"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/trainer_state.json +799 -4
- model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629791737.1385896/events.out.tfevents.1629791737.c435e1c5ee04.920.81 +3 -0
- model-bin/finetune/base/log/1629792376.705988/events.out.tfevents.1629792376.c435e1c5ee04.920.83 +3 -0
- model-bin/finetune/base/log/1629793016.5378878/events.out.tfevents.1629793016.c435e1c5ee04.920.85 +3 -0
- model-bin/finetune/base/log/1629793657.2960315/events.out.tfevents.1629793657.c435e1c5ee04.920.87 +3 -0
- model-bin/finetune/base/log/1629794296.277906/events.out.tfevents.1629794296.c435e1c5ee04.920.89 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629791737.c435e1c5ee04.920.80 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629792376.c435e1c5ee04.920.82 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629793016.c435e1c5ee04.920.84 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629793657.c435e1c5ee04.920.86 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629794296.c435e1c5ee04.920.88 +3 -0
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e16b1e5f840dae285f2923192506ae59e4e8b7cabe32f01f344dabfce86ebf8
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/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 sha256:26f4ab2149533327bf2f53b6b40c161778d6772adef0201afb903f8e0720d162
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/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:aa558066bfb3787767cc2c9de7fcea37c1b39d27c3755fd84cb44f53db08a8b7
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/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 sha256:9a8d93330377a06c01db10b78b765880cbfdaaef9acb4ce3642c308922f97869
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/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:c7091bd139e2fbbf99b5560cef7c49824907a05089a34e8b24d8616792df4e71
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -174255,11 +174255,806 @@
|
|
| 174255 |
"eval_steps_per_second": 0.677,
|
| 174256 |
"eval_wer": 0.19737984944991313,
|
| 174257 |
"step": 58863
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 174258 |
}
|
| 174259 |
],
|
| 174260 |
-
"max_steps":
|
| 174261 |
"num_train_epochs": 5000,
|
| 174262 |
-
"total_flos": 1.
|
| 174263 |
"trial_name": null,
|
| 174264 |
"trial_params": null
|
| 174265 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18631571186315712,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-55501",
|
| 4 |
+
"epoch": 479.0,
|
| 5 |
+
"global_step": 59485,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 174255 |
"eval_steps_per_second": 0.677,
|
| 174256 |
"eval_wer": 0.19737984944991313,
|
| 174257 |
"step": 58863
|
| 174258 |
+
},
|
| 174259 |
+
{
|
| 174260 |
+
"epoch": 470.02,
|
| 174261 |
+
"learning_rate": 9.073028846153847e-06,
|
| 174262 |
+
"loss": 0.3944,
|
| 174263 |
+
"step": 58865
|
| 174264 |
+
},
|
| 174265 |
+
{
|
| 174266 |
+
"epoch": 470.06,
|
| 174267 |
+
"learning_rate": 9.072948717948718e-06,
|
| 174268 |
+
"loss": 0.3238,
|
| 174269 |
+
"step": 58870
|
| 174270 |
+
},
|
| 174271 |
+
{
|
| 174272 |
+
"epoch": 470.1,
|
| 174273 |
+
"learning_rate": 9.072868589743591e-06,
|
| 174274 |
+
"loss": 0.3217,
|
| 174275 |
+
"step": 58875
|
| 174276 |
+
},
|
| 174277 |
+
{
|
| 174278 |
+
"epoch": 470.14,
|
| 174279 |
+
"learning_rate": 9.072788461538462e-06,
|
| 174280 |
+
"loss": 0.4157,
|
| 174281 |
+
"step": 58880
|
| 174282 |
+
},
|
| 174283 |
+
{
|
| 174284 |
+
"epoch": 470.18,
|
| 174285 |
+
"learning_rate": 9.072708333333334e-06,
|
| 174286 |
+
"loss": 0.6491,
|
| 174287 |
+
"step": 58885
|
| 174288 |
+
},
|
| 174289 |
+
{
|
| 174290 |
+
"epoch": 470.22,
|
| 174291 |
+
"learning_rate": 9.072628205128205e-06,
|
| 174292 |
+
"loss": 1.0585,
|
| 174293 |
+
"step": 58890
|
| 174294 |
+
},
|
| 174295 |
+
{
|
| 174296 |
+
"epoch": 470.26,
|
| 174297 |
+
"learning_rate": 9.072548076923078e-06,
|
| 174298 |
+
"loss": 0.3527,
|
| 174299 |
+
"step": 58895
|
| 174300 |
+
},
|
| 174301 |
+
{
|
| 174302 |
+
"epoch": 470.3,
|
| 174303 |
+
"learning_rate": 9.07246794871795e-06,
|
| 174304 |
+
"loss": 0.2987,
|
| 174305 |
+
"step": 58900
|
| 174306 |
+
},
|
| 174307 |
+
{
|
| 174308 |
+
"epoch": 470.34,
|
| 174309 |
+
"learning_rate": 9.072387820512821e-06,
|
| 174310 |
+
"loss": 0.4083,
|
| 174311 |
+
"step": 58905
|
| 174312 |
+
},
|
| 174313 |
+
{
|
| 174314 |
+
"epoch": 470.38,
|
| 174315 |
+
"learning_rate": 9.072307692307694e-06,
|
| 174316 |
+
"loss": 0.6874,
|
| 174317 |
+
"step": 58910
|
| 174318 |
+
},
|
| 174319 |
+
{
|
| 174320 |
+
"epoch": 470.42,
|
| 174321 |
+
"learning_rate": 9.072227564102565e-06,
|
| 174322 |
+
"loss": 1.1236,
|
| 174323 |
+
"step": 58915
|
| 174324 |
+
},
|
| 174325 |
+
{
|
| 174326 |
+
"epoch": 470.46,
|
| 174327 |
+
"learning_rate": 9.072147435897437e-06,
|
| 174328 |
+
"loss": 0.2641,
|
| 174329 |
+
"step": 58920
|
| 174330 |
+
},
|
| 174331 |
+
{
|
| 174332 |
+
"epoch": 470.5,
|
| 174333 |
+
"learning_rate": 9.072067307692308e-06,
|
| 174334 |
+
"loss": 0.329,
|
| 174335 |
+
"step": 58925
|
| 174336 |
+
},
|
| 174337 |
+
{
|
| 174338 |
+
"epoch": 470.54,
|
| 174339 |
+
"learning_rate": 9.071987179487181e-06,
|
| 174340 |
+
"loss": 0.4592,
|
| 174341 |
+
"step": 58930
|
| 174342 |
+
},
|
| 174343 |
+
{
|
| 174344 |
+
"epoch": 470.58,
|
| 174345 |
+
"learning_rate": 9.071907051282052e-06,
|
| 174346 |
+
"loss": 0.6611,
|
| 174347 |
+
"step": 58935
|
| 174348 |
+
},
|
| 174349 |
+
{
|
| 174350 |
+
"epoch": 470.62,
|
| 174351 |
+
"learning_rate": 9.071826923076924e-06,
|
| 174352 |
+
"loss": 1.1541,
|
| 174353 |
+
"step": 58940
|
| 174354 |
+
},
|
| 174355 |
+
{
|
| 174356 |
+
"epoch": 470.66,
|
| 174357 |
+
"learning_rate": 9.071746794871795e-06,
|
| 174358 |
+
"loss": 0.3752,
|
| 174359 |
+
"step": 58945
|
| 174360 |
+
},
|
| 174361 |
+
{
|
| 174362 |
+
"epoch": 470.7,
|
| 174363 |
+
"learning_rate": 9.071666666666668e-06,
|
| 174364 |
+
"loss": 0.6367,
|
| 174365 |
+
"step": 58950
|
| 174366 |
+
},
|
| 174367 |
+
{
|
| 174368 |
+
"epoch": 470.74,
|
| 174369 |
+
"learning_rate": 9.071586538461538e-06,
|
| 174370 |
+
"loss": 0.3487,
|
| 174371 |
+
"step": 58955
|
| 174372 |
+
},
|
| 174373 |
+
{
|
| 174374 |
+
"epoch": 470.78,
|
| 174375 |
+
"learning_rate": 9.071506410256411e-06,
|
| 174376 |
+
"loss": 0.6508,
|
| 174377 |
+
"step": 58960
|
| 174378 |
+
},
|
| 174379 |
+
{
|
| 174380 |
+
"epoch": 470.82,
|
| 174381 |
+
"learning_rate": 9.071426282051284e-06,
|
| 174382 |
+
"loss": 1.1212,
|
| 174383 |
+
"step": 58965
|
| 174384 |
+
},
|
| 174385 |
+
{
|
| 174386 |
+
"epoch": 470.86,
|
| 174387 |
+
"learning_rate": 9.071346153846155e-06,
|
| 174388 |
+
"loss": 0.3319,
|
| 174389 |
+
"step": 58970
|
| 174390 |
+
},
|
| 174391 |
+
{
|
| 174392 |
+
"epoch": 470.9,
|
| 174393 |
+
"learning_rate": 9.071266025641027e-06,
|
| 174394 |
+
"loss": 0.3161,
|
| 174395 |
+
"step": 58975
|
| 174396 |
+
},
|
| 174397 |
+
{
|
| 174398 |
+
"epoch": 470.94,
|
| 174399 |
+
"learning_rate": 9.071185897435898e-06,
|
| 174400 |
+
"loss": 0.4095,
|
| 174401 |
+
"step": 58980
|
| 174402 |
+
},
|
| 174403 |
+
{
|
| 174404 |
+
"epoch": 470.98,
|
| 174405 |
+
"learning_rate": 9.071105769230771e-06,
|
| 174406 |
+
"loss": 0.6715,
|
| 174407 |
+
"step": 58985
|
| 174408 |
+
},
|
| 174409 |
+
{
|
| 174410 |
+
"epoch": 471.0,
|
| 174411 |
+
"eval_loss": 0.39000481367111206,
|
| 174412 |
+
"eval_runtime": 40.6903,
|
| 174413 |
+
"eval_samples_per_second": 20.595,
|
| 174414 |
+
"eval_steps_per_second": 0.664,
|
| 174415 |
+
"eval_wer": 0.20205629283943416,
|
| 174416 |
+
"step": 58988
|
| 174417 |
+
},
|
| 174418 |
+
{
|
| 174419 |
+
"epoch": 475.02,
|
| 174420 |
+
"learning_rate": 9.07102564102564e-06,
|
| 174421 |
+
"loss": 0.3806,
|
| 174422 |
+
"step": 58990
|
| 174423 |
+
},
|
| 174424 |
+
{
|
| 174425 |
+
"epoch": 475.06,
|
| 174426 |
+
"learning_rate": 9.070945512820514e-06,
|
| 174427 |
+
"loss": 0.3352,
|
| 174428 |
+
"step": 58995
|
| 174429 |
+
},
|
| 174430 |
+
{
|
| 174431 |
+
"epoch": 475.1,
|
| 174432 |
+
"learning_rate": 9.070865384615385e-06,
|
| 174433 |
+
"loss": 0.3759,
|
| 174434 |
+
"step": 59000
|
| 174435 |
+
},
|
| 174436 |
+
{
|
| 174437 |
+
"epoch": 475.14,
|
| 174438 |
+
"learning_rate": 9.070785256410257e-06,
|
| 174439 |
+
"loss": 0.3997,
|
| 174440 |
+
"step": 59005
|
| 174441 |
+
},
|
| 174442 |
+
{
|
| 174443 |
+
"epoch": 475.18,
|
| 174444 |
+
"learning_rate": 9.07070512820513e-06,
|
| 174445 |
+
"loss": 0.6099,
|
| 174446 |
+
"step": 59010
|
| 174447 |
+
},
|
| 174448 |
+
{
|
| 174449 |
+
"epoch": 475.22,
|
| 174450 |
+
"learning_rate": 9.070625000000001e-06,
|
| 174451 |
+
"loss": 1.0184,
|
| 174452 |
+
"step": 59015
|
| 174453 |
+
},
|
| 174454 |
+
{
|
| 174455 |
+
"epoch": 475.26,
|
| 174456 |
+
"learning_rate": 9.070544871794872e-06,
|
| 174457 |
+
"loss": 0.3233,
|
| 174458 |
+
"step": 59020
|
| 174459 |
+
},
|
| 174460 |
+
{
|
| 174461 |
+
"epoch": 475.3,
|
| 174462 |
+
"learning_rate": 9.070464743589744e-06,
|
| 174463 |
+
"loss": 0.3247,
|
| 174464 |
+
"step": 59025
|
| 174465 |
+
},
|
| 174466 |
+
{
|
| 174467 |
+
"epoch": 475.34,
|
| 174468 |
+
"learning_rate": 9.070384615384617e-06,
|
| 174469 |
+
"loss": 0.4177,
|
| 174470 |
+
"step": 59030
|
| 174471 |
+
},
|
| 174472 |
+
{
|
| 174473 |
+
"epoch": 475.38,
|
| 174474 |
+
"learning_rate": 9.070304487179488e-06,
|
| 174475 |
+
"loss": 0.689,
|
| 174476 |
+
"step": 59035
|
| 174477 |
+
},
|
| 174478 |
+
{
|
| 174479 |
+
"epoch": 475.42,
|
| 174480 |
+
"learning_rate": 9.07022435897436e-06,
|
| 174481 |
+
"loss": 1.0743,
|
| 174482 |
+
"step": 59040
|
| 174483 |
+
},
|
| 174484 |
+
{
|
| 174485 |
+
"epoch": 475.46,
|
| 174486 |
+
"learning_rate": 9.07014423076923e-06,
|
| 174487 |
+
"loss": 0.4752,
|
| 174488 |
+
"step": 59045
|
| 174489 |
+
},
|
| 174490 |
+
{
|
| 174491 |
+
"epoch": 475.5,
|
| 174492 |
+
"learning_rate": 9.070064102564104e-06,
|
| 174493 |
+
"loss": 0.4128,
|
| 174494 |
+
"step": 59050
|
| 174495 |
+
},
|
| 174496 |
+
{
|
| 174497 |
+
"epoch": 475.54,
|
| 174498 |
+
"learning_rate": 9.069983974358975e-06,
|
| 174499 |
+
"loss": 0.3324,
|
| 174500 |
+
"step": 59055
|
| 174501 |
+
},
|
| 174502 |
+
{
|
| 174503 |
+
"epoch": 475.58,
|
| 174504 |
+
"learning_rate": 9.069903846153847e-06,
|
| 174505 |
+
"loss": 0.6704,
|
| 174506 |
+
"step": 59060
|
| 174507 |
+
},
|
| 174508 |
+
{
|
| 174509 |
+
"epoch": 475.62,
|
| 174510 |
+
"learning_rate": 9.06982371794872e-06,
|
| 174511 |
+
"loss": 1.1681,
|
| 174512 |
+
"step": 59065
|
| 174513 |
+
},
|
| 174514 |
+
{
|
| 174515 |
+
"epoch": 475.66,
|
| 174516 |
+
"learning_rate": 9.069743589743591e-06,
|
| 174517 |
+
"loss": 0.3644,
|
| 174518 |
+
"step": 59070
|
| 174519 |
+
},
|
| 174520 |
+
{
|
| 174521 |
+
"epoch": 475.7,
|
| 174522 |
+
"learning_rate": 9.069663461538462e-06,
|
| 174523 |
+
"loss": 0.3252,
|
| 174524 |
+
"step": 59075
|
| 174525 |
+
},
|
| 174526 |
+
{
|
| 174527 |
+
"epoch": 475.74,
|
| 174528 |
+
"learning_rate": 9.069583333333334e-06,
|
| 174529 |
+
"loss": 0.3882,
|
| 174530 |
+
"step": 59080
|
| 174531 |
+
},
|
| 174532 |
+
{
|
| 174533 |
+
"epoch": 475.78,
|
| 174534 |
+
"learning_rate": 9.069503205128207e-06,
|
| 174535 |
+
"loss": 0.6973,
|
| 174536 |
+
"step": 59085
|
| 174537 |
+
},
|
| 174538 |
+
{
|
| 174539 |
+
"epoch": 475.82,
|
| 174540 |
+
"learning_rate": 9.069423076923078e-06,
|
| 174541 |
+
"loss": 1.0712,
|
| 174542 |
+
"step": 59090
|
| 174543 |
+
},
|
| 174544 |
+
{
|
| 174545 |
+
"epoch": 475.86,
|
| 174546 |
+
"learning_rate": 9.06934294871795e-06,
|
| 174547 |
+
"loss": 0.2953,
|
| 174548 |
+
"step": 59095
|
| 174549 |
+
},
|
| 174550 |
+
{
|
| 174551 |
+
"epoch": 475.9,
|
| 174552 |
+
"learning_rate": 9.06926282051282e-06,
|
| 174553 |
+
"loss": 0.3554,
|
| 174554 |
+
"step": 59100
|
| 174555 |
+
},
|
| 174556 |
+
{
|
| 174557 |
+
"epoch": 475.94,
|
| 174558 |
+
"learning_rate": 9.069182692307694e-06,
|
| 174559 |
+
"loss": 0.3755,
|
| 174560 |
+
"step": 59105
|
| 174561 |
+
},
|
| 174562 |
+
{
|
| 174563 |
+
"epoch": 475.98,
|
| 174564 |
+
"learning_rate": 9.069102564102565e-06,
|
| 174565 |
+
"loss": 0.7179,
|
| 174566 |
+
"step": 59110
|
| 174567 |
+
},
|
| 174568 |
+
{
|
| 174569 |
+
"epoch": 476.0,
|
| 174570 |
+
"eval_loss": 0.3820476233959198,
|
| 174571 |
+
"eval_runtime": 39.9476,
|
| 174572 |
+
"eval_samples_per_second": 21.003,
|
| 174573 |
+
"eval_steps_per_second": 0.676,
|
| 174574 |
+
"eval_wer": 0.19850547499260135,
|
| 174575 |
+
"step": 59112
|
| 174576 |
+
},
|
| 174577 |
+
{
|
| 174578 |
+
"epoch": 476.02,
|
| 174579 |
+
"learning_rate": 9.069022435897437e-06,
|
| 174580 |
+
"loss": 0.3763,
|
| 174581 |
+
"step": 59115
|
| 174582 |
+
},
|
| 174583 |
+
{
|
| 174584 |
+
"epoch": 476.06,
|
| 174585 |
+
"learning_rate": 9.06894230769231e-06,
|
| 174586 |
+
"loss": 0.2986,
|
| 174587 |
+
"step": 59120
|
| 174588 |
+
},
|
| 174589 |
+
{
|
| 174590 |
+
"epoch": 476.1,
|
| 174591 |
+
"learning_rate": 9.06886217948718e-06,
|
| 174592 |
+
"loss": 0.3842,
|
| 174593 |
+
"step": 59125
|
| 174594 |
+
},
|
| 174595 |
+
{
|
| 174596 |
+
"epoch": 476.14,
|
| 174597 |
+
"learning_rate": 9.068782051282052e-06,
|
| 174598 |
+
"loss": 0.4375,
|
| 174599 |
+
"step": 59130
|
| 174600 |
+
},
|
| 174601 |
+
{
|
| 174602 |
+
"epoch": 476.18,
|
| 174603 |
+
"learning_rate": 9.068701923076924e-06,
|
| 174604 |
+
"loss": 0.8778,
|
| 174605 |
+
"step": 59135
|
| 174606 |
+
},
|
| 174607 |
+
{
|
| 174608 |
+
"epoch": 476.22,
|
| 174609 |
+
"learning_rate": 9.068621794871795e-06,
|
| 174610 |
+
"loss": 0.9418,
|
| 174611 |
+
"step": 59140
|
| 174612 |
+
},
|
| 174613 |
+
{
|
| 174614 |
+
"epoch": 476.27,
|
| 174615 |
+
"learning_rate": 9.068541666666666e-06,
|
| 174616 |
+
"loss": 0.2938,
|
| 174617 |
+
"step": 59145
|
| 174618 |
+
},
|
| 174619 |
+
{
|
| 174620 |
+
"epoch": 476.31,
|
| 174621 |
+
"learning_rate": 9.06846153846154e-06,
|
| 174622 |
+
"loss": 0.3299,
|
| 174623 |
+
"step": 59150
|
| 174624 |
+
},
|
| 174625 |
+
{
|
| 174626 |
+
"epoch": 476.35,
|
| 174627 |
+
"learning_rate": 9.06838141025641e-06,
|
| 174628 |
+
"loss": 0.3562,
|
| 174629 |
+
"step": 59155
|
| 174630 |
+
},
|
| 174631 |
+
{
|
| 174632 |
+
"epoch": 476.39,
|
| 174633 |
+
"learning_rate": 9.068301282051282e-06,
|
| 174634 |
+
"loss": 0.7895,
|
| 174635 |
+
"step": 59160
|
| 174636 |
+
},
|
| 174637 |
+
{
|
| 174638 |
+
"epoch": 476.43,
|
| 174639 |
+
"learning_rate": 9.068221153846155e-06,
|
| 174640 |
+
"loss": 0.9889,
|
| 174641 |
+
"step": 59165
|
| 174642 |
+
},
|
| 174643 |
+
{
|
| 174644 |
+
"epoch": 476.47,
|
| 174645 |
+
"learning_rate": 9.068141025641027e-06,
|
| 174646 |
+
"loss": 0.3181,
|
| 174647 |
+
"step": 59170
|
| 174648 |
+
},
|
| 174649 |
+
{
|
| 174650 |
+
"epoch": 476.51,
|
| 174651 |
+
"learning_rate": 9.068060897435898e-06,
|
| 174652 |
+
"loss": 0.3682,
|
| 174653 |
+
"step": 59175
|
| 174654 |
+
},
|
| 174655 |
+
{
|
| 174656 |
+
"epoch": 476.55,
|
| 174657 |
+
"learning_rate": 9.06798076923077e-06,
|
| 174658 |
+
"loss": 0.4526,
|
| 174659 |
+
"step": 59180
|
| 174660 |
+
},
|
| 174661 |
+
{
|
| 174662 |
+
"epoch": 476.59,
|
| 174663 |
+
"learning_rate": 9.067900641025642e-06,
|
| 174664 |
+
"loss": 0.8309,
|
| 174665 |
+
"step": 59185
|
| 174666 |
+
},
|
| 174667 |
+
{
|
| 174668 |
+
"epoch": 476.63,
|
| 174669 |
+
"learning_rate": 9.067820512820514e-06,
|
| 174670 |
+
"loss": 0.907,
|
| 174671 |
+
"step": 59190
|
| 174672 |
+
},
|
| 174673 |
+
{
|
| 174674 |
+
"epoch": 476.67,
|
| 174675 |
+
"learning_rate": 9.067740384615385e-06,
|
| 174676 |
+
"loss": 0.3164,
|
| 174677 |
+
"step": 59195
|
| 174678 |
+
},
|
| 174679 |
+
{
|
| 174680 |
+
"epoch": 476.71,
|
| 174681 |
+
"learning_rate": 9.067660256410256e-06,
|
| 174682 |
+
"loss": 0.4142,
|
| 174683 |
+
"step": 59200
|
| 174684 |
+
},
|
| 174685 |
+
{
|
| 174686 |
+
"epoch": 476.75,
|
| 174687 |
+
"learning_rate": 9.06758012820513e-06,
|
| 174688 |
+
"loss": 0.4289,
|
| 174689 |
+
"step": 59205
|
| 174690 |
+
},
|
| 174691 |
+
{
|
| 174692 |
+
"epoch": 476.79,
|
| 174693 |
+
"learning_rate": 9.067500000000001e-06,
|
| 174694 |
+
"loss": 0.8008,
|
| 174695 |
+
"step": 59210
|
| 174696 |
+
},
|
| 174697 |
+
{
|
| 174698 |
+
"epoch": 476.83,
|
| 174699 |
+
"learning_rate": 9.067419871794872e-06,
|
| 174700 |
+
"loss": 0.893,
|
| 174701 |
+
"step": 59215
|
| 174702 |
+
},
|
| 174703 |
+
{
|
| 174704 |
+
"epoch": 476.87,
|
| 174705 |
+
"learning_rate": 9.067339743589745e-06,
|
| 174706 |
+
"loss": 0.3435,
|
| 174707 |
+
"step": 59220
|
| 174708 |
+
},
|
| 174709 |
+
{
|
| 174710 |
+
"epoch": 476.91,
|
| 174711 |
+
"learning_rate": 9.067259615384617e-06,
|
| 174712 |
+
"loss": 0.3231,
|
| 174713 |
+
"step": 59225
|
| 174714 |
+
},
|
| 174715 |
+
{
|
| 174716 |
+
"epoch": 476.95,
|
| 174717 |
+
"learning_rate": 9.067179487179488e-06,
|
| 174718 |
+
"loss": 0.3646,
|
| 174719 |
+
"step": 59230
|
| 174720 |
+
},
|
| 174721 |
+
{
|
| 174722 |
+
"epoch": 476.99,
|
| 174723 |
+
"learning_rate": 9.06709935897436e-06,
|
| 174724 |
+
"loss": 1.1315,
|
| 174725 |
+
"step": 59235
|
| 174726 |
+
},
|
| 174727 |
+
{
|
| 174728 |
+
"epoch": 477.0,
|
| 174729 |
+
"eval_loss": 0.45738792419433594,
|
| 174730 |
+
"eval_runtime": 40.2604,
|
| 174731 |
+
"eval_samples_per_second": 20.814,
|
| 174732 |
+
"eval_steps_per_second": 0.671,
|
| 174733 |
+
"eval_wer": 0.19495749047200234,
|
| 174734 |
+
"step": 59236
|
| 174735 |
+
},
|
| 174736 |
+
{
|
| 174737 |
+
"epoch": 473.03,
|
| 174738 |
+
"learning_rate": 9.067019230769232e-06,
|
| 174739 |
+
"loss": 0.4845,
|
| 174740 |
+
"step": 59240
|
| 174741 |
+
},
|
| 174742 |
+
{
|
| 174743 |
+
"epoch": 473.07,
|
| 174744 |
+
"learning_rate": 9.066939102564104e-06,
|
| 174745 |
+
"loss": 0.3358,
|
| 174746 |
+
"step": 59245
|
| 174747 |
+
},
|
| 174748 |
+
{
|
| 174749 |
+
"epoch": 473.11,
|
| 174750 |
+
"learning_rate": 9.066858974358975e-06,
|
| 174751 |
+
"loss": 0.3086,
|
| 174752 |
+
"step": 59250
|
| 174753 |
+
},
|
| 174754 |
+
{
|
| 174755 |
+
"epoch": 473.15,
|
| 174756 |
+
"learning_rate": 9.066778846153846e-06,
|
| 174757 |
+
"loss": 0.483,
|
| 174758 |
+
"step": 59255
|
| 174759 |
+
},
|
| 174760 |
+
{
|
| 174761 |
+
"epoch": 473.19,
|
| 174762 |
+
"learning_rate": 9.06669871794872e-06,
|
| 174763 |
+
"loss": 1.0162,
|
| 174764 |
+
"step": 59260
|
| 174765 |
+
},
|
| 174766 |
+
{
|
| 174767 |
+
"epoch": 473.23,
|
| 174768 |
+
"learning_rate": 9.066618589743591e-06,
|
| 174769 |
+
"loss": 0.7503,
|
| 174770 |
+
"step": 59265
|
| 174771 |
+
},
|
| 174772 |
+
{
|
| 174773 |
+
"epoch": 473.27,
|
| 174774 |
+
"learning_rate": 9.066538461538462e-06,
|
| 174775 |
+
"loss": 0.4241,
|
| 174776 |
+
"step": 59270
|
| 174777 |
+
},
|
| 174778 |
+
{
|
| 174779 |
+
"epoch": 473.31,
|
| 174780 |
+
"learning_rate": 9.066458333333335e-06,
|
| 174781 |
+
"loss": 0.3694,
|
| 174782 |
+
"step": 59275
|
| 174783 |
+
},
|
| 174784 |
+
{
|
| 174785 |
+
"epoch": 473.35,
|
| 174786 |
+
"learning_rate": 9.066378205128205e-06,
|
| 174787 |
+
"loss": 0.457,
|
| 174788 |
+
"step": 59280
|
| 174789 |
+
},
|
| 174790 |
+
{
|
| 174791 |
+
"epoch": 473.39,
|
| 174792 |
+
"learning_rate": 9.066298076923078e-06,
|
| 174793 |
+
"loss": 1.1135,
|
| 174794 |
+
"step": 59285
|
| 174795 |
+
},
|
| 174796 |
+
{
|
| 174797 |
+
"epoch": 473.43,
|
| 174798 |
+
"learning_rate": 9.06621794871795e-06,
|
| 174799 |
+
"loss": 0.7058,
|
| 174800 |
+
"step": 59290
|
| 174801 |
+
},
|
| 174802 |
+
{
|
| 174803 |
+
"epoch": 473.47,
|
| 174804 |
+
"learning_rate": 9.06613782051282e-06,
|
| 174805 |
+
"loss": 0.5018,
|
| 174806 |
+
"step": 59295
|
| 174807 |
+
},
|
| 174808 |
+
{
|
| 174809 |
+
"epoch": 473.51,
|
| 174810 |
+
"learning_rate": 9.066057692307692e-06,
|
| 174811 |
+
"loss": 0.3751,
|
| 174812 |
+
"step": 59300
|
| 174813 |
+
},
|
| 174814 |
+
{
|
| 174815 |
+
"epoch": 473.55,
|
| 174816 |
+
"learning_rate": 9.065977564102565e-06,
|
| 174817 |
+
"loss": 0.4068,
|
| 174818 |
+
"step": 59305
|
| 174819 |
+
},
|
| 174820 |
+
{
|
| 174821 |
+
"epoch": 473.59,
|
| 174822 |
+
"learning_rate": 9.065897435897436e-06,
|
| 174823 |
+
"loss": 0.9338,
|
| 174824 |
+
"step": 59310
|
| 174825 |
+
},
|
| 174826 |
+
{
|
| 174827 |
+
"epoch": 473.63,
|
| 174828 |
+
"learning_rate": 9.065817307692308e-06,
|
| 174829 |
+
"loss": 0.6909,
|
| 174830 |
+
"step": 59315
|
| 174831 |
+
},
|
| 174832 |
+
{
|
| 174833 |
+
"epoch": 473.67,
|
| 174834 |
+
"learning_rate": 9.065737179487181e-06,
|
| 174835 |
+
"loss": 0.3748,
|
| 174836 |
+
"step": 59320
|
| 174837 |
+
},
|
| 174838 |
+
{
|
| 174839 |
+
"epoch": 473.71,
|
| 174840 |
+
"learning_rate": 9.065657051282052e-06,
|
| 174841 |
+
"loss": 0.3169,
|
| 174842 |
+
"step": 59325
|
| 174843 |
+
},
|
| 174844 |
+
{
|
| 174845 |
+
"epoch": 473.75,
|
| 174846 |
+
"learning_rate": 9.065576923076924e-06,
|
| 174847 |
+
"loss": 0.367,
|
| 174848 |
+
"step": 59330
|
| 174849 |
+
},
|
| 174850 |
+
{
|
| 174851 |
+
"epoch": 473.79,
|
| 174852 |
+
"learning_rate": 9.065496794871795e-06,
|
| 174853 |
+
"loss": 0.9267,
|
| 174854 |
+
"step": 59335
|
| 174855 |
+
},
|
| 174856 |
+
{
|
| 174857 |
+
"epoch": 473.83,
|
| 174858 |
+
"learning_rate": 9.065416666666668e-06,
|
| 174859 |
+
"loss": 0.7692,
|
| 174860 |
+
"step": 59340
|
| 174861 |
+
},
|
| 174862 |
+
{
|
| 174863 |
+
"epoch": 473.87,
|
| 174864 |
+
"learning_rate": 9.06533653846154e-06,
|
| 174865 |
+
"loss": 0.3793,
|
| 174866 |
+
"step": 59345
|
| 174867 |
+
},
|
| 174868 |
+
{
|
| 174869 |
+
"epoch": 473.91,
|
| 174870 |
+
"learning_rate": 9.06525641025641e-06,
|
| 174871 |
+
"loss": 0.328,
|
| 174872 |
+
"step": 59350
|
| 174873 |
+
},
|
| 174874 |
+
{
|
| 174875 |
+
"epoch": 473.95,
|
| 174876 |
+
"learning_rate": 9.065176282051282e-06,
|
| 174877 |
+
"loss": 0.4363,
|
| 174878 |
+
"step": 59355
|
| 174879 |
+
},
|
| 174880 |
+
{
|
| 174881 |
+
"epoch": 473.99,
|
| 174882 |
+
"learning_rate": 9.065096153846155e-06,
|
| 174883 |
+
"loss": 1.018,
|
| 174884 |
+
"step": 59360
|
| 174885 |
+
},
|
| 174886 |
+
{
|
| 174887 |
+
"epoch": 474.0,
|
| 174888 |
+
"eval_loss": 0.4455079734325409,
|
| 174889 |
+
"eval_runtime": 38.9041,
|
| 174890 |
+
"eval_samples_per_second": 21.54,
|
| 174891 |
+
"eval_steps_per_second": 0.694,
|
| 174892 |
+
"eval_wer": 0.19821858801197342,
|
| 174893 |
+
"step": 59361
|
| 174894 |
+
},
|
| 174895 |
+
{
|
| 174896 |
+
"epoch": 478.03,
|
| 174897 |
+
"learning_rate": 9.065016025641027e-06,
|
| 174898 |
+
"loss": 0.3379,
|
| 174899 |
+
"step": 59365
|
| 174900 |
+
},
|
| 174901 |
+
{
|
| 174902 |
+
"epoch": 478.07,
|
| 174903 |
+
"learning_rate": 9.064935897435898e-06,
|
| 174904 |
+
"loss": 0.503,
|
| 174905 |
+
"step": 59370
|
| 174906 |
+
},
|
| 174907 |
+
{
|
| 174908 |
+
"epoch": 478.11,
|
| 174909 |
+
"learning_rate": 9.064855769230771e-06,
|
| 174910 |
+
"loss": 0.3694,
|
| 174911 |
+
"step": 59375
|
| 174912 |
+
},
|
| 174913 |
+
{
|
| 174914 |
+
"epoch": 478.15,
|
| 174915 |
+
"learning_rate": 9.064775641025642e-06,
|
| 174916 |
+
"loss": 0.4163,
|
| 174917 |
+
"step": 59380
|
| 174918 |
+
},
|
| 174919 |
+
{
|
| 174920 |
+
"epoch": 478.19,
|
| 174921 |
+
"learning_rate": 9.064695512820514e-06,
|
| 174922 |
+
"loss": 0.99,
|
| 174923 |
+
"step": 59385
|
| 174924 |
+
},
|
| 174925 |
+
{
|
| 174926 |
+
"epoch": 478.23,
|
| 174927 |
+
"learning_rate": 9.064615384615385e-06,
|
| 174928 |
+
"loss": 0.8087,
|
| 174929 |
+
"step": 59390
|
| 174930 |
+
},
|
| 174931 |
+
{
|
| 174932 |
+
"epoch": 478.27,
|
| 174933 |
+
"learning_rate": 9.064535256410258e-06,
|
| 174934 |
+
"loss": 0.321,
|
| 174935 |
+
"step": 59395
|
| 174936 |
+
},
|
| 174937 |
+
{
|
| 174938 |
+
"epoch": 478.31,
|
| 174939 |
+
"learning_rate": 9.064455128205128e-06,
|
| 174940 |
+
"loss": 0.3366,
|
| 174941 |
+
"step": 59400
|
| 174942 |
+
},
|
| 174943 |
+
{
|
| 174944 |
+
"epoch": 478.35,
|
| 174945 |
+
"learning_rate": 9.064375e-06,
|
| 174946 |
+
"loss": 0.4313,
|
| 174947 |
+
"step": 59405
|
| 174948 |
+
},
|
| 174949 |
+
{
|
| 174950 |
+
"epoch": 478.4,
|
| 174951 |
+
"learning_rate": 9.064294871794872e-06,
|
| 174952 |
+
"loss": 1.168,
|
| 174953 |
+
"step": 59410
|
| 174954 |
+
},
|
| 174955 |
+
{
|
| 174956 |
+
"epoch": 478.44,
|
| 174957 |
+
"learning_rate": 9.064214743589743e-06,
|
| 174958 |
+
"loss": 0.683,
|
| 174959 |
+
"step": 59415
|
| 174960 |
+
},
|
| 174961 |
+
{
|
| 174962 |
+
"epoch": 478.48,
|
| 174963 |
+
"learning_rate": 9.064134615384617e-06,
|
| 174964 |
+
"loss": 0.3761,
|
| 174965 |
+
"step": 59420
|
| 174966 |
+
},
|
| 174967 |
+
{
|
| 174968 |
+
"epoch": 478.52,
|
| 174969 |
+
"learning_rate": 9.064054487179488e-06,
|
| 174970 |
+
"loss": 0.3645,
|
| 174971 |
+
"step": 59425
|
| 174972 |
+
},
|
| 174973 |
+
{
|
| 174974 |
+
"epoch": 478.56,
|
| 174975 |
+
"learning_rate": 9.06397435897436e-06,
|
| 174976 |
+
"loss": 0.4374,
|
| 174977 |
+
"step": 59430
|
| 174978 |
+
},
|
| 174979 |
+
{
|
| 174980 |
+
"epoch": 478.6,
|
| 174981 |
+
"learning_rate": 9.06389423076923e-06,
|
| 174982 |
+
"loss": 0.9584,
|
| 174983 |
+
"step": 59435
|
| 174984 |
+
},
|
| 174985 |
+
{
|
| 174986 |
+
"epoch": 478.64,
|
| 174987 |
+
"learning_rate": 9.063814102564104e-06,
|
| 174988 |
+
"loss": 0.7471,
|
| 174989 |
+
"step": 59440
|
| 174990 |
+
},
|
| 174991 |
+
{
|
| 174992 |
+
"epoch": 478.68,
|
| 174993 |
+
"learning_rate": 9.063733974358975e-06,
|
| 174994 |
+
"loss": 0.3021,
|
| 174995 |
+
"step": 59445
|
| 174996 |
+
},
|
| 174997 |
+
{
|
| 174998 |
+
"epoch": 478.72,
|
| 174999 |
+
"learning_rate": 9.063653846153846e-06,
|
| 175000 |
+
"loss": 0.3389,
|
| 175001 |
+
"step": 59450
|
| 175002 |
+
},
|
| 175003 |
+
{
|
| 175004 |
+
"epoch": 478.76,
|
| 175005 |
+
"learning_rate": 9.063573717948718e-06,
|
| 175006 |
+
"loss": 0.4655,
|
| 175007 |
+
"step": 59455
|
| 175008 |
+
},
|
| 175009 |
+
{
|
| 175010 |
+
"epoch": 478.8,
|
| 175011 |
+
"learning_rate": 9.06349358974359e-06,
|
| 175012 |
+
"loss": 1.0567,
|
| 175013 |
+
"step": 59460
|
| 175014 |
+
},
|
| 175015 |
+
{
|
| 175016 |
+
"epoch": 478.84,
|
| 175017 |
+
"learning_rate": 9.063413461538462e-06,
|
| 175018 |
+
"loss": 0.6921,
|
| 175019 |
+
"step": 59465
|
| 175020 |
+
},
|
| 175021 |
+
{
|
| 175022 |
+
"epoch": 478.88,
|
| 175023 |
+
"learning_rate": 9.063333333333334e-06,
|
| 175024 |
+
"loss": 0.3324,
|
| 175025 |
+
"step": 59470
|
| 175026 |
+
},
|
| 175027 |
+
{
|
| 175028 |
+
"epoch": 478.92,
|
| 175029 |
+
"learning_rate": 9.063253205128207e-06,
|
| 175030 |
+
"loss": 0.3789,
|
| 175031 |
+
"step": 59475
|
| 175032 |
+
},
|
| 175033 |
+
{
|
| 175034 |
+
"epoch": 478.96,
|
| 175035 |
+
"learning_rate": 9.063173076923078e-06,
|
| 175036 |
+
"loss": 0.4519,
|
| 175037 |
+
"step": 59480
|
| 175038 |
+
},
|
| 175039 |
+
{
|
| 175040 |
+
"epoch": 479.0,
|
| 175041 |
+
"learning_rate": 9.06309294871795e-06,
|
| 175042 |
+
"loss": 1.1864,
|
| 175043 |
+
"step": 59485
|
| 175044 |
+
},
|
| 175045 |
+
{
|
| 175046 |
+
"epoch": 479.0,
|
| 175047 |
+
"eval_loss": 0.4677662253379822,
|
| 175048 |
+
"eval_runtime": 38.993,
|
| 175049 |
+
"eval_samples_per_second": 21.491,
|
| 175050 |
+
"eval_steps_per_second": 0.692,
|
| 175051 |
+
"eval_wer": 0.1971501644135915,
|
| 175052 |
+
"step": 59485
|
| 175053 |
}
|
| 175054 |
],
|
| 175055 |
+
"max_steps": 620000,
|
| 175056 |
"num_train_epochs": 5000,
|
| 175057 |
+
"total_flos": 1.6739973479559817e+20,
|
| 175058 |
"trial_name": null,
|
| 175059 |
"trial_params": null
|
| 175060 |
}
|
model-bin/finetune/base/{checkpoint-58863 → checkpoint-59485}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629791737.1385896/events.out.tfevents.1629791737.c435e1c5ee04.920.81
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:502d5e635078a5c804571ff2e116913c8bcb4f263b00ad727623a63ca95b998c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629792376.705988/events.out.tfevents.1629792376.c435e1c5ee04.920.83
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d24e9a72c70d126232f6f75dfb7c67b4de3b3b305d2637d59723b6d1aa92306f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629793016.5378878/events.out.tfevents.1629793016.c435e1c5ee04.920.85
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:968b5d280e0fbc66bea12e77962c2d93d5b8f4717a415a1342e636bce7e9bbaf
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629793657.2960315/events.out.tfevents.1629793657.c435e1c5ee04.920.87
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ddacd00cf108dee153a1fdc6f6b92aca91a19928d19c4a79f502baeacf5b2cb1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629794296.277906/events.out.tfevents.1629794296.c435e1c5ee04.920.89
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:17ebc5e19267c38bcc6821a522866969c08e504edff48629936bafe6cff49c3e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629791737.c435e1c5ee04.920.80
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7b65409bb4952db478a732a7d551c459100ae169cbe80e43392d24ad16bd4942
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629792376.c435e1c5ee04.920.82
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:652e1b8fab9224c801dee029e61227ebaa8dcd8219f91e2ffe3bde201363fa6f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629793016.c435e1c5ee04.920.84
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:33f2808f63a80739fe4b4a70507c4a49275f16b8795116ab9934d8ac7e6b78f0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629793657.c435e1c5ee04.920.86
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dabe47f79d54fa1a333d22cb1ea2f1cd3f597a520ea353fc4c2518f88a71908e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629794296.c435e1c5ee04.920.88
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:88eef206f8e928ac9e9a2ec4a9a7732cac9904ded338b90178c78273855704d9
|
| 3 |
+
size 8622
|