"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/trainer_state.json +793 -4
- model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630252977.4699829/events.out.tfevents.1630252977.cc93b136ebf5.1086.219 +3 -0
- model-bin/finetune/base/log/1630253413.1850004/events.out.tfevents.1630253413.cc93b136ebf5.1086.221 +3 -0
- model-bin/finetune/base/log/1630253854.71079/events.out.tfevents.1630253854.cc93b136ebf5.1086.223 +3 -0
- model-bin/finetune/base/log/1630254282.042753/events.out.tfevents.1630254282.cc93b136ebf5.1086.225 +3 -0
- model-bin/finetune/base/log/1630254714.536368/events.out.tfevents.1630254714.cc93b136ebf5.1086.227 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630252977.cc93b136ebf5.1086.218 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630253413.cc93b136ebf5.1086.220 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630253854.cc93b136ebf5.1086.222 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630254282.cc93b136ebf5.1086.224 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630254714.cc93b136ebf5.1086.226 +3 -0
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/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 sha256:46ba5bc9ebc2d9cb5c039dedb6841c756b21ad0680230520f9190e7d9c8ee4d1
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/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 sha256:751bbf777839d159698b17abeaf9812e82ed6c85907fe60a3193f08e4f8a1146
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/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:eff6a839f60f322cdb5909f5073b57b6d1ded874a541816583c3d41ecfaf7177
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/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 sha256:054e7466e8875cd689e7563a044cae340ceb7830af56e188fce91c945fd5de6e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/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 sha256:11001b1c63c88e4fa645949c229ef32727a43e7b4b8e96937eec6a7dd2e7766e
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -293637,11 +293637,800 @@
|
|
| 293637 |
"eval_steps_per_second": 0.627,
|
| 293638 |
"eval_wer": 0.17048071387449626,
|
| 293639 |
"step": 152690
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 293640 |
}
|
| 293641 |
],
|
| 293642 |
-
"max_steps":
|
| 293643 |
"num_train_epochs": 5000,
|
| 293644 |
-
"total_flos": 4.
|
| 293645 |
"trial_name": null,
|
| 293646 |
"trial_params": null
|
| 293647 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1668011442822563,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-146596",
|
| 4 |
+
"epoch": 1235.995983935743,
|
| 5 |
+
"global_step": 153312,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 293637 |
"eval_steps_per_second": 0.627,
|
| 293638 |
"eval_wer": 0.17048071387449626,
|
| 293639 |
"step": 152690
|
| 293640 |
+
},
|
| 293641 |
+
{
|
| 293642 |
+
"epoch": 1231.04,
|
| 293643 |
+
"learning_rate": 7.5503069466882074e-06,
|
| 293644 |
+
"loss": 0.2499,
|
| 293645 |
+
"step": 152695
|
| 293646 |
+
},
|
| 293647 |
+
{
|
| 293648 |
+
"epoch": 1231.08,
|
| 293649 |
+
"learning_rate": 7.550226171243943e-06,
|
| 293650 |
+
"loss": 0.2478,
|
| 293651 |
+
"step": 152700
|
| 293652 |
+
},
|
| 293653 |
+
{
|
| 293654 |
+
"epoch": 1231.12,
|
| 293655 |
+
"learning_rate": 7.550145395799677e-06,
|
| 293656 |
+
"loss": 0.3172,
|
| 293657 |
+
"step": 152705
|
| 293658 |
+
},
|
| 293659 |
+
{
|
| 293660 |
+
"epoch": 1231.16,
|
| 293661 |
+
"learning_rate": 7.550064620355413e-06,
|
| 293662 |
+
"loss": 0.4319,
|
| 293663 |
+
"step": 152710
|
| 293664 |
+
},
|
| 293665 |
+
{
|
| 293666 |
+
"epoch": 1231.2,
|
| 293667 |
+
"learning_rate": 7.549983844911147e-06,
|
| 293668 |
+
"loss": 1.2308,
|
| 293669 |
+
"step": 152715
|
| 293670 |
+
},
|
| 293671 |
+
{
|
| 293672 |
+
"epoch": 1231.24,
|
| 293673 |
+
"learning_rate": 7.549903069466883e-06,
|
| 293674 |
+
"loss": 0.4007,
|
| 293675 |
+
"step": 152720
|
| 293676 |
+
},
|
| 293677 |
+
{
|
| 293678 |
+
"epoch": 1231.28,
|
| 293679 |
+
"learning_rate": 7.549822294022617e-06,
|
| 293680 |
+
"loss": 0.2828,
|
| 293681 |
+
"step": 152725
|
| 293682 |
+
},
|
| 293683 |
+
{
|
| 293684 |
+
"epoch": 1231.32,
|
| 293685 |
+
"learning_rate": 7.549741518578353e-06,
|
| 293686 |
+
"loss": 0.2824,
|
| 293687 |
+
"step": 152730
|
| 293688 |
+
},
|
| 293689 |
+
{
|
| 293690 |
+
"epoch": 1231.36,
|
| 293691 |
+
"learning_rate": 7.549660743134087e-06,
|
| 293692 |
+
"loss": 0.4561,
|
| 293693 |
+
"step": 152735
|
| 293694 |
+
},
|
| 293695 |
+
{
|
| 293696 |
+
"epoch": 1231.4,
|
| 293697 |
+
"learning_rate": 7.549579967689823e-06,
|
| 293698 |
+
"loss": 1.1774,
|
| 293699 |
+
"step": 152740
|
| 293700 |
+
},
|
| 293701 |
+
{
|
| 293702 |
+
"epoch": 1231.44,
|
| 293703 |
+
"learning_rate": 7.549499192245558e-06,
|
| 293704 |
+
"loss": 0.3069,
|
| 293705 |
+
"step": 152745
|
| 293706 |
+
},
|
| 293707 |
+
{
|
| 293708 |
+
"epoch": 1231.48,
|
| 293709 |
+
"learning_rate": 7.549418416801293e-06,
|
| 293710 |
+
"loss": 0.2546,
|
| 293711 |
+
"step": 152750
|
| 293712 |
+
},
|
| 293713 |
+
{
|
| 293714 |
+
"epoch": 1231.52,
|
| 293715 |
+
"learning_rate": 7.549337641357028e-06,
|
| 293716 |
+
"loss": 0.3069,
|
| 293717 |
+
"step": 152755
|
| 293718 |
+
},
|
| 293719 |
+
{
|
| 293720 |
+
"epoch": 1231.56,
|
| 293721 |
+
"learning_rate": 7.549256865912763e-06,
|
| 293722 |
+
"loss": 0.5205,
|
| 293723 |
+
"step": 152760
|
| 293724 |
+
},
|
| 293725 |
+
{
|
| 293726 |
+
"epoch": 1231.6,
|
| 293727 |
+
"learning_rate": 7.549176090468498e-06,
|
| 293728 |
+
"loss": 1.0464,
|
| 293729 |
+
"step": 152765
|
| 293730 |
+
},
|
| 293731 |
+
{
|
| 293732 |
+
"epoch": 1231.64,
|
| 293733 |
+
"learning_rate": 7.549095315024233e-06,
|
| 293734 |
+
"loss": 0.3116,
|
| 293735 |
+
"step": 152770
|
| 293736 |
+
},
|
| 293737 |
+
{
|
| 293738 |
+
"epoch": 1231.68,
|
| 293739 |
+
"learning_rate": 7.549014539579969e-06,
|
| 293740 |
+
"loss": 0.3058,
|
| 293741 |
+
"step": 152775
|
| 293742 |
+
},
|
| 293743 |
+
{
|
| 293744 |
+
"epoch": 1231.72,
|
| 293745 |
+
"learning_rate": 7.548933764135703e-06,
|
| 293746 |
+
"loss": 0.2748,
|
| 293747 |
+
"step": 152780
|
| 293748 |
+
},
|
| 293749 |
+
{
|
| 293750 |
+
"epoch": 1231.76,
|
| 293751 |
+
"learning_rate": 7.548852988691439e-06,
|
| 293752 |
+
"loss": 0.4329,
|
| 293753 |
+
"step": 152785
|
| 293754 |
+
},
|
| 293755 |
+
{
|
| 293756 |
+
"epoch": 1231.8,
|
| 293757 |
+
"learning_rate": 7.548772213247173e-06,
|
| 293758 |
+
"loss": 1.1518,
|
| 293759 |
+
"step": 152790
|
| 293760 |
+
},
|
| 293761 |
+
{
|
| 293762 |
+
"epoch": 1231.84,
|
| 293763 |
+
"learning_rate": 7.548691437802909e-06,
|
| 293764 |
+
"loss": 0.2424,
|
| 293765 |
+
"step": 152795
|
| 293766 |
+
},
|
| 293767 |
+
{
|
| 293768 |
+
"epoch": 1231.88,
|
| 293769 |
+
"learning_rate": 7.548610662358643e-06,
|
| 293770 |
+
"loss": 0.2967,
|
| 293771 |
+
"step": 152800
|
| 293772 |
+
},
|
| 293773 |
+
{
|
| 293774 |
+
"epoch": 1231.92,
|
| 293775 |
+
"learning_rate": 7.548529886914379e-06,
|
| 293776 |
+
"loss": 0.3013,
|
| 293777 |
+
"step": 152805
|
| 293778 |
+
},
|
| 293779 |
+
{
|
| 293780 |
+
"epoch": 1231.96,
|
| 293781 |
+
"learning_rate": 7.548449111470114e-06,
|
| 293782 |
+
"loss": 0.427,
|
| 293783 |
+
"step": 152810
|
| 293784 |
+
},
|
| 293785 |
+
{
|
| 293786 |
+
"epoch": 1232.0,
|
| 293787 |
+
"eval_loss": 0.3473464250564575,
|
| 293788 |
+
"eval_runtime": 40.4236,
|
| 293789 |
+
"eval_samples_per_second": 20.681,
|
| 293790 |
+
"eval_steps_per_second": 0.668,
|
| 293791 |
+
"eval_wer": 0.1835759611795446,
|
| 293792 |
+
"step": 152814
|
| 293793 |
+
},
|
| 293794 |
+
{
|
| 293795 |
+
"epoch": 1222.01,
|
| 293796 |
+
"learning_rate": 7.548368336025849e-06,
|
| 293797 |
+
"loss": 0.4101,
|
| 293798 |
+
"step": 152815
|
| 293799 |
+
},
|
| 293800 |
+
{
|
| 293801 |
+
"epoch": 1222.05,
|
| 293802 |
+
"learning_rate": 7.548287560581584e-06,
|
| 293803 |
+
"loss": 0.2653,
|
| 293804 |
+
"step": 152820
|
| 293805 |
+
},
|
| 293806 |
+
{
|
| 293807 |
+
"epoch": 1222.09,
|
| 293808 |
+
"learning_rate": 7.548206785137319e-06,
|
| 293809 |
+
"loss": 0.2575,
|
| 293810 |
+
"step": 152825
|
| 293811 |
+
},
|
| 293812 |
+
{
|
| 293813 |
+
"epoch": 1222.13,
|
| 293814 |
+
"learning_rate": 7.548126009693054e-06,
|
| 293815 |
+
"loss": 0.2939,
|
| 293816 |
+
"step": 152830
|
| 293817 |
+
},
|
| 293818 |
+
{
|
| 293819 |
+
"epoch": 1222.17,
|
| 293820 |
+
"learning_rate": 7.548045234248789e-06,
|
| 293821 |
+
"loss": 0.4049,
|
| 293822 |
+
"step": 152835
|
| 293823 |
+
},
|
| 293824 |
+
{
|
| 293825 |
+
"epoch": 1222.21,
|
| 293826 |
+
"learning_rate": 7.547964458804525e-06,
|
| 293827 |
+
"loss": 1.1922,
|
| 293828 |
+
"step": 152840
|
| 293829 |
+
},
|
| 293830 |
+
{
|
| 293831 |
+
"epoch": 1222.25,
|
| 293832 |
+
"learning_rate": 7.547883683360259e-06,
|
| 293833 |
+
"loss": 0.2904,
|
| 293834 |
+
"step": 152845
|
| 293835 |
+
},
|
| 293836 |
+
{
|
| 293837 |
+
"epoch": 1222.29,
|
| 293838 |
+
"learning_rate": 7.547802907915995e-06,
|
| 293839 |
+
"loss": 0.4204,
|
| 293840 |
+
"step": 152850
|
| 293841 |
+
},
|
| 293842 |
+
{
|
| 293843 |
+
"epoch": 1222.33,
|
| 293844 |
+
"learning_rate": 7.547722132471729e-06,
|
| 293845 |
+
"loss": 0.4682,
|
| 293846 |
+
"step": 152855
|
| 293847 |
+
},
|
| 293848 |
+
{
|
| 293849 |
+
"epoch": 1222.37,
|
| 293850 |
+
"learning_rate": 7.5476413570274646e-06,
|
| 293851 |
+
"loss": 0.5323,
|
| 293852 |
+
"step": 152860
|
| 293853 |
+
},
|
| 293854 |
+
{
|
| 293855 |
+
"epoch": 1222.41,
|
| 293856 |
+
"learning_rate": 7.5475605815831996e-06,
|
| 293857 |
+
"loss": 1.0863,
|
| 293858 |
+
"step": 152865
|
| 293859 |
+
},
|
| 293860 |
+
{
|
| 293861 |
+
"epoch": 1222.45,
|
| 293862 |
+
"learning_rate": 7.5474798061389345e-06,
|
| 293863 |
+
"loss": 0.3252,
|
| 293864 |
+
"step": 152870
|
| 293865 |
+
},
|
| 293866 |
+
{
|
| 293867 |
+
"epoch": 1222.49,
|
| 293868 |
+
"learning_rate": 7.5473990306946695e-06,
|
| 293869 |
+
"loss": 0.2891,
|
| 293870 |
+
"step": 152875
|
| 293871 |
+
},
|
| 293872 |
+
{
|
| 293873 |
+
"epoch": 1222.53,
|
| 293874 |
+
"learning_rate": 7.5473182552504045e-06,
|
| 293875 |
+
"loss": 0.442,
|
| 293876 |
+
"step": 152880
|
| 293877 |
+
},
|
| 293878 |
+
{
|
| 293879 |
+
"epoch": 1222.57,
|
| 293880 |
+
"learning_rate": 7.5472374798061395e-06,
|
| 293881 |
+
"loss": 0.4414,
|
| 293882 |
+
"step": 152885
|
| 293883 |
+
},
|
| 293884 |
+
{
|
| 293885 |
+
"epoch": 1222.61,
|
| 293886 |
+
"learning_rate": 7.5471567043618745e-06,
|
| 293887 |
+
"loss": 1.0553,
|
| 293888 |
+
"step": 152890
|
| 293889 |
+
},
|
| 293890 |
+
{
|
| 293891 |
+
"epoch": 1222.65,
|
| 293892 |
+
"learning_rate": 7.5470759289176095e-06,
|
| 293893 |
+
"loss": 0.2663,
|
| 293894 |
+
"step": 152895
|
| 293895 |
+
},
|
| 293896 |
+
{
|
| 293897 |
+
"epoch": 1222.69,
|
| 293898 |
+
"learning_rate": 7.5469951534733445e-06,
|
| 293899 |
+
"loss": 0.3036,
|
| 293900 |
+
"step": 152900
|
| 293901 |
+
},
|
| 293902 |
+
{
|
| 293903 |
+
"epoch": 1222.73,
|
| 293904 |
+
"learning_rate": 7.54691437802908e-06,
|
| 293905 |
+
"loss": 0.2624,
|
| 293906 |
+
"step": 152905
|
| 293907 |
+
},
|
| 293908 |
+
{
|
| 293909 |
+
"epoch": 1222.77,
|
| 293910 |
+
"learning_rate": 7.5468336025848144e-06,
|
| 293911 |
+
"loss": 0.5299,
|
| 293912 |
+
"step": 152910
|
| 293913 |
+
},
|
| 293914 |
+
{
|
| 293915 |
+
"epoch": 1222.81,
|
| 293916 |
+
"learning_rate": 7.54675282714055e-06,
|
| 293917 |
+
"loss": 1.1371,
|
| 293918 |
+
"step": 152915
|
| 293919 |
+
},
|
| 293920 |
+
{
|
| 293921 |
+
"epoch": 1222.85,
|
| 293922 |
+
"learning_rate": 7.546672051696285e-06,
|
| 293923 |
+
"loss": 0.2932,
|
| 293924 |
+
"step": 152920
|
| 293925 |
+
},
|
| 293926 |
+
{
|
| 293927 |
+
"epoch": 1222.89,
|
| 293928 |
+
"learning_rate": 7.54659127625202e-06,
|
| 293929 |
+
"loss": 0.3077,
|
| 293930 |
+
"step": 152925
|
| 293931 |
+
},
|
| 293932 |
+
{
|
| 293933 |
+
"epoch": 1222.93,
|
| 293934 |
+
"learning_rate": 7.546510500807755e-06,
|
| 293935 |
+
"loss": 0.3004,
|
| 293936 |
+
"step": 152930
|
| 293937 |
+
},
|
| 293938 |
+
{
|
| 293939 |
+
"epoch": 1222.97,
|
| 293940 |
+
"learning_rate": 7.54642972536349e-06,
|
| 293941 |
+
"loss": 0.5459,
|
| 293942 |
+
"step": 152935
|
| 293943 |
+
},
|
| 293944 |
+
{
|
| 293945 |
+
"epoch": 1223.0,
|
| 293946 |
+
"eval_loss": 0.35849061608314514,
|
| 293947 |
+
"eval_runtime": 40.5447,
|
| 293948 |
+
"eval_samples_per_second": 20.644,
|
| 293949 |
+
"eval_steps_per_second": 0.666,
|
| 293950 |
+
"eval_wer": 0.1767148683722655,
|
| 293951 |
+
"step": 152939
|
| 293952 |
+
},
|
| 293953 |
+
{
|
| 293954 |
+
"epoch": 1233.01,
|
| 293955 |
+
"learning_rate": 7.546348949919225e-06,
|
| 293956 |
+
"loss": 0.4571,
|
| 293957 |
+
"step": 152940
|
| 293958 |
+
},
|
| 293959 |
+
{
|
| 293960 |
+
"epoch": 1233.05,
|
| 293961 |
+
"learning_rate": 7.54626817447496e-06,
|
| 293962 |
+
"loss": 0.3129,
|
| 293963 |
+
"step": 152945
|
| 293964 |
+
},
|
| 293965 |
+
{
|
| 293966 |
+
"epoch": 1233.09,
|
| 293967 |
+
"learning_rate": 7.546187399030695e-06,
|
| 293968 |
+
"loss": 0.2331,
|
| 293969 |
+
"step": 152950
|
| 293970 |
+
},
|
| 293971 |
+
{
|
| 293972 |
+
"epoch": 1233.13,
|
| 293973 |
+
"learning_rate": 7.54610662358643e-06,
|
| 293974 |
+
"loss": 0.2974,
|
| 293975 |
+
"step": 152955
|
| 293976 |
+
},
|
| 293977 |
+
{
|
| 293978 |
+
"epoch": 1233.17,
|
| 293979 |
+
"learning_rate": 7.546025848142165e-06,
|
| 293980 |
+
"loss": 0.4769,
|
| 293981 |
+
"step": 152960
|
| 293982 |
+
},
|
| 293983 |
+
{
|
| 293984 |
+
"epoch": 1233.21,
|
| 293985 |
+
"learning_rate": 7.5459450726979e-06,
|
| 293986 |
+
"loss": 0.9929,
|
| 293987 |
+
"step": 152965
|
| 293988 |
+
},
|
| 293989 |
+
{
|
| 293990 |
+
"epoch": 1233.25,
|
| 293991 |
+
"learning_rate": 7.545864297253635e-06,
|
| 293992 |
+
"loss": 0.3494,
|
| 293993 |
+
"step": 152970
|
| 293994 |
+
},
|
| 293995 |
+
{
|
| 293996 |
+
"epoch": 1233.29,
|
| 293997 |
+
"learning_rate": 7.54578352180937e-06,
|
| 293998 |
+
"loss": 0.2682,
|
| 293999 |
+
"step": 152975
|
| 294000 |
+
},
|
| 294001 |
+
{
|
| 294002 |
+
"epoch": 1233.33,
|
| 294003 |
+
"learning_rate": 7.545702746365106e-06,
|
| 294004 |
+
"loss": 0.2763,
|
| 294005 |
+
"step": 152980
|
| 294006 |
+
},
|
| 294007 |
+
{
|
| 294008 |
+
"epoch": 1233.37,
|
| 294009 |
+
"learning_rate": 7.545621970920841e-06,
|
| 294010 |
+
"loss": 0.5025,
|
| 294011 |
+
"step": 152985
|
| 294012 |
+
},
|
| 294013 |
+
{
|
| 294014 |
+
"epoch": 1233.41,
|
| 294015 |
+
"learning_rate": 7.545541195476576e-06,
|
| 294016 |
+
"loss": 1.0533,
|
| 294017 |
+
"step": 152990
|
| 294018 |
+
},
|
| 294019 |
+
{
|
| 294020 |
+
"epoch": 1233.45,
|
| 294021 |
+
"learning_rate": 7.545460420032311e-06,
|
| 294022 |
+
"loss": 0.3089,
|
| 294023 |
+
"step": 152995
|
| 294024 |
+
},
|
| 294025 |
+
{
|
| 294026 |
+
"epoch": 1233.49,
|
| 294027 |
+
"learning_rate": 7.545379644588046e-06,
|
| 294028 |
+
"loss": 0.271,
|
| 294029 |
+
"step": 153000
|
| 294030 |
+
},
|
| 294031 |
+
{
|
| 294032 |
+
"epoch": 1233.53,
|
| 294033 |
+
"learning_rate": 7.545298869143781e-06,
|
| 294034 |
+
"loss": 0.2785,
|
| 294035 |
+
"step": 153005
|
| 294036 |
+
},
|
| 294037 |
+
{
|
| 294038 |
+
"epoch": 1233.57,
|
| 294039 |
+
"learning_rate": 7.545218093699516e-06,
|
| 294040 |
+
"loss": 0.4914,
|
| 294041 |
+
"step": 153010
|
| 294042 |
+
},
|
| 294043 |
+
{
|
| 294044 |
+
"epoch": 1233.61,
|
| 294045 |
+
"learning_rate": 7.545137318255251e-06,
|
| 294046 |
+
"loss": 1.182,
|
| 294047 |
+
"step": 153015
|
| 294048 |
+
},
|
| 294049 |
+
{
|
| 294050 |
+
"epoch": 1233.65,
|
| 294051 |
+
"learning_rate": 7.545056542810986e-06,
|
| 294052 |
+
"loss": 0.2868,
|
| 294053 |
+
"step": 153020
|
| 294054 |
+
},
|
| 294055 |
+
{
|
| 294056 |
+
"epoch": 1233.69,
|
| 294057 |
+
"learning_rate": 7.544975767366721e-06,
|
| 294058 |
+
"loss": 0.2879,
|
| 294059 |
+
"step": 153025
|
| 294060 |
+
},
|
| 294061 |
+
{
|
| 294062 |
+
"epoch": 1233.73,
|
| 294063 |
+
"learning_rate": 7.544894991922456e-06,
|
| 294064 |
+
"loss": 0.2492,
|
| 294065 |
+
"step": 153030
|
| 294066 |
+
},
|
| 294067 |
+
{
|
| 294068 |
+
"epoch": 1233.77,
|
| 294069 |
+
"learning_rate": 7.544814216478191e-06,
|
| 294070 |
+
"loss": 0.4884,
|
| 294071 |
+
"step": 153035
|
| 294072 |
+
},
|
| 294073 |
+
{
|
| 294074 |
+
"epoch": 1233.81,
|
| 294075 |
+
"learning_rate": 7.544733441033927e-06,
|
| 294076 |
+
"loss": 1.0584,
|
| 294077 |
+
"step": 153040
|
| 294078 |
+
},
|
| 294079 |
+
{
|
| 294080 |
+
"epoch": 1233.85,
|
| 294081 |
+
"learning_rate": 7.544652665589662e-06,
|
| 294082 |
+
"loss": 0.2708,
|
| 294083 |
+
"step": 153045
|
| 294084 |
+
},
|
| 294085 |
+
{
|
| 294086 |
+
"epoch": 1233.89,
|
| 294087 |
+
"learning_rate": 7.544571890145397e-06,
|
| 294088 |
+
"loss": 0.2559,
|
| 294089 |
+
"step": 153050
|
| 294090 |
+
},
|
| 294091 |
+
{
|
| 294092 |
+
"epoch": 1233.93,
|
| 294093 |
+
"learning_rate": 7.544491114701132e-06,
|
| 294094 |
+
"loss": 0.3759,
|
| 294095 |
+
"step": 153055
|
| 294096 |
+
},
|
| 294097 |
+
{
|
| 294098 |
+
"epoch": 1233.97,
|
| 294099 |
+
"learning_rate": 7.544410339256867e-06,
|
| 294100 |
+
"loss": 0.6144,
|
| 294101 |
+
"step": 153060
|
| 294102 |
+
},
|
| 294103 |
+
{
|
| 294104 |
+
"epoch": 1234.0,
|
| 294105 |
+
"eval_loss": 0.3701328635215759,
|
| 294106 |
+
"eval_runtime": 40.1133,
|
| 294107 |
+
"eval_samples_per_second": 20.866,
|
| 294108 |
+
"eval_steps_per_second": 0.673,
|
| 294109 |
+
"eval_wer": 0.17889839411676423,
|
| 294110 |
+
"step": 153063
|
| 294111 |
+
},
|
| 294112 |
+
{
|
| 294113 |
+
"epoch": 1224.02,
|
| 294114 |
+
"learning_rate": 7.544329563812602e-06,
|
| 294115 |
+
"loss": 0.3558,
|
| 294116 |
+
"step": 153065
|
| 294117 |
+
},
|
| 294118 |
+
{
|
| 294119 |
+
"epoch": 1224.06,
|
| 294120 |
+
"learning_rate": 7.544248788368337e-06,
|
| 294121 |
+
"loss": 0.2894,
|
| 294122 |
+
"step": 153070
|
| 294123 |
+
},
|
| 294124 |
+
{
|
| 294125 |
+
"epoch": 1224.1,
|
| 294126 |
+
"learning_rate": 7.5441680129240716e-06,
|
| 294127 |
+
"loss": 0.3037,
|
| 294128 |
+
"step": 153075
|
| 294129 |
+
},
|
| 294130 |
+
{
|
| 294131 |
+
"epoch": 1224.14,
|
| 294132 |
+
"learning_rate": 7.5440872374798066e-06,
|
| 294133 |
+
"loss": 0.2791,
|
| 294134 |
+
"step": 153080
|
| 294135 |
+
},
|
| 294136 |
+
{
|
| 294137 |
+
"epoch": 1224.18,
|
| 294138 |
+
"learning_rate": 7.5440064620355415e-06,
|
| 294139 |
+
"loss": 0.5449,
|
| 294140 |
+
"step": 153085
|
| 294141 |
+
},
|
| 294142 |
+
{
|
| 294143 |
+
"epoch": 1224.22,
|
| 294144 |
+
"learning_rate": 7.5439256865912765e-06,
|
| 294145 |
+
"loss": 0.9683,
|
| 294146 |
+
"step": 153090
|
| 294147 |
+
},
|
| 294148 |
+
{
|
| 294149 |
+
"epoch": 1224.26,
|
| 294150 |
+
"learning_rate": 7.543844911147012e-06,
|
| 294151 |
+
"loss": 0.324,
|
| 294152 |
+
"step": 153095
|
| 294153 |
+
},
|
| 294154 |
+
{
|
| 294155 |
+
"epoch": 1224.3,
|
| 294156 |
+
"learning_rate": 7.5437641357027465e-06,
|
| 294157 |
+
"loss": 0.282,
|
| 294158 |
+
"step": 153100
|
| 294159 |
+
},
|
| 294160 |
+
{
|
| 294161 |
+
"epoch": 1224.34,
|
| 294162 |
+
"learning_rate": 7.543683360258482e-06,
|
| 294163 |
+
"loss": 0.3395,
|
| 294164 |
+
"step": 153105
|
| 294165 |
+
},
|
| 294166 |
+
{
|
| 294167 |
+
"epoch": 1224.38,
|
| 294168 |
+
"learning_rate": 7.543602584814217e-06,
|
| 294169 |
+
"loss": 0.7002,
|
| 294170 |
+
"step": 153110
|
| 294171 |
+
},
|
| 294172 |
+
{
|
| 294173 |
+
"epoch": 1224.42,
|
| 294174 |
+
"learning_rate": 7.543521809369952e-06,
|
| 294175 |
+
"loss": 0.9072,
|
| 294176 |
+
"step": 153115
|
| 294177 |
+
},
|
| 294178 |
+
{
|
| 294179 |
+
"epoch": 1224.46,
|
| 294180 |
+
"learning_rate": 7.543441033925687e-06,
|
| 294181 |
+
"loss": 0.3604,
|
| 294182 |
+
"step": 153120
|
| 294183 |
+
},
|
| 294184 |
+
{
|
| 294185 |
+
"epoch": 1224.5,
|
| 294186 |
+
"learning_rate": 7.543360258481422e-06,
|
| 294187 |
+
"loss": 0.2485,
|
| 294188 |
+
"step": 153125
|
| 294189 |
+
},
|
| 294190 |
+
{
|
| 294191 |
+
"epoch": 1224.54,
|
| 294192 |
+
"learning_rate": 7.543279483037157e-06,
|
| 294193 |
+
"loss": 0.2851,
|
| 294194 |
+
"step": 153130
|
| 294195 |
+
},
|
| 294196 |
+
{
|
| 294197 |
+
"epoch": 1224.58,
|
| 294198 |
+
"learning_rate": 7.543198707592892e-06,
|
| 294199 |
+
"loss": 0.5522,
|
| 294200 |
+
"step": 153135
|
| 294201 |
+
},
|
| 294202 |
+
{
|
| 294203 |
+
"epoch": 1224.62,
|
| 294204 |
+
"learning_rate": 7.543117932148627e-06,
|
| 294205 |
+
"loss": 0.8635,
|
| 294206 |
+
"step": 153140
|
| 294207 |
+
},
|
| 294208 |
+
{
|
| 294209 |
+
"epoch": 1224.66,
|
| 294210 |
+
"learning_rate": 7.543037156704362e-06,
|
| 294211 |
+
"loss": 0.2533,
|
| 294212 |
+
"step": 153145
|
| 294213 |
+
},
|
| 294214 |
+
{
|
| 294215 |
+
"epoch": 1224.7,
|
| 294216 |
+
"learning_rate": 7.542956381260097e-06,
|
| 294217 |
+
"loss": 0.2469,
|
| 294218 |
+
"step": 153150
|
| 294219 |
+
},
|
| 294220 |
+
{
|
| 294221 |
+
"epoch": 1224.74,
|
| 294222 |
+
"learning_rate": 7.542875605815832e-06,
|
| 294223 |
+
"loss": 0.3325,
|
| 294224 |
+
"step": 153155
|
| 294225 |
+
},
|
| 294226 |
+
{
|
| 294227 |
+
"epoch": 1224.78,
|
| 294228 |
+
"learning_rate": 7.542794830371568e-06,
|
| 294229 |
+
"loss": 0.6239,
|
| 294230 |
+
"step": 153160
|
| 294231 |
+
},
|
| 294232 |
+
{
|
| 294233 |
+
"epoch": 1224.82,
|
| 294234 |
+
"learning_rate": 7.542714054927302e-06,
|
| 294235 |
+
"loss": 0.9895,
|
| 294236 |
+
"step": 153165
|
| 294237 |
+
},
|
| 294238 |
+
{
|
| 294239 |
+
"epoch": 1224.86,
|
| 294240 |
+
"learning_rate": 7.542633279483038e-06,
|
| 294241 |
+
"loss": 0.2356,
|
| 294242 |
+
"step": 153170
|
| 294243 |
+
},
|
| 294244 |
+
{
|
| 294245 |
+
"epoch": 1224.9,
|
| 294246 |
+
"learning_rate": 7.542552504038773e-06,
|
| 294247 |
+
"loss": 0.3209,
|
| 294248 |
+
"step": 153175
|
| 294249 |
+
},
|
| 294250 |
+
{
|
| 294251 |
+
"epoch": 1224.94,
|
| 294252 |
+
"learning_rate": 7.542471728594508e-06,
|
| 294253 |
+
"loss": 0.293,
|
| 294254 |
+
"step": 153180
|
| 294255 |
+
},
|
| 294256 |
+
{
|
| 294257 |
+
"epoch": 1224.98,
|
| 294258 |
+
"learning_rate": 7.542390953150243e-06,
|
| 294259 |
+
"loss": 0.5575,
|
| 294260 |
+
"step": 153185
|
| 294261 |
+
},
|
| 294262 |
+
{
|
| 294263 |
+
"epoch": 1225.0,
|
| 294264 |
+
"eval_loss": 0.4036131501197815,
|
| 294265 |
+
"eval_runtime": 41.1248,
|
| 294266 |
+
"eval_samples_per_second": 20.328,
|
| 294267 |
+
"eval_steps_per_second": 0.657,
|
| 294268 |
+
"eval_wer": 0.18242702191162555,
|
| 294269 |
+
"step": 153188
|
| 294270 |
+
},
|
| 294271 |
+
{
|
| 294272 |
+
"epoch": 1235.02,
|
| 294273 |
+
"learning_rate": 7.542310177705978e-06,
|
| 294274 |
+
"loss": 0.3768,
|
| 294275 |
+
"step": 153190
|
| 294276 |
+
},
|
| 294277 |
+
{
|
| 294278 |
+
"epoch": 1235.06,
|
| 294279 |
+
"learning_rate": 7.542229402261713e-06,
|
| 294280 |
+
"loss": 0.3256,
|
| 294281 |
+
"step": 153195
|
| 294282 |
+
},
|
| 294283 |
+
{
|
| 294284 |
+
"epoch": 1235.1,
|
| 294285 |
+
"learning_rate": 7.542148626817448e-06,
|
| 294286 |
+
"loss": 0.2672,
|
| 294287 |
+
"step": 153200
|
| 294288 |
+
},
|
| 294289 |
+
{
|
| 294290 |
+
"epoch": 1235.14,
|
| 294291 |
+
"learning_rate": 7.542067851373183e-06,
|
| 294292 |
+
"loss": 0.4272,
|
| 294293 |
+
"step": 153205
|
| 294294 |
+
},
|
| 294295 |
+
{
|
| 294296 |
+
"epoch": 1235.18,
|
| 294297 |
+
"learning_rate": 7.541987075928918e-06,
|
| 294298 |
+
"loss": 0.6648,
|
| 294299 |
+
"step": 153210
|
| 294300 |
+
},
|
| 294301 |
+
{
|
| 294302 |
+
"epoch": 1235.22,
|
| 294303 |
+
"learning_rate": 7.541906300484654e-06,
|
| 294304 |
+
"loss": 1.0655,
|
| 294305 |
+
"step": 153215
|
| 294306 |
+
},
|
| 294307 |
+
{
|
| 294308 |
+
"epoch": 1235.26,
|
| 294309 |
+
"learning_rate": 7.541825525040388e-06,
|
| 294310 |
+
"loss": 0.277,
|
| 294311 |
+
"step": 153220
|
| 294312 |
+
},
|
| 294313 |
+
{
|
| 294314 |
+
"epoch": 1235.3,
|
| 294315 |
+
"learning_rate": 7.541744749596124e-06,
|
| 294316 |
+
"loss": 0.3115,
|
| 294317 |
+
"step": 153225
|
| 294318 |
+
},
|
| 294319 |
+
{
|
| 294320 |
+
"epoch": 1235.34,
|
| 294321 |
+
"learning_rate": 7.541663974151858e-06,
|
| 294322 |
+
"loss": 0.3355,
|
| 294323 |
+
"step": 153230
|
| 294324 |
+
},
|
| 294325 |
+
{
|
| 294326 |
+
"epoch": 1235.38,
|
| 294327 |
+
"learning_rate": 7.541583198707594e-06,
|
| 294328 |
+
"loss": 0.5795,
|
| 294329 |
+
"step": 153235
|
| 294330 |
+
},
|
| 294331 |
+
{
|
| 294332 |
+
"epoch": 1235.42,
|
| 294333 |
+
"learning_rate": 7.541502423263328e-06,
|
| 294334 |
+
"loss": 0.9555,
|
| 294335 |
+
"step": 153240
|
| 294336 |
+
},
|
| 294337 |
+
{
|
| 294338 |
+
"epoch": 1235.46,
|
| 294339 |
+
"learning_rate": 7.541421647819064e-06,
|
| 294340 |
+
"loss": 0.247,
|
| 294341 |
+
"step": 153245
|
| 294342 |
+
},
|
| 294343 |
+
{
|
| 294344 |
+
"epoch": 1235.5,
|
| 294345 |
+
"learning_rate": 7.541340872374799e-06,
|
| 294346 |
+
"loss": 0.3165,
|
| 294347 |
+
"step": 153250
|
| 294348 |
+
},
|
| 294349 |
+
{
|
| 294350 |
+
"epoch": 1235.54,
|
| 294351 |
+
"learning_rate": 7.541260096930534e-06,
|
| 294352 |
+
"loss": 0.3246,
|
| 294353 |
+
"step": 153255
|
| 294354 |
+
},
|
| 294355 |
+
{
|
| 294356 |
+
"epoch": 1235.58,
|
| 294357 |
+
"learning_rate": 7.541179321486269e-06,
|
| 294358 |
+
"loss": 0.6086,
|
| 294359 |
+
"step": 153260
|
| 294360 |
+
},
|
| 294361 |
+
{
|
| 294362 |
+
"epoch": 1235.62,
|
| 294363 |
+
"learning_rate": 7.541114701130857e-06,
|
| 294364 |
+
"loss": 1.0104,
|
| 294365 |
+
"step": 153265
|
| 294366 |
+
},
|
| 294367 |
+
{
|
| 294368 |
+
"epoch": 1235.66,
|
| 294369 |
+
"learning_rate": 7.541033925686592e-06,
|
| 294370 |
+
"loss": 0.3032,
|
| 294371 |
+
"step": 153270
|
| 294372 |
+
},
|
| 294373 |
+
{
|
| 294374 |
+
"epoch": 1235.7,
|
| 294375 |
+
"learning_rate": 7.540953150242328e-06,
|
| 294376 |
+
"loss": 0.2819,
|
| 294377 |
+
"step": 153275
|
| 294378 |
+
},
|
| 294379 |
+
{
|
| 294380 |
+
"epoch": 1235.74,
|
| 294381 |
+
"learning_rate": 7.540872374798062e-06,
|
| 294382 |
+
"loss": 0.3722,
|
| 294383 |
+
"step": 153280
|
| 294384 |
+
},
|
| 294385 |
+
{
|
| 294386 |
+
"epoch": 1235.78,
|
| 294387 |
+
"learning_rate": 7.540791599353798e-06,
|
| 294388 |
+
"loss": 0.5377,
|
| 294389 |
+
"step": 153285
|
| 294390 |
+
},
|
| 294391 |
+
{
|
| 294392 |
+
"epoch": 1235.82,
|
| 294393 |
+
"learning_rate": 7.540710823909532e-06,
|
| 294394 |
+
"loss": 0.9107,
|
| 294395 |
+
"step": 153290
|
| 294396 |
+
},
|
| 294397 |
+
{
|
| 294398 |
+
"epoch": 1235.86,
|
| 294399 |
+
"learning_rate": 7.540630048465268e-06,
|
| 294400 |
+
"loss": 0.2784,
|
| 294401 |
+
"step": 153295
|
| 294402 |
+
},
|
| 294403 |
+
{
|
| 294404 |
+
"epoch": 1235.9,
|
| 294405 |
+
"learning_rate": 7.540549273021002e-06,
|
| 294406 |
+
"loss": 0.4669,
|
| 294407 |
+
"step": 153300
|
| 294408 |
+
},
|
| 294409 |
+
{
|
| 294410 |
+
"epoch": 1235.94,
|
| 294411 |
+
"learning_rate": 7.5404684975767376e-06,
|
| 294412 |
+
"loss": 0.3216,
|
| 294413 |
+
"step": 153305
|
| 294414 |
+
},
|
| 294415 |
+
{
|
| 294416 |
+
"epoch": 1235.98,
|
| 294417 |
+
"learning_rate": 7.5403877221324725e-06,
|
| 294418 |
+
"loss": 0.7395,
|
| 294419 |
+
"step": 153310
|
| 294420 |
+
},
|
| 294421 |
+
{
|
| 294422 |
+
"epoch": 1236.0,
|
| 294423 |
+
"eval_loss": 0.4267891049385071,
|
| 294424 |
+
"eval_runtime": 41.6163,
|
| 294425 |
+
"eval_samples_per_second": 20.088,
|
| 294426 |
+
"eval_steps_per_second": 0.649,
|
| 294427 |
+
"eval_wer": 0.18537154524903376,
|
| 294428 |
+
"step": 153312
|
| 294429 |
}
|
| 294430 |
],
|
| 294431 |
+
"max_steps": 620000,
|
| 294432 |
"num_train_epochs": 5000,
|
| 294433 |
+
"total_flos": 4.3143346529907927e+20,
|
| 294434 |
"trial_name": null,
|
| 294435 |
"trial_params": null
|
| 294436 |
}
|
model-bin/finetune/base/{checkpoint-152690 → checkpoint-153312}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630252977.4699829/events.out.tfevents.1630252977.cc93b136ebf5.1086.219
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fce9bb6550dccfa92fd5070248c1eaa0adea6090228719306318bd35b393a7c1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630253413.1850004/events.out.tfevents.1630253413.cc93b136ebf5.1086.221
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fcea54c33cc78347bd664ed7c545b006b39afae6d2f4c3938c518b8329b28413
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630253854.71079/events.out.tfevents.1630253854.cc93b136ebf5.1086.223
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb4e577d8a61a76383dcc972b4f6525bf89977d277aa96d3919390a96cf688c9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630254282.042753/events.out.tfevents.1630254282.cc93b136ebf5.1086.225
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cccea25948a48b81c97d539880e4bff20b7201e890f3e60e28007074868d0389
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630254714.536368/events.out.tfevents.1630254714.cc93b136ebf5.1086.227
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb5faa028afecfaee81f4f30b722dc5ea2e181d02a29038dce872c8139f4a5a0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630252977.cc93b136ebf5.1086.218
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e0dcc183616aa7c66d93997c2d0ca7ba1c5707044486378177096a43f9e5d806
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630253413.cc93b136ebf5.1086.220
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:54994f1e36b6ac1cd95dfcf6f8cdbd35f46d295cf24bfb8ca94ff3178845926a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630253854.cc93b136ebf5.1086.222
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2cafc06fc47e8533d459af37e7fdcf4c5d6ecad960d99266a01edfe7a2035280
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630254282.cc93b136ebf5.1086.224
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eccb18ba1edf2eb52cf7e2b4d06a3239eb6c88d4d0cbf250abf6a9b148936020
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630254714.cc93b136ebf5.1086.226
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:774cf42f53da5f8c1a19ed1311290dc3788db045704b96a17f3f90c4964a3fad
|
| 3 |
+
size 8622
|