"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630117908.6872168/events.out.tfevents.1630117908.86bb0ddabf9b.1042.51 +3 -0
- model-bin/finetune/base/log/1630118323.7473285/events.out.tfevents.1630118323.86bb0ddabf9b.1042.53 +3 -0
- model-bin/finetune/base/log/1630118841.0029836/events.out.tfevents.1630118841.86bb0ddabf9b.1042.55 +3 -0
- model-bin/finetune/base/log/1630119249.7913108/events.out.tfevents.1630119249.86bb0ddabf9b.1042.57 +3 -0
- model-bin/finetune/base/log/1630119657.5811403/events.out.tfevents.1630119657.86bb0ddabf9b.1042.59 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630117908.86bb0ddabf9b.1042.50 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630118323.86bb0ddabf9b.1042.52 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630118840.86bb0ddabf9b.1042.54 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630119249.86bb0ddabf9b.1042.56 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630119657.86bb0ddabf9b.1042.58 +3 -0
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/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 sha256:2f8f9d9e29f9c4357eccefce34b1788641cbf79fdb30f31166c13d2a3991cec7
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/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 sha256:8d6fd2acf7aa2ca9a155077ce484b63264f0447cd4f800916e207dcb087cf3dc
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/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 sha256:3883b24ab055ced9e5920bf2f787c0bf5fe045f5bc46d3e0a656d2fb11a7e73e
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/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 sha256:85667687b163c74870e550a93ae8876dd8fc5752af02aeaab176e91efede68c0
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/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:ee5116ffc5a6492b87e5b79cfa6a431988c523d33002a0376caef480f1a04f73
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/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,
|
|
@@ -248355,11 +248355,800 @@
|
|
| 248355 |
"eval_steps_per_second": 0.679,
|
| 248356 |
"eval_wer": 0.1851399856424982,
|
| 248357 |
"step": 117100
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 248358 |
}
|
| 248359 |
],
|
| 248360 |
"max_steps": 625000,
|
| 248361 |
"num_train_epochs": 5000,
|
| 248362 |
-
"total_flos": 3.
|
| 248363 |
"trial_name": null,
|
| 248364 |
"trial_params": null
|
| 248365 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 941.0,
|
| 5 |
+
"global_step": 117722,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 248355 |
"eval_steps_per_second": 0.679,
|
| 248356 |
"eval_wer": 0.1851399856424982,
|
| 248357 |
"step": 117100
|
| 248358 |
+
},
|
| 248359 |
+
{
|
| 248360 |
+
"epoch": 936.04,
|
| 248361 |
+
"learning_rate": 8.140096153846155e-06,
|
| 248362 |
+
"loss": 0.36,
|
| 248363 |
+
"step": 117105
|
| 248364 |
+
},
|
| 248365 |
+
{
|
| 248366 |
+
"epoch": 936.08,
|
| 248367 |
+
"learning_rate": 8.140016025641026e-06,
|
| 248368 |
+
"loss": 0.2816,
|
| 248369 |
+
"step": 117110
|
| 248370 |
+
},
|
| 248371 |
+
{
|
| 248372 |
+
"epoch": 936.12,
|
| 248373 |
+
"learning_rate": 8.139935897435897e-06,
|
| 248374 |
+
"loss": 0.302,
|
| 248375 |
+
"step": 117115
|
| 248376 |
+
},
|
| 248377 |
+
{
|
| 248378 |
+
"epoch": 936.16,
|
| 248379 |
+
"learning_rate": 8.13985576923077e-06,
|
| 248380 |
+
"loss": 0.4257,
|
| 248381 |
+
"step": 117120
|
| 248382 |
+
},
|
| 248383 |
+
{
|
| 248384 |
+
"epoch": 936.2,
|
| 248385 |
+
"learning_rate": 8.139775641025642e-06,
|
| 248386 |
+
"loss": 1.0676,
|
| 248387 |
+
"step": 117125
|
| 248388 |
+
},
|
| 248389 |
+
{
|
| 248390 |
+
"epoch": 936.24,
|
| 248391 |
+
"learning_rate": 8.139695512820513e-06,
|
| 248392 |
+
"loss": 0.3769,
|
| 248393 |
+
"step": 117130
|
| 248394 |
+
},
|
| 248395 |
+
{
|
| 248396 |
+
"epoch": 936.28,
|
| 248397 |
+
"learning_rate": 8.139615384615386e-06,
|
| 248398 |
+
"loss": 0.3151,
|
| 248399 |
+
"step": 117135
|
| 248400 |
+
},
|
| 248401 |
+
{
|
| 248402 |
+
"epoch": 936.32,
|
| 248403 |
+
"learning_rate": 8.139535256410258e-06,
|
| 248404 |
+
"loss": 0.2631,
|
| 248405 |
+
"step": 117140
|
| 248406 |
+
},
|
| 248407 |
+
{
|
| 248408 |
+
"epoch": 936.36,
|
| 248409 |
+
"learning_rate": 8.139455128205129e-06,
|
| 248410 |
+
"loss": 0.397,
|
| 248411 |
+
"step": 117145
|
| 248412 |
+
},
|
| 248413 |
+
{
|
| 248414 |
+
"epoch": 936.4,
|
| 248415 |
+
"learning_rate": 8.139375e-06,
|
| 248416 |
+
"loss": 1.1359,
|
| 248417 |
+
"step": 117150
|
| 248418 |
+
},
|
| 248419 |
+
{
|
| 248420 |
+
"epoch": 936.44,
|
| 248421 |
+
"learning_rate": 8.139294871794873e-06,
|
| 248422 |
+
"loss": 0.3287,
|
| 248423 |
+
"step": 117155
|
| 248424 |
+
},
|
| 248425 |
+
{
|
| 248426 |
+
"epoch": 936.48,
|
| 248427 |
+
"learning_rate": 8.139214743589745e-06,
|
| 248428 |
+
"loss": 0.2817,
|
| 248429 |
+
"step": 117160
|
| 248430 |
+
},
|
| 248431 |
+
{
|
| 248432 |
+
"epoch": 936.52,
|
| 248433 |
+
"learning_rate": 8.139134615384616e-06,
|
| 248434 |
+
"loss": 0.3736,
|
| 248435 |
+
"step": 117165
|
| 248436 |
+
},
|
| 248437 |
+
{
|
| 248438 |
+
"epoch": 936.56,
|
| 248439 |
+
"learning_rate": 8.139054487179487e-06,
|
| 248440 |
+
"loss": 0.4126,
|
| 248441 |
+
"step": 117170
|
| 248442 |
+
},
|
| 248443 |
+
{
|
| 248444 |
+
"epoch": 936.6,
|
| 248445 |
+
"learning_rate": 8.13897435897436e-06,
|
| 248446 |
+
"loss": 1.0384,
|
| 248447 |
+
"step": 117175
|
| 248448 |
+
},
|
| 248449 |
+
{
|
| 248450 |
+
"epoch": 936.64,
|
| 248451 |
+
"learning_rate": 8.13889423076923e-06,
|
| 248452 |
+
"loss": 0.3022,
|
| 248453 |
+
"step": 117180
|
| 248454 |
+
},
|
| 248455 |
+
{
|
| 248456 |
+
"epoch": 936.68,
|
| 248457 |
+
"learning_rate": 8.138814102564103e-06,
|
| 248458 |
+
"loss": 0.2988,
|
| 248459 |
+
"step": 117185
|
| 248460 |
+
},
|
| 248461 |
+
{
|
| 248462 |
+
"epoch": 936.72,
|
| 248463 |
+
"learning_rate": 8.138733974358976e-06,
|
| 248464 |
+
"loss": 0.3346,
|
| 248465 |
+
"step": 117190
|
| 248466 |
+
},
|
| 248467 |
+
{
|
| 248468 |
+
"epoch": 936.76,
|
| 248469 |
+
"learning_rate": 8.138653846153846e-06,
|
| 248470 |
+
"loss": 0.4354,
|
| 248471 |
+
"step": 117195
|
| 248472 |
+
},
|
| 248473 |
+
{
|
| 248474 |
+
"epoch": 936.8,
|
| 248475 |
+
"learning_rate": 8.138573717948719e-06,
|
| 248476 |
+
"loss": 1.2176,
|
| 248477 |
+
"step": 117200
|
| 248478 |
+
},
|
| 248479 |
+
{
|
| 248480 |
+
"epoch": 936.84,
|
| 248481 |
+
"learning_rate": 8.13849358974359e-06,
|
| 248482 |
+
"loss": 0.367,
|
| 248483 |
+
"step": 117205
|
| 248484 |
+
},
|
| 248485 |
+
{
|
| 248486 |
+
"epoch": 936.88,
|
| 248487 |
+
"learning_rate": 8.138413461538462e-06,
|
| 248488 |
+
"loss": 0.2743,
|
| 248489 |
+
"step": 117210
|
| 248490 |
+
},
|
| 248491 |
+
{
|
| 248492 |
+
"epoch": 936.92,
|
| 248493 |
+
"learning_rate": 8.138333333333333e-06,
|
| 248494 |
+
"loss": 0.3529,
|
| 248495 |
+
"step": 117215
|
| 248496 |
+
},
|
| 248497 |
+
{
|
| 248498 |
+
"epoch": 936.96,
|
| 248499 |
+
"learning_rate": 8.138253205128206e-06,
|
| 248500 |
+
"loss": 0.4974,
|
| 248501 |
+
"step": 117220
|
| 248502 |
+
},
|
| 248503 |
+
{
|
| 248504 |
+
"epoch": 937.0,
|
| 248505 |
+
"learning_rate": 8.138173076923077e-06,
|
| 248506 |
+
"loss": 1.2262,
|
| 248507 |
+
"step": 117225
|
| 248508 |
+
},
|
| 248509 |
+
{
|
| 248510 |
+
"epoch": 937.0,
|
| 248511 |
+
"eval_loss": 0.382039874792099,
|
| 248512 |
+
"eval_runtime": 39.12,
|
| 248513 |
+
"eval_samples_per_second": 21.575,
|
| 248514 |
+
"eval_steps_per_second": 0.69,
|
| 248515 |
+
"eval_wer": 0.18934867596075827,
|
| 248516 |
+
"step": 117225
|
| 248517 |
+
},
|
| 248518 |
+
{
|
| 248519 |
+
"epoch": 945.04,
|
| 248520 |
+
"learning_rate": 8.138092948717949e-06,
|
| 248521 |
+
"loss": 0.3032,
|
| 248522 |
+
"step": 117230
|
| 248523 |
+
},
|
| 248524 |
+
{
|
| 248525 |
+
"epoch": 945.08,
|
| 248526 |
+
"learning_rate": 8.138012820512822e-06,
|
| 248527 |
+
"loss": 0.273,
|
| 248528 |
+
"step": 117235
|
| 248529 |
+
},
|
| 248530 |
+
{
|
| 248531 |
+
"epoch": 945.12,
|
| 248532 |
+
"learning_rate": 8.137932692307693e-06,
|
| 248533 |
+
"loss": 0.3273,
|
| 248534 |
+
"step": 117240
|
| 248535 |
+
},
|
| 248536 |
+
{
|
| 248537 |
+
"epoch": 945.16,
|
| 248538 |
+
"learning_rate": 8.137852564102565e-06,
|
| 248539 |
+
"loss": 0.4947,
|
| 248540 |
+
"step": 117245
|
| 248541 |
+
},
|
| 248542 |
+
{
|
| 248543 |
+
"epoch": 945.2,
|
| 248544 |
+
"learning_rate": 8.137772435897436e-06,
|
| 248545 |
+
"loss": 1.2386,
|
| 248546 |
+
"step": 117250
|
| 248547 |
+
},
|
| 248548 |
+
{
|
| 248549 |
+
"epoch": 945.24,
|
| 248550 |
+
"learning_rate": 8.137692307692309e-06,
|
| 248551 |
+
"loss": 0.4844,
|
| 248552 |
+
"step": 117255
|
| 248553 |
+
},
|
| 248554 |
+
{
|
| 248555 |
+
"epoch": 945.28,
|
| 248556 |
+
"learning_rate": 8.13761217948718e-06,
|
| 248557 |
+
"loss": 0.2666,
|
| 248558 |
+
"step": 117260
|
| 248559 |
+
},
|
| 248560 |
+
{
|
| 248561 |
+
"epoch": 945.32,
|
| 248562 |
+
"learning_rate": 8.137532051282052e-06,
|
| 248563 |
+
"loss": 0.3561,
|
| 248564 |
+
"step": 117265
|
| 248565 |
+
},
|
| 248566 |
+
{
|
| 248567 |
+
"epoch": 945.36,
|
| 248568 |
+
"learning_rate": 8.137451923076923e-06,
|
| 248569 |
+
"loss": 0.5483,
|
| 248570 |
+
"step": 117270
|
| 248571 |
+
},
|
| 248572 |
+
{
|
| 248573 |
+
"epoch": 945.4,
|
| 248574 |
+
"learning_rate": 8.137371794871796e-06,
|
| 248575 |
+
"loss": 1.1604,
|
| 248576 |
+
"step": 117275
|
| 248577 |
+
},
|
| 248578 |
+
{
|
| 248579 |
+
"epoch": 945.44,
|
| 248580 |
+
"learning_rate": 8.137291666666667e-06,
|
| 248581 |
+
"loss": 0.4208,
|
| 248582 |
+
"step": 117280
|
| 248583 |
+
},
|
| 248584 |
+
{
|
| 248585 |
+
"epoch": 945.48,
|
| 248586 |
+
"learning_rate": 8.137211538461539e-06,
|
| 248587 |
+
"loss": 0.3461,
|
| 248588 |
+
"step": 117285
|
| 248589 |
+
},
|
| 248590 |
+
{
|
| 248591 |
+
"epoch": 945.52,
|
| 248592 |
+
"learning_rate": 8.137131410256412e-06,
|
| 248593 |
+
"loss": 0.3296,
|
| 248594 |
+
"step": 117290
|
| 248595 |
+
},
|
| 248596 |
+
{
|
| 248597 |
+
"epoch": 945.56,
|
| 248598 |
+
"learning_rate": 8.137051282051283e-06,
|
| 248599 |
+
"loss": 0.4776,
|
| 248600 |
+
"step": 117295
|
| 248601 |
+
},
|
| 248602 |
+
{
|
| 248603 |
+
"epoch": 945.6,
|
| 248604 |
+
"learning_rate": 8.136971153846155e-06,
|
| 248605 |
+
"loss": 1.2078,
|
| 248606 |
+
"step": 117300
|
| 248607 |
+
},
|
| 248608 |
+
{
|
| 248609 |
+
"epoch": 945.64,
|
| 248610 |
+
"learning_rate": 8.136891025641026e-06,
|
| 248611 |
+
"loss": 0.2939,
|
| 248612 |
+
"step": 117305
|
| 248613 |
+
},
|
| 248614 |
+
{
|
| 248615 |
+
"epoch": 945.68,
|
| 248616 |
+
"learning_rate": 8.136810897435899e-06,
|
| 248617 |
+
"loss": 0.3025,
|
| 248618 |
+
"step": 117310
|
| 248619 |
+
},
|
| 248620 |
+
{
|
| 248621 |
+
"epoch": 945.72,
|
| 248622 |
+
"learning_rate": 8.136730769230769e-06,
|
| 248623 |
+
"loss": 0.2963,
|
| 248624 |
+
"step": 117315
|
| 248625 |
+
},
|
| 248626 |
+
{
|
| 248627 |
+
"epoch": 945.76,
|
| 248628 |
+
"learning_rate": 8.136650641025642e-06,
|
| 248629 |
+
"loss": 0.474,
|
| 248630 |
+
"step": 117320
|
| 248631 |
+
},
|
| 248632 |
+
{
|
| 248633 |
+
"epoch": 945.8,
|
| 248634 |
+
"learning_rate": 8.136570512820513e-06,
|
| 248635 |
+
"loss": 1.1468,
|
| 248636 |
+
"step": 117325
|
| 248637 |
+
},
|
| 248638 |
+
{
|
| 248639 |
+
"epoch": 945.84,
|
| 248640 |
+
"learning_rate": 8.136490384615386e-06,
|
| 248641 |
+
"loss": 0.3025,
|
| 248642 |
+
"step": 117330
|
| 248643 |
+
},
|
| 248644 |
+
{
|
| 248645 |
+
"epoch": 945.88,
|
| 248646 |
+
"learning_rate": 8.136410256410258e-06,
|
| 248647 |
+
"loss": 0.2942,
|
| 248648 |
+
"step": 117335
|
| 248649 |
+
},
|
| 248650 |
+
{
|
| 248651 |
+
"epoch": 945.92,
|
| 248652 |
+
"learning_rate": 8.136330128205129e-06,
|
| 248653 |
+
"loss": 0.3408,
|
| 248654 |
+
"step": 117340
|
| 248655 |
+
},
|
| 248656 |
+
{
|
| 248657 |
+
"epoch": 945.96,
|
| 248658 |
+
"learning_rate": 8.136250000000002e-06,
|
| 248659 |
+
"loss": 0.5317,
|
| 248660 |
+
"step": 117345
|
| 248661 |
+
},
|
| 248662 |
+
{
|
| 248663 |
+
"epoch": 946.0,
|
| 248664 |
+
"eval_loss": 0.4116620719432831,
|
| 248665 |
+
"eval_runtime": 49.1871,
|
| 248666 |
+
"eval_samples_per_second": 17.139,
|
| 248667 |
+
"eval_steps_per_second": 0.549,
|
| 248668 |
+
"eval_wer": 0.1847012186169432,
|
| 248669 |
+
"step": 117349
|
| 248670 |
+
},
|
| 248671 |
+
{
|
| 248672 |
+
"epoch": 946.01,
|
| 248673 |
+
"learning_rate": 8.136169871794872e-06,
|
| 248674 |
+
"loss": 0.4503,
|
| 248675 |
+
"step": 117350
|
| 248676 |
+
},
|
| 248677 |
+
{
|
| 248678 |
+
"epoch": 946.05,
|
| 248679 |
+
"learning_rate": 8.136089743589745e-06,
|
| 248680 |
+
"loss": 0.3793,
|
| 248681 |
+
"step": 117355
|
| 248682 |
+
},
|
| 248683 |
+
{
|
| 248684 |
+
"epoch": 946.09,
|
| 248685 |
+
"learning_rate": 8.136009615384616e-06,
|
| 248686 |
+
"loss": 0.2987,
|
| 248687 |
+
"step": 117360
|
| 248688 |
+
},
|
| 248689 |
+
{
|
| 248690 |
+
"epoch": 946.13,
|
| 248691 |
+
"learning_rate": 8.135929487179487e-06,
|
| 248692 |
+
"loss": 0.3598,
|
| 248693 |
+
"step": 117365
|
| 248694 |
+
},
|
| 248695 |
+
{
|
| 248696 |
+
"epoch": 946.17,
|
| 248697 |
+
"learning_rate": 8.135849358974359e-06,
|
| 248698 |
+
"loss": 0.6339,
|
| 248699 |
+
"step": 117370
|
| 248700 |
+
},
|
| 248701 |
+
{
|
| 248702 |
+
"epoch": 946.21,
|
| 248703 |
+
"learning_rate": 8.135769230769232e-06,
|
| 248704 |
+
"loss": 1.1798,
|
| 248705 |
+
"step": 117375
|
| 248706 |
+
},
|
| 248707 |
+
{
|
| 248708 |
+
"epoch": 946.25,
|
| 248709 |
+
"learning_rate": 8.135689102564103e-06,
|
| 248710 |
+
"loss": 0.3248,
|
| 248711 |
+
"step": 117380
|
| 248712 |
+
},
|
| 248713 |
+
{
|
| 248714 |
+
"epoch": 946.29,
|
| 248715 |
+
"learning_rate": 8.135608974358974e-06,
|
| 248716 |
+
"loss": 0.4143,
|
| 248717 |
+
"step": 117385
|
| 248718 |
+
},
|
| 248719 |
+
{
|
| 248720 |
+
"epoch": 946.33,
|
| 248721 |
+
"learning_rate": 8.135528846153848e-06,
|
| 248722 |
+
"loss": 0.3344,
|
| 248723 |
+
"step": 117390
|
| 248724 |
+
},
|
| 248725 |
+
{
|
| 248726 |
+
"epoch": 946.37,
|
| 248727 |
+
"learning_rate": 8.135448717948719e-06,
|
| 248728 |
+
"loss": 0.515,
|
| 248729 |
+
"step": 117395
|
| 248730 |
+
},
|
| 248731 |
+
{
|
| 248732 |
+
"epoch": 946.41,
|
| 248733 |
+
"learning_rate": 8.13536858974359e-06,
|
| 248734 |
+
"loss": 1.1973,
|
| 248735 |
+
"step": 117400
|
| 248736 |
+
},
|
| 248737 |
+
{
|
| 248738 |
+
"epoch": 946.45,
|
| 248739 |
+
"learning_rate": 8.135288461538462e-06,
|
| 248740 |
+
"loss": 0.2734,
|
| 248741 |
+
"step": 117405
|
| 248742 |
+
},
|
| 248743 |
+
{
|
| 248744 |
+
"epoch": 946.49,
|
| 248745 |
+
"learning_rate": 8.135208333333335e-06,
|
| 248746 |
+
"loss": 0.2632,
|
| 248747 |
+
"step": 117410
|
| 248748 |
+
},
|
| 248749 |
+
{
|
| 248750 |
+
"epoch": 946.53,
|
| 248751 |
+
"learning_rate": 8.135128205128206e-06,
|
| 248752 |
+
"loss": 0.3458,
|
| 248753 |
+
"step": 117415
|
| 248754 |
+
},
|
| 248755 |
+
{
|
| 248756 |
+
"epoch": 946.57,
|
| 248757 |
+
"learning_rate": 8.135048076923077e-06,
|
| 248758 |
+
"loss": 0.5121,
|
| 248759 |
+
"step": 117420
|
| 248760 |
+
},
|
| 248761 |
+
{
|
| 248762 |
+
"epoch": 946.61,
|
| 248763 |
+
"learning_rate": 8.134967948717949e-06,
|
| 248764 |
+
"loss": 1.1818,
|
| 248765 |
+
"step": 117425
|
| 248766 |
+
},
|
| 248767 |
+
{
|
| 248768 |
+
"epoch": 946.65,
|
| 248769 |
+
"learning_rate": 8.134887820512822e-06,
|
| 248770 |
+
"loss": 0.3381,
|
| 248771 |
+
"step": 117430
|
| 248772 |
+
},
|
| 248773 |
+
{
|
| 248774 |
+
"epoch": 946.69,
|
| 248775 |
+
"learning_rate": 8.134807692307693e-06,
|
| 248776 |
+
"loss": 0.2735,
|
| 248777 |
+
"step": 117435
|
| 248778 |
+
},
|
| 248779 |
+
{
|
| 248780 |
+
"epoch": 946.73,
|
| 248781 |
+
"learning_rate": 8.134727564102565e-06,
|
| 248782 |
+
"loss": 0.3295,
|
| 248783 |
+
"step": 117440
|
| 248784 |
+
},
|
| 248785 |
+
{
|
| 248786 |
+
"epoch": 946.77,
|
| 248787 |
+
"learning_rate": 8.134647435897438e-06,
|
| 248788 |
+
"loss": 0.6012,
|
| 248789 |
+
"step": 117445
|
| 248790 |
+
},
|
| 248791 |
+
{
|
| 248792 |
+
"epoch": 946.81,
|
| 248793 |
+
"learning_rate": 8.134567307692309e-06,
|
| 248794 |
+
"loss": 1.0467,
|
| 248795 |
+
"step": 117450
|
| 248796 |
+
},
|
| 248797 |
+
{
|
| 248798 |
+
"epoch": 946.85,
|
| 248799 |
+
"learning_rate": 8.13448717948718e-06,
|
| 248800 |
+
"loss": 0.2941,
|
| 248801 |
+
"step": 117455
|
| 248802 |
+
},
|
| 248803 |
+
{
|
| 248804 |
+
"epoch": 946.89,
|
| 248805 |
+
"learning_rate": 8.134407051282052e-06,
|
| 248806 |
+
"loss": 0.488,
|
| 248807 |
+
"step": 117460
|
| 248808 |
+
},
|
| 248809 |
+
{
|
| 248810 |
+
"epoch": 946.93,
|
| 248811 |
+
"learning_rate": 8.134326923076925e-06,
|
| 248812 |
+
"loss": 0.3663,
|
| 248813 |
+
"step": 117465
|
| 248814 |
+
},
|
| 248815 |
+
{
|
| 248816 |
+
"epoch": 946.97,
|
| 248817 |
+
"learning_rate": 8.134246794871794e-06,
|
| 248818 |
+
"loss": 0.5698,
|
| 248819 |
+
"step": 117470
|
| 248820 |
+
},
|
| 248821 |
+
{
|
| 248822 |
+
"epoch": 947.0,
|
| 248823 |
+
"eval_loss": 0.3862200081348419,
|
| 248824 |
+
"eval_runtime": 39.1121,
|
| 248825 |
+
"eval_samples_per_second": 21.477,
|
| 248826 |
+
"eval_steps_per_second": 0.69,
|
| 248827 |
+
"eval_wer": 0.1748963711730056,
|
| 248828 |
+
"step": 117473
|
| 248829 |
+
},
|
| 248830 |
+
{
|
| 248831 |
+
"epoch": 947.02,
|
| 248832 |
+
"learning_rate": 8.134166666666667e-06,
|
| 248833 |
+
"loss": 0.3545,
|
| 248834 |
+
"step": 117475
|
| 248835 |
+
},
|
| 248836 |
+
{
|
| 248837 |
+
"epoch": 947.06,
|
| 248838 |
+
"learning_rate": 8.134086538461539e-06,
|
| 248839 |
+
"loss": 0.2982,
|
| 248840 |
+
"step": 117480
|
| 248841 |
+
},
|
| 248842 |
+
{
|
| 248843 |
+
"epoch": 947.1,
|
| 248844 |
+
"learning_rate": 8.13400641025641e-06,
|
| 248845 |
+
"loss": 0.2839,
|
| 248846 |
+
"step": 117485
|
| 248847 |
+
},
|
| 248848 |
+
{
|
| 248849 |
+
"epoch": 947.14,
|
| 248850 |
+
"learning_rate": 8.133926282051283e-06,
|
| 248851 |
+
"loss": 0.382,
|
| 248852 |
+
"step": 117490
|
| 248853 |
+
},
|
| 248854 |
+
{
|
| 248855 |
+
"epoch": 947.18,
|
| 248856 |
+
"learning_rate": 8.133846153846155e-06,
|
| 248857 |
+
"loss": 0.9134,
|
| 248858 |
+
"step": 117495
|
| 248859 |
+
},
|
| 248860 |
+
{
|
| 248861 |
+
"epoch": 947.22,
|
| 248862 |
+
"learning_rate": 8.133766025641026e-06,
|
| 248863 |
+
"loss": 0.9888,
|
| 248864 |
+
"step": 117500
|
| 248865 |
+
},
|
| 248866 |
+
{
|
| 248867 |
+
"epoch": 947.26,
|
| 248868 |
+
"learning_rate": 8.133685897435897e-06,
|
| 248869 |
+
"loss": 0.3261,
|
| 248870 |
+
"step": 117505
|
| 248871 |
+
},
|
| 248872 |
+
{
|
| 248873 |
+
"epoch": 947.3,
|
| 248874 |
+
"learning_rate": 8.13360576923077e-06,
|
| 248875 |
+
"loss": 0.2684,
|
| 248876 |
+
"step": 117510
|
| 248877 |
+
},
|
| 248878 |
+
{
|
| 248879 |
+
"epoch": 947.34,
|
| 248880 |
+
"learning_rate": 8.133525641025642e-06,
|
| 248881 |
+
"loss": 0.3223,
|
| 248882 |
+
"step": 117515
|
| 248883 |
+
},
|
| 248884 |
+
{
|
| 248885 |
+
"epoch": 947.38,
|
| 248886 |
+
"learning_rate": 8.133445512820513e-06,
|
| 248887 |
+
"loss": 0.534,
|
| 248888 |
+
"step": 117520
|
| 248889 |
+
},
|
| 248890 |
+
{
|
| 248891 |
+
"epoch": 947.42,
|
| 248892 |
+
"learning_rate": 8.133365384615384e-06,
|
| 248893 |
+
"loss": 1.0773,
|
| 248894 |
+
"step": 117525
|
| 248895 |
+
},
|
| 248896 |
+
{
|
| 248897 |
+
"epoch": 947.46,
|
| 248898 |
+
"learning_rate": 8.133285256410257e-06,
|
| 248899 |
+
"loss": 0.2496,
|
| 248900 |
+
"step": 117530
|
| 248901 |
+
},
|
| 248902 |
+
{
|
| 248903 |
+
"epoch": 947.5,
|
| 248904 |
+
"learning_rate": 8.133205128205129e-06,
|
| 248905 |
+
"loss": 0.3469,
|
| 248906 |
+
"step": 117535
|
| 248907 |
+
},
|
| 248908 |
+
{
|
| 248909 |
+
"epoch": 947.54,
|
| 248910 |
+
"learning_rate": 8.133125e-06,
|
| 248911 |
+
"loss": 0.3361,
|
| 248912 |
+
"step": 117540
|
| 248913 |
+
},
|
| 248914 |
+
{
|
| 248915 |
+
"epoch": 947.58,
|
| 248916 |
+
"learning_rate": 8.133044871794873e-06,
|
| 248917 |
+
"loss": 0.7251,
|
| 248918 |
+
"step": 117545
|
| 248919 |
+
},
|
| 248920 |
+
{
|
| 248921 |
+
"epoch": 947.62,
|
| 248922 |
+
"learning_rate": 8.132964743589745e-06,
|
| 248923 |
+
"loss": 1.0967,
|
| 248924 |
+
"step": 117550
|
| 248925 |
+
},
|
| 248926 |
+
{
|
| 248927 |
+
"epoch": 947.66,
|
| 248928 |
+
"learning_rate": 8.132884615384616e-06,
|
| 248929 |
+
"loss": 0.3342,
|
| 248930 |
+
"step": 117555
|
| 248931 |
+
},
|
| 248932 |
+
{
|
| 248933 |
+
"epoch": 947.7,
|
| 248934 |
+
"learning_rate": 8.132804487179487e-06,
|
| 248935 |
+
"loss": 0.2923,
|
| 248936 |
+
"step": 117560
|
| 248937 |
+
},
|
| 248938 |
+
{
|
| 248939 |
+
"epoch": 947.74,
|
| 248940 |
+
"learning_rate": 8.13272435897436e-06,
|
| 248941 |
+
"loss": 0.3418,
|
| 248942 |
+
"step": 117565
|
| 248943 |
+
},
|
| 248944 |
+
{
|
| 248945 |
+
"epoch": 947.78,
|
| 248946 |
+
"learning_rate": 8.132644230769232e-06,
|
| 248947 |
+
"loss": 0.6134,
|
| 248948 |
+
"step": 117570
|
| 248949 |
+
},
|
| 248950 |
+
{
|
| 248951 |
+
"epoch": 947.82,
|
| 248952 |
+
"learning_rate": 8.132564102564103e-06,
|
| 248953 |
+
"loss": 0.9037,
|
| 248954 |
+
"step": 117575
|
| 248955 |
+
},
|
| 248956 |
+
{
|
| 248957 |
+
"epoch": 947.86,
|
| 248958 |
+
"learning_rate": 8.132483974358974e-06,
|
| 248959 |
+
"loss": 0.3372,
|
| 248960 |
+
"step": 117580
|
| 248961 |
+
},
|
| 248962 |
+
{
|
| 248963 |
+
"epoch": 947.9,
|
| 248964 |
+
"learning_rate": 8.132403846153847e-06,
|
| 248965 |
+
"loss": 0.3168,
|
| 248966 |
+
"step": 117585
|
| 248967 |
+
},
|
| 248968 |
+
{
|
| 248969 |
+
"epoch": 947.94,
|
| 248970 |
+
"learning_rate": 8.132323717948719e-06,
|
| 248971 |
+
"loss": 0.3129,
|
| 248972 |
+
"step": 117590
|
| 248973 |
+
},
|
| 248974 |
+
{
|
| 248975 |
+
"epoch": 947.98,
|
| 248976 |
+
"learning_rate": 8.13224358974359e-06,
|
| 248977 |
+
"loss": 0.827,
|
| 248978 |
+
"step": 117595
|
| 248979 |
+
},
|
| 248980 |
+
{
|
| 248981 |
+
"epoch": 948.0,
|
| 248982 |
+
"eval_loss": 0.43474265933036804,
|
| 248983 |
+
"eval_runtime": 38.7119,
|
| 248984 |
+
"eval_samples_per_second": 21.673,
|
| 248985 |
+
"eval_steps_per_second": 0.697,
|
| 248986 |
+
"eval_wer": 0.19487402889711755,
|
| 248987 |
+
"step": 117597
|
| 248988 |
+
},
|
| 248989 |
+
{
|
| 248990 |
+
"epoch": 940.02,
|
| 248991 |
+
"learning_rate": 8.132163461538463e-06,
|
| 248992 |
+
"loss": 0.3282,
|
| 248993 |
+
"step": 117600
|
| 248994 |
+
},
|
| 248995 |
+
{
|
| 248996 |
+
"epoch": 940.06,
|
| 248997 |
+
"learning_rate": 8.132083333333335e-06,
|
| 248998 |
+
"loss": 0.3094,
|
| 248999 |
+
"step": 117605
|
| 249000 |
+
},
|
| 249001 |
+
{
|
| 249002 |
+
"epoch": 940.1,
|
| 249003 |
+
"learning_rate": 8.132003205128206e-06,
|
| 249004 |
+
"loss": 0.2889,
|
| 249005 |
+
"step": 117610
|
| 249006 |
+
},
|
| 249007 |
+
{
|
| 249008 |
+
"epoch": 940.14,
|
| 249009 |
+
"learning_rate": 8.131923076923077e-06,
|
| 249010 |
+
"loss": 0.3158,
|
| 249011 |
+
"step": 117615
|
| 249012 |
+
},
|
| 249013 |
+
{
|
| 249014 |
+
"epoch": 940.18,
|
| 249015 |
+
"learning_rate": 8.13184294871795e-06,
|
| 249016 |
+
"loss": 0.7108,
|
| 249017 |
+
"step": 117620
|
| 249018 |
+
},
|
| 249019 |
+
{
|
| 249020 |
+
"epoch": 940.22,
|
| 249021 |
+
"learning_rate": 8.13176282051282e-06,
|
| 249022 |
+
"loss": 0.8738,
|
| 249023 |
+
"step": 117625
|
| 249024 |
+
},
|
| 249025 |
+
{
|
| 249026 |
+
"epoch": 940.26,
|
| 249027 |
+
"learning_rate": 8.131682692307693e-06,
|
| 249028 |
+
"loss": 0.3388,
|
| 249029 |
+
"step": 117630
|
| 249030 |
+
},
|
| 249031 |
+
{
|
| 249032 |
+
"epoch": 940.3,
|
| 249033 |
+
"learning_rate": 8.131602564102566e-06,
|
| 249034 |
+
"loss": 0.2985,
|
| 249035 |
+
"step": 117635
|
| 249036 |
+
},
|
| 249037 |
+
{
|
| 249038 |
+
"epoch": 940.34,
|
| 249039 |
+
"learning_rate": 8.131522435897436e-06,
|
| 249040 |
+
"loss": 0.4881,
|
| 249041 |
+
"step": 117640
|
| 249042 |
+
},
|
| 249043 |
+
{
|
| 249044 |
+
"epoch": 940.38,
|
| 249045 |
+
"learning_rate": 8.131442307692309e-06,
|
| 249046 |
+
"loss": 0.7521,
|
| 249047 |
+
"step": 117645
|
| 249048 |
+
},
|
| 249049 |
+
{
|
| 249050 |
+
"epoch": 940.42,
|
| 249051 |
+
"learning_rate": 8.13136217948718e-06,
|
| 249052 |
+
"loss": 1.0756,
|
| 249053 |
+
"step": 117650
|
| 249054 |
+
},
|
| 249055 |
+
{
|
| 249056 |
+
"epoch": 940.46,
|
| 249057 |
+
"learning_rate": 8.131282051282052e-06,
|
| 249058 |
+
"loss": 0.2887,
|
| 249059 |
+
"step": 117655
|
| 249060 |
+
},
|
| 249061 |
+
{
|
| 249062 |
+
"epoch": 940.5,
|
| 249063 |
+
"learning_rate": 8.131201923076923e-06,
|
| 249064 |
+
"loss": 0.3438,
|
| 249065 |
+
"step": 117660
|
| 249066 |
+
},
|
| 249067 |
+
{
|
| 249068 |
+
"epoch": 940.54,
|
| 249069 |
+
"learning_rate": 8.131121794871796e-06,
|
| 249070 |
+
"loss": 0.3598,
|
| 249071 |
+
"step": 117665
|
| 249072 |
+
},
|
| 249073 |
+
{
|
| 249074 |
+
"epoch": 940.58,
|
| 249075 |
+
"learning_rate": 8.131041666666667e-06,
|
| 249076 |
+
"loss": 0.7274,
|
| 249077 |
+
"step": 117670
|
| 249078 |
+
},
|
| 249079 |
+
{
|
| 249080 |
+
"epoch": 940.62,
|
| 249081 |
+
"learning_rate": 8.130961538461539e-06,
|
| 249082 |
+
"loss": 0.876,
|
| 249083 |
+
"step": 117675
|
| 249084 |
+
},
|
| 249085 |
+
{
|
| 249086 |
+
"epoch": 940.66,
|
| 249087 |
+
"learning_rate": 8.13088141025641e-06,
|
| 249088 |
+
"loss": 0.2849,
|
| 249089 |
+
"step": 117680
|
| 249090 |
+
},
|
| 249091 |
+
{
|
| 249092 |
+
"epoch": 940.7,
|
| 249093 |
+
"learning_rate": 8.130801282051283e-06,
|
| 249094 |
+
"loss": 0.2736,
|
| 249095 |
+
"step": 117685
|
| 249096 |
+
},
|
| 249097 |
+
{
|
| 249098 |
+
"epoch": 940.74,
|
| 249099 |
+
"learning_rate": 8.130721153846154e-06,
|
| 249100 |
+
"loss": 0.3746,
|
| 249101 |
+
"step": 117690
|
| 249102 |
+
},
|
| 249103 |
+
{
|
| 249104 |
+
"epoch": 940.78,
|
| 249105 |
+
"learning_rate": 8.130641025641026e-06,
|
| 249106 |
+
"loss": 0.737,
|
| 249107 |
+
"step": 117695
|
| 249108 |
+
},
|
| 249109 |
+
{
|
| 249110 |
+
"epoch": 940.82,
|
| 249111 |
+
"learning_rate": 8.130560897435899e-06,
|
| 249112 |
+
"loss": 0.8436,
|
| 249113 |
+
"step": 117700
|
| 249114 |
+
},
|
| 249115 |
+
{
|
| 249116 |
+
"epoch": 940.86,
|
| 249117 |
+
"learning_rate": 8.13048076923077e-06,
|
| 249118 |
+
"loss": 0.2284,
|
| 249119 |
+
"step": 117705
|
| 249120 |
+
},
|
| 249121 |
+
{
|
| 249122 |
+
"epoch": 940.9,
|
| 249123 |
+
"learning_rate": 8.130400641025642e-06,
|
| 249124 |
+
"loss": 0.3026,
|
| 249125 |
+
"step": 117710
|
| 249126 |
+
},
|
| 249127 |
+
{
|
| 249128 |
+
"epoch": 940.94,
|
| 249129 |
+
"learning_rate": 8.130320512820513e-06,
|
| 249130 |
+
"loss": 0.426,
|
| 249131 |
+
"step": 117715
|
| 249132 |
+
},
|
| 249133 |
+
{
|
| 249134 |
+
"epoch": 940.98,
|
| 249135 |
+
"learning_rate": 8.130240384615386e-06,
|
| 249136 |
+
"loss": 0.7146,
|
| 249137 |
+
"step": 117720
|
| 249138 |
+
},
|
| 249139 |
+
{
|
| 249140 |
+
"epoch": 941.0,
|
| 249141 |
+
"eval_loss": 0.34532302618026733,
|
| 249142 |
+
"eval_runtime": 38.5318,
|
| 249143 |
+
"eval_samples_per_second": 21.774,
|
| 249144 |
+
"eval_steps_per_second": 0.701,
|
| 249145 |
+
"eval_wer": 0.18467364669191358,
|
| 249146 |
+
"step": 117722
|
| 249147 |
}
|
| 249148 |
],
|
| 249149 |
"max_steps": 625000,
|
| 249150 |
"num_train_epochs": 5000,
|
| 249151 |
+
"total_flos": 3.313008475394062e+20,
|
| 249152 |
"trial_name": null,
|
| 249153 |
"trial_params": null
|
| 249154 |
}
|
model-bin/finetune/base/{checkpoint-117100 → checkpoint-117722}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630117908.6872168/events.out.tfevents.1630117908.86bb0ddabf9b.1042.51
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8a1a623804e4583586cde6273c613a8e5ca8fe0b77e9947bcc881c674bdfb457
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630118323.7473285/events.out.tfevents.1630118323.86bb0ddabf9b.1042.53
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7f3bfba1097373b7feebc5b84cad11f506237fe60f000442bd3df78e16a4d2bc
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630118841.0029836/events.out.tfevents.1630118841.86bb0ddabf9b.1042.55
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74fe57cdf6563d18fe657f6b941c44795dcb541056449b18ebd5b5b124eb2665
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630119249.7913108/events.out.tfevents.1630119249.86bb0ddabf9b.1042.57
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4984439e0ef86c761adbbf1f111486bf5d33a618505376836a6244a0788d262e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630119657.5811403/events.out.tfevents.1630119657.86bb0ddabf9b.1042.59
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f47351033db2b4e29e30eae5441bbc8da255b1f2602bd8952e9547a59c46ff48
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630117908.86bb0ddabf9b.1042.50
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5637e932748456917dc2a740db74ad845315b0eee9eacfb686a35cd15bbc1ccf
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630118323.86bb0ddabf9b.1042.52
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:48434c5e9458dce1d2b1b175ad63f08018d916ddd57dfcfe72f568f9b60528f2
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630118840.86bb0ddabf9b.1042.54
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c4728b4ef5f812daf8e8b9549659f7eb9b5cf1aae46a768dd0ba123bdd6956ee
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630119249.86bb0ddabf9b.1042.56
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4bb9f8eb0abfa4eca6066b388a060ff81f509bc71bc76654c434e2abfe87d079
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630119657.86bb0ddabf9b.1042.58
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3afeafde1841e315916058445ba3d2199ce216a28a5e292d148f36f9b76d7f48
|
| 3 |
+
size 8622
|