
| 254977 |
+
"loss": 0.3127,
|
| 254978 |
+
"step": 122310
|
| 254979 |
+
},
|
| 254980 |
+
{
|
| 254981 |
+
"epoch": 977.89,
|
| 254982 |
+
"learning_rate": 8.05661858974359e-06,
|
| 254983 |
+
"loss": 0.3113,
|
| 254984 |
+
"step": 122315
|
| 254985 |
+
},
|
| 254986 |
+
{
|
| 254987 |
+
"epoch": 977.93,
|
| 254988 |
+
"learning_rate": 8.056538461538462e-06,
|
| 254989 |
+
"loss": 0.2507,
|
| 254990 |
+
"step": 122320
|
| 254991 |
+
},
|
| 254992 |
+
{
|
| 254993 |
+
"epoch": 977.97,
|
| 254994 |
+
"learning_rate": 8.056458333333333e-06,
|
| 254995 |
+
"loss": 0.5587,
|
| 254996 |
+
"step": 122325
|
| 254997 |
+
},
|
| 254998 |
+
{
|
| 254999 |
+
"epoch": 978.0,
|
| 255000 |
+
"eval_loss": 0.35215896368026733,
|
| 255001 |
+
"eval_runtime": 35.884,
|
| 255002 |
+
"eval_samples_per_second": 23.465,
|
| 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": 3.442586939742868e+20,
|
| 255011 |
"trial_name": null,
|
| 255012 |
"trial_params": null
|
| 255013 |
}
|
model-bin/finetune/base/{checkpoint-121707 → checkpoint-122329}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630134518.566483/events.out.tfevents.1630134518.86bb0ddabf9b.4092.61
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e612e579b37ed3f331d52a2c89d46b56a6cb14229212b5523e27c975350b9ac2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630134911.0423136/events.out.tfevents.1630134911.86bb0ddabf9b.4092.63
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:437150c6a57d50409e9dd83654e814b93fbd96cbc247badda53f8484f9649b4f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630135301.2668922/events.out.tfevents.1630135301.86bb0ddabf9b.4092.65
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e1bbb96a36c4e01cb227a07f6311fdaeb0a5cac9c8ecc28604dd66045d5567a4
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630135690.2425787/events.out.tfevents.1630135690.86bb0ddabf9b.4092.67
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ae1f703a01b139b6bfcf93aff706fec547dad2da4865432048c6da6585ace025
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630136081.7873354/events.out.tfevents.1630136081.86bb0ddabf9b.4092.69
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f35a49bc3d7eaa575e4ebf872706afa6ee96b401d163af9f4804c1b6603c6c37
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630134518.86bb0ddabf9b.4092.60
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:835f087cf9e7906ca11f62e4d8edc8a144ac59419fccc5c116ae560110d5c9e5
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630134911.86bb0ddabf9b.4092.62
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:758f52c60958962f54f9c344ee09e16b3f9c014d5329b12e18f2cc2991a589b3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630135301.86bb0ddabf9b.4092.64
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d0da5df6c8fc06509ae3f8711a1c3109fbad0fec9510e894e05ea780fac1a6c4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630135690.86bb0ddabf9b.4092.66
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:66c2d786875e9b986a2b2895342a2759d4591e5ab2cb2eea2c56bf327e595daa
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630136081.86bb0ddabf9b.4092.68
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7d97cd85878dee2434036a92620b62dc8daa43d16fc730ef9a67636ab629857f
|
| 3 |
+
size 8622
|