"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/trainer_state.json +953 -5
- model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630167382.0506334/events.out.tfevents.1630167382.86bb0ddabf9b.4092.221 +3 -0
- model-bin/finetune/base/log/1630167774.8977208/events.out.tfevents.1630167774.86bb0ddabf9b.4092.223 +3 -0
- model-bin/finetune/base/log/1630168288.1684663/events.out.tfevents.1630168288.86bb0ddabf9b.4092.225 +3 -0
- model-bin/finetune/base/log/1630168673.8962524/events.out.tfevents.1630168673.86bb0ddabf9b.4092.227 +3 -0
- model-bin/finetune/base/log/1630169063.5061305/events.out.tfevents.1630169063.86bb0ddabf9b.4092.229 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630167382.86bb0ddabf9b.4092.220 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630167774.86bb0ddabf9b.4092.222 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630168288.86bb0ddabf9b.4092.224 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630168673.86bb0ddabf9b.4092.226 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630169063.86bb0ddabf9b.4092.228 +3 -0
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/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:ccdb2fb02c8f7a62be0fc2409f506fd1f8cf9d2fc8d923400194d70adf9e242f
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/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:a3c00faa72b2cc714e800fe08cfb981372abb80878a47041a7aaa0625f56c415
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/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:af40aacf51d19533b69ca05d28c2c6385f906426263fb9be7e9af68da57b9c59
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/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:b0ef0d1df67c213aabcd2f383fde85fcd536070cc6b5c823cb5d89509deb1406
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/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 sha256:107eefd712891e15abd2fa32b03a66b6676b4b123902cabee2038cf6cd3fef9d
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/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,
|
|
@@ -266727,11 +266727,959 @@
|
|
| 266727 |
"eval_steps_per_second": 0.73,
|
| 266728 |
"eval_wer": 0.17733196661297407,
|
| 266729 |
"step": 131541
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 266730 |
}
|
| 266731 |
],
|
| 266732 |
"max_steps": 625000,
|
| 266733 |
"num_train_epochs": 5000,
|
| 266734 |
-
"total_flos": 3.
|
| 266735 |
"trial_name": null,
|
| 266736 |
"trial_params": null
|
| 266737 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.17062187276626162,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-131665",
|
| 4 |
+
"epoch": 1058.0,
|
| 5 |
+
"global_step": 132288,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 266727 |
"eval_steps_per_second": 0.73,
|
| 266728 |
"eval_wer": 0.17733196661297407,
|
| 266729 |
"step": 131541
|
| 266730 |
+
},
|
| 266731 |
+
{
|
| 266732 |
+
"epoch": 1060.03,
|
| 266733 |
+
"learning_rate": 7.908750000000001e-06,
|
| 266734 |
+
"loss": 0.2501,
|
| 266735 |
+
"step": 131545
|
| 266736 |
+
},
|
| 266737 |
+
{
|
| 266738 |
+
"epoch": 1060.07,
|
| 266739 |
+
"learning_rate": 7.908669871794873e-06,
|
| 266740 |
+
"loss": 0.2856,
|
| 266741 |
+
"step": 131550
|
| 266742 |
+
},
|
| 266743 |
+
{
|
| 266744 |
+
"epoch": 1060.11,
|
| 266745 |
+
"learning_rate": 7.908589743589744e-06,
|
| 266746 |
+
"loss": 0.4052,
|
| 266747 |
+
"step": 131555
|
| 266748 |
+
},
|
| 266749 |
+
{
|
| 266750 |
+
"epoch": 1060.15,
|
| 266751 |
+
"learning_rate": 7.908509615384617e-06,
|
| 266752 |
+
"loss": 0.391,
|
| 266753 |
+
"step": 131560
|
| 266754 |
+
},
|
| 266755 |
+
{
|
| 266756 |
+
"epoch": 1060.19,
|
| 266757 |
+
"learning_rate": 7.908429487179488e-06,
|
| 266758 |
+
"loss": 1.0021,
|
| 266759 |
+
"step": 131565
|
| 266760 |
+
},
|
| 266761 |
+
{
|
| 266762 |
+
"epoch": 1060.23,
|
| 266763 |
+
"learning_rate": 7.90834935897436e-06,
|
| 266764 |
+
"loss": 0.6628,
|
| 266765 |
+
"step": 131570
|
| 266766 |
+
},
|
| 266767 |
+
{
|
| 266768 |
+
"epoch": 1060.27,
|
| 266769 |
+
"learning_rate": 7.908269230769231e-06,
|
| 266770 |
+
"loss": 0.292,
|
| 266771 |
+
"step": 131575
|
| 266772 |
+
},
|
| 266773 |
+
{
|
| 266774 |
+
"epoch": 1060.31,
|
| 266775 |
+
"learning_rate": 7.908189102564104e-06,
|
| 266776 |
+
"loss": 0.2786,
|
| 266777 |
+
"step": 131580
|
| 266778 |
+
},
|
| 266779 |
+
{
|
| 266780 |
+
"epoch": 1060.35,
|
| 266781 |
+
"learning_rate": 7.908108974358974e-06,
|
| 266782 |
+
"loss": 0.3931,
|
| 266783 |
+
"step": 131585
|
| 266784 |
+
},
|
| 266785 |
+
{
|
| 266786 |
+
"epoch": 1060.39,
|
| 266787 |
+
"learning_rate": 7.908028846153847e-06,
|
| 266788 |
+
"loss": 0.9163,
|
| 266789 |
+
"step": 131590
|
| 266790 |
+
},
|
| 266791 |
+
{
|
| 266792 |
+
"epoch": 1060.43,
|
| 266793 |
+
"learning_rate": 7.90794871794872e-06,
|
| 266794 |
+
"loss": 0.7397,
|
| 266795 |
+
"step": 131595
|
| 266796 |
+
},
|
| 266797 |
+
{
|
| 266798 |
+
"epoch": 1060.47,
|
| 266799 |
+
"learning_rate": 7.90786858974359e-06,
|
| 266800 |
+
"loss": 0.3026,
|
| 266801 |
+
"step": 131600
|
| 266802 |
+
},
|
| 266803 |
+
{
|
| 266804 |
+
"epoch": 1060.51,
|
| 266805 |
+
"learning_rate": 7.907788461538463e-06,
|
| 266806 |
+
"loss": 0.3751,
|
| 266807 |
+
"step": 131605
|
| 266808 |
+
},
|
| 266809 |
+
{
|
| 266810 |
+
"epoch": 1060.55,
|
| 266811 |
+
"learning_rate": 7.907708333333334e-06,
|
| 266812 |
+
"loss": 0.4241,
|
| 266813 |
+
"step": 131610
|
| 266814 |
+
},
|
| 266815 |
+
{
|
| 266816 |
+
"epoch": 1060.59,
|
| 266817 |
+
"learning_rate": 7.907628205128205e-06,
|
| 266818 |
+
"loss": 0.9136,
|
| 266819 |
+
"step": 131615
|
| 266820 |
+
},
|
| 266821 |
+
{
|
| 266822 |
+
"epoch": 1060.63,
|
| 266823 |
+
"learning_rate": 7.907548076923077e-06,
|
| 266824 |
+
"loss": 0.6556,
|
| 266825 |
+
"step": 131620
|
| 266826 |
+
},
|
| 266827 |
+
{
|
| 266828 |
+
"epoch": 1060.67,
|
| 266829 |
+
"learning_rate": 7.90746794871795e-06,
|
| 266830 |
+
"loss": 0.2733,
|
| 266831 |
+
"step": 131625
|
| 266832 |
+
},
|
| 266833 |
+
{
|
| 266834 |
+
"epoch": 1060.71,
|
| 266835 |
+
"learning_rate": 7.907387820512821e-06,
|
| 266836 |
+
"loss": 0.286,
|
| 266837 |
+
"step": 131630
|
| 266838 |
+
},
|
| 266839 |
+
{
|
| 266840 |
+
"epoch": 1060.76,
|
| 266841 |
+
"learning_rate": 7.907307692307692e-06,
|
| 266842 |
+
"loss": 0.3578,
|
| 266843 |
+
"step": 131635
|
| 266844 |
+
},
|
| 266845 |
+
{
|
| 266846 |
+
"epoch": 1060.8,
|
| 266847 |
+
"learning_rate": 7.907227564102564e-06,
|
| 266848 |
+
"loss": 0.8609,
|
| 266849 |
+
"step": 131640
|
| 266850 |
+
},
|
| 266851 |
+
{
|
| 266852 |
+
"epoch": 1060.84,
|
| 266853 |
+
"learning_rate": 7.907147435897437e-06,
|
| 266854 |
+
"loss": 0.6417,
|
| 266855 |
+
"step": 131645
|
| 266856 |
+
},
|
| 266857 |
+
{
|
| 266858 |
+
"epoch": 1060.88,
|
| 266859 |
+
"learning_rate": 7.907067307692308e-06,
|
| 266860 |
+
"loss": 0.2615,
|
| 266861 |
+
"step": 131650
|
| 266862 |
+
},
|
| 266863 |
+
{
|
| 266864 |
+
"epoch": 1060.92,
|
| 266865 |
+
"learning_rate": 7.90698717948718e-06,
|
| 266866 |
+
"loss": 0.3053,
|
| 266867 |
+
"step": 131655
|
| 266868 |
+
},
|
| 266869 |
+
{
|
| 266870 |
+
"epoch": 1060.96,
|
| 266871 |
+
"learning_rate": 7.906907051282053e-06,
|
| 266872 |
+
"loss": 0.4595,
|
| 266873 |
+
"step": 131660
|
| 266874 |
+
},
|
| 266875 |
+
{
|
| 266876 |
+
"epoch": 1061.0,
|
| 266877 |
+
"learning_rate": 7.906826923076924e-06,
|
| 266878 |
+
"loss": 1.1019,
|
| 266879 |
+
"step": 131665
|
| 266880 |
+
},
|
| 266881 |
+
{
|
| 266882 |
+
"epoch": 1061.0,
|
| 266883 |
+
"eval_loss": 0.37313929200172424,
|
| 266884 |
+
"eval_runtime": 37.0356,
|
| 266885 |
+
"eval_samples_per_second": 22.573,
|
| 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 |
"max_steps": 625000,
|
| 267681 |
"num_train_epochs": 5000,
|
| 267682 |
+
"total_flos": 3.722539366657103e+20,
|
| 267683 |
"trial_name": null,
|
| 267684 |
"trial_params": null
|
| 267685 |
}
|
model-bin/finetune/base/{checkpoint-131541 → checkpoint-132288}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630167382.0506334/events.out.tfevents.1630167382.86bb0ddabf9b.4092.221
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ac1b942e95fa43d828532a47e5772cf30c293df38141d2bfe16acab83b4321fe
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630167774.8977208/events.out.tfevents.1630167774.86bb0ddabf9b.4092.223
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:71a2c4bf917e3d1cf40a0f309879873aa6982f0776fcf94a618ce18215a98830
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630168288.1684663/events.out.tfevents.1630168288.86bb0ddabf9b.4092.225
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8686f1b953ed160d6fdb2cd403335bd67f005e4734314ae5a4e7a5d67ddffe31
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630168673.8962524/events.out.tfevents.1630168673.86bb0ddabf9b.4092.227
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:afb316c1a7825334da8d92c9f36e0cb14995e466f7b033f96c17a2464992fc5d
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630169063.5061305/events.out.tfevents.1630169063.86bb0ddabf9b.4092.229
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:824e45372a3319702fc479332c1dc4fa7ecee080450bb50703585a2ba17c8eec
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630167382.86bb0ddabf9b.4092.220
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f9bc3209876892e3f4e956d6b1bc18bd1629e1e8ea39b5ae0f8e38dd677898db
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630167774.86bb0ddabf9b.4092.222
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c57333255fb5a8c5945e97e623f1133401a4ec90e701ffacc9a39ef24cf3248b
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1630168288.86bb0ddabf9b.4092.224
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7c1100022a3745048ff54f08c11a694955c220a6ccd88ce7b5f569dac95d3a60
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630168673.86bb0ddabf9b.4092.226
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b4eadc3044a7ff12224ef4bb9f614394993b36cde1aa6b42c5c144dc6f1af871
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630169063.86bb0ddabf9b.4092.228
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6a375145a0c21ae6b2f5fe3f795e3211424e9be1e0d9b6cca8cb9177759d76a0
|
| 3 |
+
size 8622
|