"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630152750.1734679/events.out.tfevents.1630152750.86bb0ddabf9b.4092.151 +3 -0
- model-bin/finetune/base/log/1630153143.2256212/events.out.tfevents.1630153143.86bb0ddabf9b.4092.153 +3 -0
- model-bin/finetune/base/log/1630153539.1576612/events.out.tfevents.1630153539.86bb0ddabf9b.4092.155 +3 -0
- model-bin/finetune/base/log/1630153932.2264154/events.out.tfevents.1630153932.86bb0ddabf9b.4092.157 +3 -0
- model-bin/finetune/base/log/1630154317.4212189/events.out.tfevents.1630154317.86bb0ddabf9b.4092.159 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630152750.86bb0ddabf9b.4092.150 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630153143.86bb0ddabf9b.4092.152 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630153539.86bb0ddabf9b.4092.154 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630153932.86bb0ddabf9b.4092.156 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630154317.86bb0ddabf9b.4092.158 +3 -0
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/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:e47e89a456f1557cb23e7cdc6f6a49b417c12bf1abbb676d2c653cec56ca19fe
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/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 sha256:012d0418cc3a9ae017ad990c5f383efc9aa00cd5828a8a9f6f9f7141abeb68d7
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/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 sha256:1172dd617e4cb3e70c4750f080ab55c6b7b5dc103d88f414dedc44c5b00e0e89
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/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:eaac8f43430a63d7b619416e88dce86fbfd5022f461e47b1923b1513a59ec850
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/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 sha256:0dffb0e6ba1ebaa12e8213c9a7b5fb48216cc731092e2f98f07382b1b835d976
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -261345,11 +261345,800 @@
|
|
| 261345 |
"eval_steps_per_second": 0.758,
|
| 261346 |
"eval_wer": 0.18508140855736463,
|
| 261347 |
"step": 127312
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 261348 |
}
|
| 261349 |
],
|
| 261350 |
-
"max_steps":
|
| 261351 |
"num_train_epochs": 5000,
|
| 261352 |
-
"total_flos": 3.
|
| 261353 |
"trial_name": null,
|
| 261354 |
"trial_params": null
|
| 261355 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17162025681719809,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-124947",
|
| 4 |
+
"epoch": 1031.0,
|
| 5 |
+
"global_step": 127934,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 261345 |
"eval_steps_per_second": 0.758,
|
| 261346 |
"eval_wer": 0.18508140855736463,
|
| 261347 |
"step": 127312
|
| 261348 |
+
},
|
| 261349 |
+
{
|
| 261350 |
+
"epoch": 1018.02,
|
| 261351 |
+
"learning_rate": 7.976522435897436e-06,
|
| 261352 |
+
"loss": 0.3539,
|
| 261353 |
+
"step": 127315
|
| 261354 |
+
},
|
| 261355 |
+
{
|
| 261356 |
+
"epoch": 1018.06,
|
| 261357 |
+
"learning_rate": 7.976442307692308e-06,
|
| 261358 |
+
"loss": 0.2478,
|
| 261359 |
+
"step": 127320
|
| 261360 |
+
},
|
| 261361 |
+
{
|
| 261362 |
+
"epoch": 1018.1,
|
| 261363 |
+
"learning_rate": 7.976362179487181e-06,
|
| 261364 |
+
"loss": 0.3601,
|
| 261365 |
+
"step": 127325
|
| 261366 |
+
},
|
| 261367 |
+
{
|
| 261368 |
+
"epoch": 1018.14,
|
| 261369 |
+
"learning_rate": 7.976282051282052e-06,
|
| 261370 |
+
"loss": 0.351,
|
| 261371 |
+
"step": 127330
|
| 261372 |
+
},
|
| 261373 |
+
{
|
| 261374 |
+
"epoch": 1018.18,
|
| 261375 |
+
"learning_rate": 7.976201923076924e-06,
|
| 261376 |
+
"loss": 0.7958,
|
| 261377 |
+
"step": 127335
|
| 261378 |
+
},
|
| 261379 |
+
{
|
| 261380 |
+
"epoch": 1018.22,
|
| 261381 |
+
"learning_rate": 7.976121794871795e-06,
|
| 261382 |
+
"loss": 0.8566,
|
| 261383 |
+
"step": 127340
|
| 261384 |
+
},
|
| 261385 |
+
{
|
| 261386 |
+
"epoch": 1018.26,
|
| 261387 |
+
"learning_rate": 7.976041666666668e-06,
|
| 261388 |
+
"loss": 0.2874,
|
| 261389 |
+
"step": 127345
|
| 261390 |
+
},
|
| 261391 |
+
{
|
| 261392 |
+
"epoch": 1018.3,
|
| 261393 |
+
"learning_rate": 7.97596153846154e-06,
|
| 261394 |
+
"loss": 0.2624,
|
| 261395 |
+
"step": 127350
|
| 261396 |
+
},
|
| 261397 |
+
{
|
| 261398 |
+
"epoch": 1018.34,
|
| 261399 |
+
"learning_rate": 7.97588141025641e-06,
|
| 261400 |
+
"loss": 0.396,
|
| 261401 |
+
"step": 127355
|
| 261402 |
+
},
|
| 261403 |
+
{
|
| 261404 |
+
"epoch": 1018.38,
|
| 261405 |
+
"learning_rate": 7.975801282051284e-06,
|
| 261406 |
+
"loss": 0.6654,
|
| 261407 |
+
"step": 127360
|
| 261408 |
+
},
|
| 261409 |
+
{
|
| 261410 |
+
"epoch": 1018.42,
|
| 261411 |
+
"learning_rate": 7.975721153846155e-06,
|
| 261412 |
+
"loss": 0.9536,
|
| 261413 |
+
"step": 127365
|
| 261414 |
+
},
|
| 261415 |
+
{
|
| 261416 |
+
"epoch": 1018.46,
|
| 261417 |
+
"learning_rate": 7.975641025641026e-06,
|
| 261418 |
+
"loss": 0.28,
|
| 261419 |
+
"step": 127370
|
| 261420 |
+
},
|
| 261421 |
+
{
|
| 261422 |
+
"epoch": 1018.5,
|
| 261423 |
+
"learning_rate": 7.975560897435898e-06,
|
| 261424 |
+
"loss": 0.3065,
|
| 261425 |
+
"step": 127375
|
| 261426 |
+
},
|
| 261427 |
+
{
|
| 261428 |
+
"epoch": 1018.54,
|
| 261429 |
+
"learning_rate": 7.975480769230771e-06,
|
| 261430 |
+
"loss": 0.3805,
|
| 261431 |
+
"step": 127380
|
| 261432 |
+
},
|
| 261433 |
+
{
|
| 261434 |
+
"epoch": 1018.58,
|
| 261435 |
+
"learning_rate": 7.97540064102564e-06,
|
| 261436 |
+
"loss": 0.7231,
|
| 261437 |
+
"step": 127385
|
| 261438 |
+
},
|
| 261439 |
+
{
|
| 261440 |
+
"epoch": 1018.62,
|
| 261441 |
+
"learning_rate": 7.975320512820514e-06,
|
| 261442 |
+
"loss": 0.9064,
|
| 261443 |
+
"step": 127390
|
| 261444 |
+
},
|
| 261445 |
+
{
|
| 261446 |
+
"epoch": 1018.66,
|
| 261447 |
+
"learning_rate": 7.975240384615385e-06,
|
| 261448 |
+
"loss": 0.2737,
|
| 261449 |
+
"step": 127395
|
| 261450 |
+
},
|
| 261451 |
+
{
|
| 261452 |
+
"epoch": 1018.7,
|
| 261453 |
+
"learning_rate": 7.975160256410256e-06,
|
| 261454 |
+
"loss": 0.3058,
|
| 261455 |
+
"step": 127400
|
| 261456 |
+
},
|
| 261457 |
+
{
|
| 261458 |
+
"epoch": 1018.74,
|
| 261459 |
+
"learning_rate": 7.97508012820513e-06,
|
| 261460 |
+
"loss": 0.3856,
|
| 261461 |
+
"step": 127405
|
| 261462 |
+
},
|
| 261463 |
+
{
|
| 261464 |
+
"epoch": 1018.78,
|
| 261465 |
+
"learning_rate": 7.975e-06,
|
| 261466 |
+
"loss": 0.6753,
|
| 261467 |
+
"step": 127410
|
| 261468 |
+
},
|
| 261469 |
+
{
|
| 261470 |
+
"epoch": 1018.82,
|
| 261471 |
+
"learning_rate": 7.974919871794872e-06,
|
| 261472 |
+
"loss": 0.8673,
|
| 261473 |
+
"step": 127415
|
| 261474 |
+
},
|
| 261475 |
+
{
|
| 261476 |
+
"epoch": 1018.86,
|
| 261477 |
+
"learning_rate": 7.974839743589743e-06,
|
| 261478 |
+
"loss": 0.2333,
|
| 261479 |
+
"step": 127420
|
| 261480 |
+
},
|
| 261481 |
+
{
|
| 261482 |
+
"epoch": 1018.9,
|
| 261483 |
+
"learning_rate": 7.974759615384616e-06,
|
| 261484 |
+
"loss": 0.3086,
|
| 261485 |
+
"step": 127425
|
| 261486 |
+
},
|
| 261487 |
+
{
|
| 261488 |
+
"epoch": 1018.94,
|
| 261489 |
+
"learning_rate": 7.974679487179488e-06,
|
| 261490 |
+
"loss": 0.4422,
|
| 261491 |
+
"step": 127430
|
| 261492 |
+
},
|
| 261493 |
+
{
|
| 261494 |
+
"epoch": 1018.98,
|
| 261495 |
+
"learning_rate": 7.97459935897436e-06,
|
| 261496 |
+
"loss": 0.7616,
|
| 261497 |
+
"step": 127435
|
| 261498 |
+
},
|
| 261499 |
+
{
|
| 261500 |
+
"epoch": 1019.0,
|
| 261501 |
+
"eval_loss": 0.40213724970817566,
|
| 261502 |
+
"eval_runtime": 37.0338,
|
| 261503 |
+
"eval_samples_per_second": 22.709,
|
| 261504 |
+
"eval_steps_per_second": 0.729,
|
| 261505 |
+
"eval_wer": 0.18384380897082306,
|
| 261506 |
+
"step": 127437
|
| 261507 |
+
},
|
| 261508 |
+
{
|
| 261509 |
+
"epoch": 1019.02,
|
| 261510 |
+
"learning_rate": 7.97451923076923e-06,
|
| 261511 |
+
"loss": 0.3438,
|
| 261512 |
+
"step": 127440
|
| 261513 |
+
},
|
| 261514 |
+
{
|
| 261515 |
+
"epoch": 1019.06,
|
| 261516 |
+
"learning_rate": 7.974439102564104e-06,
|
| 261517 |
+
"loss": 0.2793,
|
| 261518 |
+
"step": 127445
|
| 261519 |
+
},
|
| 261520 |
+
{
|
| 261521 |
+
"epoch": 1019.1,
|
| 261522 |
+
"learning_rate": 7.974358974358975e-06,
|
| 261523 |
+
"loss": 0.2579,
|
| 261524 |
+
"step": 127450
|
| 261525 |
+
},
|
| 261526 |
+
{
|
| 261527 |
+
"epoch": 1019.14,
|
| 261528 |
+
"learning_rate": 7.974278846153846e-06,
|
| 261529 |
+
"loss": 0.3732,
|
| 261530 |
+
"step": 127455
|
| 261531 |
+
},
|
| 261532 |
+
{
|
| 261533 |
+
"epoch": 1019.18,
|
| 261534 |
+
"learning_rate": 7.97419871794872e-06,
|
| 261535 |
+
"loss": 0.7385,
|
| 261536 |
+
"step": 127460
|
| 261537 |
+
},
|
| 261538 |
+
{
|
| 261539 |
+
"epoch": 1019.22,
|
| 261540 |
+
"learning_rate": 7.97411858974359e-06,
|
| 261541 |
+
"loss": 0.6752,
|
| 261542 |
+
"step": 127465
|
| 261543 |
+
},
|
| 261544 |
+
{
|
| 261545 |
+
"epoch": 1019.26,
|
| 261546 |
+
"learning_rate": 7.974038461538462e-06,
|
| 261547 |
+
"loss": 0.2907,
|
| 261548 |
+
"step": 127470
|
| 261549 |
+
},
|
| 261550 |
+
{
|
| 261551 |
+
"epoch": 1019.3,
|
| 261552 |
+
"learning_rate": 7.973958333333333e-06,
|
| 261553 |
+
"loss": 0.2569,
|
| 261554 |
+
"step": 127475
|
| 261555 |
+
},
|
| 261556 |
+
{
|
| 261557 |
+
"epoch": 1019.34,
|
| 261558 |
+
"learning_rate": 7.973878205128207e-06,
|
| 261559 |
+
"loss": 0.327,
|
| 261560 |
+
"step": 127480
|
| 261561 |
+
},
|
| 261562 |
+
{
|
| 261563 |
+
"epoch": 1019.38,
|
| 261564 |
+
"learning_rate": 7.973798076923078e-06,
|
| 261565 |
+
"loss": 0.6373,
|
| 261566 |
+
"step": 127485
|
| 261567 |
+
},
|
| 261568 |
+
{
|
| 261569 |
+
"epoch": 1019.42,
|
| 261570 |
+
"learning_rate": 7.97371794871795e-06,
|
| 261571 |
+
"loss": 0.8847,
|
| 261572 |
+
"step": 127490
|
| 261573 |
+
},
|
| 261574 |
+
{
|
| 261575 |
+
"epoch": 1019.46,
|
| 261576 |
+
"learning_rate": 7.97363782051282e-06,
|
| 261577 |
+
"loss": 0.2746,
|
| 261578 |
+
"step": 127495
|
| 261579 |
+
},
|
| 261580 |
+
{
|
| 261581 |
+
"epoch": 1019.5,
|
| 261582 |
+
"learning_rate": 7.973557692307694e-06,
|
| 261583 |
+
"loss": 0.2562,
|
| 261584 |
+
"step": 127500
|
| 261585 |
+
},
|
| 261586 |
+
{
|
| 261587 |
+
"epoch": 1019.54,
|
| 261588 |
+
"learning_rate": 7.973477564102565e-06,
|
| 261589 |
+
"loss": 0.3359,
|
| 261590 |
+
"step": 127505
|
| 261591 |
+
},
|
| 261592 |
+
{
|
| 261593 |
+
"epoch": 1019.58,
|
| 261594 |
+
"learning_rate": 7.973397435897436e-06,
|
| 261595 |
+
"loss": 0.7607,
|
| 261596 |
+
"step": 127510
|
| 261597 |
+
},
|
| 261598 |
+
{
|
| 261599 |
+
"epoch": 1019.62,
|
| 261600 |
+
"learning_rate": 7.97331730769231e-06,
|
| 261601 |
+
"loss": 0.9148,
|
| 261602 |
+
"step": 127515
|
| 261603 |
+
},
|
| 261604 |
+
{
|
| 261605 |
+
"epoch": 1019.66,
|
| 261606 |
+
"learning_rate": 7.973237179487179e-06,
|
| 261607 |
+
"loss": 0.2713,
|
| 261608 |
+
"step": 127520
|
| 261609 |
+
},
|
| 261610 |
+
{
|
| 261611 |
+
"epoch": 1019.7,
|
| 261612 |
+
"learning_rate": 7.973157051282052e-06,
|
| 261613 |
+
"loss": 0.2711,
|
| 261614 |
+
"step": 127525
|
| 261615 |
+
},
|
| 261616 |
+
{
|
| 261617 |
+
"epoch": 1019.74,
|
| 261618 |
+
"learning_rate": 7.973076923076923e-06,
|
| 261619 |
+
"loss": 0.3825,
|
| 261620 |
+
"step": 127530
|
| 261621 |
+
},
|
| 261622 |
+
{
|
| 261623 |
+
"epoch": 1019.78,
|
| 261624 |
+
"learning_rate": 7.972996794871795e-06,
|
| 261625 |
+
"loss": 0.7835,
|
| 261626 |
+
"step": 127535
|
| 261627 |
+
},
|
| 261628 |
+
{
|
| 261629 |
+
"epoch": 1019.82,
|
| 261630 |
+
"learning_rate": 7.972916666666666e-06,
|
| 261631 |
+
"loss": 0.7767,
|
| 261632 |
+
"step": 127540
|
| 261633 |
+
},
|
| 261634 |
+
{
|
| 261635 |
+
"epoch": 1019.86,
|
| 261636 |
+
"learning_rate": 7.97283653846154e-06,
|
| 261637 |
+
"loss": 0.3043,
|
| 261638 |
+
"step": 127545
|
| 261639 |
+
},
|
| 261640 |
+
{
|
| 261641 |
+
"epoch": 1019.9,
|
| 261642 |
+
"learning_rate": 7.97275641025641e-06,
|
| 261643 |
+
"loss": 0.306,
|
| 261644 |
+
"step": 127550
|
| 261645 |
+
},
|
| 261646 |
+
{
|
| 261647 |
+
"epoch": 1019.94,
|
| 261648 |
+
"learning_rate": 7.972676282051282e-06,
|
| 261649 |
+
"loss": 0.4201,
|
| 261650 |
+
"step": 127555
|
| 261651 |
+
},
|
| 261652 |
+
{
|
| 261653 |
+
"epoch": 1019.98,
|
| 261654 |
+
"learning_rate": 7.972596153846155e-06,
|
| 261655 |
+
"loss": 0.7576,
|
| 261656 |
+
"step": 127560
|
| 261657 |
+
},
|
| 261658 |
+
{
|
| 261659 |
+
"epoch": 1020.0,
|
| 261660 |
+
"eval_loss": 0.3979755938053131,
|
| 261661 |
+
"eval_runtime": 36.0151,
|
| 261662 |
+
"eval_samples_per_second": 23.351,
|
| 261663 |
+
"eval_steps_per_second": 0.75,
|
| 261664 |
+
"eval_wer": 0.177632058287796,
|
| 261665 |
+
"step": 127562
|
| 261666 |
+
},
|
| 261667 |
+
{
|
| 261668 |
+
"epoch": 1028.02,
|
| 261669 |
+
"learning_rate": 7.972516025641026e-06,
|
| 261670 |
+
"loss": 0.3316,
|
| 261671 |
+
"step": 127565
|
| 261672 |
+
},
|
| 261673 |
+
{
|
| 261674 |
+
"epoch": 1028.06,
|
| 261675 |
+
"learning_rate": 7.972435897435898e-06,
|
| 261676 |
+
"loss": 0.2653,
|
| 261677 |
+
"step": 127570
|
| 261678 |
+
},
|
| 261679 |
+
{
|
| 261680 |
+
"epoch": 1028.1,
|
| 261681 |
+
"learning_rate": 7.972355769230769e-06,
|
| 261682 |
+
"loss": 0.2628,
|
| 261683 |
+
"step": 127575
|
| 261684 |
+
},
|
| 261685 |
+
{
|
| 261686 |
+
"epoch": 1028.14,
|
| 261687 |
+
"learning_rate": 7.972275641025642e-06,
|
| 261688 |
+
"loss": 0.3052,
|
| 261689 |
+
"step": 127580
|
| 261690 |
+
},
|
| 261691 |
+
{
|
| 261692 |
+
"epoch": 1028.18,
|
| 261693 |
+
"learning_rate": 7.972195512820514e-06,
|
| 261694 |
+
"loss": 0.6711,
|
| 261695 |
+
"step": 127585
|
| 261696 |
+
},
|
| 261697 |
+
{
|
| 261698 |
+
"epoch": 1028.22,
|
| 261699 |
+
"learning_rate": 7.972115384615385e-06,
|
| 261700 |
+
"loss": 0.8237,
|
| 261701 |
+
"step": 127590
|
| 261702 |
+
},
|
| 261703 |
+
{
|
| 261704 |
+
"epoch": 1028.27,
|
| 261705 |
+
"learning_rate": 7.972035256410256e-06,
|
| 261706 |
+
"loss": 0.2567,
|
| 261707 |
+
"step": 127595
|
| 261708 |
+
},
|
| 261709 |
+
{
|
| 261710 |
+
"epoch": 1028.31,
|
| 261711 |
+
"learning_rate": 7.97195512820513e-06,
|
| 261712 |
+
"loss": 0.3521,
|
| 261713 |
+
"step": 127600
|
| 261714 |
+
},
|
| 261715 |
+
{
|
| 261716 |
+
"epoch": 1028.35,
|
| 261717 |
+
"learning_rate": 7.971875e-06,
|
| 261718 |
+
"loss": 0.3396,
|
| 261719 |
+
"step": 127605
|
| 261720 |
+
},
|
| 261721 |
+
{
|
| 261722 |
+
"epoch": 1028.39,
|
| 261723 |
+
"learning_rate": 7.971794871794872e-06,
|
| 261724 |
+
"loss": 0.64,
|
| 261725 |
+
"step": 127610
|
| 261726 |
+
},
|
| 261727 |
+
{
|
| 261728 |
+
"epoch": 1028.43,
|
| 261729 |
+
"learning_rate": 7.971714743589745e-06,
|
| 261730 |
+
"loss": 0.7791,
|
| 261731 |
+
"step": 127615
|
| 261732 |
+
},
|
| 261733 |
+
{
|
| 261734 |
+
"epoch": 1028.47,
|
| 261735 |
+
"learning_rate": 7.971634615384616e-06,
|
| 261736 |
+
"loss": 0.3231,
|
| 261737 |
+
"step": 127620
|
| 261738 |
+
},
|
| 261739 |
+
{
|
| 261740 |
+
"epoch": 1028.51,
|
| 261741 |
+
"learning_rate": 7.971554487179488e-06,
|
| 261742 |
+
"loss": 0.2828,
|
| 261743 |
+
"step": 127625
|
| 261744 |
+
},
|
| 261745 |
+
{
|
| 261746 |
+
"epoch": 1028.55,
|
| 261747 |
+
"learning_rate": 7.971474358974359e-06,
|
| 261748 |
+
"loss": 0.339,
|
| 261749 |
+
"step": 127630
|
| 261750 |
+
},
|
| 261751 |
+
{
|
| 261752 |
+
"epoch": 1028.59,
|
| 261753 |
+
"learning_rate": 7.971394230769232e-06,
|
| 261754 |
+
"loss": 0.703,
|
| 261755 |
+
"step": 127635
|
| 261756 |
+
},
|
| 261757 |
+
{
|
| 261758 |
+
"epoch": 1028.63,
|
| 261759 |
+
"learning_rate": 7.971314102564104e-06,
|
| 261760 |
+
"loss": 0.8128,
|
| 261761 |
+
"step": 127640
|
| 261762 |
+
},
|
| 261763 |
+
{
|
| 261764 |
+
"epoch": 1028.67,
|
| 261765 |
+
"learning_rate": 7.971233974358975e-06,
|
| 261766 |
+
"loss": 0.257,
|
| 261767 |
+
"step": 127645
|
| 261768 |
+
},
|
| 261769 |
+
{
|
| 261770 |
+
"epoch": 1028.71,
|
| 261771 |
+
"learning_rate": 7.971153846153848e-06,
|
| 261772 |
+
"loss": 0.2901,
|
| 261773 |
+
"step": 127650
|
| 261774 |
+
},
|
| 261775 |
+
{
|
| 261776 |
+
"epoch": 1028.75,
|
| 261777 |
+
"learning_rate": 7.97107371794872e-06,
|
| 261778 |
+
"loss": 0.3715,
|
| 261779 |
+
"step": 127655
|
| 261780 |
+
},
|
| 261781 |
+
{
|
| 261782 |
+
"epoch": 1028.79,
|
| 261783 |
+
"learning_rate": 7.97099358974359e-06,
|
| 261784 |
+
"loss": 0.7095,
|
| 261785 |
+
"step": 127660
|
| 261786 |
+
},
|
| 261787 |
+
{
|
| 261788 |
+
"epoch": 1028.83,
|
| 261789 |
+
"learning_rate": 7.970913461538462e-06,
|
| 261790 |
+
"loss": 0.7523,
|
| 261791 |
+
"step": 127665
|
| 261792 |
+
},
|
| 261793 |
+
{
|
| 261794 |
+
"epoch": 1028.87,
|
| 261795 |
+
"learning_rate": 7.970833333333335e-06,
|
| 261796 |
+
"loss": 0.2962,
|
| 261797 |
+
"step": 127670
|
| 261798 |
+
},
|
| 261799 |
+
{
|
| 261800 |
+
"epoch": 1028.91,
|
| 261801 |
+
"learning_rate": 7.970753205128205e-06,
|
| 261802 |
+
"loss": 0.3222,
|
| 261803 |
+
"step": 127675
|
| 261804 |
+
},
|
| 261805 |
+
{
|
| 261806 |
+
"epoch": 1028.95,
|
| 261807 |
+
"learning_rate": 7.970673076923078e-06,
|
| 261808 |
+
"loss": 0.346,
|
| 261809 |
+
"step": 127680
|
| 261810 |
+
},
|
| 261811 |
+
{
|
| 261812 |
+
"epoch": 1028.99,
|
| 261813 |
+
"learning_rate": 7.97059294871795e-06,
|
| 261814 |
+
"loss": 0.9398,
|
| 261815 |
+
"step": 127685
|
| 261816 |
+
},
|
| 261817 |
+
{
|
| 261818 |
+
"epoch": 1029.0,
|
| 261819 |
+
"eval_loss": 0.4858114719390869,
|
| 261820 |
+
"eval_runtime": 36.7128,
|
| 261821 |
+
"eval_samples_per_second": 22.908,
|
| 261822 |
+
"eval_steps_per_second": 0.735,
|
| 261823 |
+
"eval_wer": 0.17205701468779394,
|
| 261824 |
+
"step": 127686
|
| 261825 |
+
},
|
| 261826 |
+
{
|
| 261827 |
+
"epoch": 1029.03,
|
| 261828 |
+
"learning_rate": 7.97051282051282e-06,
|
| 261829 |
+
"loss": 0.3032,
|
| 261830 |
+
"step": 127690
|
| 261831 |
+
},
|
| 261832 |
+
{
|
| 261833 |
+
"epoch": 1029.07,
|
| 261834 |
+
"learning_rate": 7.970432692307692e-06,
|
| 261835 |
+
"loss": 0.2845,
|
| 261836 |
+
"step": 127695
|
| 261837 |
+
},
|
| 261838 |
+
{
|
| 261839 |
+
"epoch": 1029.11,
|
| 261840 |
+
"learning_rate": 7.970352564102565e-06,
|
| 261841 |
+
"loss": 0.2991,
|
| 261842 |
+
"step": 127700
|
| 261843 |
+
},
|
| 261844 |
+
{
|
| 261845 |
+
"epoch": 1029.15,
|
| 261846 |
+
"learning_rate": 7.970272435897436e-06,
|
| 261847 |
+
"loss": 0.4367,
|
| 261848 |
+
"step": 127705
|
| 261849 |
+
},
|
| 261850 |
+
{
|
| 261851 |
+
"epoch": 1029.19,
|
| 261852 |
+
"learning_rate": 7.970192307692308e-06,
|
| 261853 |
+
"loss": 0.9566,
|
| 261854 |
+
"step": 127710
|
| 261855 |
+
},
|
| 261856 |
+
{
|
| 261857 |
+
"epoch": 1029.23,
|
| 261858 |
+
"learning_rate": 7.97011217948718e-06,
|
| 261859 |
+
"loss": 0.621,
|
| 261860 |
+
"step": 127715
|
| 261861 |
+
},
|
| 261862 |
+
{
|
| 261863 |
+
"epoch": 1029.27,
|
| 261864 |
+
"learning_rate": 7.970032051282052e-06,
|
| 261865 |
+
"loss": 0.2804,
|
| 261866 |
+
"step": 127720
|
| 261867 |
+
},
|
| 261868 |
+
{
|
| 261869 |
+
"epoch": 1029.31,
|
| 261870 |
+
"learning_rate": 7.969951923076923e-06,
|
| 261871 |
+
"loss": 0.2982,
|
| 261872 |
+
"step": 127725
|
| 261873 |
+
},
|
| 261874 |
+
{
|
| 261875 |
+
"epoch": 1029.35,
|
| 261876 |
+
"learning_rate": 7.969871794871795e-06,
|
| 261877 |
+
"loss": 0.3794,
|
| 261878 |
+
"step": 127730
|
| 261879 |
+
},
|
| 261880 |
+
{
|
| 261881 |
+
"epoch": 1029.4,
|
| 261882 |
+
"learning_rate": 7.969791666666668e-06,
|
| 261883 |
+
"loss": 0.8919,
|
| 261884 |
+
"step": 127735
|
| 261885 |
+
},
|
| 261886 |
+
{
|
| 261887 |
+
"epoch": 1029.44,
|
| 261888 |
+
"learning_rate": 7.96971153846154e-06,
|
| 261889 |
+
"loss": 0.6414,
|
| 261890 |
+
"step": 127740
|
| 261891 |
+
},
|
| 261892 |
+
{
|
| 261893 |
+
"epoch": 1029.48,
|
| 261894 |
+
"learning_rate": 7.96963141025641e-06,
|
| 261895 |
+
"loss": 0.3315,
|
| 261896 |
+
"step": 127745
|
| 261897 |
+
},
|
| 261898 |
+
{
|
| 261899 |
+
"epoch": 1029.52,
|
| 261900 |
+
"learning_rate": 7.969551282051284e-06,
|
| 261901 |
+
"loss": 0.3014,
|
| 261902 |
+
"step": 127750
|
| 261903 |
+
},
|
| 261904 |
+
{
|
| 261905 |
+
"epoch": 1029.56,
|
| 261906 |
+
"learning_rate": 7.969471153846155e-06,
|
| 261907 |
+
"loss": 0.385,
|
| 261908 |
+
"step": 127755
|
| 261909 |
+
},
|
| 261910 |
+
{
|
| 261911 |
+
"epoch": 1029.6,
|
| 261912 |
+
"learning_rate": 7.969391025641026e-06,
|
| 261913 |
+
"loss": 0.913,
|
| 261914 |
+
"step": 127760
|
| 261915 |
+
},
|
| 261916 |
+
{
|
| 261917 |
+
"epoch": 1029.64,
|
| 261918 |
+
"learning_rate": 7.969310897435898e-06,
|
| 261919 |
+
"loss": 0.6987,
|
| 261920 |
+
"step": 127765
|
| 261921 |
+
},
|
| 261922 |
+
{
|
| 261923 |
+
"epoch": 1029.68,
|
| 261924 |
+
"learning_rate": 7.96923076923077e-06,
|
| 261925 |
+
"loss": 0.2962,
|
| 261926 |
+
"step": 127770
|
| 261927 |
+
},
|
| 261928 |
+
{
|
| 261929 |
+
"epoch": 1029.72,
|
| 261930 |
+
"learning_rate": 7.969150641025642e-06,
|
| 261931 |
+
"loss": 0.3002,
|
| 261932 |
+
"step": 127775
|
| 261933 |
+
},
|
| 261934 |
+
{
|
| 261935 |
+
"epoch": 1029.76,
|
| 261936 |
+
"learning_rate": 7.969070512820513e-06,
|
| 261937 |
+
"loss": 0.3731,
|
| 261938 |
+
"step": 127780
|
| 261939 |
+
},
|
| 261940 |
+
{
|
| 261941 |
+
"epoch": 1029.8,
|
| 261942 |
+
"learning_rate": 7.968990384615385e-06,
|
| 261943 |
+
"loss": 0.8354,
|
| 261944 |
+
"step": 127785
|
| 261945 |
+
},
|
| 261946 |
+
{
|
| 261947 |
+
"epoch": 1029.84,
|
| 261948 |
+
"learning_rate": 7.968910256410258e-06,
|
| 261949 |
+
"loss": 0.6015,
|
| 261950 |
+
"step": 127790
|
| 261951 |
+
},
|
| 261952 |
+
{
|
| 261953 |
+
"epoch": 1029.88,
|
| 261954 |
+
"learning_rate": 7.968830128205128e-06,
|
| 261955 |
+
"loss": 0.2902,
|
| 261956 |
+
"step": 127795
|
| 261957 |
+
},
|
| 261958 |
+
{
|
| 261959 |
+
"epoch": 1029.92,
|
| 261960 |
+
"learning_rate": 7.96875e-06,
|
| 261961 |
+
"loss": 0.2872,
|
| 261962 |
+
"step": 127800
|
| 261963 |
+
},
|
| 261964 |
+
{
|
| 261965 |
+
"epoch": 1029.96,
|
| 261966 |
+
"learning_rate": 7.968669871794874e-06,
|
| 261967 |
+
"loss": 0.388,
|
| 261968 |
+
"step": 127805
|
| 261969 |
+
},
|
| 261970 |
+
{
|
| 261971 |
+
"epoch": 1030.0,
|
| 261972 |
+
"learning_rate": 7.968589743589743e-06,
|
| 261973 |
+
"loss": 1.0202,
|
| 261974 |
+
"step": 127810
|
| 261975 |
+
},
|
| 261976 |
+
{
|
| 261977 |
+
"epoch": 1030.0,
|
| 261978 |
+
"eval_loss": 0.3965398967266083,
|
| 261979 |
+
"eval_runtime": 35.7224,
|
| 261980 |
+
"eval_samples_per_second": 23.543,
|
| 261981 |
+
"eval_steps_per_second": 0.756,
|
| 261982 |
+
"eval_wer": 0.17647491540385465,
|
| 261983 |
+
"step": 127810
|
| 261984 |
+
},
|
| 261985 |
+
{
|
| 261986 |
+
"epoch": 1030.04,
|
| 261987 |
+
"learning_rate": 7.968509615384616e-06,
|
| 261988 |
+
"loss": 0.3129,
|
| 261989 |
+
"step": 127815
|
| 261990 |
+
},
|
| 261991 |
+
{
|
| 261992 |
+
"epoch": 1030.08,
|
| 261993 |
+
"learning_rate": 7.968429487179488e-06,
|
| 261994 |
+
"loss": 0.2665,
|
| 261995 |
+
"step": 127820
|
| 261996 |
+
},
|
| 261997 |
+
{
|
| 261998 |
+
"epoch": 1030.12,
|
| 261999 |
+
"learning_rate": 7.968349358974359e-06,
|
| 262000 |
+
"loss": 0.4385,
|
| 262001 |
+
"step": 127825
|
| 262002 |
+
},
|
| 262003 |
+
{
|
| 262004 |
+
"epoch": 1030.16,
|
| 262005 |
+
"learning_rate": 7.96826923076923e-06,
|
| 262006 |
+
"loss": 0.41,
|
| 262007 |
+
"step": 127830
|
| 262008 |
+
},
|
| 262009 |
+
{
|
| 262010 |
+
"epoch": 1030.2,
|
| 262011 |
+
"learning_rate": 7.968189102564103e-06,
|
| 262012 |
+
"loss": 1.1281,
|
| 262013 |
+
"step": 127835
|
| 262014 |
+
},
|
| 262015 |
+
{
|
| 262016 |
+
"epoch": 1030.24,
|
| 262017 |
+
"learning_rate": 7.968108974358975e-06,
|
| 262018 |
+
"loss": 0.3068,
|
| 262019 |
+
"step": 127840
|
| 262020 |
+
},
|
| 262021 |
+
{
|
| 262022 |
+
"epoch": 1030.28,
|
| 262023 |
+
"learning_rate": 7.968028846153846e-06,
|
| 262024 |
+
"loss": 0.3179,
|
| 262025 |
+
"step": 127845
|
| 262026 |
+
},
|
| 262027 |
+
{
|
| 262028 |
+
"epoch": 1030.32,
|
| 262029 |
+
"learning_rate": 7.967948717948718e-06,
|
| 262030 |
+
"loss": 0.2799,
|
| 262031 |
+
"step": 127850
|
| 262032 |
+
},
|
| 262033 |
+
{
|
| 262034 |
+
"epoch": 1030.36,
|
| 262035 |
+
"learning_rate": 7.96786858974359e-06,
|
| 262036 |
+
"loss": 0.4315,
|
| 262037 |
+
"step": 127855
|
| 262038 |
+
},
|
| 262039 |
+
{
|
| 262040 |
+
"epoch": 1030.4,
|
| 262041 |
+
"learning_rate": 7.967788461538462e-06,
|
| 262042 |
+
"loss": 1.1404,
|
| 262043 |
+
"step": 127860
|
| 262044 |
+
},
|
| 262045 |
+
{
|
| 262046 |
+
"epoch": 1030.44,
|
| 262047 |
+
"learning_rate": 7.967708333333333e-06,
|
| 262048 |
+
"loss": 0.2989,
|
| 262049 |
+
"step": 127865
|
| 262050 |
+
},
|
| 262051 |
+
{
|
| 262052 |
+
"epoch": 1030.48,
|
| 262053 |
+
"learning_rate": 7.967628205128206e-06,
|
| 262054 |
+
"loss": 0.2552,
|
| 262055 |
+
"step": 127870
|
| 262056 |
+
},
|
| 262057 |
+
{
|
| 262058 |
+
"epoch": 1030.52,
|
| 262059 |
+
"learning_rate": 7.967548076923078e-06,
|
| 262060 |
+
"loss": 0.3222,
|
| 262061 |
+
"step": 127875
|
| 262062 |
+
},
|
| 262063 |
+
{
|
| 262064 |
+
"epoch": 1030.56,
|
| 262065 |
+
"learning_rate": 7.967467948717949e-06,
|
| 262066 |
+
"loss": 0.5064,
|
| 262067 |
+
"step": 127880
|
| 262068 |
+
},
|
| 262069 |
+
{
|
| 262070 |
+
"epoch": 1030.6,
|
| 262071 |
+
"learning_rate": 7.96738782051282e-06,
|
| 262072 |
+
"loss": 1.1156,
|
| 262073 |
+
"step": 127885
|
| 262074 |
+
},
|
| 262075 |
+
{
|
| 262076 |
+
"epoch": 1030.65,
|
| 262077 |
+
"learning_rate": 7.967307692307693e-06,
|
| 262078 |
+
"loss": 0.2943,
|
| 262079 |
+
"step": 127890
|
| 262080 |
+
},
|
| 262081 |
+
{
|
| 262082 |
+
"epoch": 1030.69,
|
| 262083 |
+
"learning_rate": 7.967227564102565e-06,
|
| 262084 |
+
"loss": 0.2699,
|
| 262085 |
+
"step": 127895
|
| 262086 |
+
},
|
| 262087 |
+
{
|
| 262088 |
+
"epoch": 1030.73,
|
| 262089 |
+
"learning_rate": 7.967147435897436e-06,
|
| 262090 |
+
"loss": 0.3276,
|
| 262091 |
+
"step": 127900
|
| 262092 |
+
},
|
| 262093 |
+
{
|
| 262094 |
+
"epoch": 1030.77,
|
| 262095 |
+
"learning_rate": 7.96706730769231e-06,
|
| 262096 |
+
"loss": 0.4705,
|
| 262097 |
+
"step": 127905
|
| 262098 |
+
},
|
| 262099 |
+
{
|
| 262100 |
+
"epoch": 1030.81,
|
| 262101 |
+
"learning_rate": 7.96698717948718e-06,
|
| 262102 |
+
"loss": 1.1514,
|
| 262103 |
+
"step": 127910
|
| 262104 |
+
},
|
| 262105 |
+
{
|
| 262106 |
+
"epoch": 1030.85,
|
| 262107 |
+
"learning_rate": 7.966907051282052e-06,
|
| 262108 |
+
"loss": 0.2672,
|
| 262109 |
+
"step": 127915
|
| 262110 |
+
},
|
| 262111 |
+
{
|
| 262112 |
+
"epoch": 1030.89,
|
| 262113 |
+
"learning_rate": 7.966826923076923e-06,
|
| 262114 |
+
"loss": 0.3088,
|
| 262115 |
+
"step": 127920
|
| 262116 |
+
},
|
| 262117 |
+
{
|
| 262118 |
+
"epoch": 1030.93,
|
| 262119 |
+
"learning_rate": 7.966746794871796e-06,
|
| 262120 |
+
"loss": 0.3412,
|
| 262121 |
+
"step": 127925
|
| 262122 |
+
},
|
| 262123 |
+
{
|
| 262124 |
+
"epoch": 1030.97,
|
| 262125 |
+
"learning_rate": 7.966666666666668e-06,
|
| 262126 |
+
"loss": 0.5402,
|
| 262127 |
+
"step": 127930
|
| 262128 |
+
},
|
| 262129 |
+
{
|
| 262130 |
+
"epoch": 1031.0,
|
| 262131 |
+
"eval_loss": 0.3815125524997711,
|
| 262132 |
+
"eval_runtime": 36.0201,
|
| 262133 |
+
"eval_samples_per_second": 23.348,
|
| 262134 |
+
"eval_steps_per_second": 0.75,
|
| 262135 |
+
"eval_wer": 0.18386253268584235,
|
| 262136 |
+
"step": 127934
|
| 262137 |
}
|
| 262138 |
],
|
| 262139 |
+
"max_steps": 620000,
|
| 262140 |
"num_train_epochs": 5000,
|
| 262141 |
+
"total_flos": 3.6000782787802484e+20,
|
| 262142 |
"trial_name": null,
|
| 262143 |
"trial_params": null
|
| 262144 |
}
|
model-bin/finetune/base/{checkpoint-127312 → checkpoint-127934}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630152750.1734679/events.out.tfevents.1630152750.86bb0ddabf9b.4092.151
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2b3314ede57ee18ce9240fd07a6b1613aea4b93c947239f5bfcad4054f41d5c9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630153143.2256212/events.out.tfevents.1630153143.86bb0ddabf9b.4092.153
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:18b0184a1848eaded7362fd67979b230a607b9dbcb3ef9482807c07a487275ff
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630153539.1576612/events.out.tfevents.1630153539.86bb0ddabf9b.4092.155
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cbb602326968844e8be1dfde76870e17d09b87a544c35e319ef7cc086a0dc783
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630153932.2264154/events.out.tfevents.1630153932.86bb0ddabf9b.4092.157
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6924fa7804e1e303798fd77966bb9a3c11b74017ec4bea1b765feb3bb66a48b9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630154317.4212189/events.out.tfevents.1630154317.86bb0ddabf9b.4092.159
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6199cad6fc9ca43e8aefd1eb75a1fa43c5c921e94b3900e3893a02085bfd5175
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630152750.86bb0ddabf9b.4092.150
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f173afca7c4bffe81031a9a4e975a1ff1efb465a25c6d20b2877cd7b38e2e772
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630153143.86bb0ddabf9b.4092.152
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:062708b20e08eb55b62ea11a41f79d720fff86ab5baf07895d81a146064a8886
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630153539.86bb0ddabf9b.4092.154
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dac5c8937bec75e94b8a1d6f69890c3345b46117c39dea6363da976d94378653
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630153932.86bb0ddabf9b.4092.156
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b241733e3dc8b7bf6ea89fca933e4ccf7960cf42362df7c322cee3982597178
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630154317.86bb0ddabf9b.4092.158
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:221077f087d73f2486c182467a4bf736923ef36189ae222ad0c57ae17c5d2eaa
|
| 3 |
+
size 8462
|