"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630115798.4136324/events.out.tfevents.1630115798.86bb0ddabf9b.1042.41 +3 -0
- model-bin/finetune/base/log/1630116207.9281766/events.out.tfevents.1630116207.86bb0ddabf9b.1042.43 +3 -0
- model-bin/finetune/base/log/1630116621.3491957/events.out.tfevents.1630116621.86bb0ddabf9b.1042.45 +3 -0
- model-bin/finetune/base/log/1630117036.3824139/events.out.tfevents.1630117036.86bb0ddabf9b.1042.47 +3 -0
- model-bin/finetune/base/log/1630117449.777196/events.out.tfevents.1630117449.86bb0ddabf9b.1042.49 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630115798.86bb0ddabf9b.1042.40 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630116207.86bb0ddabf9b.1042.42 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630116621.86bb0ddabf9b.1042.44 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630117036.86bb0ddabf9b.1042.46 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630117449.86bb0ddabf9b.1042.48 +3 -0
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/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:ce19a2e5b74de571f58c07d682570d2210d3daa6eb857675fefc28ee8ffe8cfe
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/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:fe6f837af1e6098ca87218f9f6f489ef7326656822686625f997aa9e4da06051
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/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 sha256:310447b7017194990c17a691e647cf4393cfa4912178125648656448f89be377
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/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:f1caec4dc20dfb31b5977c6031ec782c60b5f7b286f94be7ec1097ec7544efc9
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/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 sha256:747fd4894013b519f10367353486f379ccd36aaffe81e92b6be8bd41396f7093
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -247560,11 +247560,806 @@
|
|
| 247560 |
"eval_steps_per_second": 0.686,
|
| 247561 |
"eval_wer": 0.18822154026011984,
|
| 247562 |
"step": 116478
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 247563 |
}
|
| 247564 |
],
|
| 247565 |
"max_steps": 625000,
|
| 247566 |
"num_train_epochs": 5000,
|
| 247567 |
-
"total_flos": 3.
|
| 247568 |
"trial_name": null,
|
| 247569 |
"trial_params": null
|
| 247570 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 935.9960159362549,
|
| 5 |
+
"global_step": 117100,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 247560 |
"eval_steps_per_second": 0.686,
|
| 247561 |
"eval_wer": 0.18822154026011984,
|
| 247562 |
"step": 116478
|
| 247563 |
+
},
|
| 247564 |
+
{
|
| 247565 |
+
"epoch": 931.02,
|
| 247566 |
+
"learning_rate": 8.15011217948718e-06,
|
| 247567 |
+
"loss": 0.2933,
|
| 247568 |
+
"step": 116480
|
| 247569 |
+
},
|
| 247570 |
+
{
|
| 247571 |
+
"epoch": 931.06,
|
| 247572 |
+
"learning_rate": 8.150032051282052e-06,
|
| 247573 |
+
"loss": 0.292,
|
| 247574 |
+
"step": 116485
|
| 247575 |
+
},
|
| 247576 |
+
{
|
| 247577 |
+
"epoch": 931.1,
|
| 247578 |
+
"learning_rate": 8.149951923076923e-06,
|
| 247579 |
+
"loss": 0.2725,
|
| 247580 |
+
"step": 116490
|
| 247581 |
+
},
|
| 247582 |
+
{
|
| 247583 |
+
"epoch": 931.14,
|
| 247584 |
+
"learning_rate": 8.149871794871795e-06,
|
| 247585 |
+
"loss": 0.383,
|
| 247586 |
+
"step": 116495
|
| 247587 |
+
},
|
| 247588 |
+
{
|
| 247589 |
+
"epoch": 931.18,
|
| 247590 |
+
"learning_rate": 8.149791666666668e-06,
|
| 247591 |
+
"loss": 0.6222,
|
| 247592 |
+
"step": 116500
|
| 247593 |
+
},
|
| 247594 |
+
{
|
| 247595 |
+
"epoch": 931.22,
|
| 247596 |
+
"learning_rate": 8.14971153846154e-06,
|
| 247597 |
+
"loss": 0.9348,
|
| 247598 |
+
"step": 116505
|
| 247599 |
+
},
|
| 247600 |
+
{
|
| 247601 |
+
"epoch": 931.26,
|
| 247602 |
+
"learning_rate": 8.14963141025641e-06,
|
| 247603 |
+
"loss": 0.2812,
|
| 247604 |
+
"step": 116510
|
| 247605 |
+
},
|
| 247606 |
+
{
|
| 247607 |
+
"epoch": 931.3,
|
| 247608 |
+
"learning_rate": 8.149551282051282e-06,
|
| 247609 |
+
"loss": 0.2711,
|
| 247610 |
+
"step": 116515
|
| 247611 |
+
},
|
| 247612 |
+
{
|
| 247613 |
+
"epoch": 931.34,
|
| 247614 |
+
"learning_rate": 8.149471153846155e-06,
|
| 247615 |
+
"loss": 0.4673,
|
| 247616 |
+
"step": 116520
|
| 247617 |
+
},
|
| 247618 |
+
{
|
| 247619 |
+
"epoch": 931.38,
|
| 247620 |
+
"learning_rate": 8.149391025641026e-06,
|
| 247621 |
+
"loss": 0.7297,
|
| 247622 |
+
"step": 116525
|
| 247623 |
+
},
|
| 247624 |
+
{
|
| 247625 |
+
"epoch": 931.42,
|
| 247626 |
+
"learning_rate": 8.149310897435898e-06,
|
| 247627 |
+
"loss": 1.0439,
|
| 247628 |
+
"step": 116530
|
| 247629 |
+
},
|
| 247630 |
+
{
|
| 247631 |
+
"epoch": 931.46,
|
| 247632 |
+
"learning_rate": 8.14923076923077e-06,
|
| 247633 |
+
"loss": 0.3004,
|
| 247634 |
+
"step": 116535
|
| 247635 |
+
},
|
| 247636 |
+
{
|
| 247637 |
+
"epoch": 931.5,
|
| 247638 |
+
"learning_rate": 8.149150641025642e-06,
|
| 247639 |
+
"loss": 0.289,
|
| 247640 |
+
"step": 116540
|
| 247641 |
+
},
|
| 247642 |
+
{
|
| 247643 |
+
"epoch": 931.54,
|
| 247644 |
+
"learning_rate": 8.149070512820513e-06,
|
| 247645 |
+
"loss": 0.532,
|
| 247646 |
+
"step": 116545
|
| 247647 |
+
},
|
| 247648 |
+
{
|
| 247649 |
+
"epoch": 931.58,
|
| 247650 |
+
"learning_rate": 8.148990384615385e-06,
|
| 247651 |
+
"loss": 0.6062,
|
| 247652 |
+
"step": 116550
|
| 247653 |
+
},
|
| 247654 |
+
{
|
| 247655 |
+
"epoch": 931.62,
|
| 247656 |
+
"learning_rate": 8.148910256410258e-06,
|
| 247657 |
+
"loss": 1.037,
|
| 247658 |
+
"step": 116555
|
| 247659 |
+
},
|
| 247660 |
+
{
|
| 247661 |
+
"epoch": 931.66,
|
| 247662 |
+
"learning_rate": 8.14883012820513e-06,
|
| 247663 |
+
"loss": 0.2608,
|
| 247664 |
+
"step": 116560
|
| 247665 |
+
},
|
| 247666 |
+
{
|
| 247667 |
+
"epoch": 931.7,
|
| 247668 |
+
"learning_rate": 8.14875e-06,
|
| 247669 |
+
"loss": 0.2704,
|
| 247670 |
+
"step": 116565
|
| 247671 |
+
},
|
| 247672 |
+
{
|
| 247673 |
+
"epoch": 931.74,
|
| 247674 |
+
"learning_rate": 8.148669871794872e-06,
|
| 247675 |
+
"loss": 0.3608,
|
| 247676 |
+
"step": 116570
|
| 247677 |
+
},
|
| 247678 |
+
{
|
| 247679 |
+
"epoch": 931.78,
|
| 247680 |
+
"learning_rate": 8.148589743589745e-06,
|
| 247681 |
+
"loss": 0.615,
|
| 247682 |
+
"step": 116575
|
| 247683 |
+
},
|
| 247684 |
+
{
|
| 247685 |
+
"epoch": 931.82,
|
| 247686 |
+
"learning_rate": 8.148509615384616e-06,
|
| 247687 |
+
"loss": 0.9766,
|
| 247688 |
+
"step": 116580
|
| 247689 |
+
},
|
| 247690 |
+
{
|
| 247691 |
+
"epoch": 931.86,
|
| 247692 |
+
"learning_rate": 8.148429487179488e-06,
|
| 247693 |
+
"loss": 0.3059,
|
| 247694 |
+
"step": 116585
|
| 247695 |
+
},
|
| 247696 |
+
{
|
| 247697 |
+
"epoch": 931.9,
|
| 247698 |
+
"learning_rate": 8.14834935897436e-06,
|
| 247699 |
+
"loss": 0.2919,
|
| 247700 |
+
"step": 116590
|
| 247701 |
+
},
|
| 247702 |
+
{
|
| 247703 |
+
"epoch": 931.94,
|
| 247704 |
+
"learning_rate": 8.148269230769232e-06,
|
| 247705 |
+
"loss": 0.3559,
|
| 247706 |
+
"step": 116595
|
| 247707 |
+
},
|
| 247708 |
+
{
|
| 247709 |
+
"epoch": 931.98,
|
| 247710 |
+
"learning_rate": 8.148189102564103e-06,
|
| 247711 |
+
"loss": 0.7633,
|
| 247712 |
+
"step": 116600
|
| 247713 |
+
},
|
| 247714 |
+
{
|
| 247715 |
+
"epoch": 932.0,
|
| 247716 |
+
"eval_loss": 0.35194557905197144,
|
| 247717 |
+
"eval_runtime": 39.5509,
|
| 247718 |
+
"eval_samples_per_second": 21.314,
|
| 247719 |
+
"eval_steps_per_second": 0.683,
|
| 247720 |
+
"eval_wer": 0.18437974134816848,
|
| 247721 |
+
"step": 116603
|
| 247722 |
+
},
|
| 247723 |
+
{
|
| 247724 |
+
"epoch": 940.02,
|
| 247725 |
+
"learning_rate": 8.148108974358975e-06,
|
| 247726 |
+
"loss": 0.3737,
|
| 247727 |
+
"step": 116605
|
| 247728 |
+
},
|
| 247729 |
+
{
|
| 247730 |
+
"epoch": 940.06,
|
| 247731 |
+
"learning_rate": 8.148028846153848e-06,
|
| 247732 |
+
"loss": 0.2789,
|
| 247733 |
+
"step": 116610
|
| 247734 |
+
},
|
| 247735 |
+
{
|
| 247736 |
+
"epoch": 940.1,
|
| 247737 |
+
"learning_rate": 8.147948717948718e-06,
|
| 247738 |
+
"loss": 0.2851,
|
| 247739 |
+
"step": 116615
|
| 247740 |
+
},
|
| 247741 |
+
{
|
| 247742 |
+
"epoch": 940.14,
|
| 247743 |
+
"learning_rate": 8.14786858974359e-06,
|
| 247744 |
+
"loss": 0.3367,
|
| 247745 |
+
"step": 116620
|
| 247746 |
+
},
|
| 247747 |
+
{
|
| 247748 |
+
"epoch": 940.18,
|
| 247749 |
+
"learning_rate": 8.147788461538462e-06,
|
| 247750 |
+
"loss": 0.679,
|
| 247751 |
+
"step": 116625
|
| 247752 |
+
},
|
| 247753 |
+
{
|
| 247754 |
+
"epoch": 940.22,
|
| 247755 |
+
"learning_rate": 8.147708333333333e-06,
|
| 247756 |
+
"loss": 0.9981,
|
| 247757 |
+
"step": 116630
|
| 247758 |
+
},
|
| 247759 |
+
{
|
| 247760 |
+
"epoch": 940.26,
|
| 247761 |
+
"learning_rate": 8.147628205128206e-06,
|
| 247762 |
+
"loss": 0.3318,
|
| 247763 |
+
"step": 116635
|
| 247764 |
+
},
|
| 247765 |
+
{
|
| 247766 |
+
"epoch": 940.3,
|
| 247767 |
+
"learning_rate": 8.147548076923078e-06,
|
| 247768 |
+
"loss": 0.3239,
|
| 247769 |
+
"step": 116640
|
| 247770 |
+
},
|
| 247771 |
+
{
|
| 247772 |
+
"epoch": 940.34,
|
| 247773 |
+
"learning_rate": 8.147467948717949e-06,
|
| 247774 |
+
"loss": 0.3073,
|
| 247775 |
+
"step": 116645
|
| 247776 |
+
},
|
| 247777 |
+
{
|
| 247778 |
+
"epoch": 940.38,
|
| 247779 |
+
"learning_rate": 8.14738782051282e-06,
|
| 247780 |
+
"loss": 0.6717,
|
| 247781 |
+
"step": 116650
|
| 247782 |
+
},
|
| 247783 |
+
{
|
| 247784 |
+
"epoch": 940.42,
|
| 247785 |
+
"learning_rate": 8.147307692307693e-06,
|
| 247786 |
+
"loss": 1.1175,
|
| 247787 |
+
"step": 116655
|
| 247788 |
+
},
|
| 247789 |
+
{
|
| 247790 |
+
"epoch": 940.46,
|
| 247791 |
+
"learning_rate": 8.147227564102565e-06,
|
| 247792 |
+
"loss": 0.2822,
|
| 247793 |
+
"step": 116660
|
| 247794 |
+
},
|
| 247795 |
+
{
|
| 247796 |
+
"epoch": 940.5,
|
| 247797 |
+
"learning_rate": 8.147147435897436e-06,
|
| 247798 |
+
"loss": 0.2605,
|
| 247799 |
+
"step": 116665
|
| 247800 |
+
},
|
| 247801 |
+
{
|
| 247802 |
+
"epoch": 940.54,
|
| 247803 |
+
"learning_rate": 8.147067307692308e-06,
|
| 247804 |
+
"loss": 0.3236,
|
| 247805 |
+
"step": 116670
|
| 247806 |
+
},
|
| 247807 |
+
{
|
| 247808 |
+
"epoch": 940.58,
|
| 247809 |
+
"learning_rate": 8.14698717948718e-06,
|
| 247810 |
+
"loss": 0.6434,
|
| 247811 |
+
"step": 116675
|
| 247812 |
+
},
|
| 247813 |
+
{
|
| 247814 |
+
"epoch": 940.62,
|
| 247815 |
+
"learning_rate": 8.146907051282052e-06,
|
| 247816 |
+
"loss": 1.0242,
|
| 247817 |
+
"step": 116680
|
| 247818 |
+
},
|
| 247819 |
+
{
|
| 247820 |
+
"epoch": 940.66,
|
| 247821 |
+
"learning_rate": 8.146826923076923e-06,
|
| 247822 |
+
"loss": 0.3286,
|
| 247823 |
+
"step": 116685
|
| 247824 |
+
},
|
| 247825 |
+
{
|
| 247826 |
+
"epoch": 940.7,
|
| 247827 |
+
"learning_rate": 8.146746794871796e-06,
|
| 247828 |
+
"loss": 0.2694,
|
| 247829 |
+
"step": 116690
|
| 247830 |
+
},
|
| 247831 |
+
{
|
| 247832 |
+
"epoch": 940.74,
|
| 247833 |
+
"learning_rate": 8.146666666666668e-06,
|
| 247834 |
+
"loss": 0.2618,
|
| 247835 |
+
"step": 116695
|
| 247836 |
+
},
|
| 247837 |
+
{
|
| 247838 |
+
"epoch": 940.78,
|
| 247839 |
+
"learning_rate": 8.146586538461539e-06,
|
| 247840 |
+
"loss": 0.6336,
|
| 247841 |
+
"step": 116700
|
| 247842 |
+
},
|
| 247843 |
+
{
|
| 247844 |
+
"epoch": 940.82,
|
| 247845 |
+
"learning_rate": 8.14650641025641e-06,
|
| 247846 |
+
"loss": 0.9335,
|
| 247847 |
+
"step": 116705
|
| 247848 |
+
},
|
| 247849 |
+
{
|
| 247850 |
+
"epoch": 940.86,
|
| 247851 |
+
"learning_rate": 8.146426282051283e-06,
|
| 247852 |
+
"loss": 0.2617,
|
| 247853 |
+
"step": 116710
|
| 247854 |
+
},
|
| 247855 |
+
{
|
| 247856 |
+
"epoch": 940.9,
|
| 247857 |
+
"learning_rate": 8.146346153846155e-06,
|
| 247858 |
+
"loss": 0.3125,
|
| 247859 |
+
"step": 116715
|
| 247860 |
+
},
|
| 247861 |
+
{
|
| 247862 |
+
"epoch": 940.94,
|
| 247863 |
+
"learning_rate": 8.146266025641026e-06,
|
| 247864 |
+
"loss": 0.3524,
|
| 247865 |
+
"step": 116720
|
| 247866 |
+
},
|
| 247867 |
+
{
|
| 247868 |
+
"epoch": 940.98,
|
| 247869 |
+
"learning_rate": 8.146185897435898e-06,
|
| 247870 |
+
"loss": 0.6945,
|
| 247871 |
+
"step": 116725
|
| 247872 |
+
},
|
| 247873 |
+
{
|
| 247874 |
+
"epoch": 941.0,
|
| 247875 |
+
"eval_loss": 0.3652080297470093,
|
| 247876 |
+
"eval_runtime": 39.2072,
|
| 247877 |
+
"eval_samples_per_second": 21.501,
|
| 247878 |
+
"eval_steps_per_second": 0.689,
|
| 247879 |
+
"eval_wer": 0.18559156861329676,
|
| 247880 |
+
"step": 116727
|
| 247881 |
+
},
|
| 247882 |
+
{
|
| 247883 |
+
"epoch": 941.02,
|
| 247884 |
+
"learning_rate": 8.14610576923077e-06,
|
| 247885 |
+
"loss": 0.3172,
|
| 247886 |
+
"step": 116730
|
| 247887 |
+
},
|
| 247888 |
+
{
|
| 247889 |
+
"epoch": 941.06,
|
| 247890 |
+
"learning_rate": 8.146025641025642e-06,
|
| 247891 |
+
"loss": 0.2391,
|
| 247892 |
+
"step": 116735
|
| 247893 |
+
},
|
| 247894 |
+
{
|
| 247895 |
+
"epoch": 941.1,
|
| 247896 |
+
"learning_rate": 8.145945512820513e-06,
|
| 247897 |
+
"loss": 0.2877,
|
| 247898 |
+
"step": 116740
|
| 247899 |
+
},
|
| 247900 |
+
{
|
| 247901 |
+
"epoch": 941.14,
|
| 247902 |
+
"learning_rate": 8.145865384615386e-06,
|
| 247903 |
+
"loss": 0.3587,
|
| 247904 |
+
"step": 116745
|
| 247905 |
+
},
|
| 247906 |
+
{
|
| 247907 |
+
"epoch": 941.18,
|
| 247908 |
+
"learning_rate": 8.145785256410256e-06,
|
| 247909 |
+
"loss": 0.7811,
|
| 247910 |
+
"step": 116750
|
| 247911 |
+
},
|
| 247912 |
+
{
|
| 247913 |
+
"epoch": 941.22,
|
| 247914 |
+
"learning_rate": 8.145705128205129e-06,
|
| 247915 |
+
"loss": 0.7006,
|
| 247916 |
+
"step": 116755
|
| 247917 |
+
},
|
| 247918 |
+
{
|
| 247919 |
+
"epoch": 941.27,
|
| 247920 |
+
"learning_rate": 8.145625e-06,
|
| 247921 |
+
"loss": 0.327,
|
| 247922 |
+
"step": 116760
|
| 247923 |
+
},
|
| 247924 |
+
{
|
| 247925 |
+
"epoch": 941.31,
|
| 247926 |
+
"learning_rate": 8.145544871794872e-06,
|
| 247927 |
+
"loss": 0.3295,
|
| 247928 |
+
"step": 116765
|
| 247929 |
+
},
|
| 247930 |
+
{
|
| 247931 |
+
"epoch": 941.35,
|
| 247932 |
+
"learning_rate": 8.145464743589743e-06,
|
| 247933 |
+
"loss": 0.3501,
|
| 247934 |
+
"step": 116770
|
| 247935 |
+
},
|
| 247936 |
+
{
|
| 247937 |
+
"epoch": 941.39,
|
| 247938 |
+
"learning_rate": 8.145384615384616e-06,
|
| 247939 |
+
"loss": 0.9432,
|
| 247940 |
+
"step": 116775
|
| 247941 |
+
},
|
| 247942 |
+
{
|
| 247943 |
+
"epoch": 941.43,
|
| 247944 |
+
"learning_rate": 8.145304487179488e-06,
|
| 247945 |
+
"loss": 0.8745,
|
| 247946 |
+
"step": 116780
|
| 247947 |
+
},
|
| 247948 |
+
{
|
| 247949 |
+
"epoch": 941.47,
|
| 247950 |
+
"learning_rate": 8.145224358974359e-06,
|
| 247951 |
+
"loss": 0.2931,
|
| 247952 |
+
"step": 116785
|
| 247953 |
+
},
|
| 247954 |
+
{
|
| 247955 |
+
"epoch": 941.51,
|
| 247956 |
+
"learning_rate": 8.145144230769232e-06,
|
| 247957 |
+
"loss": 0.2615,
|
| 247958 |
+
"step": 116790
|
| 247959 |
+
},
|
| 247960 |
+
{
|
| 247961 |
+
"epoch": 941.55,
|
| 247962 |
+
"learning_rate": 8.145064102564103e-06,
|
| 247963 |
+
"loss": 0.3783,
|
| 247964 |
+
"step": 116795
|
| 247965 |
+
},
|
| 247966 |
+
{
|
| 247967 |
+
"epoch": 941.59,
|
| 247968 |
+
"learning_rate": 8.144983974358975e-06,
|
| 247969 |
+
"loss": 0.6995,
|
| 247970 |
+
"step": 116800
|
| 247971 |
+
},
|
| 247972 |
+
{
|
| 247973 |
+
"epoch": 941.63,
|
| 247974 |
+
"learning_rate": 8.144903846153846e-06,
|
| 247975 |
+
"loss": 0.8687,
|
| 247976 |
+
"step": 116805
|
| 247977 |
+
},
|
| 247978 |
+
{
|
| 247979 |
+
"epoch": 941.67,
|
| 247980 |
+
"learning_rate": 8.144823717948719e-06,
|
| 247981 |
+
"loss": 0.2578,
|
| 247982 |
+
"step": 116810
|
| 247983 |
+
},
|
| 247984 |
+
{
|
| 247985 |
+
"epoch": 941.71,
|
| 247986 |
+
"learning_rate": 8.14474358974359e-06,
|
| 247987 |
+
"loss": 0.2837,
|
| 247988 |
+
"step": 116815
|
| 247989 |
+
},
|
| 247990 |
+
{
|
| 247991 |
+
"epoch": 941.75,
|
| 247992 |
+
"learning_rate": 8.144663461538462e-06,
|
| 247993 |
+
"loss": 0.3907,
|
| 247994 |
+
"step": 116820
|
| 247995 |
+
},
|
| 247996 |
+
{
|
| 247997 |
+
"epoch": 941.79,
|
| 247998 |
+
"learning_rate": 8.144583333333333e-06,
|
| 247999 |
+
"loss": 0.6813,
|
| 248000 |
+
"step": 116825
|
| 248001 |
+
},
|
| 248002 |
+
{
|
| 248003 |
+
"epoch": 941.83,
|
| 248004 |
+
"learning_rate": 8.144503205128206e-06,
|
| 248005 |
+
"loss": 0.8454,
|
| 248006 |
+
"step": 116830
|
| 248007 |
+
},
|
| 248008 |
+
{
|
| 248009 |
+
"epoch": 941.87,
|
| 248010 |
+
"learning_rate": 8.144423076923078e-06,
|
| 248011 |
+
"loss": 0.2834,
|
| 248012 |
+
"step": 116835
|
| 248013 |
+
},
|
| 248014 |
+
{
|
| 248015 |
+
"epoch": 941.91,
|
| 248016 |
+
"learning_rate": 8.144342948717949e-06,
|
| 248017 |
+
"loss": 0.3404,
|
| 248018 |
+
"step": 116840
|
| 248019 |
+
},
|
| 248020 |
+
{
|
| 248021 |
+
"epoch": 941.95,
|
| 248022 |
+
"learning_rate": 8.144262820512822e-06,
|
| 248023 |
+
"loss": 0.3596,
|
| 248024 |
+
"step": 116845
|
| 248025 |
+
},
|
| 248026 |
+
{
|
| 248027 |
+
"epoch": 941.99,
|
| 248028 |
+
"learning_rate": 8.144182692307693e-06,
|
| 248029 |
+
"loss": 0.8048,
|
| 248030 |
+
"step": 116850
|
| 248031 |
+
},
|
| 248032 |
+
{
|
| 248033 |
+
"epoch": 942.0,
|
| 248034 |
+
"eval_loss": 0.37358829379081726,
|
| 248035 |
+
"eval_runtime": 38.8227,
|
| 248036 |
+
"eval_samples_per_second": 21.714,
|
| 248037 |
+
"eval_steps_per_second": 0.695,
|
| 248038 |
+
"eval_wer": 0.181710652768537,
|
| 248039 |
+
"step": 116851
|
| 248040 |
+
},
|
| 248041 |
+
{
|
| 248042 |
+
"epoch": 942.03,
|
| 248043 |
+
"learning_rate": 8.144102564102565e-06,
|
| 248044 |
+
"loss": 0.3532,
|
| 248045 |
+
"step": 116855
|
| 248046 |
+
},
|
| 248047 |
+
{
|
| 248048 |
+
"epoch": 942.07,
|
| 248049 |
+
"learning_rate": 8.144022435897436e-06,
|
| 248050 |
+
"loss": 0.2954,
|
| 248051 |
+
"step": 116860
|
| 248052 |
+
},
|
| 248053 |
+
{
|
| 248054 |
+
"epoch": 942.11,
|
| 248055 |
+
"learning_rate": 8.14394230769231e-06,
|
| 248056 |
+
"loss": 0.321,
|
| 248057 |
+
"step": 116865
|
| 248058 |
+
},
|
| 248059 |
+
{
|
| 248060 |
+
"epoch": 942.15,
|
| 248061 |
+
"learning_rate": 8.14386217948718e-06,
|
| 248062 |
+
"loss": 0.4422,
|
| 248063 |
+
"step": 116870
|
| 248064 |
+
},
|
| 248065 |
+
{
|
| 248066 |
+
"epoch": 942.19,
|
| 248067 |
+
"learning_rate": 8.143782051282052e-06,
|
| 248068 |
+
"loss": 0.8507,
|
| 248069 |
+
"step": 116875
|
| 248070 |
+
},
|
| 248071 |
+
{
|
| 248072 |
+
"epoch": 942.23,
|
| 248073 |
+
"learning_rate": 8.143701923076923e-06,
|
| 248074 |
+
"loss": 0.6194,
|
| 248075 |
+
"step": 116880
|
| 248076 |
+
},
|
| 248077 |
+
{
|
| 248078 |
+
"epoch": 942.27,
|
| 248079 |
+
"learning_rate": 8.143621794871796e-06,
|
| 248080 |
+
"loss": 0.3064,
|
| 248081 |
+
"step": 116885
|
| 248082 |
+
},
|
| 248083 |
+
{
|
| 248084 |
+
"epoch": 942.31,
|
| 248085 |
+
"learning_rate": 8.143541666666668e-06,
|
| 248086 |
+
"loss": 0.2845,
|
| 248087 |
+
"step": 116890
|
| 248088 |
+
},
|
| 248089 |
+
{
|
| 248090 |
+
"epoch": 942.35,
|
| 248091 |
+
"learning_rate": 8.143461538461539e-06,
|
| 248092 |
+
"loss": 0.3395,
|
| 248093 |
+
"step": 116895
|
| 248094 |
+
},
|
| 248095 |
+
{
|
| 248096 |
+
"epoch": 942.39,
|
| 248097 |
+
"learning_rate": 8.143381410256412e-06,
|
| 248098 |
+
"loss": 0.8709,
|
| 248099 |
+
"step": 116900
|
| 248100 |
+
},
|
| 248101 |
+
{
|
| 248102 |
+
"epoch": 942.43,
|
| 248103 |
+
"learning_rate": 8.143301282051282e-06,
|
| 248104 |
+
"loss": 0.5708,
|
| 248105 |
+
"step": 116905
|
| 248106 |
+
},
|
| 248107 |
+
{
|
| 248108 |
+
"epoch": 942.47,
|
| 248109 |
+
"learning_rate": 8.143221153846155e-06,
|
| 248110 |
+
"loss": 0.257,
|
| 248111 |
+
"step": 116910
|
| 248112 |
+
},
|
| 248113 |
+
{
|
| 248114 |
+
"epoch": 942.51,
|
| 248115 |
+
"learning_rate": 8.143141025641026e-06,
|
| 248116 |
+
"loss": 0.2881,
|
| 248117 |
+
"step": 116915
|
| 248118 |
+
},
|
| 248119 |
+
{
|
| 248120 |
+
"epoch": 942.55,
|
| 248121 |
+
"learning_rate": 8.143060897435897e-06,
|
| 248122 |
+
"loss": 0.3739,
|
| 248123 |
+
"step": 116920
|
| 248124 |
+
},
|
| 248125 |
+
{
|
| 248126 |
+
"epoch": 942.59,
|
| 248127 |
+
"learning_rate": 8.142980769230769e-06,
|
| 248128 |
+
"loss": 1.091,
|
| 248129 |
+
"step": 116925
|
| 248130 |
+
},
|
| 248131 |
+
{
|
| 248132 |
+
"epoch": 942.63,
|
| 248133 |
+
"learning_rate": 8.142900641025642e-06,
|
| 248134 |
+
"loss": 0.571,
|
| 248135 |
+
"step": 116930
|
| 248136 |
+
},
|
| 248137 |
+
{
|
| 248138 |
+
"epoch": 942.67,
|
| 248139 |
+
"learning_rate": 8.142820512820513e-06,
|
| 248140 |
+
"loss": 0.2523,
|
| 248141 |
+
"step": 116935
|
| 248142 |
+
},
|
| 248143 |
+
{
|
| 248144 |
+
"epoch": 942.71,
|
| 248145 |
+
"learning_rate": 8.142740384615385e-06,
|
| 248146 |
+
"loss": 0.3116,
|
| 248147 |
+
"step": 116940
|
| 248148 |
+
},
|
| 248149 |
+
{
|
| 248150 |
+
"epoch": 942.76,
|
| 248151 |
+
"learning_rate": 8.142660256410258e-06,
|
| 248152 |
+
"loss": 0.4294,
|
| 248153 |
+
"step": 116945
|
| 248154 |
+
},
|
| 248155 |
+
{
|
| 248156 |
+
"epoch": 942.8,
|
| 248157 |
+
"learning_rate": 8.142580128205129e-06,
|
| 248158 |
+
"loss": 0.8474,
|
| 248159 |
+
"step": 116950
|
| 248160 |
+
},
|
| 248161 |
+
{
|
| 248162 |
+
"epoch": 942.84,
|
| 248163 |
+
"learning_rate": 8.1425e-06,
|
| 248164 |
+
"loss": 0.6085,
|
| 248165 |
+
"step": 116955
|
| 248166 |
+
},
|
| 248167 |
+
{
|
| 248168 |
+
"epoch": 942.88,
|
| 248169 |
+
"learning_rate": 8.142419871794872e-06,
|
| 248170 |
+
"loss": 0.4619,
|
| 248171 |
+
"step": 116960
|
| 248172 |
+
},
|
| 248173 |
+
{
|
| 248174 |
+
"epoch": 942.92,
|
| 248175 |
+
"learning_rate": 8.142339743589745e-06,
|
| 248176 |
+
"loss": 0.4268,
|
| 248177 |
+
"step": 116965
|
| 248178 |
+
},
|
| 248179 |
+
{
|
| 248180 |
+
"epoch": 942.96,
|
| 248181 |
+
"learning_rate": 8.142259615384616e-06,
|
| 248182 |
+
"loss": 0.4165,
|
| 248183 |
+
"step": 116970
|
| 248184 |
+
},
|
| 248185 |
+
{
|
| 248186 |
+
"epoch": 943.0,
|
| 248187 |
+
"learning_rate": 8.142179487179488e-06,
|
| 248188 |
+
"loss": 1.148,
|
| 248189 |
+
"step": 116975
|
| 248190 |
+
},
|
| 248191 |
+
{
|
| 248192 |
+
"epoch": 943.0,
|
| 248193 |
+
"eval_loss": 0.3594907820224762,
|
| 248194 |
+
"eval_runtime": 39.3165,
|
| 248195 |
+
"eval_samples_per_second": 21.441,
|
| 248196 |
+
"eval_steps_per_second": 0.687,
|
| 248197 |
+
"eval_wer": 0.18600070348223707,
|
| 248198 |
+
"step": 116975
|
| 248199 |
+
},
|
| 248200 |
+
{
|
| 248201 |
+
"epoch": 935.04,
|
| 248202 |
+
"learning_rate": 8.142099358974359e-06,
|
| 248203 |
+
"loss": 0.323,
|
| 248204 |
+
"step": 116980
|
| 248205 |
+
},
|
| 248206 |
+
{
|
| 248207 |
+
"epoch": 935.08,
|
| 248208 |
+
"learning_rate": 8.142019230769232e-06,
|
| 248209 |
+
"loss": 0.284,
|
| 248210 |
+
"step": 116985
|
| 248211 |
+
},
|
| 248212 |
+
{
|
| 248213 |
+
"epoch": 935.12,
|
| 248214 |
+
"learning_rate": 8.141939102564103e-06,
|
| 248215 |
+
"loss": 0.2959,
|
| 248216 |
+
"step": 116990
|
| 248217 |
+
},
|
| 248218 |
+
{
|
| 248219 |
+
"epoch": 935.16,
|
| 248220 |
+
"learning_rate": 8.141858974358975e-06,
|
| 248221 |
+
"loss": 0.4164,
|
| 248222 |
+
"step": 116995
|
| 248223 |
+
},
|
| 248224 |
+
{
|
| 248225 |
+
"epoch": 935.2,
|
| 248226 |
+
"learning_rate": 8.141778846153848e-06,
|
| 248227 |
+
"loss": 1.1566,
|
| 248228 |
+
"step": 117000
|
| 248229 |
+
},
|
| 248230 |
+
{
|
| 248231 |
+
"epoch": 935.24,
|
| 248232 |
+
"learning_rate": 8.141698717948719e-06,
|
| 248233 |
+
"loss": 0.3514,
|
| 248234 |
+
"step": 117005
|
| 248235 |
+
},
|
| 248236 |
+
{
|
| 248237 |
+
"epoch": 935.28,
|
| 248238 |
+
"learning_rate": 8.14161858974359e-06,
|
| 248239 |
+
"loss": 0.2887,
|
| 248240 |
+
"step": 117010
|
| 248241 |
+
},
|
| 248242 |
+
{
|
| 248243 |
+
"epoch": 935.32,
|
| 248244 |
+
"learning_rate": 8.141538461538462e-06,
|
| 248245 |
+
"loss": 0.3248,
|
| 248246 |
+
"step": 117015
|
| 248247 |
+
},
|
| 248248 |
+
{
|
| 248249 |
+
"epoch": 935.36,
|
| 248250 |
+
"learning_rate": 8.141458333333335e-06,
|
| 248251 |
+
"loss": 0.5457,
|
| 248252 |
+
"step": 117020
|
| 248253 |
+
},
|
| 248254 |
+
{
|
| 248255 |
+
"epoch": 935.4,
|
| 248256 |
+
"learning_rate": 8.141378205128204e-06,
|
| 248257 |
+
"loss": 1.2681,
|
| 248258 |
+
"step": 117025
|
| 248259 |
+
},
|
| 248260 |
+
{
|
| 248261 |
+
"epoch": 935.44,
|
| 248262 |
+
"learning_rate": 8.141298076923078e-06,
|
| 248263 |
+
"loss": 0.2972,
|
| 248264 |
+
"step": 117030
|
| 248265 |
+
},
|
| 248266 |
+
{
|
| 248267 |
+
"epoch": 935.48,
|
| 248268 |
+
"learning_rate": 8.14121794871795e-06,
|
| 248269 |
+
"loss": 0.2916,
|
| 248270 |
+
"step": 117035
|
| 248271 |
+
},
|
| 248272 |
+
{
|
| 248273 |
+
"epoch": 935.52,
|
| 248274 |
+
"learning_rate": 8.14113782051282e-06,
|
| 248275 |
+
"loss": 0.3855,
|
| 248276 |
+
"step": 117040
|
| 248277 |
+
},
|
| 248278 |
+
{
|
| 248279 |
+
"epoch": 935.56,
|
| 248280 |
+
"learning_rate": 8.141057692307693e-06,
|
| 248281 |
+
"loss": 0.3866,
|
| 248282 |
+
"step": 117045
|
| 248283 |
+
},
|
| 248284 |
+
{
|
| 248285 |
+
"epoch": 935.6,
|
| 248286 |
+
"learning_rate": 8.140977564102565e-06,
|
| 248287 |
+
"loss": 1.0996,
|
| 248288 |
+
"step": 117050
|
| 248289 |
+
},
|
| 248290 |
+
{
|
| 248291 |
+
"epoch": 935.64,
|
| 248292 |
+
"learning_rate": 8.140897435897436e-06,
|
| 248293 |
+
"loss": 0.3514,
|
| 248294 |
+
"step": 117055
|
| 248295 |
+
},
|
| 248296 |
+
{
|
| 248297 |
+
"epoch": 935.68,
|
| 248298 |
+
"learning_rate": 8.140817307692307e-06,
|
| 248299 |
+
"loss": 0.2344,
|
| 248300 |
+
"step": 117060
|
| 248301 |
+
},
|
| 248302 |
+
{
|
| 248303 |
+
"epoch": 935.72,
|
| 248304 |
+
"learning_rate": 8.14073717948718e-06,
|
| 248305 |
+
"loss": 0.3085,
|
| 248306 |
+
"step": 117065
|
| 248307 |
+
},
|
| 248308 |
+
{
|
| 248309 |
+
"epoch": 935.76,
|
| 248310 |
+
"learning_rate": 8.140657051282052e-06,
|
| 248311 |
+
"loss": 0.413,
|
| 248312 |
+
"step": 117070
|
| 248313 |
+
},
|
| 248314 |
+
{
|
| 248315 |
+
"epoch": 935.8,
|
| 248316 |
+
"learning_rate": 8.140576923076923e-06,
|
| 248317 |
+
"loss": 1.332,
|
| 248318 |
+
"step": 117075
|
| 248319 |
+
},
|
| 248320 |
+
{
|
| 248321 |
+
"epoch": 935.84,
|
| 248322 |
+
"learning_rate": 8.140496794871795e-06,
|
| 248323 |
+
"loss": 0.301,
|
| 248324 |
+
"step": 117080
|
| 248325 |
+
},
|
| 248326 |
+
{
|
| 248327 |
+
"epoch": 935.88,
|
| 248328 |
+
"learning_rate": 8.140416666666668e-06,
|
| 248329 |
+
"loss": 0.2786,
|
| 248330 |
+
"step": 117085
|
| 248331 |
+
},
|
| 248332 |
+
{
|
| 248333 |
+
"epoch": 935.92,
|
| 248334 |
+
"learning_rate": 8.140336538461539e-06,
|
| 248335 |
+
"loss": 0.3301,
|
| 248336 |
+
"step": 117090
|
| 248337 |
+
},
|
| 248338 |
+
{
|
| 248339 |
+
"epoch": 935.96,
|
| 248340 |
+
"learning_rate": 8.14025641025641e-06,
|
| 248341 |
+
"loss": 0.557,
|
| 248342 |
+
"step": 117095
|
| 248343 |
+
},
|
| 248344 |
+
{
|
| 248345 |
+
"epoch": 936.0,
|
| 248346 |
+
"learning_rate": 8.140176282051283e-06,
|
| 248347 |
+
"loss": 1.4011,
|
| 248348 |
+
"step": 117100
|
| 248349 |
+
},
|
| 248350 |
+
{
|
| 248351 |
+
"epoch": 936.0,
|
| 248352 |
+
"eval_loss": 0.34906426072120667,
|
| 248353 |
+
"eval_runtime": 39.7868,
|
| 248354 |
+
"eval_samples_per_second": 21.188,
|
| 248355 |
+
"eval_steps_per_second": 0.679,
|
| 248356 |
+
"eval_wer": 0.1851399856424982,
|
| 248357 |
+
"step": 117100
|
| 248358 |
}
|
| 248359 |
],
|
| 248360 |
"max_steps": 625000,
|
| 248361 |
"num_train_epochs": 5000,
|
| 248362 |
+
"total_flos": 3.29550208819677e+20,
|
| 248363 |
"trial_name": null,
|
| 248364 |
"trial_params": null
|
| 248365 |
}
|
model-bin/finetune/base/{checkpoint-116478 → checkpoint-117100}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630115798.4136324/events.out.tfevents.1630115798.86bb0ddabf9b.1042.41
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d1b074a755f2daa10761fb515413ac8941c898fbb3c07592e5fda37594331551
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630116207.9281766/events.out.tfevents.1630116207.86bb0ddabf9b.1042.43
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:68f3715bd0ab3d6f1ef9528a3b33c373e6085f271f727d16032a3c8f800f5c32
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630116621.3491957/events.out.tfevents.1630116621.86bb0ddabf9b.1042.45
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b61e75b6b5eedd92c5c3ec7d2c6390e778c14598efdc20e5064d0427e336885a
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630117036.3824139/events.out.tfevents.1630117036.86bb0ddabf9b.1042.47
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:88ab210bdc5285d925d2e47a8308f6851a592c9f596915f05ffec0a75f9de165
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630117449.777196/events.out.tfevents.1630117449.86bb0ddabf9b.1042.49
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c8abf16a78934443be86f9165efeb04aef6c69ed81bc56bc9d7f82f4abe26ba9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630115798.86bb0ddabf9b.1042.40
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:11e43862968b4f8a128db20b0bdfb5c095f3a586a6dd95906ead603ea7a306a3
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630116207.86bb0ddabf9b.1042.42
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8b0d26ec2fa4bd9832fa8d496a62c8dded8ab637811b0ff8e9731ec6b08accc1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630116621.86bb0ddabf9b.1042.44
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fe9d89f47418ab426a2df3d71b21fbae52017b55817c17fbc00d6fd3dfbf7344
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630117036.86bb0ddabf9b.1042.46
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ace9ba494ce37d956652650695ce1ffb4ec148d656c10db037ae03ec311edb80
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630117449.86bb0ddabf9b.1042.48
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:251d71865605240570d7ac2db4f9cd67a49e2ad8ffbe1d36f89a65c721298fec
|
| 3 |
+
size 8622
|