"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132786}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132786}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132786}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132786}/trainer_state.json +639 -3
- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/training_args.bin +0 -0
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132910}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132910}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132910}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-131665 → checkpoint-132910}/trainer_state.json +1589 -5
- model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630169499.9794455/events.out.tfevents.1630169499.86bb0ddabf9b.4092.231 +3 -0
- model-bin/finetune/base/log/1630169887.7954714/events.out.tfevents.1630169887.86bb0ddabf9b.4092.233 +3 -0
- model-bin/finetune/base/log/1630170280.3267503/events.out.tfevents.1630170280.86bb0ddabf9b.4092.235 +3 -0
- model-bin/finetune/base/log/1630170670.464988/events.out.tfevents.1630170670.86bb0ddabf9b.4092.237 +3 -0
- model-bin/finetune/base/log/1630171053.185435/events.out.tfevents.1630171053.86bb0ddabf9b.4092.239 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630169499.86bb0ddabf9b.4092.230 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630169887.86bb0ddabf9b.4092.232 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630170280.86bb0ddabf9b.4092.234 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630170670.86bb0ddabf9b.4092.236 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630171053.86bb0ddabf9b.4092.238 +3 -0
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/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 sha256:51b13415c4cf5cf4cf520e1b8071a72b26c08263016d8735a793416828e064ea
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132786}/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:d7a1fa6198f9700ab7e1d873b3fb9d5ac3af310b88ba214798c9f692a4d74a61
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132786}/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:c88467082976a02e20231815824795350bcda166a59d72227ed994c426d37238
|
| 3 |
+
size 14439
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132786}/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 sha256:03a362456e0ece54c924c9cc7ffe141eb2a7a3a957183267e18b8ead00a752c8
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/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 sha256:47857352ca6f8a4f7a59ec1ed2d9a5619d17c69643fc867991f0eedb8af98a94
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132786}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17062187276626162,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-131665",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -267675,11 +267675,647 @@
|
|
| 267675 |
"eval_steps_per_second": 0.737,
|
| 267676 |
"eval_wer": 0.18639436214583632,
|
| 267677 |
"step": 132288
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 267678 |
}
|
| 267679 |
],
|
| 267680 |
"max_steps": 625000,
|
| 267681 |
"num_train_epochs": 5000,
|
| 267682 |
-
"total_flos": 3.
|
| 267683 |
"trial_name": null,
|
| 267684 |
"trial_params": null
|
| 267685 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17062187276626162,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-131665",
|
| 4 |
+
"epoch": 1062.0,
|
| 5 |
+
"global_step": 132786,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 267675 |
"eval_steps_per_second": 0.737,
|
| 267676 |
"eval_wer": 0.18639436214583632,
|
| 267677 |
"step": 132288
|
| 267678 |
+
},
|
| 267679 |
+
{
|
| 267680 |
+
"epoch": 1066.02,
|
| 267681 |
+
"learning_rate": 7.896826923076924e-06,
|
| 267682 |
+
"loss": 0.362,
|
| 267683 |
+
"step": 132290
|
| 267684 |
+
},
|
| 267685 |
+
{
|
| 267686 |
+
"epoch": 1066.06,
|
| 267687 |
+
"learning_rate": 7.896746794871795e-06,
|
| 267688 |
+
"loss": 0.4653,
|
| 267689 |
+
"step": 132295
|
| 267690 |
+
},
|
| 267691 |
+
{
|
| 267692 |
+
"epoch": 1066.1,
|
| 267693 |
+
"learning_rate": 7.896666666666667e-06,
|
| 267694 |
+
"loss": 0.3126,
|
| 267695 |
+
"step": 132300
|
| 267696 |
+
},
|
| 267697 |
+
{
|
| 267698 |
+
"epoch": 1066.14,
|
| 267699 |
+
"learning_rate": 7.89658653846154e-06,
|
| 267700 |
+
"loss": 0.4131,
|
| 267701 |
+
"step": 132305
|
| 267702 |
+
},
|
| 267703 |
+
{
|
| 267704 |
+
"epoch": 1066.18,
|
| 267705 |
+
"learning_rate": 7.896506410256411e-06,
|
| 267706 |
+
"loss": 0.671,
|
| 267707 |
+
"step": 132310
|
| 267708 |
+
},
|
| 267709 |
+
{
|
| 267710 |
+
"epoch": 1066.22,
|
| 267711 |
+
"learning_rate": 7.896426282051282e-06,
|
| 267712 |
+
"loss": 1.0433,
|
| 267713 |
+
"step": 132315
|
| 267714 |
+
},
|
| 267715 |
+
{
|
| 267716 |
+
"epoch": 1066.26,
|
| 267717 |
+
"learning_rate": 7.896346153846156e-06,
|
| 267718 |
+
"loss": 0.2792,
|
| 267719 |
+
"step": 132320
|
| 267720 |
+
},
|
| 267721 |
+
{
|
| 267722 |
+
"epoch": 1066.3,
|
| 267723 |
+
"learning_rate": 7.896266025641027e-06,
|
| 267724 |
+
"loss": 0.2856,
|
| 267725 |
+
"step": 132325
|
| 267726 |
+
},
|
| 267727 |
+
{
|
| 267728 |
+
"epoch": 1066.34,
|
| 267729 |
+
"learning_rate": 7.896185897435898e-06,
|
| 267730 |
+
"loss": 0.3185,
|
| 267731 |
+
"step": 132330
|
| 267732 |
+
},
|
| 267733 |
+
{
|
| 267734 |
+
"epoch": 1066.38,
|
| 267735 |
+
"learning_rate": 7.89610576923077e-06,
|
| 267736 |
+
"loss": 0.5152,
|
| 267737 |
+
"step": 132335
|
| 267738 |
+
},
|
| 267739 |
+
{
|
| 267740 |
+
"epoch": 1066.42,
|
| 267741 |
+
"learning_rate": 7.896025641025643e-06,
|
| 267742 |
+
"loss": 1.2072,
|
| 267743 |
+
"step": 132340
|
| 267744 |
+
},
|
| 267745 |
+
{
|
| 267746 |
+
"epoch": 1066.46,
|
| 267747 |
+
"learning_rate": 7.895945512820512e-06,
|
| 267748 |
+
"loss": 0.2893,
|
| 267749 |
+
"step": 132345
|
| 267750 |
+
},
|
| 267751 |
+
{
|
| 267752 |
+
"epoch": 1066.5,
|
| 267753 |
+
"learning_rate": 7.895865384615385e-06,
|
| 267754 |
+
"loss": 0.3008,
|
| 267755 |
+
"step": 132350
|
| 267756 |
+
},
|
| 267757 |
+
{
|
| 267758 |
+
"epoch": 1066.54,
|
| 267759 |
+
"learning_rate": 7.895785256410257e-06,
|
| 267760 |
+
"loss": 0.3702,
|
| 267761 |
+
"step": 132355
|
| 267762 |
+
},
|
| 267763 |
+
{
|
| 267764 |
+
"epoch": 1066.58,
|
| 267765 |
+
"learning_rate": 7.895705128205128e-06,
|
| 267766 |
+
"loss": 0.5822,
|
| 267767 |
+
"step": 132360
|
| 267768 |
+
},
|
| 267769 |
+
{
|
| 267770 |
+
"epoch": 1066.62,
|
| 267771 |
+
"learning_rate": 7.895641025641027e-06,
|
| 267772 |
+
"loss": 1.1017,
|
| 267773 |
+
"step": 132365
|
| 267774 |
+
},
|
| 267775 |
+
{
|
| 267776 |
+
"epoch": 1066.66,
|
| 267777 |
+
"learning_rate": 7.895560897435897e-06,
|
| 267778 |
+
"loss": 0.2476,
|
| 267779 |
+
"step": 132370
|
| 267780 |
+
},
|
| 267781 |
+
{
|
| 267782 |
+
"epoch": 1066.7,
|
| 267783 |
+
"learning_rate": 7.89548076923077e-06,
|
| 267784 |
+
"loss": 0.4423,
|
| 267785 |
+
"step": 132375
|
| 267786 |
+
},
|
| 267787 |
+
{
|
| 267788 |
+
"epoch": 1066.74,
|
| 267789 |
+
"learning_rate": 7.895400641025641e-06,
|
| 267790 |
+
"loss": 0.371,
|
| 267791 |
+
"step": 132380
|
| 267792 |
+
},
|
| 267793 |
+
{
|
| 267794 |
+
"epoch": 1066.78,
|
| 267795 |
+
"learning_rate": 7.895320512820513e-06,
|
| 267796 |
+
"loss": 0.6137,
|
| 267797 |
+
"step": 132385
|
| 267798 |
+
},
|
| 267799 |
+
{
|
| 267800 |
+
"epoch": 1066.82,
|
| 267801 |
+
"learning_rate": 7.895240384615386e-06,
|
| 267802 |
+
"loss": 1.058,
|
| 267803 |
+
"step": 132390
|
| 267804 |
+
},
|
| 267805 |
+
{
|
| 267806 |
+
"epoch": 1066.86,
|
| 267807 |
+
"learning_rate": 7.895160256410257e-06,
|
| 267808 |
+
"loss": 0.3168,
|
| 267809 |
+
"step": 132395
|
| 267810 |
+
},
|
| 267811 |
+
{
|
| 267812 |
+
"epoch": 1066.9,
|
| 267813 |
+
"learning_rate": 7.895080128205128e-06,
|
| 267814 |
+
"loss": 0.3264,
|
| 267815 |
+
"step": 132400
|
| 267816 |
+
},
|
| 267817 |
+
{
|
| 267818 |
+
"epoch": 1066.94,
|
| 267819 |
+
"learning_rate": 7.895e-06,
|
| 267820 |
+
"loss": 0.3443,
|
| 267821 |
+
"step": 132405
|
| 267822 |
+
},
|
| 267823 |
+
{
|
| 267824 |
+
"epoch": 1066.98,
|
| 267825 |
+
"learning_rate": 7.894919871794873e-06,
|
| 267826 |
+
"loss": 0.5947,
|
| 267827 |
+
"step": 132410
|
| 267828 |
+
},
|
| 267829 |
+
{
|
| 267830 |
+
"epoch": 1067.0,
|
| 267831 |
+
"eval_loss": 0.3375318646430969,
|
| 267832 |
+
"eval_runtime": 36.5033,
|
| 267833 |
+
"eval_samples_per_second": 22.875,
|
| 267834 |
+
"eval_steps_per_second": 0.74,
|
| 267835 |
+
"eval_wer": 0.17405509764609786,
|
| 267836 |
+
"step": 132412
|
| 267837 |
+
},
|
| 267838 |
+
{
|
| 267839 |
+
"epoch": 1059.02,
|
| 267840 |
+
"learning_rate": 7.894839743589744e-06,
|
| 267841 |
+
"loss": 0.3428,
|
| 267842 |
+
"step": 132415
|
| 267843 |
+
},
|
| 267844 |
+
{
|
| 267845 |
+
"epoch": 1059.06,
|
| 267846 |
+
"learning_rate": 7.894759615384616e-06,
|
| 267847 |
+
"loss": 0.2433,
|
| 267848 |
+
"step": 132420
|
| 267849 |
+
},
|
| 267850 |
+
{
|
| 267851 |
+
"epoch": 1059.1,
|
| 267852 |
+
"learning_rate": 7.894679487179487e-06,
|
| 267853 |
+
"loss": 0.2733,
|
| 267854 |
+
"step": 132425
|
| 267855 |
+
},
|
| 267856 |
+
{
|
| 267857 |
+
"epoch": 1059.14,
|
| 267858 |
+
"learning_rate": 7.89459935897436e-06,
|
| 267859 |
+
"loss": 0.471,
|
| 267860 |
+
"step": 132430
|
| 267861 |
+
},
|
| 267862 |
+
{
|
| 267863 |
+
"epoch": 1059.18,
|
| 267864 |
+
"learning_rate": 7.894519230769231e-06,
|
| 267865 |
+
"loss": 0.8245,
|
| 267866 |
+
"step": 132435
|
| 267867 |
+
},
|
| 267868 |
+
{
|
| 267869 |
+
"epoch": 1059.22,
|
| 267870 |
+
"learning_rate": 7.894439102564103e-06,
|
| 267871 |
+
"loss": 0.874,
|
| 267872 |
+
"step": 132440
|
| 267873 |
+
},
|
| 267874 |
+
{
|
| 267875 |
+
"epoch": 1059.26,
|
| 267876 |
+
"learning_rate": 7.894358974358976e-06,
|
| 267877 |
+
"loss": 0.2662,
|
| 267878 |
+
"step": 132445
|
| 267879 |
+
},
|
| 267880 |
+
{
|
| 267881 |
+
"epoch": 1059.3,
|
| 267882 |
+
"learning_rate": 7.894278846153847e-06,
|
| 267883 |
+
"loss": 0.3069,
|
| 267884 |
+
"step": 132450
|
| 267885 |
+
},
|
| 267886 |
+
{
|
| 267887 |
+
"epoch": 1059.34,
|
| 267888 |
+
"learning_rate": 7.894198717948718e-06,
|
| 267889 |
+
"loss": 0.5776,
|
| 267890 |
+
"step": 132455
|
| 267891 |
+
},
|
| 267892 |
+
{
|
| 267893 |
+
"epoch": 1059.38,
|
| 267894 |
+
"learning_rate": 7.89411858974359e-06,
|
| 267895 |
+
"loss": 0.6466,
|
| 267896 |
+
"step": 132460
|
| 267897 |
+
},
|
| 267898 |
+
{
|
| 267899 |
+
"epoch": 1059.42,
|
| 267900 |
+
"learning_rate": 7.894038461538463e-06,
|
| 267901 |
+
"loss": 0.8247,
|
| 267902 |
+
"step": 132465
|
| 267903 |
+
},
|
| 267904 |
+
{
|
| 267905 |
+
"epoch": 1059.46,
|
| 267906 |
+
"learning_rate": 7.893958333333334e-06,
|
| 267907 |
+
"loss": 0.2522,
|
| 267908 |
+
"step": 132470
|
| 267909 |
+
},
|
| 267910 |
+
{
|
| 267911 |
+
"epoch": 1059.5,
|
| 267912 |
+
"learning_rate": 7.893878205128206e-06,
|
| 267913 |
+
"loss": 0.3294,
|
| 267914 |
+
"step": 132475
|
| 267915 |
+
},
|
| 267916 |
+
{
|
| 267917 |
+
"epoch": 1059.54,
|
| 267918 |
+
"learning_rate": 7.893798076923077e-06,
|
| 267919 |
+
"loss": 0.2562,
|
| 267920 |
+
"step": 132480
|
| 267921 |
+
},
|
| 267922 |
+
{
|
| 267923 |
+
"epoch": 1059.58,
|
| 267924 |
+
"learning_rate": 7.89371794871795e-06,
|
| 267925 |
+
"loss": 0.7071,
|
| 267926 |
+
"step": 132485
|
| 267927 |
+
},
|
| 267928 |
+
{
|
| 267929 |
+
"epoch": 1059.62,
|
| 267930 |
+
"learning_rate": 7.893637820512821e-06,
|
| 267931 |
+
"loss": 0.9708,
|
| 267932 |
+
"step": 132490
|
| 267933 |
+
},
|
| 267934 |
+
{
|
| 267935 |
+
"epoch": 1059.66,
|
| 267936 |
+
"learning_rate": 7.893557692307693e-06,
|
| 267937 |
+
"loss": 0.247,
|
| 267938 |
+
"step": 132495
|
| 267939 |
+
},
|
| 267940 |
+
{
|
| 267941 |
+
"epoch": 1059.7,
|
| 267942 |
+
"learning_rate": 7.893477564102566e-06,
|
| 267943 |
+
"loss": 0.3189,
|
| 267944 |
+
"step": 132500
|
| 267945 |
+
},
|
| 267946 |
+
{
|
| 267947 |
+
"epoch": 1059.74,
|
| 267948 |
+
"learning_rate": 7.893397435897437e-06,
|
| 267949 |
+
"loss": 0.3609,
|
| 267950 |
+
"step": 132505
|
| 267951 |
+
},
|
| 267952 |
+
{
|
| 267953 |
+
"epoch": 1059.78,
|
| 267954 |
+
"learning_rate": 7.893317307692308e-06,
|
| 267955 |
+
"loss": 0.7373,
|
| 267956 |
+
"step": 132510
|
| 267957 |
+
},
|
| 267958 |
+
{
|
| 267959 |
+
"epoch": 1059.82,
|
| 267960 |
+
"learning_rate": 7.89323717948718e-06,
|
| 267961 |
+
"loss": 0.8497,
|
| 267962 |
+
"step": 132515
|
| 267963 |
+
},
|
| 267964 |
+
{
|
| 267965 |
+
"epoch": 1059.86,
|
| 267966 |
+
"learning_rate": 7.893157051282053e-06,
|
| 267967 |
+
"loss": 0.2729,
|
| 267968 |
+
"step": 132520
|
| 267969 |
+
},
|
| 267970 |
+
{
|
| 267971 |
+
"epoch": 1059.9,
|
| 267972 |
+
"learning_rate": 7.893076923076923e-06,
|
| 267973 |
+
"loss": 0.2637,
|
| 267974 |
+
"step": 132525
|
| 267975 |
+
},
|
| 267976 |
+
{
|
| 267977 |
+
"epoch": 1059.94,
|
| 267978 |
+
"learning_rate": 7.892996794871796e-06,
|
| 267979 |
+
"loss": 0.3606,
|
| 267980 |
+
"step": 132530
|
| 267981 |
+
},
|
| 267982 |
+
{
|
| 267983 |
+
"epoch": 1059.98,
|
| 267984 |
+
"learning_rate": 7.892916666666667e-06,
|
| 267985 |
+
"loss": 0.6808,
|
| 267986 |
+
"step": 132535
|
| 267987 |
+
},
|
| 267988 |
+
{
|
| 267989 |
+
"epoch": 1060.0,
|
| 267990 |
+
"eval_loss": 0.3099425137042999,
|
| 267991 |
+
"eval_runtime": 36.2358,
|
| 267992 |
+
"eval_samples_per_second": 23.043,
|
| 267993 |
+
"eval_steps_per_second": 0.745,
|
| 267994 |
+
"eval_wer": 0.17826874320898226,
|
| 267995 |
+
"step": 132537
|
| 267996 |
+
},
|
| 267997 |
+
{
|
| 267998 |
+
"epoch": 1068.02,
|
| 267999 |
+
"learning_rate": 7.892836538461538e-06,
|
| 268000 |
+
"loss": 0.3121,
|
| 268001 |
+
"step": 132540
|
| 268002 |
+
},
|
| 268003 |
+
{
|
| 268004 |
+
"epoch": 1068.06,
|
| 268005 |
+
"learning_rate": 7.892756410256411e-06,
|
| 268006 |
+
"loss": 0.2593,
|
| 268007 |
+
"step": 132545
|
| 268008 |
+
},
|
| 268009 |
+
{
|
| 268010 |
+
"epoch": 1068.1,
|
| 268011 |
+
"learning_rate": 7.892676282051283e-06,
|
| 268012 |
+
"loss": 0.3408,
|
| 268013 |
+
"step": 132550
|
| 268014 |
+
},
|
| 268015 |
+
{
|
| 268016 |
+
"epoch": 1068.15,
|
| 268017 |
+
"learning_rate": 7.892596153846154e-06,
|
| 268018 |
+
"loss": 0.3439,
|
| 268019 |
+
"step": 132555
|
| 268020 |
+
},
|
| 268021 |
+
{
|
| 268022 |
+
"epoch": 1068.19,
|
| 268023 |
+
"learning_rate": 7.892516025641025e-06,
|
| 268024 |
+
"loss": 0.6745,
|
| 268025 |
+
"step": 132560
|
| 268026 |
+
},
|
| 268027 |
+
{
|
| 268028 |
+
"epoch": 1068.23,
|
| 268029 |
+
"learning_rate": 7.892435897435899e-06,
|
| 268030 |
+
"loss": 0.786,
|
| 268031 |
+
"step": 132565
|
| 268032 |
+
},
|
| 268033 |
+
{
|
| 268034 |
+
"epoch": 1068.27,
|
| 268035 |
+
"learning_rate": 7.89235576923077e-06,
|
| 268036 |
+
"loss": 0.2751,
|
| 268037 |
+
"step": 132570
|
| 268038 |
+
},
|
| 268039 |
+
{
|
| 268040 |
+
"epoch": 1068.31,
|
| 268041 |
+
"learning_rate": 7.892275641025641e-06,
|
| 268042 |
+
"loss": 0.269,
|
| 268043 |
+
"step": 132575
|
| 268044 |
+
},
|
| 268045 |
+
{
|
| 268046 |
+
"epoch": 1068.35,
|
| 268047 |
+
"learning_rate": 7.892195512820513e-06,
|
| 268048 |
+
"loss": 0.3516,
|
| 268049 |
+
"step": 132580
|
| 268050 |
+
},
|
| 268051 |
+
{
|
| 268052 |
+
"epoch": 1068.39,
|
| 268053 |
+
"learning_rate": 7.892115384615386e-06,
|
| 268054 |
+
"loss": 0.7291,
|
| 268055 |
+
"step": 132585
|
| 268056 |
+
},
|
| 268057 |
+
{
|
| 268058 |
+
"epoch": 1068.43,
|
| 268059 |
+
"learning_rate": 7.892035256410257e-06,
|
| 268060 |
+
"loss": 0.8608,
|
| 268061 |
+
"step": 132590
|
| 268062 |
+
},
|
| 268063 |
+
{
|
| 268064 |
+
"epoch": 1068.47,
|
| 268065 |
+
"learning_rate": 7.891955128205128e-06,
|
| 268066 |
+
"loss": 0.2534,
|
| 268067 |
+
"step": 132595
|
| 268068 |
+
},
|
| 268069 |
+
{
|
| 268070 |
+
"epoch": 1068.51,
|
| 268071 |
+
"learning_rate": 7.891875000000001e-06,
|
| 268072 |
+
"loss": 0.2694,
|
| 268073 |
+
"step": 132600
|
| 268074 |
+
},
|
| 268075 |
+
{
|
| 268076 |
+
"epoch": 1068.55,
|
| 268077 |
+
"learning_rate": 7.891794871794873e-06,
|
| 268078 |
+
"loss": 0.3579,
|
| 268079 |
+
"step": 132605
|
| 268080 |
+
},
|
| 268081 |
+
{
|
| 268082 |
+
"epoch": 1068.59,
|
| 268083 |
+
"learning_rate": 7.891714743589744e-06,
|
| 268084 |
+
"loss": 0.7137,
|
| 268085 |
+
"step": 132610
|
| 268086 |
+
},
|
| 268087 |
+
{
|
| 268088 |
+
"epoch": 1068.63,
|
| 268089 |
+
"learning_rate": 7.891634615384615e-06,
|
| 268090 |
+
"loss": 0.9326,
|
| 268091 |
+
"step": 132615
|
| 268092 |
+
},
|
| 268093 |
+
{
|
| 268094 |
+
"epoch": 1068.67,
|
| 268095 |
+
"learning_rate": 7.891554487179489e-06,
|
| 268096 |
+
"loss": 0.255,
|
| 268097 |
+
"step": 132620
|
| 268098 |
+
},
|
| 268099 |
+
{
|
| 268100 |
+
"epoch": 1068.71,
|
| 268101 |
+
"learning_rate": 7.89147435897436e-06,
|
| 268102 |
+
"loss": 0.2923,
|
| 268103 |
+
"step": 132625
|
| 268104 |
+
},
|
| 268105 |
+
{
|
| 268106 |
+
"epoch": 1068.75,
|
| 268107 |
+
"learning_rate": 7.891394230769231e-06,
|
| 268108 |
+
"loss": 0.3931,
|
| 268109 |
+
"step": 132630
|
| 268110 |
+
},
|
| 268111 |
+
{
|
| 268112 |
+
"epoch": 1068.79,
|
| 268113 |
+
"learning_rate": 7.891314102564103e-06,
|
| 268114 |
+
"loss": 0.7453,
|
| 268115 |
+
"step": 132635
|
| 268116 |
+
},
|
| 268117 |
+
{
|
| 268118 |
+
"epoch": 1068.83,
|
| 268119 |
+
"learning_rate": 7.891233974358976e-06,
|
| 268120 |
+
"loss": 0.7781,
|
| 268121 |
+
"step": 132640
|
| 268122 |
+
},
|
| 268123 |
+
{
|
| 268124 |
+
"epoch": 1068.87,
|
| 268125 |
+
"learning_rate": 7.891153846153847e-06,
|
| 268126 |
+
"loss": 0.2954,
|
| 268127 |
+
"step": 132645
|
| 268128 |
+
},
|
| 268129 |
+
{
|
| 268130 |
+
"epoch": 1068.91,
|
| 268131 |
+
"learning_rate": 7.891073717948718e-06,
|
| 268132 |
+
"loss": 0.2961,
|
| 268133 |
+
"step": 132650
|
| 268134 |
+
},
|
| 268135 |
+
{
|
| 268136 |
+
"epoch": 1068.95,
|
| 268137 |
+
"learning_rate": 7.890993589743591e-06,
|
| 268138 |
+
"loss": 0.3804,
|
| 268139 |
+
"step": 132655
|
| 268140 |
+
},
|
| 268141 |
+
{
|
| 268142 |
+
"epoch": 1068.99,
|
| 268143 |
+
"learning_rate": 7.890913461538461e-06,
|
| 268144 |
+
"loss": 0.8052,
|
| 268145 |
+
"step": 132660
|
| 268146 |
+
},
|
| 268147 |
+
{
|
| 268148 |
+
"epoch": 1069.0,
|
| 268149 |
+
"eval_loss": 0.3759481608867645,
|
| 268150 |
+
"eval_runtime": 35.5265,
|
| 268151 |
+
"eval_samples_per_second": 23.504,
|
| 268152 |
+
"eval_steps_per_second": 0.76,
|
| 268153 |
+
"eval_wer": 0.17900044490582825,
|
| 268154 |
+
"step": 132661
|
| 268155 |
+
},
|
| 268156 |
+
{
|
| 268157 |
+
"epoch": 1061.03,
|
| 268158 |
+
"learning_rate": 7.890833333333334e-06,
|
| 268159 |
+
"loss": 0.3047,
|
| 268160 |
+
"step": 132665
|
| 268161 |
+
},
|
| 268162 |
+
{
|
| 268163 |
+
"epoch": 1061.07,
|
| 268164 |
+
"learning_rate": 7.890753205128206e-06,
|
| 268165 |
+
"loss": 0.262,
|
| 268166 |
+
"step": 132670
|
| 268167 |
+
},
|
| 268168 |
+
{
|
| 268169 |
+
"epoch": 1061.11,
|
| 268170 |
+
"learning_rate": 7.890673076923077e-06,
|
| 268171 |
+
"loss": 0.3535,
|
| 268172 |
+
"step": 132675
|
| 268173 |
+
},
|
| 268174 |
+
{
|
| 268175 |
+
"epoch": 1061.15,
|
| 268176 |
+
"learning_rate": 7.890592948717948e-06,
|
| 268177 |
+
"loss": 0.3661,
|
| 268178 |
+
"step": 132680
|
| 268179 |
+
},
|
| 268180 |
+
{
|
| 268181 |
+
"epoch": 1061.19,
|
| 268182 |
+
"learning_rate": 7.890512820512821e-06,
|
| 268183 |
+
"loss": 0.9716,
|
| 268184 |
+
"step": 132685
|
| 268185 |
+
},
|
| 268186 |
+
{
|
| 268187 |
+
"epoch": 1061.23,
|
| 268188 |
+
"learning_rate": 7.890432692307693e-06,
|
| 268189 |
+
"loss": 0.6551,
|
| 268190 |
+
"step": 132690
|
| 268191 |
+
},
|
| 268192 |
+
{
|
| 268193 |
+
"epoch": 1061.27,
|
| 268194 |
+
"learning_rate": 7.890352564102564e-06,
|
| 268195 |
+
"loss": 0.2745,
|
| 268196 |
+
"step": 132695
|
| 268197 |
+
},
|
| 268198 |
+
{
|
| 268199 |
+
"epoch": 1061.31,
|
| 268200 |
+
"learning_rate": 7.890272435897437e-06,
|
| 268201 |
+
"loss": 0.3181,
|
| 268202 |
+
"step": 132700
|
| 268203 |
+
},
|
| 268204 |
+
{
|
| 268205 |
+
"epoch": 1061.35,
|
| 268206 |
+
"learning_rate": 7.890192307692308e-06,
|
| 268207 |
+
"loss": 0.3705,
|
| 268208 |
+
"step": 132705
|
| 268209 |
+
},
|
| 268210 |
+
{
|
| 268211 |
+
"epoch": 1061.39,
|
| 268212 |
+
"learning_rate": 7.89011217948718e-06,
|
| 268213 |
+
"loss": 0.8255,
|
| 268214 |
+
"step": 132710
|
| 268215 |
+
},
|
| 268216 |
+
{
|
| 268217 |
+
"epoch": 1061.43,
|
| 268218 |
+
"learning_rate": 7.890032051282051e-06,
|
| 268219 |
+
"loss": 0.6142,
|
| 268220 |
+
"step": 132715
|
| 268221 |
+
},
|
| 268222 |
+
{
|
| 268223 |
+
"epoch": 1061.47,
|
| 268224 |
+
"learning_rate": 7.889951923076924e-06,
|
| 268225 |
+
"loss": 0.266,
|
| 268226 |
+
"step": 132720
|
| 268227 |
+
},
|
| 268228 |
+
{
|
| 268229 |
+
"epoch": 1061.51,
|
| 268230 |
+
"learning_rate": 7.889871794871796e-06,
|
| 268231 |
+
"loss": 0.2973,
|
| 268232 |
+
"step": 132725
|
| 268233 |
+
},
|
| 268234 |
+
{
|
| 268235 |
+
"epoch": 1061.55,
|
| 268236 |
+
"learning_rate": 7.889791666666667e-06,
|
| 268237 |
+
"loss": 0.4102,
|
| 268238 |
+
"step": 132730
|
| 268239 |
+
},
|
| 268240 |
+
{
|
| 268241 |
+
"epoch": 1061.59,
|
| 268242 |
+
"learning_rate": 7.889711538461538e-06,
|
| 268243 |
+
"loss": 0.863,
|
| 268244 |
+
"step": 132735
|
| 268245 |
+
},
|
| 268246 |
+
{
|
| 268247 |
+
"epoch": 1061.63,
|
| 268248 |
+
"learning_rate": 7.889631410256411e-06,
|
| 268249 |
+
"loss": 0.6402,
|
| 268250 |
+
"step": 132740
|
| 268251 |
+
},
|
| 268252 |
+
{
|
| 268253 |
+
"epoch": 1061.67,
|
| 268254 |
+
"learning_rate": 7.889551282051283e-06,
|
| 268255 |
+
"loss": 0.2461,
|
| 268256 |
+
"step": 132745
|
| 268257 |
+
},
|
| 268258 |
+
{
|
| 268259 |
+
"epoch": 1061.71,
|
| 268260 |
+
"learning_rate": 7.889471153846154e-06,
|
| 268261 |
+
"loss": 0.3107,
|
| 268262 |
+
"step": 132750
|
| 268263 |
+
},
|
| 268264 |
+
{
|
| 268265 |
+
"epoch": 1061.75,
|
| 268266 |
+
"learning_rate": 7.889391025641027e-06,
|
| 268267 |
+
"loss": 0.4564,
|
| 268268 |
+
"step": 132755
|
| 268269 |
+
},
|
| 268270 |
+
{
|
| 268271 |
+
"epoch": 1061.79,
|
| 268272 |
+
"learning_rate": 7.889310897435898e-06,
|
| 268273 |
+
"loss": 0.8667,
|
| 268274 |
+
"step": 132760
|
| 268275 |
+
},
|
| 268276 |
+
{
|
| 268277 |
+
"epoch": 1061.83,
|
| 268278 |
+
"learning_rate": 7.88923076923077e-06,
|
| 268279 |
+
"loss": 0.6114,
|
| 268280 |
+
"step": 132765
|
| 268281 |
+
},
|
| 268282 |
+
{
|
| 268283 |
+
"epoch": 1061.87,
|
| 268284 |
+
"learning_rate": 7.889150641025641e-06,
|
| 268285 |
+
"loss": 0.2584,
|
| 268286 |
+
"step": 132770
|
| 268287 |
+
},
|
| 268288 |
+
{
|
| 268289 |
+
"epoch": 1061.91,
|
| 268290 |
+
"learning_rate": 7.889070512820514e-06,
|
| 268291 |
+
"loss": 0.2723,
|
| 268292 |
+
"step": 132775
|
| 268293 |
+
},
|
| 268294 |
+
{
|
| 268295 |
+
"epoch": 1061.95,
|
| 268296 |
+
"learning_rate": 7.888990384615386e-06,
|
| 268297 |
+
"loss": 0.3697,
|
| 268298 |
+
"step": 132780
|
| 268299 |
+
},
|
| 268300 |
+
{
|
| 268301 |
+
"epoch": 1061.99,
|
| 268302 |
+
"learning_rate": 7.888910256410257e-06,
|
| 268303 |
+
"loss": 0.9063,
|
| 268304 |
+
"step": 132785
|
| 268305 |
+
},
|
| 268306 |
+
{
|
| 268307 |
+
"epoch": 1062.0,
|
| 268308 |
+
"eval_loss": 0.35875725746154785,
|
| 268309 |
+
"eval_runtime": 35.2957,
|
| 268310 |
+
"eval_samples_per_second": 23.657,
|
| 268311 |
+
"eval_steps_per_second": 0.765,
|
| 268312 |
+
"eval_wer": 0.18351754320896077,
|
| 268313 |
+
"step": 132786
|
| 268314 |
}
|
| 268315 |
],
|
| 268316 |
"max_steps": 625000,
|
| 268317 |
"num_train_epochs": 5000,
|
| 268318 |
+
"total_flos": 3.736502338216323e+20,
|
| 268319 |
"trial_name": null,
|
| 268320 |
"trial_params": null
|
| 268321 |
}
|
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132786}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/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 sha256:7646d4658c7fb581271e3f3a4cd882f46c4e9eaff0c1d3ece8c4a2f60d2fe9b3
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132910}/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:ab98858ccc2a45c3fd5e7878c3e771f33185de7cf38be497d83aacd345b6bc15
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132910}/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:fbf964a925c16111bd00a1243d38524b685492690de876b21aa0b98f4a2bdd2b
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132910}/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 sha256:72f28b93e2cec7a5a1f715c7d28d98c2d935e7a77f9f7576f9c998aa09c5d067
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/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:b87ec296eb9fdd22d42513ded3472eba2b1440f44b4d69537e0b23fe4befe58a
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-131665 → checkpoint-132910}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -266886,11 +266886,1595 @@
|
|
| 266886 |
"eval_steps_per_second": 0.729,
|
| 266887 |
"eval_wer": 0.17062187276626162,
|
| 266888 |
"step": 131665
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 266889 |
}
|
| 266890 |
],
|
| 266891 |
"max_steps": 620000,
|
| 266892 |
"num_train_epochs": 5000,
|
| 266893 |
-
"total_flos": 3.
|
| 266894 |
"trial_name": null,
|
| 266895 |
"trial_params": null
|
| 266896 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.1689111747851003,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-132910",
|
| 4 |
+
"epoch": 1071.0,
|
| 5 |
+
"global_step": 132910,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 266886 |
"eval_steps_per_second": 0.729,
|
| 266887 |
"eval_wer": 0.17062187276626162,
|
| 266888 |
"step": 131665
|
| 266889 |
+
},
|
| 266890 |
+
{
|
| 266891 |
+
"epoch": 1053.04,
|
| 266892 |
+
"learning_rate": 7.906746794871795e-06,
|
| 266893 |
+
"loss": 0.3188,
|
| 266894 |
+
"step": 131670
|
| 266895 |
+
},
|
| 266896 |
+
{
|
| 266897 |
+
"epoch": 1053.08,
|
| 266898 |
+
"learning_rate": 7.906666666666667e-06,
|
| 266899 |
+
"loss": 0.2718,
|
| 266900 |
+
"step": 131675
|
| 266901 |
+
},
|
| 266902 |
+
{
|
| 266903 |
+
"epoch": 1053.12,
|
| 266904 |
+
"learning_rate": 7.90658653846154e-06,
|
| 266905 |
+
"loss": 0.3764,
|
| 266906 |
+
"step": 131680
|
| 266907 |
+
},
|
| 266908 |
+
{
|
| 266909 |
+
"epoch": 1053.16,
|
| 266910 |
+
"learning_rate": 7.906506410256411e-06,
|
| 266911 |
+
"loss": 0.4363,
|
| 266912 |
+
"step": 131685
|
| 266913 |
+
},
|
| 266914 |
+
{
|
| 266915 |
+
"epoch": 1053.2,
|
| 266916 |
+
"learning_rate": 7.906426282051282e-06,
|
| 266917 |
+
"loss": 1.2935,
|
| 266918 |
+
"step": 131690
|
| 266919 |
+
},
|
| 266920 |
+
{
|
| 266921 |
+
"epoch": 1053.24,
|
| 266922 |
+
"learning_rate": 7.906346153846155e-06,
|
| 266923 |
+
"loss": 0.4497,
|
| 266924 |
+
"step": 131695
|
| 266925 |
+
},
|
| 266926 |
+
{
|
| 266927 |
+
"epoch": 1053.28,
|
| 266928 |
+
"learning_rate": 7.906266025641027e-06,
|
| 266929 |
+
"loss": 0.2773,
|
| 266930 |
+
"step": 131700
|
| 266931 |
+
},
|
| 266932 |
+
{
|
| 266933 |
+
"epoch": 1053.32,
|
| 266934 |
+
"learning_rate": 7.906185897435898e-06,
|
| 266935 |
+
"loss": 0.3123,
|
| 266936 |
+
"step": 131705
|
| 266937 |
+
},
|
| 266938 |
+
{
|
| 266939 |
+
"epoch": 1053.36,
|
| 266940 |
+
"learning_rate": 7.90610576923077e-06,
|
| 266941 |
+
"loss": 0.4918,
|
| 266942 |
+
"step": 131710
|
| 266943 |
+
},
|
| 266944 |
+
{
|
| 266945 |
+
"epoch": 1053.4,
|
| 266946 |
+
"learning_rate": 7.906025641025643e-06,
|
| 266947 |
+
"loss": 1.0791,
|
| 266948 |
+
"step": 131715
|
| 266949 |
+
},
|
| 266950 |
+
{
|
| 266951 |
+
"epoch": 1053.44,
|
| 266952 |
+
"learning_rate": 7.905945512820514e-06,
|
| 266953 |
+
"loss": 0.5006,
|
| 266954 |
+
"step": 131720
|
| 266955 |
+
},
|
| 266956 |
+
{
|
| 266957 |
+
"epoch": 1053.48,
|
| 266958 |
+
"learning_rate": 7.905865384615385e-06,
|
| 266959 |
+
"loss": 0.2765,
|
| 266960 |
+
"step": 131725
|
| 266961 |
+
},
|
| 266962 |
+
{
|
| 266963 |
+
"epoch": 1053.52,
|
| 266964 |
+
"learning_rate": 7.905785256410257e-06,
|
| 266965 |
+
"loss": 0.2707,
|
| 266966 |
+
"step": 131730
|
| 266967 |
+
},
|
| 266968 |
+
{
|
| 266969 |
+
"epoch": 1053.56,
|
| 266970 |
+
"learning_rate": 7.90570512820513e-06,
|
| 266971 |
+
"loss": 0.8728,
|
| 266972 |
+
"step": 131735
|
| 266973 |
+
},
|
| 266974 |
+
{
|
| 266975 |
+
"epoch": 1053.6,
|
| 266976 |
+
"learning_rate": 7.905625e-06,
|
| 266977 |
+
"loss": 1.1669,
|
| 266978 |
+
"step": 131740
|
| 266979 |
+
},
|
| 266980 |
+
{
|
| 266981 |
+
"epoch": 1053.64,
|
| 266982 |
+
"learning_rate": 7.905544871794872e-06,
|
| 266983 |
+
"loss": 0.2836,
|
| 266984 |
+
"step": 131745
|
| 266985 |
+
},
|
| 266986 |
+
{
|
| 266987 |
+
"epoch": 1053.68,
|
| 266988 |
+
"learning_rate": 7.905464743589745e-06,
|
| 266989 |
+
"loss": 0.2357,
|
| 266990 |
+
"step": 131750
|
| 266991 |
+
},
|
| 266992 |
+
{
|
| 266993 |
+
"epoch": 1053.72,
|
| 266994 |
+
"learning_rate": 7.905400641025641e-06,
|
| 266995 |
+
"loss": 0.2992,
|
| 266996 |
+
"step": 131755
|
| 266997 |
+
},
|
| 266998 |
+
{
|
| 266999 |
+
"epoch": 1053.76,
|
| 267000 |
+
"learning_rate": 7.905320512820514e-06,
|
| 267001 |
+
"loss": 0.4702,
|
| 267002 |
+
"step": 131760
|
| 267003 |
+
},
|
| 267004 |
+
{
|
| 267005 |
+
"epoch": 1053.8,
|
| 267006 |
+
"learning_rate": 7.905240384615386e-06,
|
| 267007 |
+
"loss": 1.2597,
|
| 267008 |
+
"step": 131765
|
| 267009 |
+
},
|
| 267010 |
+
{
|
| 267011 |
+
"epoch": 1053.84,
|
| 267012 |
+
"learning_rate": 7.905160256410257e-06,
|
| 267013 |
+
"loss": 0.3311,
|
| 267014 |
+
"step": 131770
|
| 267015 |
+
},
|
| 267016 |
+
{
|
| 267017 |
+
"epoch": 1053.88,
|
| 267018 |
+
"learning_rate": 7.90508012820513e-06,
|
| 267019 |
+
"loss": 0.2991,
|
| 267020 |
+
"step": 131775
|
| 267021 |
+
},
|
| 267022 |
+
{
|
| 267023 |
+
"epoch": 1053.92,
|
| 267024 |
+
"learning_rate": 7.905e-06,
|
| 267025 |
+
"loss": 0.2988,
|
| 267026 |
+
"step": 131780
|
| 267027 |
+
},
|
| 267028 |
+
{
|
| 267029 |
+
"epoch": 1053.96,
|
| 267030 |
+
"learning_rate": 7.904919871794873e-06,
|
| 267031 |
+
"loss": 0.4901,
|
| 267032 |
+
"step": 131785
|
| 267033 |
+
},
|
| 267034 |
+
{
|
| 267035 |
+
"epoch": 1054.0,
|
| 267036 |
+
"learning_rate": 7.904839743589744e-06,
|
| 267037 |
+
"loss": 1.5886,
|
| 267038 |
+
"step": 131790
|
| 267039 |
+
},
|
| 267040 |
+
{
|
| 267041 |
+
"epoch": 1054.0,
|
| 267042 |
+
"eval_loss": 0.35416796803474426,
|
| 267043 |
+
"eval_runtime": 34.5583,
|
| 267044 |
+
"eval_samples_per_second": 24.191,
|
| 267045 |
+
"eval_steps_per_second": 0.781,
|
| 267046 |
+
"eval_wer": 0.18272500566764907,
|
| 267047 |
+
"step": 131790
|
| 267048 |
+
},
|
| 267049 |
+
{
|
| 267050 |
+
"epoch": 1062.04,
|
| 267051 |
+
"learning_rate": 7.904759615384615e-06,
|
| 267052 |
+
"loss": 0.3037,
|
| 267053 |
+
"step": 131795
|
| 267054 |
+
},
|
| 267055 |
+
{
|
| 267056 |
+
"epoch": 1062.08,
|
| 267057 |
+
"learning_rate": 7.904679487179487e-06,
|
| 267058 |
+
"loss": 0.3027,
|
| 267059 |
+
"step": 131800
|
| 267060 |
+
},
|
| 267061 |
+
{
|
| 267062 |
+
"epoch": 1062.12,
|
| 267063 |
+
"learning_rate": 7.90459935897436e-06,
|
| 267064 |
+
"loss": 0.2598,
|
| 267065 |
+
"step": 131805
|
| 267066 |
+
},
|
| 267067 |
+
{
|
| 267068 |
+
"epoch": 1062.16,
|
| 267069 |
+
"learning_rate": 7.904519230769231e-06,
|
| 267070 |
+
"loss": 0.5055,
|
| 267071 |
+
"step": 131810
|
| 267072 |
+
},
|
| 267073 |
+
{
|
| 267074 |
+
"epoch": 1062.2,
|
| 267075 |
+
"learning_rate": 7.904439102564103e-06,
|
| 267076 |
+
"loss": 1.1921,
|
| 267077 |
+
"step": 131815
|
| 267078 |
+
},
|
| 267079 |
+
{
|
| 267080 |
+
"epoch": 1062.24,
|
| 267081 |
+
"learning_rate": 7.904358974358976e-06,
|
| 267082 |
+
"loss": 0.2803,
|
| 267083 |
+
"step": 131820
|
| 267084 |
+
},
|
| 267085 |
+
{
|
| 267086 |
+
"epoch": 1062.28,
|
| 267087 |
+
"learning_rate": 7.904278846153847e-06,
|
| 267088 |
+
"loss": 0.2656,
|
| 267089 |
+
"step": 131825
|
| 267090 |
+
},
|
| 267091 |
+
{
|
| 267092 |
+
"epoch": 1062.32,
|
| 267093 |
+
"learning_rate": 7.904198717948718e-06,
|
| 267094 |
+
"loss": 0.3037,
|
| 267095 |
+
"step": 131830
|
| 267096 |
+
},
|
| 267097 |
+
{
|
| 267098 |
+
"epoch": 1062.36,
|
| 267099 |
+
"learning_rate": 7.90411858974359e-06,
|
| 267100 |
+
"loss": 0.4523,
|
| 267101 |
+
"step": 131835
|
| 267102 |
+
},
|
| 267103 |
+
{
|
| 267104 |
+
"epoch": 1062.4,
|
| 267105 |
+
"learning_rate": 7.904038461538463e-06,
|
| 267106 |
+
"loss": 1.2989,
|
| 267107 |
+
"step": 131840
|
| 267108 |
+
},
|
| 267109 |
+
{
|
| 267110 |
+
"epoch": 1062.44,
|
| 267111 |
+
"learning_rate": 7.903958333333334e-06,
|
| 267112 |
+
"loss": 0.2874,
|
| 267113 |
+
"step": 131845
|
| 267114 |
+
},
|
| 267115 |
+
{
|
| 267116 |
+
"epoch": 1062.48,
|
| 267117 |
+
"learning_rate": 7.903878205128205e-06,
|
| 267118 |
+
"loss": 0.2579,
|
| 267119 |
+
"step": 131850
|
| 267120 |
+
},
|
| 267121 |
+
{
|
| 267122 |
+
"epoch": 1062.52,
|
| 267123 |
+
"learning_rate": 7.903798076923077e-06,
|
| 267124 |
+
"loss": 0.3465,
|
| 267125 |
+
"step": 131855
|
| 267126 |
+
},
|
| 267127 |
+
{
|
| 267128 |
+
"epoch": 1062.56,
|
| 267129 |
+
"learning_rate": 7.90371794871795e-06,
|
| 267130 |
+
"loss": 0.4392,
|
| 267131 |
+
"step": 131860
|
| 267132 |
+
},
|
| 267133 |
+
{
|
| 267134 |
+
"epoch": 1062.6,
|
| 267135 |
+
"learning_rate": 7.903637820512821e-06,
|
| 267136 |
+
"loss": 1.1929,
|
| 267137 |
+
"step": 131865
|
| 267138 |
+
},
|
| 267139 |
+
{
|
| 267140 |
+
"epoch": 1062.64,
|
| 267141 |
+
"learning_rate": 7.903557692307693e-06,
|
| 267142 |
+
"loss": 0.2869,
|
| 267143 |
+
"step": 131870
|
| 267144 |
+
},
|
| 267145 |
+
{
|
| 267146 |
+
"epoch": 1062.68,
|
| 267147 |
+
"learning_rate": 7.903477564102566e-06,
|
| 267148 |
+
"loss": 0.2637,
|
| 267149 |
+
"step": 131875
|
| 267150 |
+
},
|
| 267151 |
+
{
|
| 267152 |
+
"epoch": 1062.72,
|
| 267153 |
+
"learning_rate": 7.903397435897437e-06,
|
| 267154 |
+
"loss": 0.2865,
|
| 267155 |
+
"step": 131880
|
| 267156 |
+
},
|
| 267157 |
+
{
|
| 267158 |
+
"epoch": 1062.76,
|
| 267159 |
+
"learning_rate": 7.903317307692308e-06,
|
| 267160 |
+
"loss": 0.4224,
|
| 267161 |
+
"step": 131885
|
| 267162 |
+
},
|
| 267163 |
+
{
|
| 267164 |
+
"epoch": 1062.8,
|
| 267165 |
+
"learning_rate": 7.90323717948718e-06,
|
| 267166 |
+
"loss": 1.1498,
|
| 267167 |
+
"step": 131890
|
| 267168 |
+
},
|
| 267169 |
+
{
|
| 267170 |
+
"epoch": 1062.84,
|
| 267171 |
+
"learning_rate": 7.903157051282053e-06,
|
| 267172 |
+
"loss": 0.3456,
|
| 267173 |
+
"step": 131895
|
| 267174 |
+
},
|
| 267175 |
+
{
|
| 267176 |
+
"epoch": 1062.88,
|
| 267177 |
+
"learning_rate": 7.903076923076922e-06,
|
| 267178 |
+
"loss": 0.2872,
|
| 267179 |
+
"step": 131900
|
| 267180 |
+
},
|
| 267181 |
+
{
|
| 267182 |
+
"epoch": 1062.92,
|
| 267183 |
+
"learning_rate": 7.902996794871796e-06,
|
| 267184 |
+
"loss": 0.3246,
|
| 267185 |
+
"step": 131905
|
| 267186 |
+
},
|
| 267187 |
+
{
|
| 267188 |
+
"epoch": 1062.96,
|
| 267189 |
+
"learning_rate": 7.902916666666667e-06,
|
| 267190 |
+
"loss": 0.5289,
|
| 267191 |
+
"step": 131910
|
| 267192 |
+
},
|
| 267193 |
+
{
|
| 267194 |
+
"epoch": 1063.0,
|
| 267195 |
+
"eval_loss": 0.3880179524421692,
|
| 267196 |
+
"eval_runtime": 49.3292,
|
| 267197 |
+
"eval_samples_per_second": 16.927,
|
| 267198 |
+
"eval_steps_per_second": 0.547,
|
| 267199 |
+
"eval_wer": 0.1791111111111111,
|
| 267200 |
+
"step": 131914
|
| 267201 |
+
},
|
| 267202 |
+
{
|
| 267203 |
+
"epoch": 1063.01,
|
| 267204 |
+
"learning_rate": 7.90283653846154e-06,
|
| 267205 |
+
"loss": 0.4208,
|
| 267206 |
+
"step": 131915
|
| 267207 |
+
},
|
| 267208 |
+
{
|
| 267209 |
+
"epoch": 1063.05,
|
| 267210 |
+
"learning_rate": 7.902756410256411e-06,
|
| 267211 |
+
"loss": 0.262,
|
| 267212 |
+
"step": 131920
|
| 267213 |
+
},
|
| 267214 |
+
{
|
| 267215 |
+
"epoch": 1063.09,
|
| 267216 |
+
"learning_rate": 7.902676282051283e-06,
|
| 267217 |
+
"loss": 0.2644,
|
| 267218 |
+
"step": 131925
|
| 267219 |
+
},
|
| 267220 |
+
{
|
| 267221 |
+
"epoch": 1063.13,
|
| 267222 |
+
"learning_rate": 7.902596153846156e-06,
|
| 267223 |
+
"loss": 0.3166,
|
| 267224 |
+
"step": 131930
|
| 267225 |
+
},
|
| 267226 |
+
{
|
| 267227 |
+
"epoch": 1063.17,
|
| 267228 |
+
"learning_rate": 7.902516025641025e-06,
|
| 267229 |
+
"loss": 0.5168,
|
| 267230 |
+
"step": 131935
|
| 267231 |
+
},
|
| 267232 |
+
{
|
| 267233 |
+
"epoch": 1063.21,
|
| 267234 |
+
"learning_rate": 7.902435897435898e-06,
|
| 267235 |
+
"loss": 1.076,
|
| 267236 |
+
"step": 131940
|
| 267237 |
+
},
|
| 267238 |
+
{
|
| 267239 |
+
"epoch": 1063.25,
|
| 267240 |
+
"learning_rate": 7.90235576923077e-06,
|
| 267241 |
+
"loss": 0.3121,
|
| 267242 |
+
"step": 131945
|
| 267243 |
+
},
|
| 267244 |
+
{
|
| 267245 |
+
"epoch": 1063.29,
|
| 267246 |
+
"learning_rate": 7.902275641025641e-06,
|
| 267247 |
+
"loss": 0.3081,
|
| 267248 |
+
"step": 131950
|
| 267249 |
+
},
|
| 267250 |
+
{
|
| 267251 |
+
"epoch": 1063.33,
|
| 267252 |
+
"learning_rate": 7.902195512820512e-06,
|
| 267253 |
+
"loss": 0.2783,
|
| 267254 |
+
"step": 131955
|
| 267255 |
+
},
|
| 267256 |
+
{
|
| 267257 |
+
"epoch": 1063.37,
|
| 267258 |
+
"learning_rate": 7.902115384615386e-06,
|
| 267259 |
+
"loss": 0.5739,
|
| 267260 |
+
"step": 131960
|
| 267261 |
+
},
|
| 267262 |
+
{
|
| 267263 |
+
"epoch": 1063.41,
|
| 267264 |
+
"learning_rate": 7.902035256410257e-06,
|
| 267265 |
+
"loss": 0.8893,
|
| 267266 |
+
"step": 131965
|
| 267267 |
+
},
|
| 267268 |
+
{
|
| 267269 |
+
"epoch": 1063.45,
|
| 267270 |
+
"learning_rate": 7.901955128205128e-06,
|
| 267271 |
+
"loss": 0.2758,
|
| 267272 |
+
"step": 131970
|
| 267273 |
+
},
|
| 267274 |
+
{
|
| 267275 |
+
"epoch": 1063.49,
|
| 267276 |
+
"learning_rate": 7.901875000000001e-06,
|
| 267277 |
+
"loss": 0.2949,
|
| 267278 |
+
"step": 131975
|
| 267279 |
+
},
|
| 267280 |
+
{
|
| 267281 |
+
"epoch": 1063.53,
|
| 267282 |
+
"learning_rate": 7.901794871794873e-06,
|
| 267283 |
+
"loss": 0.3229,
|
| 267284 |
+
"step": 131980
|
| 267285 |
+
},
|
| 267286 |
+
{
|
| 267287 |
+
"epoch": 1063.57,
|
| 267288 |
+
"learning_rate": 7.901714743589744e-06,
|
| 267289 |
+
"loss": 0.4443,
|
| 267290 |
+
"step": 131985
|
| 267291 |
+
},
|
| 267292 |
+
{
|
| 267293 |
+
"epoch": 1063.61,
|
| 267294 |
+
"learning_rate": 7.901634615384615e-06,
|
| 267295 |
+
"loss": 1.0208,
|
| 267296 |
+
"step": 131990
|
| 267297 |
+
},
|
| 267298 |
+
{
|
| 267299 |
+
"epoch": 1063.65,
|
| 267300 |
+
"learning_rate": 7.901554487179488e-06,
|
| 267301 |
+
"loss": 0.3214,
|
| 267302 |
+
"step": 131995
|
| 267303 |
+
},
|
| 267304 |
+
{
|
| 267305 |
+
"epoch": 1063.69,
|
| 267306 |
+
"learning_rate": 7.90147435897436e-06,
|
| 267307 |
+
"loss": 0.2543,
|
| 267308 |
+
"step": 132000
|
| 267309 |
+
},
|
| 267310 |
+
{
|
| 267311 |
+
"epoch": 1063.73,
|
| 267312 |
+
"learning_rate": 7.901394230769231e-06,
|
| 267313 |
+
"loss": 0.3166,
|
| 267314 |
+
"step": 132005
|
| 267315 |
+
},
|
| 267316 |
+
{
|
| 267317 |
+
"epoch": 1063.77,
|
| 267318 |
+
"learning_rate": 7.901314102564103e-06,
|
| 267319 |
+
"loss": 0.5529,
|
| 267320 |
+
"step": 132010
|
| 267321 |
+
},
|
| 267322 |
+
{
|
| 267323 |
+
"epoch": 1063.81,
|
| 267324 |
+
"learning_rate": 7.901233974358976e-06,
|
| 267325 |
+
"loss": 1.1172,
|
| 267326 |
+
"step": 132015
|
| 267327 |
+
},
|
| 267328 |
+
{
|
| 267329 |
+
"epoch": 1063.85,
|
| 267330 |
+
"learning_rate": 7.901153846153847e-06,
|
| 267331 |
+
"loss": 0.2745,
|
| 267332 |
+
"step": 132020
|
| 267333 |
+
},
|
| 267334 |
+
{
|
| 267335 |
+
"epoch": 1063.9,
|
| 267336 |
+
"learning_rate": 7.901073717948718e-06,
|
| 267337 |
+
"loss": 0.2381,
|
| 267338 |
+
"step": 132025
|
| 267339 |
+
},
|
| 267340 |
+
{
|
| 267341 |
+
"epoch": 1063.94,
|
| 267342 |
+
"learning_rate": 7.900993589743591e-06,
|
| 267343 |
+
"loss": 0.3242,
|
| 267344 |
+
"step": 132030
|
| 267345 |
+
},
|
| 267346 |
+
{
|
| 267347 |
+
"epoch": 1063.98,
|
| 267348 |
+
"learning_rate": 7.900913461538463e-06,
|
| 267349 |
+
"loss": 0.5418,
|
| 267350 |
+
"step": 132035
|
| 267351 |
+
},
|
| 267352 |
+
{
|
| 267353 |
+
"epoch": 1064.0,
|
| 267354 |
+
"eval_loss": 0.33710455894470215,
|
| 267355 |
+
"eval_runtime": 36.422,
|
| 267356 |
+
"eval_samples_per_second": 22.926,
|
| 267357 |
+
"eval_steps_per_second": 0.741,
|
| 267358 |
+
"eval_wer": 0.17304985554485516,
|
| 267359 |
+
"step": 132038
|
| 267360 |
+
},
|
| 267361 |
+
{
|
| 267362 |
+
"epoch": 1056.02,
|
| 267363 |
+
"learning_rate": 7.900833333333334e-06,
|
| 267364 |
+
"loss": 0.312,
|
| 267365 |
+
"step": 132040
|
| 267366 |
+
},
|
| 267367 |
+
{
|
| 267368 |
+
"epoch": 1056.06,
|
| 267369 |
+
"learning_rate": 7.900753205128205e-06,
|
| 267370 |
+
"loss": 0.2922,
|
| 267371 |
+
"step": 132045
|
| 267372 |
+
},
|
| 267373 |
+
{
|
| 267374 |
+
"epoch": 1056.1,
|
| 267375 |
+
"learning_rate": 7.900673076923078e-06,
|
| 267376 |
+
"loss": 0.2696,
|
| 267377 |
+
"step": 132050
|
| 267378 |
+
},
|
| 267379 |
+
{
|
| 267380 |
+
"epoch": 1056.14,
|
| 267381 |
+
"learning_rate": 7.900592948717948e-06,
|
| 267382 |
+
"loss": 0.3133,
|
| 267383 |
+
"step": 132055
|
| 267384 |
+
},
|
| 267385 |
+
{
|
| 267386 |
+
"epoch": 1056.18,
|
| 267387 |
+
"learning_rate": 7.900512820512821e-06,
|
| 267388 |
+
"loss": 0.5465,
|
| 267389 |
+
"step": 132060
|
| 267390 |
+
},
|
| 267391 |
+
{
|
| 267392 |
+
"epoch": 1056.22,
|
| 267393 |
+
"learning_rate": 7.900432692307694e-06,
|
| 267394 |
+
"loss": 1.1344,
|
| 267395 |
+
"step": 132065
|
| 267396 |
+
},
|
| 267397 |
+
{
|
| 267398 |
+
"epoch": 1056.25,
|
| 267399 |
+
"learning_rate": 7.900352564102564e-06,
|
| 267400 |
+
"loss": 0.2817,
|
| 267401 |
+
"step": 132070
|
| 267402 |
+
},
|
| 267403 |
+
{
|
| 267404 |
+
"epoch": 1056.29,
|
| 267405 |
+
"learning_rate": 7.900272435897437e-06,
|
| 267406 |
+
"loss": 0.2982,
|
| 267407 |
+
"step": 132075
|
| 267408 |
+
},
|
| 267409 |
+
{
|
| 267410 |
+
"epoch": 1056.33,
|
| 267411 |
+
"learning_rate": 7.900192307692308e-06,
|
| 267412 |
+
"loss": 0.3226,
|
| 267413 |
+
"step": 132080
|
| 267414 |
+
},
|
| 267415 |
+
{
|
| 267416 |
+
"epoch": 1056.37,
|
| 267417 |
+
"learning_rate": 7.90011217948718e-06,
|
| 267418 |
+
"loss": 0.626,
|
| 267419 |
+
"step": 132085
|
| 267420 |
+
},
|
| 267421 |
+
{
|
| 267422 |
+
"epoch": 1056.41,
|
| 267423 |
+
"learning_rate": 7.900032051282051e-06,
|
| 267424 |
+
"loss": 1.0718,
|
| 267425 |
+
"step": 132090
|
| 267426 |
+
},
|
| 267427 |
+
{
|
| 267428 |
+
"epoch": 1056.45,
|
| 267429 |
+
"learning_rate": 7.899951923076924e-06,
|
| 267430 |
+
"loss": 0.3002,
|
| 267431 |
+
"step": 132095
|
| 267432 |
+
},
|
| 267433 |
+
{
|
| 267434 |
+
"epoch": 1056.49,
|
| 267435 |
+
"learning_rate": 7.899871794871795e-06,
|
| 267436 |
+
"loss": 0.3234,
|
| 267437 |
+
"step": 132100
|
| 267438 |
+
},
|
| 267439 |
+
{
|
| 267440 |
+
"epoch": 1056.53,
|
| 267441 |
+
"learning_rate": 7.899791666666667e-06,
|
| 267442 |
+
"loss": 0.3039,
|
| 267443 |
+
"step": 132105
|
| 267444 |
+
},
|
| 267445 |
+
{
|
| 267446 |
+
"epoch": 1056.57,
|
| 267447 |
+
"learning_rate": 7.899711538461538e-06,
|
| 267448 |
+
"loss": 0.715,
|
| 267449 |
+
"step": 132110
|
| 267450 |
+
},
|
| 267451 |
+
{
|
| 267452 |
+
"epoch": 1056.61,
|
| 267453 |
+
"learning_rate": 7.899631410256411e-06,
|
| 267454 |
+
"loss": 1.0403,
|
| 267455 |
+
"step": 132115
|
| 267456 |
+
},
|
| 267457 |
+
{
|
| 267458 |
+
"epoch": 1056.65,
|
| 267459 |
+
"learning_rate": 7.899551282051283e-06,
|
| 267460 |
+
"loss": 0.3533,
|
| 267461 |
+
"step": 132120
|
| 267462 |
+
},
|
| 267463 |
+
{
|
| 267464 |
+
"epoch": 1056.69,
|
| 267465 |
+
"learning_rate": 7.899471153846154e-06,
|
| 267466 |
+
"loss": 0.2698,
|
| 267467 |
+
"step": 132125
|
| 267468 |
+
},
|
| 267469 |
+
{
|
| 267470 |
+
"epoch": 1056.73,
|
| 267471 |
+
"learning_rate": 7.899391025641027e-06,
|
| 267472 |
+
"loss": 0.342,
|
| 267473 |
+
"step": 132130
|
| 267474 |
+
},
|
| 267475 |
+
{
|
| 267476 |
+
"epoch": 1056.77,
|
| 267477 |
+
"learning_rate": 7.899310897435898e-06,
|
| 267478 |
+
"loss": 0.6652,
|
| 267479 |
+
"step": 132135
|
| 267480 |
+
},
|
| 267481 |
+
{
|
| 267482 |
+
"epoch": 1056.81,
|
| 267483 |
+
"learning_rate": 7.89923076923077e-06,
|
| 267484 |
+
"loss": 1.2066,
|
| 267485 |
+
"step": 132140
|
| 267486 |
+
},
|
| 267487 |
+
{
|
| 267488 |
+
"epoch": 1056.85,
|
| 267489 |
+
"learning_rate": 7.899150641025641e-06,
|
| 267490 |
+
"loss": 0.269,
|
| 267491 |
+
"step": 132145
|
| 267492 |
+
},
|
| 267493 |
+
{
|
| 267494 |
+
"epoch": 1056.89,
|
| 267495 |
+
"learning_rate": 7.899070512820514e-06,
|
| 267496 |
+
"loss": 0.287,
|
| 267497 |
+
"step": 132150
|
| 267498 |
+
},
|
| 267499 |
+
{
|
| 267500 |
+
"epoch": 1056.93,
|
| 267501 |
+
"learning_rate": 7.898990384615385e-06,
|
| 267502 |
+
"loss": 0.306,
|
| 267503 |
+
"step": 132155
|
| 267504 |
+
},
|
| 267505 |
+
{
|
| 267506 |
+
"epoch": 1056.97,
|
| 267507 |
+
"learning_rate": 7.898910256410257e-06,
|
| 267508 |
+
"loss": 0.5773,
|
| 267509 |
+
"step": 132160
|
| 267510 |
+
},
|
| 267511 |
+
{
|
| 267512 |
+
"epoch": 1057.0,
|
| 267513 |
+
"eval_loss": 0.3560391068458557,
|
| 267514 |
+
"eval_runtime": 36.9602,
|
| 267515 |
+
"eval_samples_per_second": 22.619,
|
| 267516 |
+
"eval_steps_per_second": 0.731,
|
| 267517 |
+
"eval_wer": 0.18133865564188467,
|
| 267518 |
+
"step": 132163
|
| 267519 |
+
},
|
| 267520 |
+
{
|
| 267521 |
+
"epoch": 1057.02,
|
| 267522 |
+
"learning_rate": 7.89883012820513e-06,
|
| 267523 |
+
"loss": 0.3881,
|
| 267524 |
+
"step": 132165
|
| 267525 |
+
},
|
| 267526 |
+
{
|
| 267527 |
+
"epoch": 1057.06,
|
| 267528 |
+
"learning_rate": 7.898750000000001e-06,
|
| 267529 |
+
"loss": 0.3016,
|
| 267530 |
+
"step": 132170
|
| 267531 |
+
},
|
| 267532 |
+
{
|
| 267533 |
+
"epoch": 1057.1,
|
| 267534 |
+
"learning_rate": 7.898669871794873e-06,
|
| 267535 |
+
"loss": 0.2899,
|
| 267536 |
+
"step": 132175
|
| 267537 |
+
},
|
| 267538 |
+
{
|
| 267539 |
+
"epoch": 1057.14,
|
| 267540 |
+
"learning_rate": 7.898589743589744e-06,
|
| 267541 |
+
"loss": 0.3101,
|
| 267542 |
+
"step": 132180
|
| 267543 |
+
},
|
| 267544 |
+
{
|
| 267545 |
+
"epoch": 1057.18,
|
| 267546 |
+
"learning_rate": 7.898509615384617e-06,
|
| 267547 |
+
"loss": 0.561,
|
| 267548 |
+
"step": 132185
|
| 267549 |
+
},
|
| 267550 |
+
{
|
| 267551 |
+
"epoch": 1057.22,
|
| 267552 |
+
"learning_rate": 7.898429487179488e-06,
|
| 267553 |
+
"loss": 0.9643,
|
| 267554 |
+
"step": 132190
|
| 267555 |
+
},
|
| 267556 |
+
{
|
| 267557 |
+
"epoch": 1057.26,
|
| 267558 |
+
"learning_rate": 7.89834935897436e-06,
|
| 267559 |
+
"loss": 0.2841,
|
| 267560 |
+
"step": 132195
|
| 267561 |
+
},
|
| 267562 |
+
{
|
| 267563 |
+
"epoch": 1057.3,
|
| 267564 |
+
"learning_rate": 7.898269230769231e-06,
|
| 267565 |
+
"loss": 0.2687,
|
| 267566 |
+
"step": 132200
|
| 267567 |
+
},
|
| 267568 |
+
{
|
| 267569 |
+
"epoch": 1057.34,
|
| 267570 |
+
"learning_rate": 7.898189102564104e-06,
|
| 267571 |
+
"loss": 0.3573,
|
| 267572 |
+
"step": 132205
|
| 267573 |
+
},
|
| 267574 |
+
{
|
| 267575 |
+
"epoch": 1057.38,
|
| 267576 |
+
"learning_rate": 7.898108974358974e-06,
|
| 267577 |
+
"loss": 0.5993,
|
| 267578 |
+
"step": 132210
|
| 267579 |
+
},
|
| 267580 |
+
{
|
| 267581 |
+
"epoch": 1057.42,
|
| 267582 |
+
"learning_rate": 7.898028846153847e-06,
|
| 267583 |
+
"loss": 0.9853,
|
| 267584 |
+
"step": 132215
|
| 267585 |
+
},
|
| 267586 |
+
{
|
| 267587 |
+
"epoch": 1057.46,
|
| 267588 |
+
"learning_rate": 7.89794871794872e-06,
|
| 267589 |
+
"loss": 0.2776,
|
| 267590 |
+
"step": 132220
|
| 267591 |
+
},
|
| 267592 |
+
{
|
| 267593 |
+
"epoch": 1057.5,
|
| 267594 |
+
"learning_rate": 7.89786858974359e-06,
|
| 267595 |
+
"loss": 0.256,
|
| 267596 |
+
"step": 132225
|
| 267597 |
+
},
|
| 267598 |
+
{
|
| 267599 |
+
"epoch": 1057.54,
|
| 267600 |
+
"learning_rate": 7.897788461538463e-06,
|
| 267601 |
+
"loss": 0.3488,
|
| 267602 |
+
"step": 132230
|
| 267603 |
+
},
|
| 267604 |
+
{
|
| 267605 |
+
"epoch": 1057.58,
|
| 267606 |
+
"learning_rate": 7.897708333333334e-06,
|
| 267607 |
+
"loss": 0.6541,
|
| 267608 |
+
"step": 132235
|
| 267609 |
+
},
|
| 267610 |
+
{
|
| 267611 |
+
"epoch": 1057.62,
|
| 267612 |
+
"learning_rate": 7.897628205128205e-06,
|
| 267613 |
+
"loss": 1.0331,
|
| 267614 |
+
"step": 132240
|
| 267615 |
+
},
|
| 267616 |
+
{
|
| 267617 |
+
"epoch": 1057.66,
|
| 267618 |
+
"learning_rate": 7.897548076923077e-06,
|
| 267619 |
+
"loss": 0.244,
|
| 267620 |
+
"step": 132245
|
| 267621 |
+
},
|
| 267622 |
+
{
|
| 267623 |
+
"epoch": 1057.7,
|
| 267624 |
+
"learning_rate": 7.89746794871795e-06,
|
| 267625 |
+
"loss": 0.2768,
|
| 267626 |
+
"step": 132250
|
| 267627 |
+
},
|
| 267628 |
+
{
|
| 267629 |
+
"epoch": 1057.74,
|
| 267630 |
+
"learning_rate": 7.897387820512821e-06,
|
| 267631 |
+
"loss": 0.3966,
|
| 267632 |
+
"step": 132255
|
| 267633 |
+
},
|
| 267634 |
+
{
|
| 267635 |
+
"epoch": 1057.78,
|
| 267636 |
+
"learning_rate": 7.897307692307692e-06,
|
| 267637 |
+
"loss": 0.6784,
|
| 267638 |
+
"step": 132260
|
| 267639 |
+
},
|
| 267640 |
+
{
|
| 267641 |
+
"epoch": 1057.82,
|
| 267642 |
+
"learning_rate": 7.897227564102566e-06,
|
| 267643 |
+
"loss": 0.9715,
|
| 267644 |
+
"step": 132265
|
| 267645 |
+
},
|
| 267646 |
+
{
|
| 267647 |
+
"epoch": 1057.86,
|
| 267648 |
+
"learning_rate": 7.897147435897437e-06,
|
| 267649 |
+
"loss": 0.3056,
|
| 267650 |
+
"step": 132270
|
| 267651 |
+
},
|
| 267652 |
+
{
|
| 267653 |
+
"epoch": 1057.9,
|
| 267654 |
+
"learning_rate": 7.897067307692308e-06,
|
| 267655 |
+
"loss": 0.3011,
|
| 267656 |
+
"step": 132275
|
| 267657 |
+
},
|
| 267658 |
+
{
|
| 267659 |
+
"epoch": 1057.94,
|
| 267660 |
+
"learning_rate": 7.89698717948718e-06,
|
| 267661 |
+
"loss": 0.3895,
|
| 267662 |
+
"step": 132280
|
| 267663 |
+
},
|
| 267664 |
+
{
|
| 267665 |
+
"epoch": 1057.98,
|
| 267666 |
+
"learning_rate": 7.896907051282053e-06,
|
| 267667 |
+
"loss": 0.6871,
|
| 267668 |
+
"step": 132285
|
| 267669 |
+
},
|
| 267670 |
+
{
|
| 267671 |
+
"epoch": 1058.0,
|
| 267672 |
+
"eval_loss": 0.3775459825992584,
|
| 267673 |
+
"eval_runtime": 36.6315,
|
| 267674 |
+
"eval_samples_per_second": 22.795,
|
| 267675 |
+
"eval_steps_per_second": 0.737,
|
| 267676 |
+
"eval_wer": 0.18639436214583632,
|
| 267677 |
+
"step": 132288
|
| 267678 |
+
},
|
| 267679 |
+
{
|
| 267680 |
+
"epoch": 1066.02,
|
| 267681 |
+
"learning_rate": 7.896826923076924e-06,
|
| 267682 |
+
"loss": 0.362,
|
| 267683 |
+
"step": 132290
|
| 267684 |
+
},
|
| 267685 |
+
{
|
| 267686 |
+
"epoch": 1066.06,
|
| 267687 |
+
"learning_rate": 7.896746794871795e-06,
|
| 267688 |
+
"loss": 0.4653,
|
| 267689 |
+
"step": 132295
|
| 267690 |
+
},
|
| 267691 |
+
{
|
| 267692 |
+
"epoch": 1066.1,
|
| 267693 |
+
"learning_rate": 7.896666666666667e-06,
|
| 267694 |
+
"loss": 0.3126,
|
| 267695 |
+
"step": 132300
|
| 267696 |
+
},
|
| 267697 |
+
{
|
| 267698 |
+
"epoch": 1066.14,
|
| 267699 |
+
"learning_rate": 7.89658653846154e-06,
|
| 267700 |
+
"loss": 0.4131,
|
| 267701 |
+
"step": 132305
|
| 267702 |
+
},
|
| 267703 |
+
{
|
| 267704 |
+
"epoch": 1066.18,
|
| 267705 |
+
"learning_rate": 7.896506410256411e-06,
|
| 267706 |
+
"loss": 0.671,
|
| 267707 |
+
"step": 132310
|
| 267708 |
+
},
|
| 267709 |
+
{
|
| 267710 |
+
"epoch": 1066.22,
|
| 267711 |
+
"learning_rate": 7.896426282051282e-06,
|
| 267712 |
+
"loss": 1.0433,
|
| 267713 |
+
"step": 132315
|
| 267714 |
+
},
|
| 267715 |
+
{
|
| 267716 |
+
"epoch": 1066.26,
|
| 267717 |
+
"learning_rate": 7.896346153846156e-06,
|
| 267718 |
+
"loss": 0.2792,
|
| 267719 |
+
"step": 132320
|
| 267720 |
+
},
|
| 267721 |
+
{
|
| 267722 |
+
"epoch": 1066.3,
|
| 267723 |
+
"learning_rate": 7.896266025641027e-06,
|
| 267724 |
+
"loss": 0.2856,
|
| 267725 |
+
"step": 132325
|
| 267726 |
+
},
|
| 267727 |
+
{
|
| 267728 |
+
"epoch": 1066.34,
|
| 267729 |
+
"learning_rate": 7.896185897435898e-06,
|
| 267730 |
+
"loss": 0.3185,
|
| 267731 |
+
"step": 132330
|
| 267732 |
+
},
|
| 267733 |
+
{
|
| 267734 |
+
"epoch": 1066.38,
|
| 267735 |
+
"learning_rate": 7.89610576923077e-06,
|
| 267736 |
+
"loss": 0.5152,
|
| 267737 |
+
"step": 132335
|
| 267738 |
+
},
|
| 267739 |
+
{
|
| 267740 |
+
"epoch": 1066.42,
|
| 267741 |
+
"learning_rate": 7.896025641025643e-06,
|
| 267742 |
+
"loss": 1.2072,
|
| 267743 |
+
"step": 132340
|
| 267744 |
+
},
|
| 267745 |
+
{
|
| 267746 |
+
"epoch": 1066.46,
|
| 267747 |
+
"learning_rate": 7.895945512820512e-06,
|
| 267748 |
+
"loss": 0.2893,
|
| 267749 |
+
"step": 132345
|
| 267750 |
+
},
|
| 267751 |
+
{
|
| 267752 |
+
"epoch": 1066.5,
|
| 267753 |
+
"learning_rate": 7.895865384615385e-06,
|
| 267754 |
+
"loss": 0.3008,
|
| 267755 |
+
"step": 132350
|
| 267756 |
+
},
|
| 267757 |
+
{
|
| 267758 |
+
"epoch": 1066.54,
|
| 267759 |
+
"learning_rate": 7.895785256410257e-06,
|
| 267760 |
+
"loss": 0.3702,
|
| 267761 |
+
"step": 132355
|
| 267762 |
+
},
|
| 267763 |
+
{
|
| 267764 |
+
"epoch": 1066.58,
|
| 267765 |
+
"learning_rate": 7.895705128205128e-06,
|
| 267766 |
+
"loss": 0.5822,
|
| 267767 |
+
"step": 132360
|
| 267768 |
+
},
|
| 267769 |
+
{
|
| 267770 |
+
"epoch": 1066.62,
|
| 267771 |
+
"learning_rate": 7.895641025641027e-06,
|
| 267772 |
+
"loss": 1.1017,
|
| 267773 |
+
"step": 132365
|
| 267774 |
+
},
|
| 267775 |
+
{
|
| 267776 |
+
"epoch": 1066.66,
|
| 267777 |
+
"learning_rate": 7.895560897435897e-06,
|
| 267778 |
+
"loss": 0.2476,
|
| 267779 |
+
"step": 132370
|
| 267780 |
+
},
|
| 267781 |
+
{
|
| 267782 |
+
"epoch": 1066.7,
|
| 267783 |
+
"learning_rate": 7.89548076923077e-06,
|
| 267784 |
+
"loss": 0.4423,
|
| 267785 |
+
"step": 132375
|
| 267786 |
+
},
|
| 267787 |
+
{
|
| 267788 |
+
"epoch": 1066.74,
|
| 267789 |
+
"learning_rate": 7.895400641025641e-06,
|
| 267790 |
+
"loss": 0.371,
|
| 267791 |
+
"step": 132380
|
| 267792 |
+
},
|
| 267793 |
+
{
|
| 267794 |
+
"epoch": 1066.78,
|
| 267795 |
+
"learning_rate": 7.895320512820513e-06,
|
| 267796 |
+
"loss": 0.6137,
|
| 267797 |
+
"step": 132385
|
| 267798 |
+
},
|
| 267799 |
+
{
|
| 267800 |
+
"epoch": 1066.82,
|
| 267801 |
+
"learning_rate": 7.895240384615386e-06,
|
| 267802 |
+
"loss": 1.058,
|
| 267803 |
+
"step": 132390
|
| 267804 |
+
},
|
| 267805 |
+
{
|
| 267806 |
+
"epoch": 1066.86,
|
| 267807 |
+
"learning_rate": 7.895160256410257e-06,
|
| 267808 |
+
"loss": 0.3168,
|
| 267809 |
+
"step": 132395
|
| 267810 |
+
},
|
| 267811 |
+
{
|
| 267812 |
+
"epoch": 1066.9,
|
| 267813 |
+
"learning_rate": 7.895080128205128e-06,
|
| 267814 |
+
"loss": 0.3264,
|
| 267815 |
+
"step": 132400
|
| 267816 |
+
},
|
| 267817 |
+
{
|
| 267818 |
+
"epoch": 1066.94,
|
| 267819 |
+
"learning_rate": 7.895e-06,
|
| 267820 |
+
"loss": 0.3443,
|
| 267821 |
+
"step": 132405
|
| 267822 |
+
},
|
| 267823 |
+
{
|
| 267824 |
+
"epoch": 1066.98,
|
| 267825 |
+
"learning_rate": 7.894919871794873e-06,
|
| 267826 |
+
"loss": 0.5947,
|
| 267827 |
+
"step": 132410
|
| 267828 |
+
},
|
| 267829 |
+
{
|
| 267830 |
+
"epoch": 1067.0,
|
| 267831 |
+
"eval_loss": 0.3375318646430969,
|
| 267832 |
+
"eval_runtime": 36.5033,
|
| 267833 |
+
"eval_samples_per_second": 22.875,
|
| 267834 |
+
"eval_steps_per_second": 0.74,
|
| 267835 |
+
"eval_wer": 0.17405509764609786,
|
| 267836 |
+
"step": 132412
|
| 267837 |
+
},
|
| 267838 |
+
{
|
| 267839 |
+
"epoch": 1059.02,
|
| 267840 |
+
"learning_rate": 7.894839743589744e-06,
|
| 267841 |
+
"loss": 0.3428,
|
| 267842 |
+
"step": 132415
|
| 267843 |
+
},
|
| 267844 |
+
{
|
| 267845 |
+
"epoch": 1059.06,
|
| 267846 |
+
"learning_rate": 7.894759615384616e-06,
|
| 267847 |
+
"loss": 0.2433,
|
| 267848 |
+
"step": 132420
|
| 267849 |
+
},
|
| 267850 |
+
{
|
| 267851 |
+
"epoch": 1059.1,
|
| 267852 |
+
"learning_rate": 7.894679487179487e-06,
|
| 267853 |
+
"loss": 0.2733,
|
| 267854 |
+
"step": 132425
|
| 267855 |
+
},
|
| 267856 |
+
{
|
| 267857 |
+
"epoch": 1059.14,
|
| 267858 |
+
"learning_rate": 7.89459935897436e-06,
|
| 267859 |
+
"loss": 0.471,
|
| 267860 |
+
"step": 132430
|
| 267861 |
+
},
|
| 267862 |
+
{
|
| 267863 |
+
"epoch": 1059.18,
|
| 267864 |
+
"learning_rate": 7.894519230769231e-06,
|
| 267865 |
+
"loss": 0.8245,
|
| 267866 |
+
"step": 132435
|
| 267867 |
+
},
|
| 267868 |
+
{
|
| 267869 |
+
"epoch": 1059.22,
|
| 267870 |
+
"learning_rate": 7.894439102564103e-06,
|
| 267871 |
+
"loss": 0.874,
|
| 267872 |
+
"step": 132440
|
| 267873 |
+
},
|
| 267874 |
+
{
|
| 267875 |
+
"epoch": 1059.26,
|
| 267876 |
+
"learning_rate": 7.894358974358976e-06,
|
| 267877 |
+
"loss": 0.2662,
|
| 267878 |
+
"step": 132445
|
| 267879 |
+
},
|
| 267880 |
+
{
|
| 267881 |
+
"epoch": 1059.3,
|
| 267882 |
+
"learning_rate": 7.894278846153847e-06,
|
| 267883 |
+
"loss": 0.3069,
|
| 267884 |
+
"step": 132450
|
| 267885 |
+
},
|
| 267886 |
+
{
|
| 267887 |
+
"epoch": 1059.34,
|
| 267888 |
+
"learning_rate": 7.894198717948718e-06,
|
| 267889 |
+
"loss": 0.5776,
|
| 267890 |
+
"step": 132455
|
| 267891 |
+
},
|
| 267892 |
+
{
|
| 267893 |
+
"epoch": 1059.38,
|
| 267894 |
+
"learning_rate": 7.89411858974359e-06,
|
| 267895 |
+
"loss": 0.6466,
|
| 267896 |
+
"step": 132460
|
| 267897 |
+
},
|
| 267898 |
+
{
|
| 267899 |
+
"epoch": 1059.42,
|
| 267900 |
+
"learning_rate": 7.894038461538463e-06,
|
| 267901 |
+
"loss": 0.8247,
|
| 267902 |
+
"step": 132465
|
| 267903 |
+
},
|
| 267904 |
+
{
|
| 267905 |
+
"epoch": 1059.46,
|
| 267906 |
+
"learning_rate": 7.893958333333334e-06,
|
| 267907 |
+
"loss": 0.2522,
|
| 267908 |
+
"step": 132470
|
| 267909 |
+
},
|
| 267910 |
+
{
|
| 267911 |
+
"epoch": 1059.5,
|
| 267912 |
+
"learning_rate": 7.893878205128206e-06,
|
| 267913 |
+
"loss": 0.3294,
|
| 267914 |
+
"step": 132475
|
| 267915 |
+
},
|
| 267916 |
+
{
|
| 267917 |
+
"epoch": 1059.54,
|
| 267918 |
+
"learning_rate": 7.893798076923077e-06,
|
| 267919 |
+
"loss": 0.2562,
|
| 267920 |
+
"step": 132480
|
| 267921 |
+
},
|
| 267922 |
+
{
|
| 267923 |
+
"epoch": 1059.58,
|
| 267924 |
+
"learning_rate": 7.89371794871795e-06,
|
| 267925 |
+
"loss": 0.7071,
|
| 267926 |
+
"step": 132485
|
| 267927 |
+
},
|
| 267928 |
+
{
|
| 267929 |
+
"epoch": 1059.62,
|
| 267930 |
+
"learning_rate": 7.893637820512821e-06,
|
| 267931 |
+
"loss": 0.9708,
|
| 267932 |
+
"step": 132490
|
| 267933 |
+
},
|
| 267934 |
+
{
|
| 267935 |
+
"epoch": 1059.66,
|
| 267936 |
+
"learning_rate": 7.893557692307693e-06,
|
| 267937 |
+
"loss": 0.247,
|
| 267938 |
+
"step": 132495
|
| 267939 |
+
},
|
| 267940 |
+
{
|
| 267941 |
+
"epoch": 1059.7,
|
| 267942 |
+
"learning_rate": 7.893477564102566e-06,
|
| 267943 |
+
"loss": 0.3189,
|
| 267944 |
+
"step": 132500
|
| 267945 |
+
},
|
| 267946 |
+
{
|
| 267947 |
+
"epoch": 1059.74,
|
| 267948 |
+
"learning_rate": 7.893397435897437e-06,
|
| 267949 |
+
"loss": 0.3609,
|
| 267950 |
+
"step": 132505
|
| 267951 |
+
},
|
| 267952 |
+
{
|
| 267953 |
+
"epoch": 1059.78,
|
| 267954 |
+
"learning_rate": 7.893317307692308e-06,
|
| 267955 |
+
"loss": 0.7373,
|
| 267956 |
+
"step": 132510
|
| 267957 |
+
},
|
| 267958 |
+
{
|
| 267959 |
+
"epoch": 1059.82,
|
| 267960 |
+
"learning_rate": 7.89323717948718e-06,
|
| 267961 |
+
"loss": 0.8497,
|
| 267962 |
+
"step": 132515
|
| 267963 |
+
},
|
| 267964 |
+
{
|
| 267965 |
+
"epoch": 1059.86,
|
| 267966 |
+
"learning_rate": 7.893157051282053e-06,
|
| 267967 |
+
"loss": 0.2729,
|
| 267968 |
+
"step": 132520
|
| 267969 |
+
},
|
| 267970 |
+
{
|
| 267971 |
+
"epoch": 1059.9,
|
| 267972 |
+
"learning_rate": 7.893076923076923e-06,
|
| 267973 |
+
"loss": 0.2637,
|
| 267974 |
+
"step": 132525
|
| 267975 |
+
},
|
| 267976 |
+
{
|
| 267977 |
+
"epoch": 1059.94,
|
| 267978 |
+
"learning_rate": 7.892996794871796e-06,
|
| 267979 |
+
"loss": 0.3606,
|
| 267980 |
+
"step": 132530
|
| 267981 |
+
},
|
| 267982 |
+
{
|
| 267983 |
+
"epoch": 1059.98,
|
| 267984 |
+
"learning_rate": 7.892916666666667e-06,
|
| 267985 |
+
"loss": 0.6808,
|
| 267986 |
+
"step": 132535
|
| 267987 |
+
},
|
| 267988 |
+
{
|
| 267989 |
+
"epoch": 1060.0,
|
| 267990 |
+
"eval_loss": 0.3099425137042999,
|
| 267991 |
+
"eval_runtime": 36.2358,
|
| 267992 |
+
"eval_samples_per_second": 23.043,
|
| 267993 |
+
"eval_steps_per_second": 0.745,
|
| 267994 |
+
"eval_wer": 0.17826874320898226,
|
| 267995 |
+
"step": 132537
|
| 267996 |
+
},
|
| 267997 |
+
{
|
| 267998 |
+
"epoch": 1068.02,
|
| 267999 |
+
"learning_rate": 7.892836538461538e-06,
|
| 268000 |
+
"loss": 0.3121,
|
| 268001 |
+
"step": 132540
|
| 268002 |
+
},
|
| 268003 |
+
{
|
| 268004 |
+
"epoch": 1068.06,
|
| 268005 |
+
"learning_rate": 7.892756410256411e-06,
|
| 268006 |
+
"loss": 0.2593,
|
| 268007 |
+
"step": 132545
|
| 268008 |
+
},
|
| 268009 |
+
{
|
| 268010 |
+
"epoch": 1068.1,
|
| 268011 |
+
"learning_rate": 7.892676282051283e-06,
|
| 268012 |
+
"loss": 0.3408,
|
| 268013 |
+
"step": 132550
|
| 268014 |
+
},
|
| 268015 |
+
{
|
| 268016 |
+
"epoch": 1068.15,
|
| 268017 |
+
"learning_rate": 7.892596153846154e-06,
|
| 268018 |
+
"loss": 0.3439,
|
| 268019 |
+
"step": 132555
|
| 268020 |
+
},
|
| 268021 |
+
{
|
| 268022 |
+
"epoch": 1068.19,
|
| 268023 |
+
"learning_rate": 7.892516025641025e-06,
|
| 268024 |
+
"loss": 0.6745,
|
| 268025 |
+
"step": 132560
|
| 268026 |
+
},
|
| 268027 |
+
{
|
| 268028 |
+
"epoch": 1068.23,
|
| 268029 |
+
"learning_rate": 7.892435897435899e-06,
|
| 268030 |
+
"loss": 0.786,
|
| 268031 |
+
"step": 132565
|
| 268032 |
+
},
|
| 268033 |
+
{
|
| 268034 |
+
"epoch": 1068.27,
|
| 268035 |
+
"learning_rate": 7.89235576923077e-06,
|
| 268036 |
+
"loss": 0.2751,
|
| 268037 |
+
"step": 132570
|
| 268038 |
+
},
|
| 268039 |
+
{
|
| 268040 |
+
"epoch": 1068.31,
|
| 268041 |
+
"learning_rate": 7.892275641025641e-06,
|
| 268042 |
+
"loss": 0.269,
|
| 268043 |
+
"step": 132575
|
| 268044 |
+
},
|
| 268045 |
+
{
|
| 268046 |
+
"epoch": 1068.35,
|
| 268047 |
+
"learning_rate": 7.892195512820513e-06,
|
| 268048 |
+
"loss": 0.3516,
|
| 268049 |
+
"step": 132580
|
| 268050 |
+
},
|
| 268051 |
+
{
|
| 268052 |
+
"epoch": 1068.39,
|
| 268053 |
+
"learning_rate": 7.892115384615386e-06,
|
| 268054 |
+
"loss": 0.7291,
|
| 268055 |
+
"step": 132585
|
| 268056 |
+
},
|
| 268057 |
+
{
|
| 268058 |
+
"epoch": 1068.43,
|
| 268059 |
+
"learning_rate": 7.892035256410257e-06,
|
| 268060 |
+
"loss": 0.8608,
|
| 268061 |
+
"step": 132590
|
| 268062 |
+
},
|
| 268063 |
+
{
|
| 268064 |
+
"epoch": 1068.47,
|
| 268065 |
+
"learning_rate": 7.891955128205128e-06,
|
| 268066 |
+
"loss": 0.2534,
|
| 268067 |
+
"step": 132595
|
| 268068 |
+
},
|
| 268069 |
+
{
|
| 268070 |
+
"epoch": 1068.51,
|
| 268071 |
+
"learning_rate": 7.891875000000001e-06,
|
| 268072 |
+
"loss": 0.2694,
|
| 268073 |
+
"step": 132600
|
| 268074 |
+
},
|
| 268075 |
+
{
|
| 268076 |
+
"epoch": 1068.55,
|
| 268077 |
+
"learning_rate": 7.891794871794873e-06,
|
| 268078 |
+
"loss": 0.3579,
|
| 268079 |
+
"step": 132605
|
| 268080 |
+
},
|
| 268081 |
+
{
|
| 268082 |
+
"epoch": 1068.59,
|
| 268083 |
+
"learning_rate": 7.891714743589744e-06,
|
| 268084 |
+
"loss": 0.7137,
|
| 268085 |
+
"step": 132610
|
| 268086 |
+
},
|
| 268087 |
+
{
|
| 268088 |
+
"epoch": 1068.63,
|
| 268089 |
+
"learning_rate": 7.891634615384615e-06,
|
| 268090 |
+
"loss": 0.9326,
|
| 268091 |
+
"step": 132615
|
| 268092 |
+
},
|
| 268093 |
+
{
|
| 268094 |
+
"epoch": 1068.67,
|
| 268095 |
+
"learning_rate": 7.891554487179489e-06,
|
| 268096 |
+
"loss": 0.255,
|
| 268097 |
+
"step": 132620
|
| 268098 |
+
},
|
| 268099 |
+
{
|
| 268100 |
+
"epoch": 1068.71,
|
| 268101 |
+
"learning_rate": 7.89147435897436e-06,
|
| 268102 |
+
"loss": 0.2923,
|
| 268103 |
+
"step": 132625
|
| 268104 |
+
},
|
| 268105 |
+
{
|
| 268106 |
+
"epoch": 1068.75,
|
| 268107 |
+
"learning_rate": 7.891394230769231e-06,
|
| 268108 |
+
"loss": 0.3931,
|
| 268109 |
+
"step": 132630
|
| 268110 |
+
},
|
| 268111 |
+
{
|
| 268112 |
+
"epoch": 1068.79,
|
| 268113 |
+
"learning_rate": 7.891314102564103e-06,
|
| 268114 |
+
"loss": 0.7453,
|
| 268115 |
+
"step": 132635
|
| 268116 |
+
},
|
| 268117 |
+
{
|
| 268118 |
+
"epoch": 1068.83,
|
| 268119 |
+
"learning_rate": 7.891233974358976e-06,
|
| 268120 |
+
"loss": 0.7781,
|
| 268121 |
+
"step": 132640
|
| 268122 |
+
},
|
| 268123 |
+
{
|
| 268124 |
+
"epoch": 1068.87,
|
| 268125 |
+
"learning_rate": 7.891153846153847e-06,
|
| 268126 |
+
"loss": 0.2954,
|
| 268127 |
+
"step": 132645
|
| 268128 |
+
},
|
| 268129 |
+
{
|
| 268130 |
+
"epoch": 1068.91,
|
| 268131 |
+
"learning_rate": 7.891073717948718e-06,
|
| 268132 |
+
"loss": 0.2961,
|
| 268133 |
+
"step": 132650
|
| 268134 |
+
},
|
| 268135 |
+
{
|
| 268136 |
+
"epoch": 1068.95,
|
| 268137 |
+
"learning_rate": 7.890993589743591e-06,
|
| 268138 |
+
"loss": 0.3804,
|
| 268139 |
+
"step": 132655
|
| 268140 |
+
},
|
| 268141 |
+
{
|
| 268142 |
+
"epoch": 1068.99,
|
| 268143 |
+
"learning_rate": 7.890913461538461e-06,
|
| 268144 |
+
"loss": 0.8052,
|
| 268145 |
+
"step": 132660
|
| 268146 |
+
},
|
| 268147 |
+
{
|
| 268148 |
+
"epoch": 1069.0,
|
| 268149 |
+
"eval_loss": 0.3759481608867645,
|
| 268150 |
+
"eval_runtime": 35.5265,
|
| 268151 |
+
"eval_samples_per_second": 23.504,
|
| 268152 |
+
"eval_steps_per_second": 0.76,
|
| 268153 |
+
"eval_wer": 0.17900044490582825,
|
| 268154 |
+
"step": 132661
|
| 268155 |
+
},
|
| 268156 |
+
{
|
| 268157 |
+
"epoch": 1061.03,
|
| 268158 |
+
"learning_rate": 7.890833333333334e-06,
|
| 268159 |
+
"loss": 0.3047,
|
| 268160 |
+
"step": 132665
|
| 268161 |
+
},
|
| 268162 |
+
{
|
| 268163 |
+
"epoch": 1061.07,
|
| 268164 |
+
"learning_rate": 7.890753205128206e-06,
|
| 268165 |
+
"loss": 0.262,
|
| 268166 |
+
"step": 132670
|
| 268167 |
+
},
|
| 268168 |
+
{
|
| 268169 |
+
"epoch": 1061.11,
|
| 268170 |
+
"learning_rate": 7.890673076923077e-06,
|
| 268171 |
+
"loss": 0.3535,
|
| 268172 |
+
"step": 132675
|
| 268173 |
+
},
|
| 268174 |
+
{
|
| 268175 |
+
"epoch": 1061.15,
|
| 268176 |
+
"learning_rate": 7.890592948717948e-06,
|
| 268177 |
+
"loss": 0.3661,
|
| 268178 |
+
"step": 132680
|
| 268179 |
+
},
|
| 268180 |
+
{
|
| 268181 |
+
"epoch": 1061.19,
|
| 268182 |
+
"learning_rate": 7.890512820512821e-06,
|
| 268183 |
+
"loss": 0.9716,
|
| 268184 |
+
"step": 132685
|
| 268185 |
+
},
|
| 268186 |
+
{
|
| 268187 |
+
"epoch": 1061.23,
|
| 268188 |
+
"learning_rate": 7.890432692307693e-06,
|
| 268189 |
+
"loss": 0.6551,
|
| 268190 |
+
"step": 132690
|
| 268191 |
+
},
|
| 268192 |
+
{
|
| 268193 |
+
"epoch": 1061.27,
|
| 268194 |
+
"learning_rate": 7.890352564102564e-06,
|
| 268195 |
+
"loss": 0.2745,
|
| 268196 |
+
"step": 132695
|
| 268197 |
+
},
|
| 268198 |
+
{
|
| 268199 |
+
"epoch": 1061.31,
|
| 268200 |
+
"learning_rate": 7.890272435897437e-06,
|
| 268201 |
+
"loss": 0.3181,
|
| 268202 |
+
"step": 132700
|
| 268203 |
+
},
|
| 268204 |
+
{
|
| 268205 |
+
"epoch": 1061.35,
|
| 268206 |
+
"learning_rate": 7.890192307692308e-06,
|
| 268207 |
+
"loss": 0.3705,
|
| 268208 |
+
"step": 132705
|
| 268209 |
+
},
|
| 268210 |
+
{
|
| 268211 |
+
"epoch": 1061.39,
|
| 268212 |
+
"learning_rate": 7.89011217948718e-06,
|
| 268213 |
+
"loss": 0.8255,
|
| 268214 |
+
"step": 132710
|
| 268215 |
+
},
|
| 268216 |
+
{
|
| 268217 |
+
"epoch": 1061.43,
|
| 268218 |
+
"learning_rate": 7.890032051282051e-06,
|
| 268219 |
+
"loss": 0.6142,
|
| 268220 |
+
"step": 132715
|
| 268221 |
+
},
|
| 268222 |
+
{
|
| 268223 |
+
"epoch": 1061.47,
|
| 268224 |
+
"learning_rate": 7.889951923076924e-06,
|
| 268225 |
+
"loss": 0.266,
|
| 268226 |
+
"step": 132720
|
| 268227 |
+
},
|
| 268228 |
+
{
|
| 268229 |
+
"epoch": 1061.51,
|
| 268230 |
+
"learning_rate": 7.889871794871796e-06,
|
| 268231 |
+
"loss": 0.2973,
|
| 268232 |
+
"step": 132725
|
| 268233 |
+
},
|
| 268234 |
+
{
|
| 268235 |
+
"epoch": 1061.55,
|
| 268236 |
+
"learning_rate": 7.889791666666667e-06,
|
| 268237 |
+
"loss": 0.4102,
|
| 268238 |
+
"step": 132730
|
| 268239 |
+
},
|
| 268240 |
+
{
|
| 268241 |
+
"epoch": 1061.59,
|
| 268242 |
+
"learning_rate": 7.889711538461538e-06,
|
| 268243 |
+
"loss": 0.863,
|
| 268244 |
+
"step": 132735
|
| 268245 |
+
},
|
| 268246 |
+
{
|
| 268247 |
+
"epoch": 1061.63,
|
| 268248 |
+
"learning_rate": 7.889631410256411e-06,
|
| 268249 |
+
"loss": 0.6402,
|
| 268250 |
+
"step": 132740
|
| 268251 |
+
},
|
| 268252 |
+
{
|
| 268253 |
+
"epoch": 1061.67,
|
| 268254 |
+
"learning_rate": 7.889551282051283e-06,
|
| 268255 |
+
"loss": 0.2461,
|
| 268256 |
+
"step": 132745
|
| 268257 |
+
},
|
| 268258 |
+
{
|
| 268259 |
+
"epoch": 1061.71,
|
| 268260 |
+
"learning_rate": 7.889471153846154e-06,
|
| 268261 |
+
"loss": 0.3107,
|
| 268262 |
+
"step": 132750
|
| 268263 |
+
},
|
| 268264 |
+
{
|
| 268265 |
+
"epoch": 1061.75,
|
| 268266 |
+
"learning_rate": 7.889391025641027e-06,
|
| 268267 |
+
"loss": 0.4564,
|
| 268268 |
+
"step": 132755
|
| 268269 |
+
},
|
| 268270 |
+
{
|
| 268271 |
+
"epoch": 1061.79,
|
| 268272 |
+
"learning_rate": 7.889310897435898e-06,
|
| 268273 |
+
"loss": 0.8667,
|
| 268274 |
+
"step": 132760
|
| 268275 |
+
},
|
| 268276 |
+
{
|
| 268277 |
+
"epoch": 1061.83,
|
| 268278 |
+
"learning_rate": 7.88923076923077e-06,
|
| 268279 |
+
"loss": 0.6114,
|
| 268280 |
+
"step": 132765
|
| 268281 |
+
},
|
| 268282 |
+
{
|
| 268283 |
+
"epoch": 1061.87,
|
| 268284 |
+
"learning_rate": 7.889150641025641e-06,
|
| 268285 |
+
"loss": 0.2584,
|
| 268286 |
+
"step": 132770
|
| 268287 |
+
},
|
| 268288 |
+
{
|
| 268289 |
+
"epoch": 1061.91,
|
| 268290 |
+
"learning_rate": 7.889070512820514e-06,
|
| 268291 |
+
"loss": 0.2723,
|
| 268292 |
+
"step": 132775
|
| 268293 |
+
},
|
| 268294 |
+
{
|
| 268295 |
+
"epoch": 1061.95,
|
| 268296 |
+
"learning_rate": 7.888990384615386e-06,
|
| 268297 |
+
"loss": 0.3697,
|
| 268298 |
+
"step": 132780
|
| 268299 |
+
},
|
| 268300 |
+
{
|
| 268301 |
+
"epoch": 1061.99,
|
| 268302 |
+
"learning_rate": 7.888910256410257e-06,
|
| 268303 |
+
"loss": 0.9063,
|
| 268304 |
+
"step": 132785
|
| 268305 |
+
},
|
| 268306 |
+
{
|
| 268307 |
+
"epoch": 1062.0,
|
| 268308 |
+
"eval_loss": 0.35875725746154785,
|
| 268309 |
+
"eval_runtime": 35.2957,
|
| 268310 |
+
"eval_samples_per_second": 23.657,
|
| 268311 |
+
"eval_steps_per_second": 0.765,
|
| 268312 |
+
"eval_wer": 0.18351754320896077,
|
| 268313 |
+
"step": 132786
|
| 268314 |
+
},
|
| 268315 |
+
{
|
| 268316 |
+
"epoch": 1070.03,
|
| 268317 |
+
"learning_rate": 7.88883012820513e-06,
|
| 268318 |
+
"loss": 0.3289,
|
| 268319 |
+
"step": 132790
|
| 268320 |
+
},
|
| 268321 |
+
{
|
| 268322 |
+
"epoch": 1070.07,
|
| 268323 |
+
"learning_rate": 7.888750000000001e-06,
|
| 268324 |
+
"loss": 0.3066,
|
| 268325 |
+
"step": 132795
|
| 268326 |
+
},
|
| 268327 |
+
{
|
| 268328 |
+
"epoch": 1070.11,
|
| 268329 |
+
"learning_rate": 7.888669871794873e-06,
|
| 268330 |
+
"loss": 0.3182,
|
| 268331 |
+
"step": 132800
|
| 268332 |
+
},
|
| 268333 |
+
{
|
| 268334 |
+
"epoch": 1070.15,
|
| 268335 |
+
"learning_rate": 7.888589743589744e-06,
|
| 268336 |
+
"loss": 0.3897,
|
| 268337 |
+
"step": 132805
|
| 268338 |
+
},
|
| 268339 |
+
{
|
| 268340 |
+
"epoch": 1070.19,
|
| 268341 |
+
"learning_rate": 7.888509615384617e-06,
|
| 268342 |
+
"loss": 0.9667,
|
| 268343 |
+
"step": 132810
|
| 268344 |
+
},
|
| 268345 |
+
{
|
| 268346 |
+
"epoch": 1070.23,
|
| 268347 |
+
"learning_rate": 7.888429487179487e-06,
|
| 268348 |
+
"loss": 0.651,
|
| 268349 |
+
"step": 132815
|
| 268350 |
+
},
|
| 268351 |
+
{
|
| 268352 |
+
"epoch": 1070.27,
|
| 268353 |
+
"learning_rate": 7.88834935897436e-06,
|
| 268354 |
+
"loss": 0.2818,
|
| 268355 |
+
"step": 132820
|
| 268356 |
+
},
|
| 268357 |
+
{
|
| 268358 |
+
"epoch": 1070.31,
|
| 268359 |
+
"learning_rate": 7.888269230769231e-06,
|
| 268360 |
+
"loss": 0.2908,
|
| 268361 |
+
"step": 132825
|
| 268362 |
+
},
|
| 268363 |
+
{
|
| 268364 |
+
"epoch": 1070.35,
|
| 268365 |
+
"learning_rate": 7.888189102564103e-06,
|
| 268366 |
+
"loss": 0.4134,
|
| 268367 |
+
"step": 132830
|
| 268368 |
+
},
|
| 268369 |
+
{
|
| 268370 |
+
"epoch": 1070.4,
|
| 268371 |
+
"learning_rate": 7.888108974358974e-06,
|
| 268372 |
+
"loss": 1.0368,
|
| 268373 |
+
"step": 132835
|
| 268374 |
+
},
|
| 268375 |
+
{
|
| 268376 |
+
"epoch": 1070.44,
|
| 268377 |
+
"learning_rate": 7.888028846153847e-06,
|
| 268378 |
+
"loss": 0.6221,
|
| 268379 |
+
"step": 132840
|
| 268380 |
+
},
|
| 268381 |
+
{
|
| 268382 |
+
"epoch": 1070.48,
|
| 268383 |
+
"learning_rate": 7.887948717948718e-06,
|
| 268384 |
+
"loss": 0.2551,
|
| 268385 |
+
"step": 132845
|
| 268386 |
+
},
|
| 268387 |
+
{
|
| 268388 |
+
"epoch": 1070.52,
|
| 268389 |
+
"learning_rate": 7.88786858974359e-06,
|
| 268390 |
+
"loss": 0.3402,
|
| 268391 |
+
"step": 132850
|
| 268392 |
+
},
|
| 268393 |
+
{
|
| 268394 |
+
"epoch": 1070.56,
|
| 268395 |
+
"learning_rate": 7.887788461538463e-06,
|
| 268396 |
+
"loss": 0.377,
|
| 268397 |
+
"step": 132855
|
| 268398 |
+
},
|
| 268399 |
+
{
|
| 268400 |
+
"epoch": 1070.6,
|
| 268401 |
+
"learning_rate": 7.887708333333334e-06,
|
| 268402 |
+
"loss": 0.9088,
|
| 268403 |
+
"step": 132860
|
| 268404 |
+
},
|
| 268405 |
+
{
|
| 268406 |
+
"epoch": 1070.64,
|
| 268407 |
+
"learning_rate": 7.887628205128205e-06,
|
| 268408 |
+
"loss": 0.5489,
|
| 268409 |
+
"step": 132865
|
| 268410 |
+
},
|
| 268411 |
+
{
|
| 268412 |
+
"epoch": 1070.68,
|
| 268413 |
+
"learning_rate": 7.887548076923077e-06,
|
| 268414 |
+
"loss": 0.3574,
|
| 268415 |
+
"step": 132870
|
| 268416 |
+
},
|
| 268417 |
+
{
|
| 268418 |
+
"epoch": 1070.72,
|
| 268419 |
+
"learning_rate": 7.88746794871795e-06,
|
| 268420 |
+
"loss": 0.2859,
|
| 268421 |
+
"step": 132875
|
| 268422 |
+
},
|
| 268423 |
+
{
|
| 268424 |
+
"epoch": 1070.76,
|
| 268425 |
+
"learning_rate": 7.887387820512821e-06,
|
| 268426 |
+
"loss": 0.3736,
|
| 268427 |
+
"step": 132880
|
| 268428 |
+
},
|
| 268429 |
+
{
|
| 268430 |
+
"epoch": 1070.8,
|
| 268431 |
+
"learning_rate": 7.887307692307693e-06,
|
| 268432 |
+
"loss": 0.9012,
|
| 268433 |
+
"step": 132885
|
| 268434 |
+
},
|
| 268435 |
+
{
|
| 268436 |
+
"epoch": 1070.84,
|
| 268437 |
+
"learning_rate": 7.887227564102566e-06,
|
| 268438 |
+
"loss": 0.6423,
|
| 268439 |
+
"step": 132890
|
| 268440 |
+
},
|
| 268441 |
+
{
|
| 268442 |
+
"epoch": 1070.88,
|
| 268443 |
+
"learning_rate": 7.887147435897437e-06,
|
| 268444 |
+
"loss": 0.2914,
|
| 268445 |
+
"step": 132895
|
| 268446 |
+
},
|
| 268447 |
+
{
|
| 268448 |
+
"epoch": 1070.92,
|
| 268449 |
+
"learning_rate": 7.887067307692308e-06,
|
| 268450 |
+
"loss": 0.3114,
|
| 268451 |
+
"step": 132900
|
| 268452 |
+
},
|
| 268453 |
+
{
|
| 268454 |
+
"epoch": 1070.96,
|
| 268455 |
+
"learning_rate": 7.88698717948718e-06,
|
| 268456 |
+
"loss": 0.4223,
|
| 268457 |
+
"step": 132905
|
| 268458 |
+
},
|
| 268459 |
+
{
|
| 268460 |
+
"epoch": 1071.0,
|
| 268461 |
+
"learning_rate": 7.886907051282053e-06,
|
| 268462 |
+
"loss": 1.1008,
|
| 268463 |
+
"step": 132910
|
| 268464 |
+
},
|
| 268465 |
+
{
|
| 268466 |
+
"epoch": 1071.0,
|
| 268467 |
+
"eval_loss": 0.39602500200271606,
|
| 268468 |
+
"eval_runtime": 36.0699,
|
| 268469 |
+
"eval_samples_per_second": 23.15,
|
| 268470 |
+
"eval_steps_per_second": 0.749,
|
| 268471 |
+
"eval_wer": 0.1689111747851003,
|
| 268472 |
+
"step": 132910
|
| 268473 |
}
|
| 268474 |
],
|
| 268475 |
"max_steps": 620000,
|
| 268476 |
"num_train_epochs": 5000,
|
| 268477 |
+
"total_flos": 3.739984126331327e+20,
|
| 268478 |
"trial_name": null,
|
| 268479 |
"trial_params": null
|
| 268480 |
}
|
model-bin/finetune/base/{checkpoint-132288 → checkpoint-132910}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630169499.9794455/events.out.tfevents.1630169499.86bb0ddabf9b.4092.231
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cb94fb78cc5291d6ed73aed19e845ed72507274a13590773a13fba334e273d65
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630169887.7954714/events.out.tfevents.1630169887.86bb0ddabf9b.4092.233
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bf8a7f58841a431d4683b8614895539eaefbd8e9b7fe7aa5bbc6279db05ef909
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630170280.3267503/events.out.tfevents.1630170280.86bb0ddabf9b.4092.235
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:90959cbf6d4d056b1c845dff9a0d049630e1f1570f5d97811fd282ef2e84f397
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630170670.464988/events.out.tfevents.1630170670.86bb0ddabf9b.4092.237
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2e7e948a632da86fa4490db73f7490ad89281dc8286b8d7c4cf82be53f8f6d3e
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630171053.185435/events.out.tfevents.1630171053.86bb0ddabf9b.4092.239
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2ae40200e0c98e112f4b99b6e9801537d7e1ceee6dc697141f8b5f70ccb677b3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630169499.86bb0ddabf9b.4092.230
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:adb44b33fde4c4e7879d17af993fa5ba1d6b9263bf70dc1990d26173b631111c
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630169887.86bb0ddabf9b.4092.232
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a1305551f590d663de5201823ba1fff615fd550b39e264b3aa79fcd78e1e626e
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630170280.86bb0ddabf9b.4092.234
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bcc2a597c15648a9642b9b456057536a6f19c4fa97167f68955470d361a447e7
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630170670.86bb0ddabf9b.4092.236
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e947bc8c335596f14eb5fcc2ad8626d0fa93f005c891ee78a52ea36530ab6e24
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630171053.86bb0ddabf9b.4092.238
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:02bfef2f06f36f911f3222ad415a048e76a71616a09d83319adb6656ab7d483b
|
| 3 |
+
size 8622
|