"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630043182.9793327/events.out.tfevents.1630043182.52f5c7e305a3.886.61 +3 -0
- model-bin/finetune/base/log/1630043660.3624449/events.out.tfevents.1630043660.52f5c7e305a3.886.63 +3 -0
- model-bin/finetune/base/log/1630044141.0074263/events.out.tfevents.1630044141.52f5c7e305a3.886.65 +3 -0
- model-bin/finetune/base/log/1630044614.4908102/events.out.tfevents.1630044614.52f5c7e305a3.886.67 +3 -0
- model-bin/finetune/base/log/1630045101.34449/events.out.tfevents.1630045101.52f5c7e305a3.886.69 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630043182.52f5c7e305a3.886.60 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630043660.52f5c7e305a3.886.62 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630044141.52f5c7e305a3.886.64 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630044614.52f5c7e305a3.886.66 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630045101.52f5c7e305a3.886.68 +3 -0
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/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 sha256:8ad06371735a046dd7c9ef9132035682f9c5de2a72ee91eadd07f7a4003b5e3e
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/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 sha256:5a8eb7a6778d47cb003b93cbbdd60345b9fb33fdffcedb8fb466ca7f8f337a22
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/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:f1bab8a8b45fe33a8814faa874970aaa38052e7827b5642f8165984321edc61a
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/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 sha256:01a18ea4d20203ec9fd1ae2b3ac1f3ea23edcb584eb282e1b66ac38040ce9b63
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/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 sha256:6819449b28fc88148155c2c7a3ee8e6199cb0db9b034e04f8a104713b88847e3
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/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,
|
|
@@ -239493,11 +239493,800 @@
|
|
| 239493 |
"eval_steps_per_second": 0.626,
|
| 239494 |
"eval_wer": 0.17663838502047982,
|
| 239495 |
"step": 110137
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 239496 |
}
|
| 239497 |
],
|
| 239498 |
"max_steps": 620000,
|
| 239499 |
"num_train_epochs": 5000,
|
| 239500 |
-
"total_flos": 3.
|
| 239501 |
"trial_name": null,
|
| 239502 |
"trial_params": null
|
| 239503 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 892.995983935743,
|
| 5 |
+
"global_step": 110758,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 239493 |
"eval_steps_per_second": 0.626,
|
| 239494 |
"eval_wer": 0.17663838502047982,
|
| 239495 |
"step": 110137
|
| 239496 |
+
},
|
| 239497 |
+
{
|
| 239498 |
+
"epoch": 888.02,
|
| 239499 |
+
"learning_rate": 8.237528271405494e-06,
|
| 239500 |
+
"loss": 0.3629,
|
| 239501 |
+
"step": 110140
|
| 239502 |
+
},
|
| 239503 |
+
{
|
| 239504 |
+
"epoch": 888.06,
|
| 239505 |
+
"learning_rate": 8.237447495961228e-06,
|
| 239506 |
+
"loss": 0.2722,
|
| 239507 |
+
"step": 110145
|
| 239508 |
+
},
|
| 239509 |
+
{
|
| 239510 |
+
"epoch": 888.1,
|
| 239511 |
+
"learning_rate": 8.237366720516964e-06,
|
| 239512 |
+
"loss": 0.3,
|
| 239513 |
+
"step": 110150
|
| 239514 |
+
},
|
| 239515 |
+
{
|
| 239516 |
+
"epoch": 888.15,
|
| 239517 |
+
"learning_rate": 8.237285945072698e-06,
|
| 239518 |
+
"loss": 0.3672,
|
| 239519 |
+
"step": 110155
|
| 239520 |
+
},
|
| 239521 |
+
{
|
| 239522 |
+
"epoch": 888.19,
|
| 239523 |
+
"learning_rate": 8.237205169628434e-06,
|
| 239524 |
+
"loss": 0.7683,
|
| 239525 |
+
"step": 110160
|
| 239526 |
+
},
|
| 239527 |
+
{
|
| 239528 |
+
"epoch": 888.23,
|
| 239529 |
+
"learning_rate": 8.237124394184168e-06,
|
| 239530 |
+
"loss": 0.8823,
|
| 239531 |
+
"step": 110165
|
| 239532 |
+
},
|
| 239533 |
+
{
|
| 239534 |
+
"epoch": 888.27,
|
| 239535 |
+
"learning_rate": 8.237043618739904e-06,
|
| 239536 |
+
"loss": 0.314,
|
| 239537 |
+
"step": 110170
|
| 239538 |
+
},
|
| 239539 |
+
{
|
| 239540 |
+
"epoch": 888.31,
|
| 239541 |
+
"learning_rate": 8.236962843295638e-06,
|
| 239542 |
+
"loss": 0.3071,
|
| 239543 |
+
"step": 110175
|
| 239544 |
+
},
|
| 239545 |
+
{
|
| 239546 |
+
"epoch": 888.35,
|
| 239547 |
+
"learning_rate": 8.236882067851374e-06,
|
| 239548 |
+
"loss": 0.3577,
|
| 239549 |
+
"step": 110180
|
| 239550 |
+
},
|
| 239551 |
+
{
|
| 239552 |
+
"epoch": 888.39,
|
| 239553 |
+
"learning_rate": 8.23680129240711e-06,
|
| 239554 |
+
"loss": 0.697,
|
| 239555 |
+
"step": 110185
|
| 239556 |
+
},
|
| 239557 |
+
{
|
| 239558 |
+
"epoch": 888.43,
|
| 239559 |
+
"learning_rate": 8.236720516962844e-06,
|
| 239560 |
+
"loss": 0.9972,
|
| 239561 |
+
"step": 110190
|
| 239562 |
+
},
|
| 239563 |
+
{
|
| 239564 |
+
"epoch": 888.47,
|
| 239565 |
+
"learning_rate": 8.23663974151858e-06,
|
| 239566 |
+
"loss": 0.2665,
|
| 239567 |
+
"step": 110195
|
| 239568 |
+
},
|
| 239569 |
+
{
|
| 239570 |
+
"epoch": 888.51,
|
| 239571 |
+
"learning_rate": 8.236558966074314e-06,
|
| 239572 |
+
"loss": 0.2957,
|
| 239573 |
+
"step": 110200
|
| 239574 |
+
},
|
| 239575 |
+
{
|
| 239576 |
+
"epoch": 888.55,
|
| 239577 |
+
"learning_rate": 8.23647819063005e-06,
|
| 239578 |
+
"loss": 0.4025,
|
| 239579 |
+
"step": 110205
|
| 239580 |
+
},
|
| 239581 |
+
{
|
| 239582 |
+
"epoch": 888.59,
|
| 239583 |
+
"learning_rate": 8.236397415185784e-06,
|
| 239584 |
+
"loss": 0.7967,
|
| 239585 |
+
"step": 110210
|
| 239586 |
+
},
|
| 239587 |
+
{
|
| 239588 |
+
"epoch": 888.63,
|
| 239589 |
+
"learning_rate": 8.23631663974152e-06,
|
| 239590 |
+
"loss": 0.8779,
|
| 239591 |
+
"step": 110215
|
| 239592 |
+
},
|
| 239593 |
+
{
|
| 239594 |
+
"epoch": 888.67,
|
| 239595 |
+
"learning_rate": 8.236235864297254e-06,
|
| 239596 |
+
"loss": 0.3245,
|
| 239597 |
+
"step": 110220
|
| 239598 |
+
},
|
| 239599 |
+
{
|
| 239600 |
+
"epoch": 888.71,
|
| 239601 |
+
"learning_rate": 8.23615508885299e-06,
|
| 239602 |
+
"loss": 0.2896,
|
| 239603 |
+
"step": 110225
|
| 239604 |
+
},
|
| 239605 |
+
{
|
| 239606 |
+
"epoch": 888.75,
|
| 239607 |
+
"learning_rate": 8.236074313408724e-06,
|
| 239608 |
+
"loss": 0.3919,
|
| 239609 |
+
"step": 110230
|
| 239610 |
+
},
|
| 239611 |
+
{
|
| 239612 |
+
"epoch": 888.79,
|
| 239613 |
+
"learning_rate": 8.23599353796446e-06,
|
| 239614 |
+
"loss": 0.8848,
|
| 239615 |
+
"step": 110235
|
| 239616 |
+
},
|
| 239617 |
+
{
|
| 239618 |
+
"epoch": 888.83,
|
| 239619 |
+
"learning_rate": 8.235912762520194e-06,
|
| 239620 |
+
"loss": 0.8322,
|
| 239621 |
+
"step": 110240
|
| 239622 |
+
},
|
| 239623 |
+
{
|
| 239624 |
+
"epoch": 888.87,
|
| 239625 |
+
"learning_rate": 8.23583198707593e-06,
|
| 239626 |
+
"loss": 0.241,
|
| 239627 |
+
"step": 110245
|
| 239628 |
+
},
|
| 239629 |
+
{
|
| 239630 |
+
"epoch": 888.91,
|
| 239631 |
+
"learning_rate": 8.235751211631666e-06,
|
| 239632 |
+
"loss": 0.281,
|
| 239633 |
+
"step": 110250
|
| 239634 |
+
},
|
| 239635 |
+
{
|
| 239636 |
+
"epoch": 888.95,
|
| 239637 |
+
"learning_rate": 8.2356704361874e-06,
|
| 239638 |
+
"loss": 0.3819,
|
| 239639 |
+
"step": 110255
|
| 239640 |
+
},
|
| 239641 |
+
{
|
| 239642 |
+
"epoch": 888.99,
|
| 239643 |
+
"learning_rate": 8.235589660743136e-06,
|
| 239644 |
+
"loss": 0.9705,
|
| 239645 |
+
"step": 110260
|
| 239646 |
+
},
|
| 239647 |
+
{
|
| 239648 |
+
"epoch": 889.0,
|
| 239649 |
+
"eval_loss": 0.42611831426620483,
|
| 239650 |
+
"eval_runtime": 42.4355,
|
| 239651 |
+
"eval_samples_per_second": 19.748,
|
| 239652 |
+
"eval_steps_per_second": 0.636,
|
| 239653 |
+
"eval_wer": 0.18801652892561985,
|
| 239654 |
+
"step": 110261
|
| 239655 |
+
},
|
| 239656 |
+
{
|
| 239657 |
+
"epoch": 889.03,
|
| 239658 |
+
"learning_rate": 8.23550888529887e-06,
|
| 239659 |
+
"loss": 0.288,
|
| 239660 |
+
"step": 110265
|
| 239661 |
+
},
|
| 239662 |
+
{
|
| 239663 |
+
"epoch": 889.07,
|
| 239664 |
+
"learning_rate": 8.235428109854606e-06,
|
| 239665 |
+
"loss": 0.2627,
|
| 239666 |
+
"step": 110270
|
| 239667 |
+
},
|
| 239668 |
+
{
|
| 239669 |
+
"epoch": 889.11,
|
| 239670 |
+
"learning_rate": 8.23534733441034e-06,
|
| 239671 |
+
"loss": 0.3018,
|
| 239672 |
+
"step": 110275
|
| 239673 |
+
},
|
| 239674 |
+
{
|
| 239675 |
+
"epoch": 889.15,
|
| 239676 |
+
"learning_rate": 8.235266558966076e-06,
|
| 239677 |
+
"loss": 0.4068,
|
| 239678 |
+
"step": 110280
|
| 239679 |
+
},
|
| 239680 |
+
{
|
| 239681 |
+
"epoch": 889.19,
|
| 239682 |
+
"learning_rate": 8.23518578352181e-06,
|
| 239683 |
+
"loss": 0.9867,
|
| 239684 |
+
"step": 110285
|
| 239685 |
+
},
|
| 239686 |
+
{
|
| 239687 |
+
"epoch": 889.23,
|
| 239688 |
+
"learning_rate": 8.235105008077546e-06,
|
| 239689 |
+
"loss": 0.7438,
|
| 239690 |
+
"step": 110290
|
| 239691 |
+
},
|
| 239692 |
+
{
|
| 239693 |
+
"epoch": 889.27,
|
| 239694 |
+
"learning_rate": 8.23502423263328e-06,
|
| 239695 |
+
"loss": 0.2609,
|
| 239696 |
+
"step": 110295
|
| 239697 |
+
},
|
| 239698 |
+
{
|
| 239699 |
+
"epoch": 889.31,
|
| 239700 |
+
"learning_rate": 8.234943457189016e-06,
|
| 239701 |
+
"loss": 0.2882,
|
| 239702 |
+
"step": 110300
|
| 239703 |
+
},
|
| 239704 |
+
{
|
| 239705 |
+
"epoch": 889.35,
|
| 239706 |
+
"learning_rate": 8.234862681744751e-06,
|
| 239707 |
+
"loss": 0.341,
|
| 239708 |
+
"step": 110305
|
| 239709 |
+
},
|
| 239710 |
+
{
|
| 239711 |
+
"epoch": 889.39,
|
| 239712 |
+
"learning_rate": 8.234781906300485e-06,
|
| 239713 |
+
"loss": 0.8645,
|
| 239714 |
+
"step": 110310
|
| 239715 |
+
},
|
| 239716 |
+
{
|
| 239717 |
+
"epoch": 889.43,
|
| 239718 |
+
"learning_rate": 8.234701130856221e-06,
|
| 239719 |
+
"loss": 0.7435,
|
| 239720 |
+
"step": 110315
|
| 239721 |
+
},
|
| 239722 |
+
{
|
| 239723 |
+
"epoch": 889.47,
|
| 239724 |
+
"learning_rate": 8.234620355411955e-06,
|
| 239725 |
+
"loss": 0.2552,
|
| 239726 |
+
"step": 110320
|
| 239727 |
+
},
|
| 239728 |
+
{
|
| 239729 |
+
"epoch": 889.51,
|
| 239730 |
+
"learning_rate": 8.234539579967691e-06,
|
| 239731 |
+
"loss": 0.3273,
|
| 239732 |
+
"step": 110325
|
| 239733 |
+
},
|
| 239734 |
+
{
|
| 239735 |
+
"epoch": 889.55,
|
| 239736 |
+
"learning_rate": 8.234458804523425e-06,
|
| 239737 |
+
"loss": 0.4116,
|
| 239738 |
+
"step": 110330
|
| 239739 |
+
},
|
| 239740 |
+
{
|
| 239741 |
+
"epoch": 889.59,
|
| 239742 |
+
"learning_rate": 8.234378029079161e-06,
|
| 239743 |
+
"loss": 0.9143,
|
| 239744 |
+
"step": 110335
|
| 239745 |
+
},
|
| 239746 |
+
{
|
| 239747 |
+
"epoch": 889.63,
|
| 239748 |
+
"learning_rate": 8.234297253634895e-06,
|
| 239749 |
+
"loss": 0.6231,
|
| 239750 |
+
"step": 110340
|
| 239751 |
+
},
|
| 239752 |
+
{
|
| 239753 |
+
"epoch": 889.67,
|
| 239754 |
+
"learning_rate": 8.234216478190631e-06,
|
| 239755 |
+
"loss": 0.2671,
|
| 239756 |
+
"step": 110345
|
| 239757 |
+
},
|
| 239758 |
+
{
|
| 239759 |
+
"epoch": 889.71,
|
| 239760 |
+
"learning_rate": 8.234135702746365e-06,
|
| 239761 |
+
"loss": 0.3068,
|
| 239762 |
+
"step": 110350
|
| 239763 |
+
},
|
| 239764 |
+
{
|
| 239765 |
+
"epoch": 889.76,
|
| 239766 |
+
"learning_rate": 8.234054927302101e-06,
|
| 239767 |
+
"loss": 0.4126,
|
| 239768 |
+
"step": 110355
|
| 239769 |
+
},
|
| 239770 |
+
{
|
| 239771 |
+
"epoch": 889.8,
|
| 239772 |
+
"learning_rate": 8.233974151857837e-06,
|
| 239773 |
+
"loss": 0.9831,
|
| 239774 |
+
"step": 110360
|
| 239775 |
+
},
|
| 239776 |
+
{
|
| 239777 |
+
"epoch": 889.84,
|
| 239778 |
+
"learning_rate": 8.233893376413571e-06,
|
| 239779 |
+
"loss": 0.6819,
|
| 239780 |
+
"step": 110365
|
| 239781 |
+
},
|
| 239782 |
+
{
|
| 239783 |
+
"epoch": 889.88,
|
| 239784 |
+
"learning_rate": 8.233812600969307e-06,
|
| 239785 |
+
"loss": 0.3047,
|
| 239786 |
+
"step": 110370
|
| 239787 |
+
},
|
| 239788 |
+
{
|
| 239789 |
+
"epoch": 889.92,
|
| 239790 |
+
"learning_rate": 8.233731825525041e-06,
|
| 239791 |
+
"loss": 0.3253,
|
| 239792 |
+
"step": 110375
|
| 239793 |
+
},
|
| 239794 |
+
{
|
| 239795 |
+
"epoch": 889.96,
|
| 239796 |
+
"learning_rate": 8.233651050080777e-06,
|
| 239797 |
+
"loss": 0.3878,
|
| 239798 |
+
"step": 110380
|
| 239799 |
+
},
|
| 239800 |
+
{
|
| 239801 |
+
"epoch": 890.0,
|
| 239802 |
+
"learning_rate": 8.233570274636511e-06,
|
| 239803 |
+
"loss": 0.9747,
|
| 239804 |
+
"step": 110385
|
| 239805 |
+
},
|
| 239806 |
+
{
|
| 239807 |
+
"epoch": 890.0,
|
| 239808 |
+
"eval_loss": 0.3920542299747467,
|
| 239809 |
+
"eval_runtime": 43.9158,
|
| 239810 |
+
"eval_samples_per_second": 19.082,
|
| 239811 |
+
"eval_steps_per_second": 0.615,
|
| 239812 |
+
"eval_wer": 0.17994765159226406,
|
| 239813 |
+
"step": 110385
|
| 239814 |
+
},
|
| 239815 |
+
{
|
| 239816 |
+
"epoch": 890.04,
|
| 239817 |
+
"learning_rate": 8.233489499192247e-06,
|
| 239818 |
+
"loss": 0.3625,
|
| 239819 |
+
"step": 110390
|
| 239820 |
+
},
|
| 239821 |
+
{
|
| 239822 |
+
"epoch": 890.08,
|
| 239823 |
+
"learning_rate": 8.233408723747981e-06,
|
| 239824 |
+
"loss": 0.2793,
|
| 239825 |
+
"step": 110395
|
| 239826 |
+
},
|
| 239827 |
+
{
|
| 239828 |
+
"epoch": 890.12,
|
| 239829 |
+
"learning_rate": 8.233327948303717e-06,
|
| 239830 |
+
"loss": 0.303,
|
| 239831 |
+
"step": 110400
|
| 239832 |
+
},
|
| 239833 |
+
{
|
| 239834 |
+
"epoch": 890.16,
|
| 239835 |
+
"learning_rate": 8.233247172859451e-06,
|
| 239836 |
+
"loss": 0.404,
|
| 239837 |
+
"step": 110405
|
| 239838 |
+
},
|
| 239839 |
+
{
|
| 239840 |
+
"epoch": 890.2,
|
| 239841 |
+
"learning_rate": 8.23318255250404e-06,
|
| 239842 |
+
"loss": 1.3125,
|
| 239843 |
+
"step": 110410
|
| 239844 |
+
},
|
| 239845 |
+
{
|
| 239846 |
+
"epoch": 890.24,
|
| 239847 |
+
"learning_rate": 8.233101777059774e-06,
|
| 239848 |
+
"loss": 0.4036,
|
| 239849 |
+
"step": 110415
|
| 239850 |
+
},
|
| 239851 |
+
{
|
| 239852 |
+
"epoch": 890.28,
|
| 239853 |
+
"learning_rate": 8.23302100161551e-06,
|
| 239854 |
+
"loss": 0.3023,
|
| 239855 |
+
"step": 110420
|
| 239856 |
+
},
|
| 239857 |
+
{
|
| 239858 |
+
"epoch": 890.32,
|
| 239859 |
+
"learning_rate": 8.232940226171244e-06,
|
| 239860 |
+
"loss": 0.331,
|
| 239861 |
+
"step": 110425
|
| 239862 |
+
},
|
| 239863 |
+
{
|
| 239864 |
+
"epoch": 890.36,
|
| 239865 |
+
"learning_rate": 8.23285945072698e-06,
|
| 239866 |
+
"loss": 0.4793,
|
| 239867 |
+
"step": 110430
|
| 239868 |
+
},
|
| 239869 |
+
{
|
| 239870 |
+
"epoch": 890.4,
|
| 239871 |
+
"learning_rate": 8.232778675282714e-06,
|
| 239872 |
+
"loss": 1.0189,
|
| 239873 |
+
"step": 110435
|
| 239874 |
+
},
|
| 239875 |
+
{
|
| 239876 |
+
"epoch": 890.44,
|
| 239877 |
+
"learning_rate": 8.23269789983845e-06,
|
| 239878 |
+
"loss": 0.278,
|
| 239879 |
+
"step": 110440
|
| 239880 |
+
},
|
| 239881 |
+
{
|
| 239882 |
+
"epoch": 890.48,
|
| 239883 |
+
"learning_rate": 8.232617124394184e-06,
|
| 239884 |
+
"loss": 0.2914,
|
| 239885 |
+
"step": 110445
|
| 239886 |
+
},
|
| 239887 |
+
{
|
| 239888 |
+
"epoch": 890.52,
|
| 239889 |
+
"learning_rate": 8.23253634894992e-06,
|
| 239890 |
+
"loss": 0.267,
|
| 239891 |
+
"step": 110450
|
| 239892 |
+
},
|
| 239893 |
+
{
|
| 239894 |
+
"epoch": 890.56,
|
| 239895 |
+
"learning_rate": 8.232455573505656e-06,
|
| 239896 |
+
"loss": 0.4424,
|
| 239897 |
+
"step": 110455
|
| 239898 |
+
},
|
| 239899 |
+
{
|
| 239900 |
+
"epoch": 890.6,
|
| 239901 |
+
"learning_rate": 8.23237479806139e-06,
|
| 239902 |
+
"loss": 1.4394,
|
| 239903 |
+
"step": 110460
|
| 239904 |
+
},
|
| 239905 |
+
{
|
| 239906 |
+
"epoch": 890.64,
|
| 239907 |
+
"learning_rate": 8.232294022617126e-06,
|
| 239908 |
+
"loss": 0.3042,
|
| 239909 |
+
"step": 110465
|
| 239910 |
+
},
|
| 239911 |
+
{
|
| 239912 |
+
"epoch": 890.68,
|
| 239913 |
+
"learning_rate": 8.23221324717286e-06,
|
| 239914 |
+
"loss": 0.2659,
|
| 239915 |
+
"step": 110470
|
| 239916 |
+
},
|
| 239917 |
+
{
|
| 239918 |
+
"epoch": 890.72,
|
| 239919 |
+
"learning_rate": 8.232132471728596e-06,
|
| 239920 |
+
"loss": 0.3487,
|
| 239921 |
+
"step": 110475
|
| 239922 |
+
},
|
| 239923 |
+
{
|
| 239924 |
+
"epoch": 890.76,
|
| 239925 |
+
"learning_rate": 8.23205169628433e-06,
|
| 239926 |
+
"loss": 0.4529,
|
| 239927 |
+
"step": 110480
|
| 239928 |
+
},
|
| 239929 |
+
{
|
| 239930 |
+
"epoch": 890.8,
|
| 239931 |
+
"learning_rate": 8.231970920840066e-06,
|
| 239932 |
+
"loss": 1.1291,
|
| 239933 |
+
"step": 110485
|
| 239934 |
+
},
|
| 239935 |
+
{
|
| 239936 |
+
"epoch": 890.84,
|
| 239937 |
+
"learning_rate": 8.2318901453958e-06,
|
| 239938 |
+
"loss": 0.3086,
|
| 239939 |
+
"step": 110490
|
| 239940 |
+
},
|
| 239941 |
+
{
|
| 239942 |
+
"epoch": 890.88,
|
| 239943 |
+
"learning_rate": 8.231809369951536e-06,
|
| 239944 |
+
"loss": 0.4488,
|
| 239945 |
+
"step": 110495
|
| 239946 |
+
},
|
| 239947 |
+
{
|
| 239948 |
+
"epoch": 890.92,
|
| 239949 |
+
"learning_rate": 8.23172859450727e-06,
|
| 239950 |
+
"loss": 0.3459,
|
| 239951 |
+
"step": 110500
|
| 239952 |
+
},
|
| 239953 |
+
{
|
| 239954 |
+
"epoch": 890.96,
|
| 239955 |
+
"learning_rate": 8.231647819063006e-06,
|
| 239956 |
+
"loss": 0.4926,
|
| 239957 |
+
"step": 110505
|
| 239958 |
+
},
|
| 239959 |
+
{
|
| 239960 |
+
"epoch": 891.0,
|
| 239961 |
+
"eval_loss": 0.43205010890960693,
|
| 239962 |
+
"eval_runtime": 42.0036,
|
| 239963 |
+
"eval_samples_per_second": 19.951,
|
| 239964 |
+
"eval_steps_per_second": 0.643,
|
| 239965 |
+
"eval_wer": 0.18726401394133024,
|
| 239966 |
+
"step": 110509
|
| 239967 |
+
},
|
| 239968 |
+
{
|
| 239969 |
+
"epoch": 884.01,
|
| 239970 |
+
"learning_rate": 8.23156704361874e-06,
|
| 239971 |
+
"loss": 0.465,
|
| 239972 |
+
"step": 110510
|
| 239973 |
+
},
|
| 239974 |
+
{
|
| 239975 |
+
"epoch": 884.05,
|
| 239976 |
+
"learning_rate": 8.231486268174476e-06,
|
| 239977 |
+
"loss": 0.3706,
|
| 239978 |
+
"step": 110515
|
| 239979 |
+
},
|
| 239980 |
+
{
|
| 239981 |
+
"epoch": 884.09,
|
| 239982 |
+
"learning_rate": 8.231405492730212e-06,
|
| 239983 |
+
"loss": 0.3154,
|
| 239984 |
+
"step": 110520
|
| 239985 |
+
},
|
| 239986 |
+
{
|
| 239987 |
+
"epoch": 884.13,
|
| 239988 |
+
"learning_rate": 8.231324717285946e-06,
|
| 239989 |
+
"loss": 0.2977,
|
| 239990 |
+
"step": 110525
|
| 239991 |
+
},
|
| 239992 |
+
{
|
| 239993 |
+
"epoch": 884.17,
|
| 239994 |
+
"learning_rate": 8.231243941841682e-06,
|
| 239995 |
+
"loss": 0.5078,
|
| 239996 |
+
"step": 110530
|
| 239997 |
+
},
|
| 239998 |
+
{
|
| 239999 |
+
"epoch": 884.21,
|
| 240000 |
+
"learning_rate": 8.231163166397416e-06,
|
| 240001 |
+
"loss": 1.1521,
|
| 240002 |
+
"step": 110535
|
| 240003 |
+
},
|
| 240004 |
+
{
|
| 240005 |
+
"epoch": 884.25,
|
| 240006 |
+
"learning_rate": 8.231082390953151e-06,
|
| 240007 |
+
"loss": 0.3121,
|
| 240008 |
+
"step": 110540
|
| 240009 |
+
},
|
| 240010 |
+
{
|
| 240011 |
+
"epoch": 884.29,
|
| 240012 |
+
"learning_rate": 8.231001615508886e-06,
|
| 240013 |
+
"loss": 0.273,
|
| 240014 |
+
"step": 110545
|
| 240015 |
+
},
|
| 240016 |
+
{
|
| 240017 |
+
"epoch": 884.33,
|
| 240018 |
+
"learning_rate": 8.230920840064621e-06,
|
| 240019 |
+
"loss": 0.321,
|
| 240020 |
+
"step": 110550
|
| 240021 |
+
},
|
| 240022 |
+
{
|
| 240023 |
+
"epoch": 884.37,
|
| 240024 |
+
"learning_rate": 8.230840064620356e-06,
|
| 240025 |
+
"loss": 0.5071,
|
| 240026 |
+
"step": 110555
|
| 240027 |
+
},
|
| 240028 |
+
{
|
| 240029 |
+
"epoch": 884.41,
|
| 240030 |
+
"learning_rate": 8.230759289176091e-06,
|
| 240031 |
+
"loss": 1.0542,
|
| 240032 |
+
"step": 110560
|
| 240033 |
+
},
|
| 240034 |
+
{
|
| 240035 |
+
"epoch": 884.45,
|
| 240036 |
+
"learning_rate": 8.230678513731826e-06,
|
| 240037 |
+
"loss": 0.3164,
|
| 240038 |
+
"step": 110565
|
| 240039 |
+
},
|
| 240040 |
+
{
|
| 240041 |
+
"epoch": 884.49,
|
| 240042 |
+
"learning_rate": 8.230597738287561e-06,
|
| 240043 |
+
"loss": 0.2888,
|
| 240044 |
+
"step": 110570
|
| 240045 |
+
},
|
| 240046 |
+
{
|
| 240047 |
+
"epoch": 884.53,
|
| 240048 |
+
"learning_rate": 8.230516962843297e-06,
|
| 240049 |
+
"loss": 0.3305,
|
| 240050 |
+
"step": 110575
|
| 240051 |
+
},
|
| 240052 |
+
{
|
| 240053 |
+
"epoch": 884.57,
|
| 240054 |
+
"learning_rate": 8.230436187399031e-06,
|
| 240055 |
+
"loss": 0.5111,
|
| 240056 |
+
"step": 110580
|
| 240057 |
+
},
|
| 240058 |
+
{
|
| 240059 |
+
"epoch": 884.61,
|
| 240060 |
+
"learning_rate": 8.230355411954767e-06,
|
| 240061 |
+
"loss": 1.1807,
|
| 240062 |
+
"step": 110585
|
| 240063 |
+
},
|
| 240064 |
+
{
|
| 240065 |
+
"epoch": 884.65,
|
| 240066 |
+
"learning_rate": 8.230274636510501e-06,
|
| 240067 |
+
"loss": 0.3135,
|
| 240068 |
+
"step": 110590
|
| 240069 |
+
},
|
| 240070 |
+
{
|
| 240071 |
+
"epoch": 884.69,
|
| 240072 |
+
"learning_rate": 8.230193861066237e-06,
|
| 240073 |
+
"loss": 0.3553,
|
| 240074 |
+
"step": 110595
|
| 240075 |
+
},
|
| 240076 |
+
{
|
| 240077 |
+
"epoch": 884.73,
|
| 240078 |
+
"learning_rate": 8.230113085621971e-06,
|
| 240079 |
+
"loss": 0.2976,
|
| 240080 |
+
"step": 110600
|
| 240081 |
+
},
|
| 240082 |
+
{
|
| 240083 |
+
"epoch": 884.77,
|
| 240084 |
+
"learning_rate": 8.230032310177707e-06,
|
| 240085 |
+
"loss": 0.5609,
|
| 240086 |
+
"step": 110605
|
| 240087 |
+
},
|
| 240088 |
+
{
|
| 240089 |
+
"epoch": 884.81,
|
| 240090 |
+
"learning_rate": 8.229951534733441e-06,
|
| 240091 |
+
"loss": 1.2691,
|
| 240092 |
+
"step": 110610
|
| 240093 |
+
},
|
| 240094 |
+
{
|
| 240095 |
+
"epoch": 884.85,
|
| 240096 |
+
"learning_rate": 8.229870759289177e-06,
|
| 240097 |
+
"loss": 0.2722,
|
| 240098 |
+
"step": 110615
|
| 240099 |
+
},
|
| 240100 |
+
{
|
| 240101 |
+
"epoch": 884.89,
|
| 240102 |
+
"learning_rate": 8.229789983844911e-06,
|
| 240103 |
+
"loss": 0.2593,
|
| 240104 |
+
"step": 110620
|
| 240105 |
+
},
|
| 240106 |
+
{
|
| 240107 |
+
"epoch": 884.93,
|
| 240108 |
+
"learning_rate": 8.229709208400647e-06,
|
| 240109 |
+
"loss": 0.3413,
|
| 240110 |
+
"step": 110625
|
| 240111 |
+
},
|
| 240112 |
+
{
|
| 240113 |
+
"epoch": 884.97,
|
| 240114 |
+
"learning_rate": 8.229628432956383e-06,
|
| 240115 |
+
"loss": 0.5821,
|
| 240116 |
+
"step": 110630
|
| 240117 |
+
},
|
| 240118 |
+
{
|
| 240119 |
+
"epoch": 885.0,
|
| 240120 |
+
"eval_loss": 0.3840881586074829,
|
| 240121 |
+
"eval_runtime": 42.9084,
|
| 240122 |
+
"eval_samples_per_second": 19.53,
|
| 240123 |
+
"eval_steps_per_second": 0.629,
|
| 240124 |
+
"eval_wer": 0.1874064091045223,
|
| 240125 |
+
"step": 110634
|
| 240126 |
+
},
|
| 240127 |
+
{
|
| 240128 |
+
"epoch": 892.01,
|
| 240129 |
+
"learning_rate": 8.229547657512117e-06,
|
| 240130 |
+
"loss": 0.343,
|
| 240131 |
+
"step": 110635
|
| 240132 |
+
},
|
| 240133 |
+
{
|
| 240134 |
+
"epoch": 892.05,
|
| 240135 |
+
"learning_rate": 8.229466882067853e-06,
|
| 240136 |
+
"loss": 0.361,
|
| 240137 |
+
"step": 110640
|
| 240138 |
+
},
|
| 240139 |
+
{
|
| 240140 |
+
"epoch": 892.09,
|
| 240141 |
+
"learning_rate": 8.229386106623587e-06,
|
| 240142 |
+
"loss": 0.2761,
|
| 240143 |
+
"step": 110645
|
| 240144 |
+
},
|
| 240145 |
+
{
|
| 240146 |
+
"epoch": 892.13,
|
| 240147 |
+
"learning_rate": 8.229305331179323e-06,
|
| 240148 |
+
"loss": 0.3229,
|
| 240149 |
+
"step": 110650
|
| 240150 |
+
},
|
| 240151 |
+
{
|
| 240152 |
+
"epoch": 892.17,
|
| 240153 |
+
"learning_rate": 8.229224555735057e-06,
|
| 240154 |
+
"loss": 0.6241,
|
| 240155 |
+
"step": 110655
|
| 240156 |
+
},
|
| 240157 |
+
{
|
| 240158 |
+
"epoch": 892.21,
|
| 240159 |
+
"learning_rate": 8.229143780290793e-06,
|
| 240160 |
+
"loss": 1.1542,
|
| 240161 |
+
"step": 110660
|
| 240162 |
+
},
|
| 240163 |
+
{
|
| 240164 |
+
"epoch": 892.25,
|
| 240165 |
+
"learning_rate": 8.229063004846527e-06,
|
| 240166 |
+
"loss": 0.3145,
|
| 240167 |
+
"step": 110665
|
| 240168 |
+
},
|
| 240169 |
+
{
|
| 240170 |
+
"epoch": 892.29,
|
| 240171 |
+
"learning_rate": 8.228982229402263e-06,
|
| 240172 |
+
"loss": 0.2945,
|
| 240173 |
+
"step": 110670
|
| 240174 |
+
},
|
| 240175 |
+
{
|
| 240176 |
+
"epoch": 892.33,
|
| 240177 |
+
"learning_rate": 8.228901453957997e-06,
|
| 240178 |
+
"loss": 0.3749,
|
| 240179 |
+
"step": 110675
|
| 240180 |
+
},
|
| 240181 |
+
{
|
| 240182 |
+
"epoch": 892.37,
|
| 240183 |
+
"learning_rate": 8.228820678513733e-06,
|
| 240184 |
+
"loss": 0.6141,
|
| 240185 |
+
"step": 110680
|
| 240186 |
+
},
|
| 240187 |
+
{
|
| 240188 |
+
"epoch": 892.41,
|
| 240189 |
+
"learning_rate": 8.228739903069467e-06,
|
| 240190 |
+
"loss": 1.1746,
|
| 240191 |
+
"step": 110685
|
| 240192 |
+
},
|
| 240193 |
+
{
|
| 240194 |
+
"epoch": 892.45,
|
| 240195 |
+
"learning_rate": 8.228659127625203e-06,
|
| 240196 |
+
"loss": 0.3046,
|
| 240197 |
+
"step": 110690
|
| 240198 |
+
},
|
| 240199 |
+
{
|
| 240200 |
+
"epoch": 892.49,
|
| 240201 |
+
"learning_rate": 8.228578352180939e-06,
|
| 240202 |
+
"loss": 0.2851,
|
| 240203 |
+
"step": 110695
|
| 240204 |
+
},
|
| 240205 |
+
{
|
| 240206 |
+
"epoch": 892.53,
|
| 240207 |
+
"learning_rate": 8.228497576736673e-06,
|
| 240208 |
+
"loss": 0.4006,
|
| 240209 |
+
"step": 110700
|
| 240210 |
+
},
|
| 240211 |
+
{
|
| 240212 |
+
"epoch": 892.57,
|
| 240213 |
+
"learning_rate": 8.228416801292409e-06,
|
| 240214 |
+
"loss": 0.4473,
|
| 240215 |
+
"step": 110705
|
| 240216 |
+
},
|
| 240217 |
+
{
|
| 240218 |
+
"epoch": 892.61,
|
| 240219 |
+
"learning_rate": 8.228336025848143e-06,
|
| 240220 |
+
"loss": 1.119,
|
| 240221 |
+
"step": 110710
|
| 240222 |
+
},
|
| 240223 |
+
{
|
| 240224 |
+
"epoch": 892.65,
|
| 240225 |
+
"learning_rate": 8.228255250403879e-06,
|
| 240226 |
+
"loss": 0.2655,
|
| 240227 |
+
"step": 110715
|
| 240228 |
+
},
|
| 240229 |
+
{
|
| 240230 |
+
"epoch": 892.69,
|
| 240231 |
+
"learning_rate": 8.228174474959613e-06,
|
| 240232 |
+
"loss": 0.2907,
|
| 240233 |
+
"step": 110720
|
| 240234 |
+
},
|
| 240235 |
+
{
|
| 240236 |
+
"epoch": 892.73,
|
| 240237 |
+
"learning_rate": 8.228093699515349e-06,
|
| 240238 |
+
"loss": 0.349,
|
| 240239 |
+
"step": 110725
|
| 240240 |
+
},
|
| 240241 |
+
{
|
| 240242 |
+
"epoch": 892.77,
|
| 240243 |
+
"learning_rate": 8.228012924071083e-06,
|
| 240244 |
+
"loss": 0.5328,
|
| 240245 |
+
"step": 110730
|
| 240246 |
+
},
|
| 240247 |
+
{
|
| 240248 |
+
"epoch": 892.81,
|
| 240249 |
+
"learning_rate": 8.227932148626819e-06,
|
| 240250 |
+
"loss": 1.2069,
|
| 240251 |
+
"step": 110735
|
| 240252 |
+
},
|
| 240253 |
+
{
|
| 240254 |
+
"epoch": 892.85,
|
| 240255 |
+
"learning_rate": 8.227851373182553e-06,
|
| 240256 |
+
"loss": 0.3149,
|
| 240257 |
+
"step": 110740
|
| 240258 |
+
},
|
| 240259 |
+
{
|
| 240260 |
+
"epoch": 892.89,
|
| 240261 |
+
"learning_rate": 8.227770597738289e-06,
|
| 240262 |
+
"loss": 0.2972,
|
| 240263 |
+
"step": 110745
|
| 240264 |
+
},
|
| 240265 |
+
{
|
| 240266 |
+
"epoch": 892.93,
|
| 240267 |
+
"learning_rate": 8.227689822294024e-06,
|
| 240268 |
+
"loss": 0.3142,
|
| 240269 |
+
"step": 110750
|
| 240270 |
+
},
|
| 240271 |
+
{
|
| 240272 |
+
"epoch": 892.97,
|
| 240273 |
+
"learning_rate": 8.227609046849758e-06,
|
| 240274 |
+
"loss": 0.5553,
|
| 240275 |
+
"step": 110755
|
| 240276 |
+
},
|
| 240277 |
+
{
|
| 240278 |
+
"epoch": 893.0,
|
| 240279 |
+
"eval_loss": 0.4035487771034241,
|
| 240280 |
+
"eval_runtime": 42.7896,
|
| 240281 |
+
"eval_samples_per_second": 19.561,
|
| 240282 |
+
"eval_steps_per_second": 0.631,
|
| 240283 |
+
"eval_wer": 0.19120699881376038,
|
| 240284 |
+
"step": 110758
|
| 240285 |
}
|
| 240286 |
],
|
| 240287 |
"max_steps": 620000,
|
| 240288 |
"num_train_epochs": 5000,
|
| 240289 |
+
"total_flos": 3.116997433357764e+20,
|
| 240290 |
"trial_name": null,
|
| 240291 |
"trial_params": null
|
| 240292 |
}
|
model-bin/finetune/base/{checkpoint-110137 → checkpoint-110758}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630043182.9793327/events.out.tfevents.1630043182.52f5c7e305a3.886.61
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e1f00bb5a396015c03550536597f9426facee5524b2058e21eef5a9a9ad67f5c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630043660.3624449/events.out.tfevents.1630043660.52f5c7e305a3.886.63
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3463fd5d9db3f95d5f021713cb17b26cb432db6ca5e492235f7288c5a6950c07
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630044141.0074263/events.out.tfevents.1630044141.52f5c7e305a3.886.65
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e45edba365f8e1dbb05b657fc77cabbd8e5ef19a9ddf7fd244375b2e71e6a56b
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630044614.4908102/events.out.tfevents.1630044614.52f5c7e305a3.886.67
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6e11f0f401b134564312eff1cb83a38b548e4acf57680ac5df891e74e5b7b4fe
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630045101.34449/events.out.tfevents.1630045101.52f5c7e305a3.886.69
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:82b80481d3c970eb85038dccc20c4fed87b3c74816df3b8c9214051dacfa8966
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630043182.52f5c7e305a3.886.60
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b39e70f29657ea9db2972c5f1ef5ba50d2da4f9dc1af961feb14e5f4ccc6a06e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630043660.52f5c7e305a3.886.62
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:22c9c5a419e0d25bff8f2289c2f4742c59eba61ae063658b165858b29931284a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630044141.52f5c7e305a3.886.64
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5dacc211fa503c671832b66ca3b056afc6fc8aa1e1a49d512814ed03e45e2a53
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630044614.52f5c7e305a3.886.66
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fcb4057910eb1c706c1330447351f4d77faa624e67b2f7a5bddc9fa0cdbda40b
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630045101.52f5c7e305a3.886.68
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fe202661207e4c2a7f9c064ab6a463111048e6c37f8d3aae6ec11268075c54e7
|
| 3 |
+
size 8622
|