"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629755733.4601321/events.out.tfevents.1629755733.74272264b15c.932.233 +3 -0
- model-bin/finetune/base/log/1629756373.9871938/events.out.tfevents.1629756373.74272264b15c.932.235 +3 -0
- model-bin/finetune/base/log/1629757025.4645097/events.out.tfevents.1629757025.74272264b15c.932.237 +3 -0
- model-bin/finetune/base/log/1629757671.1643252/events.out.tfevents.1629757671.74272264b15c.932.239 +3 -0
- model-bin/finetune/base/log/1629758322.377062/events.out.tfevents.1629758322.74272264b15c.932.241 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629755733.74272264b15c.932.232 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629756373.74272264b15c.932.234 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629757025.74272264b15c.932.236 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629757671.74272264b15c.932.238 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629758322.74272264b15c.932.240 +3 -0
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/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 sha256:626b3f50af4abb59b8ff7f6804e06113fb806913351046af8543ea49da10c813
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/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:d352b6d04a55702a9681be729eddc009d2ea5243b18a85abd21f667f39d49109
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/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:dbb325d6f8bea8b4f4941152996b4536b066af419a1d56f45a2e5ec2f1f384ba
|
| 3 |
+
size 14439
|
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/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:a219303497a68a287afddd4ed7c81f20be58bcdfd986786df2e94c5cc582f93f
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/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:cfc84a5ed027e14acf0de020be6f00867fccccd474c7cf821d3fac65edf38749
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18992848189928482,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-52515",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -166335,11 +166335,806 @@
|
|
| 166335 |
"eval_steps_per_second": 0.667,
|
| 166336 |
"eval_wer": 0.20118429709774105,
|
| 166337 |
"step": 52639
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 166338 |
}
|
| 166339 |
],
|
| 166340 |
"max_steps": 620000,
|
| 166341 |
"num_train_epochs": 5000,
|
| 166342 |
-
"total_flos": 1.
|
| 166343 |
"trial_name": null,
|
| 166344 |
"trial_params": null
|
| 166345 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18992848189928482,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-52515",
|
| 4 |
+
"epoch": 429.0,
|
| 5 |
+
"global_step": 53261,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 166335 |
"eval_steps_per_second": 0.667,
|
| 166336 |
"eval_wer": 0.20118429709774105,
|
| 166337 |
"step": 52639
|
| 166338 |
+
},
|
| 166339 |
+
{
|
| 166340 |
+
"epoch": 421.01,
|
| 166341 |
+
"learning_rate": 9.172740384615385e-06,
|
| 166342 |
+
"loss": 0.3764,
|
| 166343 |
+
"step": 52640
|
| 166344 |
+
},
|
| 166345 |
+
{
|
| 166346 |
+
"epoch": 421.05,
|
| 166347 |
+
"learning_rate": 9.172660256410258e-06,
|
| 166348 |
+
"loss": 0.4723,
|
| 166349 |
+
"step": 52645
|
| 166350 |
+
},
|
| 166351 |
+
{
|
| 166352 |
+
"epoch": 421.09,
|
| 166353 |
+
"learning_rate": 9.17258012820513e-06,
|
| 166354 |
+
"loss": 0.3263,
|
| 166355 |
+
"step": 52650
|
| 166356 |
+
},
|
| 166357 |
+
{
|
| 166358 |
+
"epoch": 421.13,
|
| 166359 |
+
"learning_rate": 9.1725e-06,
|
| 166360 |
+
"loss": 0.3414,
|
| 166361 |
+
"step": 52655
|
| 166362 |
+
},
|
| 166363 |
+
{
|
| 166364 |
+
"epoch": 421.17,
|
| 166365 |
+
"learning_rate": 9.172419871794872e-06,
|
| 166366 |
+
"loss": 0.5096,
|
| 166367 |
+
"step": 52660
|
| 166368 |
+
},
|
| 166369 |
+
{
|
| 166370 |
+
"epoch": 421.21,
|
| 166371 |
+
"learning_rate": 9.172339743589745e-06,
|
| 166372 |
+
"loss": 1.3892,
|
| 166373 |
+
"step": 52665
|
| 166374 |
+
},
|
| 166375 |
+
{
|
| 166376 |
+
"epoch": 421.25,
|
| 166377 |
+
"learning_rate": 9.172259615384616e-06,
|
| 166378 |
+
"loss": 0.3895,
|
| 166379 |
+
"step": 52670
|
| 166380 |
+
},
|
| 166381 |
+
{
|
| 166382 |
+
"epoch": 421.29,
|
| 166383 |
+
"learning_rate": 9.172179487179488e-06,
|
| 166384 |
+
"loss": 0.3716,
|
| 166385 |
+
"step": 52675
|
| 166386 |
+
},
|
| 166387 |
+
{
|
| 166388 |
+
"epoch": 421.33,
|
| 166389 |
+
"learning_rate": 9.172099358974361e-06,
|
| 166390 |
+
"loss": 0.3533,
|
| 166391 |
+
"step": 52680
|
| 166392 |
+
},
|
| 166393 |
+
{
|
| 166394 |
+
"epoch": 421.37,
|
| 166395 |
+
"learning_rate": 9.17201923076923e-06,
|
| 166396 |
+
"loss": 0.6364,
|
| 166397 |
+
"step": 52685
|
| 166398 |
+
},
|
| 166399 |
+
{
|
| 166400 |
+
"epoch": 421.41,
|
| 166401 |
+
"learning_rate": 9.171939102564104e-06,
|
| 166402 |
+
"loss": 1.2984,
|
| 166403 |
+
"step": 52690
|
| 166404 |
+
},
|
| 166405 |
+
{
|
| 166406 |
+
"epoch": 421.45,
|
| 166407 |
+
"learning_rate": 9.171858974358975e-06,
|
| 166408 |
+
"loss": 0.367,
|
| 166409 |
+
"step": 52695
|
| 166410 |
+
},
|
| 166411 |
+
{
|
| 166412 |
+
"epoch": 421.49,
|
| 166413 |
+
"learning_rate": 9.171778846153846e-06,
|
| 166414 |
+
"loss": 0.3319,
|
| 166415 |
+
"step": 52700
|
| 166416 |
+
},
|
| 166417 |
+
{
|
| 166418 |
+
"epoch": 421.53,
|
| 166419 |
+
"learning_rate": 9.171698717948718e-06,
|
| 166420 |
+
"loss": 0.3717,
|
| 166421 |
+
"step": 52705
|
| 166422 |
+
},
|
| 166423 |
+
{
|
| 166424 |
+
"epoch": 421.57,
|
| 166425 |
+
"learning_rate": 9.17161858974359e-06,
|
| 166426 |
+
"loss": 0.5638,
|
| 166427 |
+
"step": 52710
|
| 166428 |
+
},
|
| 166429 |
+
{
|
| 166430 |
+
"epoch": 421.61,
|
| 166431 |
+
"learning_rate": 9.171538461538462e-06,
|
| 166432 |
+
"loss": 1.1557,
|
| 166433 |
+
"step": 52715
|
| 166434 |
+
},
|
| 166435 |
+
{
|
| 166436 |
+
"epoch": 421.65,
|
| 166437 |
+
"learning_rate": 9.171458333333333e-06,
|
| 166438 |
+
"loss": 0.4361,
|
| 166439 |
+
"step": 52720
|
| 166440 |
+
},
|
| 166441 |
+
{
|
| 166442 |
+
"epoch": 421.69,
|
| 166443 |
+
"learning_rate": 9.171378205128206e-06,
|
| 166444 |
+
"loss": 0.3131,
|
| 166445 |
+
"step": 52725
|
| 166446 |
+
},
|
| 166447 |
+
{
|
| 166448 |
+
"epoch": 421.73,
|
| 166449 |
+
"learning_rate": 9.171298076923078e-06,
|
| 166450 |
+
"loss": 0.3745,
|
| 166451 |
+
"step": 52730
|
| 166452 |
+
},
|
| 166453 |
+
{
|
| 166454 |
+
"epoch": 421.77,
|
| 166455 |
+
"learning_rate": 9.17121794871795e-06,
|
| 166456 |
+
"loss": 0.5119,
|
| 166457 |
+
"step": 52735
|
| 166458 |
+
},
|
| 166459 |
+
{
|
| 166460 |
+
"epoch": 421.81,
|
| 166461 |
+
"learning_rate": 9.17113782051282e-06,
|
| 166462 |
+
"loss": 1.1258,
|
| 166463 |
+
"step": 52740
|
| 166464 |
+
},
|
| 166465 |
+
{
|
| 166466 |
+
"epoch": 421.85,
|
| 166467 |
+
"learning_rate": 9.171057692307694e-06,
|
| 166468 |
+
"loss": 0.3384,
|
| 166469 |
+
"step": 52745
|
| 166470 |
+
},
|
| 166471 |
+
{
|
| 166472 |
+
"epoch": 421.89,
|
| 166473 |
+
"learning_rate": 9.170977564102565e-06,
|
| 166474 |
+
"loss": 0.3044,
|
| 166475 |
+
"step": 52750
|
| 166476 |
+
},
|
| 166477 |
+
{
|
| 166478 |
+
"epoch": 421.93,
|
| 166479 |
+
"learning_rate": 9.170897435897436e-06,
|
| 166480 |
+
"loss": 0.4152,
|
| 166481 |
+
"step": 52755
|
| 166482 |
+
},
|
| 166483 |
+
{
|
| 166484 |
+
"epoch": 421.97,
|
| 166485 |
+
"learning_rate": 9.170817307692308e-06,
|
| 166486 |
+
"loss": 0.605,
|
| 166487 |
+
"step": 52760
|
| 166488 |
+
},
|
| 166489 |
+
{
|
| 166490 |
+
"epoch": 422.0,
|
| 166491 |
+
"eval_loss": 0.4416767954826355,
|
| 166492 |
+
"eval_runtime": 38.8014,
|
| 166493 |
+
"eval_samples_per_second": 21.649,
|
| 166494 |
+
"eval_steps_per_second": 0.696,
|
| 166495 |
+
"eval_wer": 0.20573733751680862,
|
| 166496 |
+
"step": 52764
|
| 166497 |
+
},
|
| 166498 |
+
{
|
| 166499 |
+
"epoch": 425.01,
|
| 166500 |
+
"learning_rate": 9.17073717948718e-06,
|
| 166501 |
+
"loss": 0.4205,
|
| 166502 |
+
"step": 52765
|
| 166503 |
+
},
|
| 166504 |
+
{
|
| 166505 |
+
"epoch": 425.05,
|
| 166506 |
+
"learning_rate": 9.170657051282052e-06,
|
| 166507 |
+
"loss": 0.3089,
|
| 166508 |
+
"step": 52770
|
| 166509 |
+
},
|
| 166510 |
+
{
|
| 166511 |
+
"epoch": 425.09,
|
| 166512 |
+
"learning_rate": 9.170576923076923e-06,
|
| 166513 |
+
"loss": 0.3435,
|
| 166514 |
+
"step": 52775
|
| 166515 |
+
},
|
| 166516 |
+
{
|
| 166517 |
+
"epoch": 425.13,
|
| 166518 |
+
"learning_rate": 9.170496794871797e-06,
|
| 166519 |
+
"loss": 0.3688,
|
| 166520 |
+
"step": 52780
|
| 166521 |
+
},
|
| 166522 |
+
{
|
| 166523 |
+
"epoch": 425.17,
|
| 166524 |
+
"learning_rate": 9.170416666666668e-06,
|
| 166525 |
+
"loss": 0.6443,
|
| 166526 |
+
"step": 52785
|
| 166527 |
+
},
|
| 166528 |
+
{
|
| 166529 |
+
"epoch": 425.21,
|
| 166530 |
+
"learning_rate": 9.17033653846154e-06,
|
| 166531 |
+
"loss": 1.2337,
|
| 166532 |
+
"step": 52790
|
| 166533 |
+
},
|
| 166534 |
+
{
|
| 166535 |
+
"epoch": 425.25,
|
| 166536 |
+
"learning_rate": 9.17025641025641e-06,
|
| 166537 |
+
"loss": 0.3467,
|
| 166538 |
+
"step": 52795
|
| 166539 |
+
},
|
| 166540 |
+
{
|
| 166541 |
+
"epoch": 425.29,
|
| 166542 |
+
"learning_rate": 9.170176282051284e-06,
|
| 166543 |
+
"loss": 0.3054,
|
| 166544 |
+
"step": 52800
|
| 166545 |
+
},
|
| 166546 |
+
{
|
| 166547 |
+
"epoch": 425.33,
|
| 166548 |
+
"learning_rate": 9.170096153846153e-06,
|
| 166549 |
+
"loss": 0.3231,
|
| 166550 |
+
"step": 52805
|
| 166551 |
+
},
|
| 166552 |
+
{
|
| 166553 |
+
"epoch": 425.37,
|
| 166554 |
+
"learning_rate": 9.170016025641026e-06,
|
| 166555 |
+
"loss": 0.5788,
|
| 166556 |
+
"step": 52810
|
| 166557 |
+
},
|
| 166558 |
+
{
|
| 166559 |
+
"epoch": 425.41,
|
| 166560 |
+
"learning_rate": 9.1699358974359e-06,
|
| 166561 |
+
"loss": 1.4097,
|
| 166562 |
+
"step": 52815
|
| 166563 |
+
},
|
| 166564 |
+
{
|
| 166565 |
+
"epoch": 425.45,
|
| 166566 |
+
"learning_rate": 9.169855769230769e-06,
|
| 166567 |
+
"loss": 0.3334,
|
| 166568 |
+
"step": 52820
|
| 166569 |
+
},
|
| 166570 |
+
{
|
| 166571 |
+
"epoch": 425.49,
|
| 166572 |
+
"learning_rate": 9.169775641025642e-06,
|
| 166573 |
+
"loss": 0.3256,
|
| 166574 |
+
"step": 52825
|
| 166575 |
+
},
|
| 166576 |
+
{
|
| 166577 |
+
"epoch": 425.53,
|
| 166578 |
+
"learning_rate": 9.169695512820513e-06,
|
| 166579 |
+
"loss": 0.3063,
|
| 166580 |
+
"step": 52830
|
| 166581 |
+
},
|
| 166582 |
+
{
|
| 166583 |
+
"epoch": 425.57,
|
| 166584 |
+
"learning_rate": 9.169615384615387e-06,
|
| 166585 |
+
"loss": 0.6159,
|
| 166586 |
+
"step": 52835
|
| 166587 |
+
},
|
| 166588 |
+
{
|
| 166589 |
+
"epoch": 425.61,
|
| 166590 |
+
"learning_rate": 9.169535256410256e-06,
|
| 166591 |
+
"loss": 1.2831,
|
| 166592 |
+
"step": 52840
|
| 166593 |
+
},
|
| 166594 |
+
{
|
| 166595 |
+
"epoch": 425.65,
|
| 166596 |
+
"learning_rate": 9.16945512820513e-06,
|
| 166597 |
+
"loss": 0.3114,
|
| 166598 |
+
"step": 52845
|
| 166599 |
+
},
|
| 166600 |
+
{
|
| 166601 |
+
"epoch": 425.69,
|
| 166602 |
+
"learning_rate": 9.169375e-06,
|
| 166603 |
+
"loss": 0.3027,
|
| 166604 |
+
"step": 52850
|
| 166605 |
+
},
|
| 166606 |
+
{
|
| 166607 |
+
"epoch": 425.73,
|
| 166608 |
+
"learning_rate": 9.169294871794872e-06,
|
| 166609 |
+
"loss": 0.4015,
|
| 166610 |
+
"step": 52855
|
| 166611 |
+
},
|
| 166612 |
+
{
|
| 166613 |
+
"epoch": 425.77,
|
| 166614 |
+
"learning_rate": 9.169214743589743e-06,
|
| 166615 |
+
"loss": 0.6918,
|
| 166616 |
+
"step": 52860
|
| 166617 |
+
},
|
| 166618 |
+
{
|
| 166619 |
+
"epoch": 425.81,
|
| 166620 |
+
"learning_rate": 9.169134615384616e-06,
|
| 166621 |
+
"loss": 1.3139,
|
| 166622 |
+
"step": 52865
|
| 166623 |
+
},
|
| 166624 |
+
{
|
| 166625 |
+
"epoch": 425.85,
|
| 166626 |
+
"learning_rate": 9.169054487179488e-06,
|
| 166627 |
+
"loss": 0.3231,
|
| 166628 |
+
"step": 52870
|
| 166629 |
+
},
|
| 166630 |
+
{
|
| 166631 |
+
"epoch": 425.89,
|
| 166632 |
+
"learning_rate": 9.168974358974359e-06,
|
| 166633 |
+
"loss": 0.3526,
|
| 166634 |
+
"step": 52875
|
| 166635 |
+
},
|
| 166636 |
+
{
|
| 166637 |
+
"epoch": 425.93,
|
| 166638 |
+
"learning_rate": 9.168894230769232e-06,
|
| 166639 |
+
"loss": 0.3667,
|
| 166640 |
+
"step": 52880
|
| 166641 |
+
},
|
| 166642 |
+
{
|
| 166643 |
+
"epoch": 425.97,
|
| 166644 |
+
"learning_rate": 9.168814102564104e-06,
|
| 166645 |
+
"loss": 0.6799,
|
| 166646 |
+
"step": 52885
|
| 166647 |
+
},
|
| 166648 |
+
{
|
| 166649 |
+
"epoch": 426.0,
|
| 166650 |
+
"eval_loss": 0.4513387680053711,
|
| 166651 |
+
"eval_runtime": 39.7893,
|
| 166652 |
+
"eval_samples_per_second": 21.086,
|
| 166653 |
+
"eval_steps_per_second": 0.679,
|
| 166654 |
+
"eval_wer": 0.20593128390596746,
|
| 166655 |
+
"step": 52888
|
| 166656 |
+
},
|
| 166657 |
+
{
|
| 166658 |
+
"epoch": 423.02,
|
| 166659 |
+
"learning_rate": 9.168733974358975e-06,
|
| 166660 |
+
"loss": 0.5276,
|
| 166661 |
+
"step": 52890
|
| 166662 |
+
},
|
| 166663 |
+
{
|
| 166664 |
+
"epoch": 423.06,
|
| 166665 |
+
"learning_rate": 9.168653846153846e-06,
|
| 166666 |
+
"loss": 0.3482,
|
| 166667 |
+
"step": 52895
|
| 166668 |
+
},
|
| 166669 |
+
{
|
| 166670 |
+
"epoch": 423.1,
|
| 166671 |
+
"learning_rate": 9.16857371794872e-06,
|
| 166672 |
+
"loss": 0.3158,
|
| 166673 |
+
"step": 52900
|
| 166674 |
+
},
|
| 166675 |
+
{
|
| 166676 |
+
"epoch": 423.14,
|
| 166677 |
+
"learning_rate": 9.16849358974359e-06,
|
| 166678 |
+
"loss": 0.3622,
|
| 166679 |
+
"step": 52905
|
| 166680 |
+
},
|
| 166681 |
+
{
|
| 166682 |
+
"epoch": 423.18,
|
| 166683 |
+
"learning_rate": 9.168413461538462e-06,
|
| 166684 |
+
"loss": 0.7214,
|
| 166685 |
+
"step": 52910
|
| 166686 |
+
},
|
| 166687 |
+
{
|
| 166688 |
+
"epoch": 423.22,
|
| 166689 |
+
"learning_rate": 9.168333333333333e-06,
|
| 166690 |
+
"loss": 1.0821,
|
| 166691 |
+
"step": 52915
|
| 166692 |
+
},
|
| 166693 |
+
{
|
| 166694 |
+
"epoch": 423.26,
|
| 166695 |
+
"learning_rate": 9.168253205128206e-06,
|
| 166696 |
+
"loss": 0.3709,
|
| 166697 |
+
"step": 52920
|
| 166698 |
+
},
|
| 166699 |
+
{
|
| 166700 |
+
"epoch": 423.3,
|
| 166701 |
+
"learning_rate": 9.168173076923078e-06,
|
| 166702 |
+
"loss": 0.3363,
|
| 166703 |
+
"step": 52925
|
| 166704 |
+
},
|
| 166705 |
+
{
|
| 166706 |
+
"epoch": 423.34,
|
| 166707 |
+
"learning_rate": 9.168092948717949e-06,
|
| 166708 |
+
"loss": 0.3673,
|
| 166709 |
+
"step": 52930
|
| 166710 |
+
},
|
| 166711 |
+
{
|
| 166712 |
+
"epoch": 423.38,
|
| 166713 |
+
"learning_rate": 9.168012820512822e-06,
|
| 166714 |
+
"loss": 0.7023,
|
| 166715 |
+
"step": 52935
|
| 166716 |
+
},
|
| 166717 |
+
{
|
| 166718 |
+
"epoch": 423.42,
|
| 166719 |
+
"learning_rate": 9.167932692307694e-06,
|
| 166720 |
+
"loss": 1.1365,
|
| 166721 |
+
"step": 52940
|
| 166722 |
+
},
|
| 166723 |
+
{
|
| 166724 |
+
"epoch": 423.46,
|
| 166725 |
+
"learning_rate": 9.167852564102565e-06,
|
| 166726 |
+
"loss": 0.3282,
|
| 166727 |
+
"step": 52945
|
| 166728 |
+
},
|
| 166729 |
+
{
|
| 166730 |
+
"epoch": 423.5,
|
| 166731 |
+
"learning_rate": 9.167772435897436e-06,
|
| 166732 |
+
"loss": 0.3293,
|
| 166733 |
+
"step": 52950
|
| 166734 |
+
},
|
| 166735 |
+
{
|
| 166736 |
+
"epoch": 423.54,
|
| 166737 |
+
"learning_rate": 9.16769230769231e-06,
|
| 166738 |
+
"loss": 0.4172,
|
| 166739 |
+
"step": 52955
|
| 166740 |
+
},
|
| 166741 |
+
{
|
| 166742 |
+
"epoch": 423.58,
|
| 166743 |
+
"learning_rate": 9.167612179487179e-06,
|
| 166744 |
+
"loss": 0.6418,
|
| 166745 |
+
"step": 52960
|
| 166746 |
+
},
|
| 166747 |
+
{
|
| 166748 |
+
"epoch": 423.62,
|
| 166749 |
+
"learning_rate": 9.167532051282052e-06,
|
| 166750 |
+
"loss": 1.0752,
|
| 166751 |
+
"step": 52965
|
| 166752 |
+
},
|
| 166753 |
+
{
|
| 166754 |
+
"epoch": 423.66,
|
| 166755 |
+
"learning_rate": 9.167451923076925e-06,
|
| 166756 |
+
"loss": 0.365,
|
| 166757 |
+
"step": 52970
|
| 166758 |
+
},
|
| 166759 |
+
{
|
| 166760 |
+
"epoch": 423.7,
|
| 166761 |
+
"learning_rate": 9.167371794871795e-06,
|
| 166762 |
+
"loss": 0.3403,
|
| 166763 |
+
"step": 52975
|
| 166764 |
+
},
|
| 166765 |
+
{
|
| 166766 |
+
"epoch": 423.74,
|
| 166767 |
+
"learning_rate": 9.167291666666668e-06,
|
| 166768 |
+
"loss": 0.3715,
|
| 166769 |
+
"step": 52980
|
| 166770 |
+
},
|
| 166771 |
+
{
|
| 166772 |
+
"epoch": 423.78,
|
| 166773 |
+
"learning_rate": 9.16721153846154e-06,
|
| 166774 |
+
"loss": 0.6747,
|
| 166775 |
+
"step": 52985
|
| 166776 |
+
},
|
| 166777 |
+
{
|
| 166778 |
+
"epoch": 423.82,
|
| 166779 |
+
"learning_rate": 9.16713141025641e-06,
|
| 166780 |
+
"loss": 1.2286,
|
| 166781 |
+
"step": 52990
|
| 166782 |
+
},
|
| 166783 |
+
{
|
| 166784 |
+
"epoch": 423.86,
|
| 166785 |
+
"learning_rate": 9.167051282051282e-06,
|
| 166786 |
+
"loss": 0.3437,
|
| 166787 |
+
"step": 52995
|
| 166788 |
+
},
|
| 166789 |
+
{
|
| 166790 |
+
"epoch": 423.9,
|
| 166791 |
+
"learning_rate": 9.166971153846155e-06,
|
| 166792 |
+
"loss": 0.3155,
|
| 166793 |
+
"step": 53000
|
| 166794 |
+
},
|
| 166795 |
+
{
|
| 166796 |
+
"epoch": 423.94,
|
| 166797 |
+
"learning_rate": 9.166891025641026e-06,
|
| 166798 |
+
"loss": 0.3946,
|
| 166799 |
+
"step": 53005
|
| 166800 |
+
},
|
| 166801 |
+
{
|
| 166802 |
+
"epoch": 423.98,
|
| 166803 |
+
"learning_rate": 9.166810897435898e-06,
|
| 166804 |
+
"loss": 0.7243,
|
| 166805 |
+
"step": 53010
|
| 166806 |
+
},
|
| 166807 |
+
{
|
| 166808 |
+
"epoch": 424.0,
|
| 166809 |
+
"eval_loss": 0.37052494287490845,
|
| 166810 |
+
"eval_runtime": 39.814,
|
| 166811 |
+
"eval_samples_per_second": 21.073,
|
| 166812 |
+
"eval_steps_per_second": 0.678,
|
| 166813 |
+
"eval_wer": 0.19977989728539985,
|
| 166814 |
+
"step": 53013
|
| 166815 |
+
},
|
| 166816 |
+
{
|
| 166817 |
+
"epoch": 427.02,
|
| 166818 |
+
"learning_rate": 9.166730769230769e-06,
|
| 166819 |
+
"loss": 0.4042,
|
| 166820 |
+
"step": 53015
|
| 166821 |
+
},
|
| 166822 |
+
{
|
| 166823 |
+
"epoch": 427.06,
|
| 166824 |
+
"learning_rate": 9.166650641025642e-06,
|
| 166825 |
+
"loss": 0.3449,
|
| 166826 |
+
"step": 53020
|
| 166827 |
+
},
|
| 166828 |
+
{
|
| 166829 |
+
"epoch": 427.1,
|
| 166830 |
+
"learning_rate": 9.166570512820513e-06,
|
| 166831 |
+
"loss": 0.3405,
|
| 166832 |
+
"step": 53025
|
| 166833 |
+
},
|
| 166834 |
+
{
|
| 166835 |
+
"epoch": 427.14,
|
| 166836 |
+
"learning_rate": 9.166490384615385e-06,
|
| 166837 |
+
"loss": 0.411,
|
| 166838 |
+
"step": 53030
|
| 166839 |
+
},
|
| 166840 |
+
{
|
| 166841 |
+
"epoch": 427.18,
|
| 166842 |
+
"learning_rate": 9.166410256410258e-06,
|
| 166843 |
+
"loss": 0.5731,
|
| 166844 |
+
"step": 53035
|
| 166845 |
+
},
|
| 166846 |
+
{
|
| 166847 |
+
"epoch": 427.22,
|
| 166848 |
+
"learning_rate": 9.16633012820513e-06,
|
| 166849 |
+
"loss": 1.0737,
|
| 166850 |
+
"step": 53040
|
| 166851 |
+
},
|
| 166852 |
+
{
|
| 166853 |
+
"epoch": 427.26,
|
| 166854 |
+
"learning_rate": 9.16625e-06,
|
| 166855 |
+
"loss": 0.3384,
|
| 166856 |
+
"step": 53045
|
| 166857 |
+
},
|
| 166858 |
+
{
|
| 166859 |
+
"epoch": 427.3,
|
| 166860 |
+
"learning_rate": 9.166169871794872e-06,
|
| 166861 |
+
"loss": 0.3478,
|
| 166862 |
+
"step": 53050
|
| 166863 |
+
},
|
| 166864 |
+
{
|
| 166865 |
+
"epoch": 427.34,
|
| 166866 |
+
"learning_rate": 9.166089743589745e-06,
|
| 166867 |
+
"loss": 0.3982,
|
| 166868 |
+
"step": 53055
|
| 166869 |
+
},
|
| 166870 |
+
{
|
| 166871 |
+
"epoch": 427.38,
|
| 166872 |
+
"learning_rate": 9.166009615384616e-06,
|
| 166873 |
+
"loss": 0.6412,
|
| 166874 |
+
"step": 53060
|
| 166875 |
+
},
|
| 166876 |
+
{
|
| 166877 |
+
"epoch": 427.42,
|
| 166878 |
+
"learning_rate": 9.165929487179488e-06,
|
| 166879 |
+
"loss": 1.1081,
|
| 166880 |
+
"step": 53065
|
| 166881 |
+
},
|
| 166882 |
+
{
|
| 166883 |
+
"epoch": 427.46,
|
| 166884 |
+
"learning_rate": 9.16584935897436e-06,
|
| 166885 |
+
"loss": 0.3373,
|
| 166886 |
+
"step": 53070
|
| 166887 |
+
},
|
| 166888 |
+
{
|
| 166889 |
+
"epoch": 427.5,
|
| 166890 |
+
"learning_rate": 9.165769230769232e-06,
|
| 166891 |
+
"loss": 0.3116,
|
| 166892 |
+
"step": 53075
|
| 166893 |
+
},
|
| 166894 |
+
{
|
| 166895 |
+
"epoch": 427.54,
|
| 166896 |
+
"learning_rate": 9.165689102564103e-06,
|
| 166897 |
+
"loss": 0.482,
|
| 166898 |
+
"step": 53080
|
| 166899 |
+
},
|
| 166900 |
+
{
|
| 166901 |
+
"epoch": 427.58,
|
| 166902 |
+
"learning_rate": 9.165608974358975e-06,
|
| 166903 |
+
"loss": 0.6877,
|
| 166904 |
+
"step": 53085
|
| 166905 |
+
},
|
| 166906 |
+
{
|
| 166907 |
+
"epoch": 427.62,
|
| 166908 |
+
"learning_rate": 9.165528846153848e-06,
|
| 166909 |
+
"loss": 0.9501,
|
| 166910 |
+
"step": 53090
|
| 166911 |
+
},
|
| 166912 |
+
{
|
| 166913 |
+
"epoch": 427.66,
|
| 166914 |
+
"learning_rate": 9.16544871794872e-06,
|
| 166915 |
+
"loss": 0.4583,
|
| 166916 |
+
"step": 53095
|
| 166917 |
+
},
|
| 166918 |
+
{
|
| 166919 |
+
"epoch": 427.7,
|
| 166920 |
+
"learning_rate": 9.16536858974359e-06,
|
| 166921 |
+
"loss": 0.4234,
|
| 166922 |
+
"step": 53100
|
| 166923 |
+
},
|
| 166924 |
+
{
|
| 166925 |
+
"epoch": 427.74,
|
| 166926 |
+
"learning_rate": 9.165288461538462e-06,
|
| 166927 |
+
"loss": 0.4676,
|
| 166928 |
+
"step": 53105
|
| 166929 |
+
},
|
| 166930 |
+
{
|
| 166931 |
+
"epoch": 427.78,
|
| 166932 |
+
"learning_rate": 9.165208333333335e-06,
|
| 166933 |
+
"loss": 0.7411,
|
| 166934 |
+
"step": 53110
|
| 166935 |
+
},
|
| 166936 |
+
{
|
| 166937 |
+
"epoch": 427.82,
|
| 166938 |
+
"learning_rate": 9.165128205128205e-06,
|
| 166939 |
+
"loss": 1.0585,
|
| 166940 |
+
"step": 53115
|
| 166941 |
+
},
|
| 166942 |
+
{
|
| 166943 |
+
"epoch": 427.86,
|
| 166944 |
+
"learning_rate": 9.165048076923078e-06,
|
| 166945 |
+
"loss": 0.3383,
|
| 166946 |
+
"step": 53120
|
| 166947 |
+
},
|
| 166948 |
+
{
|
| 166949 |
+
"epoch": 427.9,
|
| 166950 |
+
"learning_rate": 9.16496794871795e-06,
|
| 166951 |
+
"loss": 0.3084,
|
| 166952 |
+
"step": 53125
|
| 166953 |
+
},
|
| 166954 |
+
{
|
| 166955 |
+
"epoch": 427.94,
|
| 166956 |
+
"learning_rate": 9.16488782051282e-06,
|
| 166957 |
+
"loss": 0.3699,
|
| 166958 |
+
"step": 53130
|
| 166959 |
+
},
|
| 166960 |
+
{
|
| 166961 |
+
"epoch": 427.98,
|
| 166962 |
+
"learning_rate": 9.164807692307693e-06,
|
| 166963 |
+
"loss": 0.7039,
|
| 166964 |
+
"step": 53135
|
| 166965 |
+
},
|
| 166966 |
+
{
|
| 166967 |
+
"epoch": 428.0,
|
| 166968 |
+
"eval_loss": 0.46915048360824585,
|
| 166969 |
+
"eval_runtime": 39.008,
|
| 166970 |
+
"eval_samples_per_second": 21.534,
|
| 166971 |
+
"eval_steps_per_second": 0.692,
|
| 166972 |
+
"eval_wer": 0.20150507781106158,
|
| 166973 |
+
"step": 53137
|
| 166974 |
+
},
|
| 166975 |
+
{
|
| 166976 |
+
"epoch": 428.02,
|
| 166977 |
+
"learning_rate": 9.164727564102565e-06,
|
| 166978 |
+
"loss": 0.4169,
|
| 166979 |
+
"step": 53140
|
| 166980 |
+
},
|
| 166981 |
+
{
|
| 166982 |
+
"epoch": 428.06,
|
| 166983 |
+
"learning_rate": 9.164647435897436e-06,
|
| 166984 |
+
"loss": 0.4001,
|
| 166985 |
+
"step": 53145
|
| 166986 |
+
},
|
| 166987 |
+
{
|
| 166988 |
+
"epoch": 428.1,
|
| 166989 |
+
"learning_rate": 9.164567307692308e-06,
|
| 166990 |
+
"loss": 0.3144,
|
| 166991 |
+
"step": 53150
|
| 166992 |
+
},
|
| 166993 |
+
{
|
| 166994 |
+
"epoch": 428.15,
|
| 166995 |
+
"learning_rate": 9.16448717948718e-06,
|
| 166996 |
+
"loss": 0.375,
|
| 166997 |
+
"step": 53155
|
| 166998 |
+
},
|
| 166999 |
+
{
|
| 167000 |
+
"epoch": 428.19,
|
| 167001 |
+
"learning_rate": 9.164407051282052e-06,
|
| 167002 |
+
"loss": 0.7092,
|
| 167003 |
+
"step": 53160
|
| 167004 |
+
},
|
| 167005 |
+
{
|
| 167006 |
+
"epoch": 428.23,
|
| 167007 |
+
"learning_rate": 9.164326923076923e-06,
|
| 167008 |
+
"loss": 0.8503,
|
| 167009 |
+
"step": 53165
|
| 167010 |
+
},
|
| 167011 |
+
{
|
| 167012 |
+
"epoch": 428.27,
|
| 167013 |
+
"learning_rate": 9.164246794871796e-06,
|
| 167014 |
+
"loss": 0.3192,
|
| 167015 |
+
"step": 53170
|
| 167016 |
+
},
|
| 167017 |
+
{
|
| 167018 |
+
"epoch": 428.31,
|
| 167019 |
+
"learning_rate": 9.164166666666668e-06,
|
| 167020 |
+
"loss": 0.3571,
|
| 167021 |
+
"step": 53175
|
| 167022 |
+
},
|
| 167023 |
+
{
|
| 167024 |
+
"epoch": 428.35,
|
| 167025 |
+
"learning_rate": 9.164086538461539e-06,
|
| 167026 |
+
"loss": 0.3936,
|
| 167027 |
+
"step": 53180
|
| 167028 |
+
},
|
| 167029 |
+
{
|
| 167030 |
+
"epoch": 428.39,
|
| 167031 |
+
"learning_rate": 9.16400641025641e-06,
|
| 167032 |
+
"loss": 0.8202,
|
| 167033 |
+
"step": 53185
|
| 167034 |
+
},
|
| 167035 |
+
{
|
| 167036 |
+
"epoch": 428.43,
|
| 167037 |
+
"learning_rate": 9.163926282051283e-06,
|
| 167038 |
+
"loss": 0.9116,
|
| 167039 |
+
"step": 53190
|
| 167040 |
+
},
|
| 167041 |
+
{
|
| 167042 |
+
"epoch": 428.47,
|
| 167043 |
+
"learning_rate": 9.163846153846155e-06,
|
| 167044 |
+
"loss": 0.301,
|
| 167045 |
+
"step": 53195
|
| 167046 |
+
},
|
| 167047 |
+
{
|
| 167048 |
+
"epoch": 428.51,
|
| 167049 |
+
"learning_rate": 9.163766025641026e-06,
|
| 167050 |
+
"loss": 0.3834,
|
| 167051 |
+
"step": 53200
|
| 167052 |
+
},
|
| 167053 |
+
{
|
| 167054 |
+
"epoch": 428.55,
|
| 167055 |
+
"learning_rate": 9.163685897435898e-06,
|
| 167056 |
+
"loss": 0.3736,
|
| 167057 |
+
"step": 53205
|
| 167058 |
+
},
|
| 167059 |
+
{
|
| 167060 |
+
"epoch": 428.59,
|
| 167061 |
+
"learning_rate": 9.16360576923077e-06,
|
| 167062 |
+
"loss": 0.8737,
|
| 167063 |
+
"step": 53210
|
| 167064 |
+
},
|
| 167065 |
+
{
|
| 167066 |
+
"epoch": 428.63,
|
| 167067 |
+
"learning_rate": 9.163525641025642e-06,
|
| 167068 |
+
"loss": 0.8672,
|
| 167069 |
+
"step": 53215
|
| 167070 |
+
},
|
| 167071 |
+
{
|
| 167072 |
+
"epoch": 428.67,
|
| 167073 |
+
"learning_rate": 9.163445512820513e-06,
|
| 167074 |
+
"loss": 0.3897,
|
| 167075 |
+
"step": 53220
|
| 167076 |
+
},
|
| 167077 |
+
{
|
| 167078 |
+
"epoch": 428.71,
|
| 167079 |
+
"learning_rate": 9.163365384615386e-06,
|
| 167080 |
+
"loss": 0.3164,
|
| 167081 |
+
"step": 53225
|
| 167082 |
+
},
|
| 167083 |
+
{
|
| 167084 |
+
"epoch": 428.75,
|
| 167085 |
+
"learning_rate": 9.163285256410258e-06,
|
| 167086 |
+
"loss": 0.4835,
|
| 167087 |
+
"step": 53230
|
| 167088 |
+
},
|
| 167089 |
+
{
|
| 167090 |
+
"epoch": 428.79,
|
| 167091 |
+
"learning_rate": 9.163205128205129e-06,
|
| 167092 |
+
"loss": 0.8753,
|
| 167093 |
+
"step": 53235
|
| 167094 |
+
},
|
| 167095 |
+
{
|
| 167096 |
+
"epoch": 428.83,
|
| 167097 |
+
"learning_rate": 9.163125e-06,
|
| 167098 |
+
"loss": 0.9728,
|
| 167099 |
+
"step": 53240
|
| 167100 |
+
},
|
| 167101 |
+
{
|
| 167102 |
+
"epoch": 428.87,
|
| 167103 |
+
"learning_rate": 9.163044871794874e-06,
|
| 167104 |
+
"loss": 0.3156,
|
| 167105 |
+
"step": 53245
|
| 167106 |
+
},
|
| 167107 |
+
{
|
| 167108 |
+
"epoch": 428.91,
|
| 167109 |
+
"learning_rate": 9.162964743589743e-06,
|
| 167110 |
+
"loss": 0.3698,
|
| 167111 |
+
"step": 53250
|
| 167112 |
+
},
|
| 167113 |
+
{
|
| 167114 |
+
"epoch": 428.95,
|
| 167115 |
+
"learning_rate": 9.162884615384616e-06,
|
| 167116 |
+
"loss": 0.379,
|
| 167117 |
+
"step": 53255
|
| 167118 |
+
},
|
| 167119 |
+
{
|
| 167120 |
+
"epoch": 428.99,
|
| 167121 |
+
"learning_rate": 9.162804487179488e-06,
|
| 167122 |
+
"loss": 1.0544,
|
| 167123 |
+
"step": 53260
|
| 167124 |
+
},
|
| 167125 |
+
{
|
| 167126 |
+
"epoch": 429.0,
|
| 167127 |
+
"eval_loss": 0.4316280484199524,
|
| 167128 |
+
"eval_runtime": 39.1434,
|
| 167129 |
+
"eval_samples_per_second": 21.46,
|
| 167130 |
+
"eval_steps_per_second": 0.69,
|
| 167131 |
+
"eval_wer": 0.19746457867263237,
|
| 167132 |
+
"step": 53261
|
| 167133 |
}
|
| 167134 |
],
|
| 167135 |
"max_steps": 620000,
|
| 167136 |
"num_train_epochs": 5000,
|
| 167137 |
+
"total_flos": 1.498716444928101e+20,
|
| 167138 |
"trial_name": null,
|
| 167139 |
"trial_params": null
|
| 167140 |
}
|
model-bin/finetune/base/{checkpoint-52639 → checkpoint-53261}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629755733.4601321/events.out.tfevents.1629755733.74272264b15c.932.233
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:31a2cc8832c32300c14e1c7695f363cd13e957299554e670eb49f14023ac079b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629756373.9871938/events.out.tfevents.1629756373.74272264b15c.932.235
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:814894ef5c62f2fb00d650766a42586232ef0d382039b2b647b59a28a4c0850e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629757025.4645097/events.out.tfevents.1629757025.74272264b15c.932.237
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8cfb58a0693ae9cf653e41c51f06492a94b461b18a9bd841579a031ce25084a6
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629757671.1643252/events.out.tfevents.1629757671.74272264b15c.932.239
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ac23d0a40134686a9af9b0bbaac0bdf8a1fb2581809a68792161951ef8a6647
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629758322.377062/events.out.tfevents.1629758322.74272264b15c.932.241
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d39fbd365d92c4ab3a4dcac679912124bcce98231b062dfb0ff607e86fbe9070
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629755733.74272264b15c.932.232
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a3f9696917d24b62b30dbb895a58e5230bdda96f526dfefdfcbb66248e3c57fe
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629756373.74272264b15c.932.234
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:40220232a15a53218d39df1591397d40f50e7252924f0aa2653f9dc36ee2f299
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629757025.74272264b15c.932.236
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:367cf700b25526169e911fd3f7e9cf143902d8dfba1fb6e5f7d4c386664246bc
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629757671.74272264b15c.932.238
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c0a22573378421f2eeb2d4e991e84078f549c4e34d0d3468267273dab2044fa8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629758322.74272264b15c.932.240
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e87abe2dd0fbb1d817cf267d0a288d0c6da529cf7f182ad79d5aa7e1fbd454ca
|
| 3 |
+
size 8622
|