"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630033662.7831442/events.out.tfevents.1630033662.52f5c7e305a3.886.21 +3 -0
- model-bin/finetune/base/log/1630034124.1396604/events.out.tfevents.1630034124.52f5c7e305a3.886.23 +3 -0
- model-bin/finetune/base/log/1630034675.0/events.out.tfevents.1630034675.52f5c7e305a3.886.25 +3 -0
- model-bin/finetune/base/log/1630035174.274609/events.out.tfevents.1630035174.52f5c7e305a3.886.27 +3 -0
- model-bin/finetune/base/log/1630035652.7388/events.out.tfevents.1630035652.52f5c7e305a3.886.29 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630033662.52f5c7e305a3.886.20 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630034124.52f5c7e305a3.886.22 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630034673.52f5c7e305a3.886.24 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630035174.52f5c7e305a3.886.26 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630035652.52f5c7e305a3.886.28 +3 -0
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/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 sha256:75c6b4c22adf6941b4a07955497e1418455eae7fa0ecaa5770709e321673a877
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/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 sha256:4d7d5b9d0fcdfa9d9129d65f510db2b943dbfe87757f59ad6921419ea80149c6
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/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 sha256:5afc42f10650a3da66fe14af063a8daec2da1423bd18f11524a4151a630892be
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/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 sha256:50b3b42714dc7cff659fc5adea3b07cd7bba81e408274970aa93fa977e8d5ab7
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/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 sha256:5924eb55e74d44fc679e7deaa8ccc722220c05a53978020687e6471f8417c999
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/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,
|
|
@@ -236325,11 +236325,806 @@
|
|
| 236325 |
"eval_steps_per_second": 0.638,
|
| 236326 |
"eval_wer": 0.18214181923859343,
|
| 236327 |
"step": 107649
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 236328 |
}
|
| 236329 |
],
|
| 236330 |
"max_steps": 620000,
|
| 236331 |
"num_train_epochs": 5000,
|
| 236332 |
-
"total_flos": 3.
|
| 236333 |
"trial_name": null,
|
| 236334 |
"trial_params": null
|
| 236335 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 872.995983935743,
|
| 5 |
+
"global_step": 108271,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 236325 |
"eval_steps_per_second": 0.638,
|
| 236326 |
"eval_wer": 0.18214181923859343,
|
| 236327 |
"step": 107649
|
| 236328 |
+
},
|
| 236329 |
+
{
|
| 236330 |
+
"epoch": 868.01,
|
| 236331 |
+
"learning_rate": 8.277754442649435e-06,
|
| 236332 |
+
"loss": 0.3764,
|
| 236333 |
+
"step": 107650
|
| 236334 |
+
},
|
| 236335 |
+
{
|
| 236336 |
+
"epoch": 868.05,
|
| 236337 |
+
"learning_rate": 8.27767366720517e-06,
|
| 236338 |
+
"loss": 0.3286,
|
| 236339 |
+
"step": 107655
|
| 236340 |
+
},
|
| 236341 |
+
{
|
| 236342 |
+
"epoch": 868.09,
|
| 236343 |
+
"learning_rate": 8.277592891760905e-06,
|
| 236344 |
+
"loss": 0.3103,
|
| 236345 |
+
"step": 107660
|
| 236346 |
+
},
|
| 236347 |
+
{
|
| 236348 |
+
"epoch": 868.13,
|
| 236349 |
+
"learning_rate": 8.27751211631664e-06,
|
| 236350 |
+
"loss": 0.351,
|
| 236351 |
+
"step": 107665
|
| 236352 |
+
},
|
| 236353 |
+
{
|
| 236354 |
+
"epoch": 868.17,
|
| 236355 |
+
"learning_rate": 8.277431340872375e-06,
|
| 236356 |
+
"loss": 0.5124,
|
| 236357 |
+
"step": 107670
|
| 236358 |
+
},
|
| 236359 |
+
{
|
| 236360 |
+
"epoch": 868.21,
|
| 236361 |
+
"learning_rate": 8.27735056542811e-06,
|
| 236362 |
+
"loss": 1.3094,
|
| 236363 |
+
"step": 107675
|
| 236364 |
+
},
|
| 236365 |
+
{
|
| 236366 |
+
"epoch": 868.25,
|
| 236367 |
+
"learning_rate": 8.277269789983846e-06,
|
| 236368 |
+
"loss": 0.3836,
|
| 236369 |
+
"step": 107680
|
| 236370 |
+
},
|
| 236371 |
+
{
|
| 236372 |
+
"epoch": 868.29,
|
| 236373 |
+
"learning_rate": 8.27718901453958e-06,
|
| 236374 |
+
"loss": 0.2594,
|
| 236375 |
+
"step": 107685
|
| 236376 |
+
},
|
| 236377 |
+
{
|
| 236378 |
+
"epoch": 868.33,
|
| 236379 |
+
"learning_rate": 8.277108239095316e-06,
|
| 236380 |
+
"loss": 0.3062,
|
| 236381 |
+
"step": 107690
|
| 236382 |
+
},
|
| 236383 |
+
{
|
| 236384 |
+
"epoch": 868.37,
|
| 236385 |
+
"learning_rate": 8.27702746365105e-06,
|
| 236386 |
+
"loss": 0.4898,
|
| 236387 |
+
"step": 107695
|
| 236388 |
+
},
|
| 236389 |
+
{
|
| 236390 |
+
"epoch": 868.41,
|
| 236391 |
+
"learning_rate": 8.276946688206786e-06,
|
| 236392 |
+
"loss": 1.1542,
|
| 236393 |
+
"step": 107700
|
| 236394 |
+
},
|
| 236395 |
+
{
|
| 236396 |
+
"epoch": 868.45,
|
| 236397 |
+
"learning_rate": 8.27686591276252e-06,
|
| 236398 |
+
"loss": 0.3539,
|
| 236399 |
+
"step": 107705
|
| 236400 |
+
},
|
| 236401 |
+
{
|
| 236402 |
+
"epoch": 868.49,
|
| 236403 |
+
"learning_rate": 8.276785137318256e-06,
|
| 236404 |
+
"loss": 0.3048,
|
| 236405 |
+
"step": 107710
|
| 236406 |
+
},
|
| 236407 |
+
{
|
| 236408 |
+
"epoch": 868.53,
|
| 236409 |
+
"learning_rate": 8.27670436187399e-06,
|
| 236410 |
+
"loss": 0.338,
|
| 236411 |
+
"step": 107715
|
| 236412 |
+
},
|
| 236413 |
+
{
|
| 236414 |
+
"epoch": 868.57,
|
| 236415 |
+
"learning_rate": 8.276623586429726e-06,
|
| 236416 |
+
"loss": 0.4273,
|
| 236417 |
+
"step": 107720
|
| 236418 |
+
},
|
| 236419 |
+
{
|
| 236420 |
+
"epoch": 868.61,
|
| 236421 |
+
"learning_rate": 8.27654281098546e-06,
|
| 236422 |
+
"loss": 1.0331,
|
| 236423 |
+
"step": 107725
|
| 236424 |
+
},
|
| 236425 |
+
{
|
| 236426 |
+
"epoch": 868.65,
|
| 236427 |
+
"learning_rate": 8.276462035541196e-06,
|
| 236428 |
+
"loss": 0.2924,
|
| 236429 |
+
"step": 107730
|
| 236430 |
+
},
|
| 236431 |
+
{
|
| 236432 |
+
"epoch": 868.69,
|
| 236433 |
+
"learning_rate": 8.27638126009693e-06,
|
| 236434 |
+
"loss": 0.2745,
|
| 236435 |
+
"step": 107735
|
| 236436 |
+
},
|
| 236437 |
+
{
|
| 236438 |
+
"epoch": 868.73,
|
| 236439 |
+
"learning_rate": 8.276300484652666e-06,
|
| 236440 |
+
"loss": 0.3977,
|
| 236441 |
+
"step": 107740
|
| 236442 |
+
},
|
| 236443 |
+
{
|
| 236444 |
+
"epoch": 868.77,
|
| 236445 |
+
"learning_rate": 8.276219709208402e-06,
|
| 236446 |
+
"loss": 0.5079,
|
| 236447 |
+
"step": 107745
|
| 236448 |
+
},
|
| 236449 |
+
{
|
| 236450 |
+
"epoch": 868.81,
|
| 236451 |
+
"learning_rate": 8.276138933764136e-06,
|
| 236452 |
+
"loss": 1.1278,
|
| 236453 |
+
"step": 107750
|
| 236454 |
+
},
|
| 236455 |
+
{
|
| 236456 |
+
"epoch": 868.85,
|
| 236457 |
+
"learning_rate": 8.276058158319872e-06,
|
| 236458 |
+
"loss": 0.2706,
|
| 236459 |
+
"step": 107755
|
| 236460 |
+
},
|
| 236461 |
+
{
|
| 236462 |
+
"epoch": 868.89,
|
| 236463 |
+
"learning_rate": 8.275977382875606e-06,
|
| 236464 |
+
"loss": 0.2768,
|
| 236465 |
+
"step": 107760
|
| 236466 |
+
},
|
| 236467 |
+
{
|
| 236468 |
+
"epoch": 868.93,
|
| 236469 |
+
"learning_rate": 8.275896607431342e-06,
|
| 236470 |
+
"loss": 0.3489,
|
| 236471 |
+
"step": 107765
|
| 236472 |
+
},
|
| 236473 |
+
{
|
| 236474 |
+
"epoch": 868.97,
|
| 236475 |
+
"learning_rate": 8.275815831987076e-06,
|
| 236476 |
+
"loss": 0.5729,
|
| 236477 |
+
"step": 107770
|
| 236478 |
+
},
|
| 236479 |
+
{
|
| 236480 |
+
"epoch": 869.0,
|
| 236481 |
+
"eval_loss": 0.35712718963623047,
|
| 236482 |
+
"eval_runtime": 42.0074,
|
| 236483 |
+
"eval_samples_per_second": 19.949,
|
| 236484 |
+
"eval_steps_per_second": 0.643,
|
| 236485 |
+
"eval_wer": 0.18226816130441112,
|
| 236486 |
+
"step": 107773
|
| 236487 |
+
},
|
| 236488 |
+
{
|
| 236489 |
+
"epoch": 862.02,
|
| 236490 |
+
"learning_rate": 8.275735056542812e-06,
|
| 236491 |
+
"loss": 0.3065,
|
| 236492 |
+
"step": 107775
|
| 236493 |
+
},
|
| 236494 |
+
{
|
| 236495 |
+
"epoch": 862.06,
|
| 236496 |
+
"learning_rate": 8.275654281098546e-06,
|
| 236497 |
+
"loss": 0.2829,
|
| 236498 |
+
"step": 107780
|
| 236499 |
+
},
|
| 236500 |
+
{
|
| 236501 |
+
"epoch": 862.1,
|
| 236502 |
+
"learning_rate": 8.275573505654282e-06,
|
| 236503 |
+
"loss": 0.2927,
|
| 236504 |
+
"step": 107785
|
| 236505 |
+
},
|
| 236506 |
+
{
|
| 236507 |
+
"epoch": 862.14,
|
| 236508 |
+
"learning_rate": 8.275492730210016e-06,
|
| 236509 |
+
"loss": 0.3984,
|
| 236510 |
+
"step": 107790
|
| 236511 |
+
},
|
| 236512 |
+
{
|
| 236513 |
+
"epoch": 862.18,
|
| 236514 |
+
"learning_rate": 8.275411954765752e-06,
|
| 236515 |
+
"loss": 0.508,
|
| 236516 |
+
"step": 107795
|
| 236517 |
+
},
|
| 236518 |
+
{
|
| 236519 |
+
"epoch": 862.22,
|
| 236520 |
+
"learning_rate": 8.275331179321488e-06,
|
| 236521 |
+
"loss": 0.9641,
|
| 236522 |
+
"step": 107800
|
| 236523 |
+
},
|
| 236524 |
+
{
|
| 236525 |
+
"epoch": 862.26,
|
| 236526 |
+
"learning_rate": 8.275250403877222e-06,
|
| 236527 |
+
"loss": 0.3067,
|
| 236528 |
+
"step": 107805
|
| 236529 |
+
},
|
| 236530 |
+
{
|
| 236531 |
+
"epoch": 862.3,
|
| 236532 |
+
"learning_rate": 8.275169628432958e-06,
|
| 236533 |
+
"loss": 0.2628,
|
| 236534 |
+
"step": 107810
|
| 236535 |
+
},
|
| 236536 |
+
{
|
| 236537 |
+
"epoch": 862.34,
|
| 236538 |
+
"learning_rate": 8.275088852988692e-06,
|
| 236539 |
+
"loss": 0.3289,
|
| 236540 |
+
"step": 107815
|
| 236541 |
+
},
|
| 236542 |
+
{
|
| 236543 |
+
"epoch": 862.38,
|
| 236544 |
+
"learning_rate": 8.275008077544428e-06,
|
| 236545 |
+
"loss": 0.5552,
|
| 236546 |
+
"step": 107820
|
| 236547 |
+
},
|
| 236548 |
+
{
|
| 236549 |
+
"epoch": 862.42,
|
| 236550 |
+
"learning_rate": 8.274927302100162e-06,
|
| 236551 |
+
"loss": 1.0026,
|
| 236552 |
+
"step": 107825
|
| 236553 |
+
},
|
| 236554 |
+
{
|
| 236555 |
+
"epoch": 862.46,
|
| 236556 |
+
"learning_rate": 8.274846526655898e-06,
|
| 236557 |
+
"loss": 0.3078,
|
| 236558 |
+
"step": 107830
|
| 236559 |
+
},
|
| 236560 |
+
{
|
| 236561 |
+
"epoch": 862.5,
|
| 236562 |
+
"learning_rate": 8.274765751211632e-06,
|
| 236563 |
+
"loss": 0.3241,
|
| 236564 |
+
"step": 107835
|
| 236565 |
+
},
|
| 236566 |
+
{
|
| 236567 |
+
"epoch": 862.54,
|
| 236568 |
+
"learning_rate": 8.274684975767368e-06,
|
| 236569 |
+
"loss": 0.3367,
|
| 236570 |
+
"step": 107840
|
| 236571 |
+
},
|
| 236572 |
+
{
|
| 236573 |
+
"epoch": 862.58,
|
| 236574 |
+
"learning_rate": 8.274604200323102e-06,
|
| 236575 |
+
"loss": 0.5786,
|
| 236576 |
+
"step": 107845
|
| 236577 |
+
},
|
| 236578 |
+
{
|
| 236579 |
+
"epoch": 862.62,
|
| 236580 |
+
"learning_rate": 8.274523424878838e-06,
|
| 236581 |
+
"loss": 0.9167,
|
| 236582 |
+
"step": 107850
|
| 236583 |
+
},
|
| 236584 |
+
{
|
| 236585 |
+
"epoch": 862.66,
|
| 236586 |
+
"learning_rate": 8.274442649434574e-06,
|
| 236587 |
+
"loss": 0.372,
|
| 236588 |
+
"step": 107855
|
| 236589 |
+
},
|
| 236590 |
+
{
|
| 236591 |
+
"epoch": 862.7,
|
| 236592 |
+
"learning_rate": 8.274361873990308e-06,
|
| 236593 |
+
"loss": 0.2746,
|
| 236594 |
+
"step": 107860
|
| 236595 |
+
},
|
| 236596 |
+
{
|
| 236597 |
+
"epoch": 862.74,
|
| 236598 |
+
"learning_rate": 8.274281098546044e-06,
|
| 236599 |
+
"loss": 0.3219,
|
| 236600 |
+
"step": 107865
|
| 236601 |
+
},
|
| 236602 |
+
{
|
| 236603 |
+
"epoch": 862.78,
|
| 236604 |
+
"learning_rate": 8.274200323101778e-06,
|
| 236605 |
+
"loss": 0.5105,
|
| 236606 |
+
"step": 107870
|
| 236607 |
+
},
|
| 236608 |
+
{
|
| 236609 |
+
"epoch": 862.82,
|
| 236610 |
+
"learning_rate": 8.274119547657514e-06,
|
| 236611 |
+
"loss": 0.9359,
|
| 236612 |
+
"step": 107875
|
| 236613 |
+
},
|
| 236614 |
+
{
|
| 236615 |
+
"epoch": 862.86,
|
| 236616 |
+
"learning_rate": 8.274038772213248e-06,
|
| 236617 |
+
"loss": 0.276,
|
| 236618 |
+
"step": 107880
|
| 236619 |
+
},
|
| 236620 |
+
{
|
| 236621 |
+
"epoch": 862.9,
|
| 236622 |
+
"learning_rate": 8.273957996768984e-06,
|
| 236623 |
+
"loss": 0.3328,
|
| 236624 |
+
"step": 107885
|
| 236625 |
+
},
|
| 236626 |
+
{
|
| 236627 |
+
"epoch": 862.94,
|
| 236628 |
+
"learning_rate": 8.273877221324718e-06,
|
| 236629 |
+
"loss": 0.3753,
|
| 236630 |
+
"step": 107890
|
| 236631 |
+
},
|
| 236632 |
+
{
|
| 236633 |
+
"epoch": 862.98,
|
| 236634 |
+
"learning_rate": 8.273796445880453e-06,
|
| 236635 |
+
"loss": 0.5561,
|
| 236636 |
+
"step": 107895
|
| 236637 |
+
},
|
| 236638 |
+
{
|
| 236639 |
+
"epoch": 863.0,
|
| 236640 |
+
"eval_loss": 0.3618251383304596,
|
| 236641 |
+
"eval_runtime": 43.5968,
|
| 236642 |
+
"eval_samples_per_second": 19.222,
|
| 236643 |
+
"eval_steps_per_second": 0.619,
|
| 236644 |
+
"eval_wer": 0.18531263728217895,
|
| 236645 |
+
"step": 107898
|
| 236646 |
+
},
|
| 236647 |
+
{
|
| 236648 |
+
"epoch": 870.02,
|
| 236649 |
+
"learning_rate": 8.273715670436188e-06,
|
| 236650 |
+
"loss": 0.3881,
|
| 236651 |
+
"step": 107900
|
| 236652 |
+
},
|
| 236653 |
+
{
|
| 236654 |
+
"epoch": 870.06,
|
| 236655 |
+
"learning_rate": 8.273634894991923e-06,
|
| 236656 |
+
"loss": 0.2775,
|
| 236657 |
+
"step": 107905
|
| 236658 |
+
},
|
| 236659 |
+
{
|
| 236660 |
+
"epoch": 870.1,
|
| 236661 |
+
"learning_rate": 8.273554119547658e-06,
|
| 236662 |
+
"loss": 0.3118,
|
| 236663 |
+
"step": 107910
|
| 236664 |
+
},
|
| 236665 |
+
{
|
| 236666 |
+
"epoch": 870.14,
|
| 236667 |
+
"learning_rate": 8.273473344103393e-06,
|
| 236668 |
+
"loss": 0.3418,
|
| 236669 |
+
"step": 107915
|
| 236670 |
+
},
|
| 236671 |
+
{
|
| 236672 |
+
"epoch": 870.18,
|
| 236673 |
+
"learning_rate": 8.27339256865913e-06,
|
| 236674 |
+
"loss": 0.6577,
|
| 236675 |
+
"step": 107920
|
| 236676 |
+
},
|
| 236677 |
+
{
|
| 236678 |
+
"epoch": 870.22,
|
| 236679 |
+
"learning_rate": 8.273311793214863e-06,
|
| 236680 |
+
"loss": 0.9149,
|
| 236681 |
+
"step": 107925
|
| 236682 |
+
},
|
| 236683 |
+
{
|
| 236684 |
+
"epoch": 870.26,
|
| 236685 |
+
"learning_rate": 8.2732310177706e-06,
|
| 236686 |
+
"loss": 0.285,
|
| 236687 |
+
"step": 107930
|
| 236688 |
+
},
|
| 236689 |
+
{
|
| 236690 |
+
"epoch": 870.3,
|
| 236691 |
+
"learning_rate": 8.273150242326333e-06,
|
| 236692 |
+
"loss": 0.2842,
|
| 236693 |
+
"step": 107935
|
| 236694 |
+
},
|
| 236695 |
+
{
|
| 236696 |
+
"epoch": 870.34,
|
| 236697 |
+
"learning_rate": 8.27306946688207e-06,
|
| 236698 |
+
"loss": 0.3287,
|
| 236699 |
+
"step": 107940
|
| 236700 |
+
},
|
| 236701 |
+
{
|
| 236702 |
+
"epoch": 870.38,
|
| 236703 |
+
"learning_rate": 8.272988691437803e-06,
|
| 236704 |
+
"loss": 0.6697,
|
| 236705 |
+
"step": 107945
|
| 236706 |
+
},
|
| 236707 |
+
{
|
| 236708 |
+
"epoch": 870.42,
|
| 236709 |
+
"learning_rate": 8.27290791599354e-06,
|
| 236710 |
+
"loss": 0.9871,
|
| 236711 |
+
"step": 107950
|
| 236712 |
+
},
|
| 236713 |
+
{
|
| 236714 |
+
"epoch": 870.46,
|
| 236715 |
+
"learning_rate": 8.272827140549273e-06,
|
| 236716 |
+
"loss": 0.3111,
|
| 236717 |
+
"step": 107955
|
| 236718 |
+
},
|
| 236719 |
+
{
|
| 236720 |
+
"epoch": 870.5,
|
| 236721 |
+
"learning_rate": 8.27274636510501e-06,
|
| 236722 |
+
"loss": 0.2628,
|
| 236723 |
+
"step": 107960
|
| 236724 |
+
},
|
| 236725 |
+
{
|
| 236726 |
+
"epoch": 870.54,
|
| 236727 |
+
"learning_rate": 8.272665589660743e-06,
|
| 236728 |
+
"loss": 0.3619,
|
| 236729 |
+
"step": 107965
|
| 236730 |
+
},
|
| 236731 |
+
{
|
| 236732 |
+
"epoch": 870.58,
|
| 236733 |
+
"learning_rate": 8.272584814216479e-06,
|
| 236734 |
+
"loss": 0.604,
|
| 236735 |
+
"step": 107970
|
| 236736 |
+
},
|
| 236737 |
+
{
|
| 236738 |
+
"epoch": 870.62,
|
| 236739 |
+
"learning_rate": 8.272504038772215e-06,
|
| 236740 |
+
"loss": 1.0354,
|
| 236741 |
+
"step": 107975
|
| 236742 |
+
},
|
| 236743 |
+
{
|
| 236744 |
+
"epoch": 870.66,
|
| 236745 |
+
"learning_rate": 8.272423263327949e-06,
|
| 236746 |
+
"loss": 0.2997,
|
| 236747 |
+
"step": 107980
|
| 236748 |
+
},
|
| 236749 |
+
{
|
| 236750 |
+
"epoch": 870.7,
|
| 236751 |
+
"learning_rate": 8.272342487883685e-06,
|
| 236752 |
+
"loss": 0.3036,
|
| 236753 |
+
"step": 107985
|
| 236754 |
+
},
|
| 236755 |
+
{
|
| 236756 |
+
"epoch": 870.74,
|
| 236757 |
+
"learning_rate": 8.272261712439419e-06,
|
| 236758 |
+
"loss": 0.3245,
|
| 236759 |
+
"step": 107990
|
| 236760 |
+
},
|
| 236761 |
+
{
|
| 236762 |
+
"epoch": 870.78,
|
| 236763 |
+
"learning_rate": 8.272180936995155e-06,
|
| 236764 |
+
"loss": 0.5752,
|
| 236765 |
+
"step": 107995
|
| 236766 |
+
},
|
| 236767 |
+
{
|
| 236768 |
+
"epoch": 870.82,
|
| 236769 |
+
"learning_rate": 8.272100161550889e-06,
|
| 236770 |
+
"loss": 1.0438,
|
| 236771 |
+
"step": 108000
|
| 236772 |
+
},
|
| 236773 |
+
{
|
| 236774 |
+
"epoch": 870.86,
|
| 236775 |
+
"learning_rate": 8.272019386106625e-06,
|
| 236776 |
+
"loss": 0.2946,
|
| 236777 |
+
"step": 108005
|
| 236778 |
+
},
|
| 236779 |
+
{
|
| 236780 |
+
"epoch": 870.9,
|
| 236781 |
+
"learning_rate": 8.271938610662359e-06,
|
| 236782 |
+
"loss": 0.337,
|
| 236783 |
+
"step": 108010
|
| 236784 |
+
},
|
| 236785 |
+
{
|
| 236786 |
+
"epoch": 870.94,
|
| 236787 |
+
"learning_rate": 8.271857835218095e-06,
|
| 236788 |
+
"loss": 0.3651,
|
| 236789 |
+
"step": 108015
|
| 236790 |
+
},
|
| 236791 |
+
{
|
| 236792 |
+
"epoch": 870.98,
|
| 236793 |
+
"learning_rate": 8.271777059773829e-06,
|
| 236794 |
+
"loss": 0.6707,
|
| 236795 |
+
"step": 108020
|
| 236796 |
+
},
|
| 236797 |
+
{
|
| 236798 |
+
"epoch": 871.0,
|
| 236799 |
+
"eval_loss": 0.3929585814476013,
|
| 236800 |
+
"eval_runtime": 43.9649,
|
| 236801 |
+
"eval_samples_per_second": 19.038,
|
| 236802 |
+
"eval_steps_per_second": 0.614,
|
| 236803 |
+
"eval_wer": 0.19029392648227053,
|
| 236804 |
+
"step": 108022
|
| 236805 |
+
},
|
| 236806 |
+
{
|
| 236807 |
+
"epoch": 864.02,
|
| 236808 |
+
"learning_rate": 8.271696284329565e-06,
|
| 236809 |
+
"loss": 0.357,
|
| 236810 |
+
"step": 108025
|
| 236811 |
+
},
|
| 236812 |
+
{
|
| 236813 |
+
"epoch": 864.06,
|
| 236814 |
+
"learning_rate": 8.2716155088853e-06,
|
| 236815 |
+
"loss": 0.3147,
|
| 236816 |
+
"step": 108030
|
| 236817 |
+
},
|
| 236818 |
+
{
|
| 236819 |
+
"epoch": 864.1,
|
| 236820 |
+
"learning_rate": 8.271534733441035e-06,
|
| 236821 |
+
"loss": 0.3054,
|
| 236822 |
+
"step": 108035
|
| 236823 |
+
},
|
| 236824 |
+
{
|
| 236825 |
+
"epoch": 864.14,
|
| 236826 |
+
"learning_rate": 8.27145395799677e-06,
|
| 236827 |
+
"loss": 0.3213,
|
| 236828 |
+
"step": 108040
|
| 236829 |
+
},
|
| 236830 |
+
{
|
| 236831 |
+
"epoch": 864.18,
|
| 236832 |
+
"learning_rate": 8.271373182552505e-06,
|
| 236833 |
+
"loss": 0.8225,
|
| 236834 |
+
"step": 108045
|
| 236835 |
+
},
|
| 236836 |
+
{
|
| 236837 |
+
"epoch": 864.22,
|
| 236838 |
+
"learning_rate": 8.27129240710824e-06,
|
| 236839 |
+
"loss": 0.7424,
|
| 236840 |
+
"step": 108050
|
| 236841 |
+
},
|
| 236842 |
+
{
|
| 236843 |
+
"epoch": 864.26,
|
| 236844 |
+
"learning_rate": 8.271211631663975e-06,
|
| 236845 |
+
"loss": 0.3202,
|
| 236846 |
+
"step": 108055
|
| 236847 |
+
},
|
| 236848 |
+
{
|
| 236849 |
+
"epoch": 864.3,
|
| 236850 |
+
"learning_rate": 8.27113085621971e-06,
|
| 236851 |
+
"loss": 0.3282,
|
| 236852 |
+
"step": 108060
|
| 236853 |
+
},
|
| 236854 |
+
{
|
| 236855 |
+
"epoch": 864.34,
|
| 236856 |
+
"learning_rate": 8.271050080775445e-06,
|
| 236857 |
+
"loss": 0.3521,
|
| 236858 |
+
"step": 108065
|
| 236859 |
+
},
|
| 236860 |
+
{
|
| 236861 |
+
"epoch": 864.38,
|
| 236862 |
+
"learning_rate": 8.27096930533118e-06,
|
| 236863 |
+
"loss": 0.875,
|
| 236864 |
+
"step": 108070
|
| 236865 |
+
},
|
| 236866 |
+
{
|
| 236867 |
+
"epoch": 864.42,
|
| 236868 |
+
"learning_rate": 8.270888529886915e-06,
|
| 236869 |
+
"loss": 0.8975,
|
| 236870 |
+
"step": 108075
|
| 236871 |
+
},
|
| 236872 |
+
{
|
| 236873 |
+
"epoch": 864.46,
|
| 236874 |
+
"learning_rate": 8.27080775444265e-06,
|
| 236875 |
+
"loss": 0.2981,
|
| 236876 |
+
"step": 108080
|
| 236877 |
+
},
|
| 236878 |
+
{
|
| 236879 |
+
"epoch": 864.5,
|
| 236880 |
+
"learning_rate": 8.270726978998385e-06,
|
| 236881 |
+
"loss": 0.2764,
|
| 236882 |
+
"step": 108085
|
| 236883 |
+
},
|
| 236884 |
+
{
|
| 236885 |
+
"epoch": 864.54,
|
| 236886 |
+
"learning_rate": 8.27064620355412e-06,
|
| 236887 |
+
"loss": 0.3825,
|
| 236888 |
+
"step": 108090
|
| 236889 |
+
},
|
| 236890 |
+
{
|
| 236891 |
+
"epoch": 864.58,
|
| 236892 |
+
"learning_rate": 8.270565428109856e-06,
|
| 236893 |
+
"loss": 0.7116,
|
| 236894 |
+
"step": 108095
|
| 236895 |
+
},
|
| 236896 |
+
{
|
| 236897 |
+
"epoch": 864.62,
|
| 236898 |
+
"learning_rate": 8.27048465266559e-06,
|
| 236899 |
+
"loss": 0.763,
|
| 236900 |
+
"step": 108100
|
| 236901 |
+
},
|
| 236902 |
+
{
|
| 236903 |
+
"epoch": 864.66,
|
| 236904 |
+
"learning_rate": 8.270403877221326e-06,
|
| 236905 |
+
"loss": 0.2727,
|
| 236906 |
+
"step": 108105
|
| 236907 |
+
},
|
| 236908 |
+
{
|
| 236909 |
+
"epoch": 864.7,
|
| 236910 |
+
"learning_rate": 8.27032310177706e-06,
|
| 236911 |
+
"loss": 0.2901,
|
| 236912 |
+
"step": 108110
|
| 236913 |
+
},
|
| 236914 |
+
{
|
| 236915 |
+
"epoch": 864.74,
|
| 236916 |
+
"learning_rate": 8.270242326332796e-06,
|
| 236917 |
+
"loss": 0.3345,
|
| 236918 |
+
"step": 108115
|
| 236919 |
+
},
|
| 236920 |
+
{
|
| 236921 |
+
"epoch": 864.78,
|
| 236922 |
+
"learning_rate": 8.27016155088853e-06,
|
| 236923 |
+
"loss": 0.6988,
|
| 236924 |
+
"step": 108120
|
| 236925 |
+
},
|
| 236926 |
+
{
|
| 236927 |
+
"epoch": 864.82,
|
| 236928 |
+
"learning_rate": 8.270080775444266e-06,
|
| 236929 |
+
"loss": 0.88,
|
| 236930 |
+
"step": 108125
|
| 236931 |
+
},
|
| 236932 |
+
{
|
| 236933 |
+
"epoch": 864.86,
|
| 236934 |
+
"learning_rate": 8.27e-06,
|
| 236935 |
+
"loss": 0.2457,
|
| 236936 |
+
"step": 108130
|
| 236937 |
+
},
|
| 236938 |
+
{
|
| 236939 |
+
"epoch": 864.9,
|
| 236940 |
+
"learning_rate": 8.269919224555736e-06,
|
| 236941 |
+
"loss": 0.3383,
|
| 236942 |
+
"step": 108135
|
| 236943 |
+
},
|
| 236944 |
+
{
|
| 236945 |
+
"epoch": 864.94,
|
| 236946 |
+
"learning_rate": 8.26983844911147e-06,
|
| 236947 |
+
"loss": 0.3584,
|
| 236948 |
+
"step": 108140
|
| 236949 |
+
},
|
| 236950 |
+
{
|
| 236951 |
+
"epoch": 864.98,
|
| 236952 |
+
"learning_rate": 8.269757673667206e-06,
|
| 236953 |
+
"loss": 0.7838,
|
| 236954 |
+
"step": 108145
|
| 236955 |
+
},
|
| 236956 |
+
{
|
| 236957 |
+
"epoch": 865.0,
|
| 236958 |
+
"eval_loss": 0.48220154643058777,
|
| 236959 |
+
"eval_runtime": 43.8282,
|
| 236960 |
+
"eval_samples_per_second": 19.12,
|
| 236961 |
+
"eval_steps_per_second": 0.616,
|
| 236962 |
+
"eval_wer": 0.17802808660035108,
|
| 236963 |
+
"step": 108147
|
| 236964 |
+
},
|
| 236965 |
+
{
|
| 236966 |
+
"epoch": 872.02,
|
| 236967 |
+
"learning_rate": 8.269676898222942e-06,
|
| 236968 |
+
"loss": 0.3217,
|
| 236969 |
+
"step": 108150
|
| 236970 |
+
},
|
| 236971 |
+
{
|
| 236972 |
+
"epoch": 872.06,
|
| 236973 |
+
"learning_rate": 8.269596122778676e-06,
|
| 236974 |
+
"loss": 0.3162,
|
| 236975 |
+
"step": 108155
|
| 236976 |
+
},
|
| 236977 |
+
{
|
| 236978 |
+
"epoch": 872.1,
|
| 236979 |
+
"learning_rate": 8.269515347334412e-06,
|
| 236980 |
+
"loss": 0.3366,
|
| 236981 |
+
"step": 108160
|
| 236982 |
+
},
|
| 236983 |
+
{
|
| 236984 |
+
"epoch": 872.14,
|
| 236985 |
+
"learning_rate": 8.269434571890146e-06,
|
| 236986 |
+
"loss": 0.3481,
|
| 236987 |
+
"step": 108165
|
| 236988 |
+
},
|
| 236989 |
+
{
|
| 236990 |
+
"epoch": 872.18,
|
| 236991 |
+
"learning_rate": 8.269353796445882e-06,
|
| 236992 |
+
"loss": 0.6536,
|
| 236993 |
+
"step": 108170
|
| 236994 |
+
},
|
| 236995 |
+
{
|
| 236996 |
+
"epoch": 872.22,
|
| 236997 |
+
"learning_rate": 8.269273021001616e-06,
|
| 236998 |
+
"loss": 0.959,
|
| 236999 |
+
"step": 108175
|
| 237000 |
+
},
|
| 237001 |
+
{
|
| 237002 |
+
"epoch": 872.27,
|
| 237003 |
+
"learning_rate": 8.269192245557352e-06,
|
| 237004 |
+
"loss": 0.3007,
|
| 237005 |
+
"step": 108180
|
| 237006 |
+
},
|
| 237007 |
+
{
|
| 237008 |
+
"epoch": 872.31,
|
| 237009 |
+
"learning_rate": 8.269111470113086e-06,
|
| 237010 |
+
"loss": 0.3131,
|
| 237011 |
+
"step": 108185
|
| 237012 |
+
},
|
| 237013 |
+
{
|
| 237014 |
+
"epoch": 872.35,
|
| 237015 |
+
"learning_rate": 8.269030694668822e-06,
|
| 237016 |
+
"loss": 0.3185,
|
| 237017 |
+
"step": 108190
|
| 237018 |
+
},
|
| 237019 |
+
{
|
| 237020 |
+
"epoch": 872.39,
|
| 237021 |
+
"learning_rate": 8.268949919224556e-06,
|
| 237022 |
+
"loss": 0.738,
|
| 237023 |
+
"step": 108195
|
| 237024 |
+
},
|
| 237025 |
+
{
|
| 237026 |
+
"epoch": 872.43,
|
| 237027 |
+
"learning_rate": 8.268869143780292e-06,
|
| 237028 |
+
"loss": 0.7973,
|
| 237029 |
+
"step": 108200
|
| 237030 |
+
},
|
| 237031 |
+
{
|
| 237032 |
+
"epoch": 872.47,
|
| 237033 |
+
"learning_rate": 8.268788368336028e-06,
|
| 237034 |
+
"loss": 0.2688,
|
| 237035 |
+
"step": 108205
|
| 237036 |
+
},
|
| 237037 |
+
{
|
| 237038 |
+
"epoch": 872.51,
|
| 237039 |
+
"learning_rate": 8.268707592891762e-06,
|
| 237040 |
+
"loss": 0.2801,
|
| 237041 |
+
"step": 108210
|
| 237042 |
+
},
|
| 237043 |
+
{
|
| 237044 |
+
"epoch": 872.55,
|
| 237045 |
+
"learning_rate": 8.268626817447498e-06,
|
| 237046 |
+
"loss": 0.3582,
|
| 237047 |
+
"step": 108215
|
| 237048 |
+
},
|
| 237049 |
+
{
|
| 237050 |
+
"epoch": 872.59,
|
| 237051 |
+
"learning_rate": 8.268546042003232e-06,
|
| 237052 |
+
"loss": 1.0277,
|
| 237053 |
+
"step": 108220
|
| 237054 |
+
},
|
| 237055 |
+
{
|
| 237056 |
+
"epoch": 872.63,
|
| 237057 |
+
"learning_rate": 8.268465266558968e-06,
|
| 237058 |
+
"loss": 0.9334,
|
| 237059 |
+
"step": 108225
|
| 237060 |
+
},
|
| 237061 |
+
{
|
| 237062 |
+
"epoch": 872.67,
|
| 237063 |
+
"learning_rate": 8.268384491114702e-06,
|
| 237064 |
+
"loss": 0.2694,
|
| 237065 |
+
"step": 108230
|
| 237066 |
+
},
|
| 237067 |
+
{
|
| 237068 |
+
"epoch": 872.71,
|
| 237069 |
+
"learning_rate": 8.268303715670438e-06,
|
| 237070 |
+
"loss": 0.2765,
|
| 237071 |
+
"step": 108235
|
| 237072 |
+
},
|
| 237073 |
+
{
|
| 237074 |
+
"epoch": 872.75,
|
| 237075 |
+
"learning_rate": 8.268222940226172e-06,
|
| 237076 |
+
"loss": 0.3387,
|
| 237077 |
+
"step": 108240
|
| 237078 |
+
},
|
| 237079 |
+
{
|
| 237080 |
+
"epoch": 872.79,
|
| 237081 |
+
"learning_rate": 8.268142164781908e-06,
|
| 237082 |
+
"loss": 0.7583,
|
| 237083 |
+
"step": 108245
|
| 237084 |
+
},
|
| 237085 |
+
{
|
| 237086 |
+
"epoch": 872.83,
|
| 237087 |
+
"learning_rate": 8.268061389337642e-06,
|
| 237088 |
+
"loss": 0.8668,
|
| 237089 |
+
"step": 108250
|
| 237090 |
+
},
|
| 237091 |
+
{
|
| 237092 |
+
"epoch": 872.87,
|
| 237093 |
+
"learning_rate": 8.267980613893378e-06,
|
| 237094 |
+
"loss": 0.2671,
|
| 237095 |
+
"step": 108255
|
| 237096 |
+
},
|
| 237097 |
+
{
|
| 237098 |
+
"epoch": 872.91,
|
| 237099 |
+
"learning_rate": 8.267899838449112e-06,
|
| 237100 |
+
"loss": 0.3526,
|
| 237101 |
+
"step": 108260
|
| 237102 |
+
},
|
| 237103 |
+
{
|
| 237104 |
+
"epoch": 872.95,
|
| 237105 |
+
"learning_rate": 8.267819063004848e-06,
|
| 237106 |
+
"loss": 0.3374,
|
| 237107 |
+
"step": 108265
|
| 237108 |
+
},
|
| 237109 |
+
{
|
| 237110 |
+
"epoch": 872.99,
|
| 237111 |
+
"learning_rate": 8.267738287560583e-06,
|
| 237112 |
+
"loss": 0.8303,
|
| 237113 |
+
"step": 108270
|
| 237114 |
+
},
|
| 237115 |
+
{
|
| 237116 |
+
"epoch": 873.0,
|
| 237117 |
+
"eval_loss": 0.41857171058654785,
|
| 237118 |
+
"eval_runtime": 42.4687,
|
| 237119 |
+
"eval_samples_per_second": 19.732,
|
| 237120 |
+
"eval_steps_per_second": 0.636,
|
| 237121 |
+
"eval_wer": 0.1901564344746163,
|
| 237122 |
+
"step": 108271
|
| 237123 |
}
|
| 237124 |
],
|
| 237125 |
"max_steps": 620000,
|
| 237126 |
"num_train_epochs": 5000,
|
| 237127 |
+
"total_flos": 3.046909280261572e+20,
|
| 237128 |
"trial_name": null,
|
| 237129 |
"trial_params": null
|
| 237130 |
}
|
model-bin/finetune/base/{checkpoint-107649 → checkpoint-108271}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630033662.7831442/events.out.tfevents.1630033662.52f5c7e305a3.886.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:42a2cc3997e1371baf0af119d881c6a01f91aab68979d64f3f96e6615dd4e6f5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630034124.1396604/events.out.tfevents.1630034124.52f5c7e305a3.886.23
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c5fbde1ca932c7fae4903ca3ae3f9b5f7ff50369431b2a4351a97080bf9dcfc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630034675.0/events.out.tfevents.1630034675.52f5c7e305a3.886.25
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:39e5d3fd3f23c4f000af36c4b3d193339411018e1731b7b6d176e2074f25b90e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630035174.274609/events.out.tfevents.1630035174.52f5c7e305a3.886.27
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:682a5351a4aaf3f1963c798329b6a4bb3b166a0dbf85eb577e4b8267dfa2f15d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630035652.7388/events.out.tfevents.1630035652.52f5c7e305a3.886.29
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c5d1003f9cec94203ecdb5761aedf77d16b45334cb22012cb9c830b350235791
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630033662.52f5c7e305a3.886.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:30fa37afa1d82e480dd414adf8330a7d4bcabaf055495557d136112f0ae45cd0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630034124.52f5c7e305a3.886.22
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ccf314c084cb12bd0cbaeb4a265645200b90ae3a2b1569a644869453b9825641
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630034673.52f5c7e305a3.886.24
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:24abafda70392fa10a2b0603f69c893d51fef6e3a78812dfb8ddb3c8e9413596
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630035174.52f5c7e305a3.886.26
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8a0e43d84f26abbaa4869047b0a96fed37870f05cbac82b6c5ae382bc60460d4
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630035652.52f5c7e305a3.886.28
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9bb084ddcd321ecfb7236267fc9a3cc1bc5477aeabb5443adcff3dacab8c725d
|
| 3 |
+
size 8622
|