"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/trainer_state.json +951 -3
- model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629885731.4375427/events.out.tfevents.1629885731.7e498afd5545.905.103 +3 -0
- model-bin/finetune/base/log/1629886673.7174067/events.out.tfevents.1629886673.7e498afd5545.6884.1 +3 -0
- model-bin/finetune/base/log/1629887166.966957/events.out.tfevents.1629887166.7e498afd5545.6884.3 +3 -0
- model-bin/finetune/base/log/1629887754.1185539/events.out.tfevents.1629887754.7e498afd5545.6884.5 +3 -0
- model-bin/finetune/base/log/1629888265.905986/events.out.tfevents.1629888265.7e498afd5545.6884.7 +3 -0
- model-bin/finetune/base/log/1629888670.3141375/events.out.tfevents.1629888670.7e498afd5545.7645.1 +3 -0
- model-bin/finetune/base/log/1629889158.7460623/events.out.tfevents.1629889158.7e498afd5545.7645.3 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629885731.7e498afd5545.905.102 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629886673.7e498afd5545.6884.0 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629887166.7e498afd5545.6884.2 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629887754.7e498afd5545.6884.4 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629888265.7e498afd5545.6884.6 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629888670.7e498afd5545.7645.0 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629889158.7e498afd5545.7645.2 +3 -0
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/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:d3050cf20869e3a9448541918d2c5f60759d8b463489abec7d2351b162043b44
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/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:d64d095173484c67aac73bbcda03020e34b97526f9fe227094721e85f2e05439
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/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 sha256:2f3897e9bd2592aac309ca2470aa4e1dbf7e9bfe7ea5bc705b40d20586c37fd3
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/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:ae3fdca3ff26c556338acdb54399668f539904fe5603e88e8af61c26dd7b7c6d
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/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 sha256:5e90f7e3da290c47dbc970fd03b66691cf691563f4a2d2ab877317c750af9d81
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -195789,11 +195789,959 @@
|
|
| 195789 |
"eval_steps_per_second": 0.662,
|
| 195790 |
"eval_wer": 0.1924281241104469,
|
| 195791 |
"step": 75787
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 195792 |
}
|
| 195793 |
],
|
| 195794 |
"max_steps": 625000,
|
| 195795 |
"num_train_epochs": 5000,
|
| 195796 |
-
"total_flos": 2.
|
| 195797 |
"trial_name": null,
|
| 195798 |
"trial_params": null
|
| 195799 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.18412114350410416,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-69565",
|
| 4 |
+
"epoch": 612.0,
|
| 5 |
+
"global_step": 76532,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 195789 |
"eval_steps_per_second": 0.662,
|
| 195790 |
"eval_wer": 0.1924281241104469,
|
| 195791 |
"step": 75787
|
| 195792 |
+
},
|
| 195793 |
+
{
|
| 195794 |
+
"epoch": 611.02,
|
| 195795 |
+
"learning_rate": 8.801907051282052e-06,
|
| 195796 |
+
"loss": 0.3247,
|
| 195797 |
+
"step": 75790
|
| 195798 |
+
},
|
| 195799 |
+
{
|
| 195800 |
+
"epoch": 611.06,
|
| 195801 |
+
"learning_rate": 8.801826923076923e-06,
|
| 195802 |
+
"loss": 0.3314,
|
| 195803 |
+
"step": 75795
|
| 195804 |
+
},
|
| 195805 |
+
{
|
| 195806 |
+
"epoch": 611.1,
|
| 195807 |
+
"learning_rate": 8.801746794871796e-06,
|
| 195808 |
+
"loss": 0.3502,
|
| 195809 |
+
"step": 75800
|
| 195810 |
+
},
|
| 195811 |
+
{
|
| 195812 |
+
"epoch": 611.14,
|
| 195813 |
+
"learning_rate": 8.801666666666667e-06,
|
| 195814 |
+
"loss": 0.3793,
|
| 195815 |
+
"step": 75805
|
| 195816 |
+
},
|
| 195817 |
+
{
|
| 195818 |
+
"epoch": 611.18,
|
| 195819 |
+
"learning_rate": 8.801586538461539e-06,
|
| 195820 |
+
"loss": 0.7933,
|
| 195821 |
+
"step": 75810
|
| 195822 |
+
},
|
| 195823 |
+
{
|
| 195824 |
+
"epoch": 611.22,
|
| 195825 |
+
"learning_rate": 8.80150641025641e-06,
|
| 195826 |
+
"loss": 0.9573,
|
| 195827 |
+
"step": 75815
|
| 195828 |
+
},
|
| 195829 |
+
{
|
| 195830 |
+
"epoch": 611.27,
|
| 195831 |
+
"learning_rate": 8.801426282051283e-06,
|
| 195832 |
+
"loss": 0.2859,
|
| 195833 |
+
"step": 75820
|
| 195834 |
+
},
|
| 195835 |
+
{
|
| 195836 |
+
"epoch": 611.31,
|
| 195837 |
+
"learning_rate": 8.801346153846155e-06,
|
| 195838 |
+
"loss": 0.4085,
|
| 195839 |
+
"step": 75825
|
| 195840 |
+
},
|
| 195841 |
+
{
|
| 195842 |
+
"epoch": 611.35,
|
| 195843 |
+
"learning_rate": 8.801266025641026e-06,
|
| 195844 |
+
"loss": 0.3864,
|
| 195845 |
+
"step": 75830
|
| 195846 |
+
},
|
| 195847 |
+
{
|
| 195848 |
+
"epoch": 611.39,
|
| 195849 |
+
"learning_rate": 8.801185897435897e-06,
|
| 195850 |
+
"loss": 0.8069,
|
| 195851 |
+
"step": 75835
|
| 195852 |
+
},
|
| 195853 |
+
{
|
| 195854 |
+
"epoch": 611.43,
|
| 195855 |
+
"learning_rate": 8.80110576923077e-06,
|
| 195856 |
+
"loss": 0.7998,
|
| 195857 |
+
"step": 75840
|
| 195858 |
+
},
|
| 195859 |
+
{
|
| 195860 |
+
"epoch": 611.47,
|
| 195861 |
+
"learning_rate": 8.801025641025642e-06,
|
| 195862 |
+
"loss": 0.3127,
|
| 195863 |
+
"step": 75845
|
| 195864 |
+
},
|
| 195865 |
+
{
|
| 195866 |
+
"epoch": 611.51,
|
| 195867 |
+
"learning_rate": 8.800945512820513e-06,
|
| 195868 |
+
"loss": 0.4849,
|
| 195869 |
+
"step": 75850
|
| 195870 |
+
},
|
| 195871 |
+
{
|
| 195872 |
+
"epoch": 611.55,
|
| 195873 |
+
"learning_rate": 8.800865384615386e-06,
|
| 195874 |
+
"loss": 0.3793,
|
| 195875 |
+
"step": 75855
|
| 195876 |
+
},
|
| 195877 |
+
{
|
| 195878 |
+
"epoch": 611.59,
|
| 195879 |
+
"learning_rate": 8.800785256410257e-06,
|
| 195880 |
+
"loss": 0.7649,
|
| 195881 |
+
"step": 75860
|
| 195882 |
+
},
|
| 195883 |
+
{
|
| 195884 |
+
"epoch": 611.63,
|
| 195885 |
+
"learning_rate": 8.800705128205129e-06,
|
| 195886 |
+
"loss": 0.9171,
|
| 195887 |
+
"step": 75865
|
| 195888 |
+
},
|
| 195889 |
+
{
|
| 195890 |
+
"epoch": 611.67,
|
| 195891 |
+
"learning_rate": 8.800625e-06,
|
| 195892 |
+
"loss": 0.3034,
|
| 195893 |
+
"step": 75870
|
| 195894 |
+
},
|
| 195895 |
+
{
|
| 195896 |
+
"epoch": 611.71,
|
| 195897 |
+
"learning_rate": 8.800544871794873e-06,
|
| 195898 |
+
"loss": 0.3238,
|
| 195899 |
+
"step": 75875
|
| 195900 |
+
},
|
| 195901 |
+
{
|
| 195902 |
+
"epoch": 611.75,
|
| 195903 |
+
"learning_rate": 8.800464743589745e-06,
|
| 195904 |
+
"loss": 0.3996,
|
| 195905 |
+
"step": 75880
|
| 195906 |
+
},
|
| 195907 |
+
{
|
| 195908 |
+
"epoch": 611.79,
|
| 195909 |
+
"learning_rate": 8.800384615384616e-06,
|
| 195910 |
+
"loss": 0.8372,
|
| 195911 |
+
"step": 75885
|
| 195912 |
+
},
|
| 195913 |
+
{
|
| 195914 |
+
"epoch": 611.83,
|
| 195915 |
+
"learning_rate": 8.800304487179487e-06,
|
| 195916 |
+
"loss": 0.8874,
|
| 195917 |
+
"step": 75890
|
| 195918 |
+
},
|
| 195919 |
+
{
|
| 195920 |
+
"epoch": 611.87,
|
| 195921 |
+
"learning_rate": 8.80022435897436e-06,
|
| 195922 |
+
"loss": 0.3511,
|
| 195923 |
+
"step": 75895
|
| 195924 |
+
},
|
| 195925 |
+
{
|
| 195926 |
+
"epoch": 611.91,
|
| 195927 |
+
"learning_rate": 8.800144230769232e-06,
|
| 195928 |
+
"loss": 0.5591,
|
| 195929 |
+
"step": 75900
|
| 195930 |
+
},
|
| 195931 |
+
{
|
| 195932 |
+
"epoch": 611.95,
|
| 195933 |
+
"learning_rate": 8.800064102564103e-06,
|
| 195934 |
+
"loss": 0.411,
|
| 195935 |
+
"step": 75905
|
| 195936 |
+
},
|
| 195937 |
+
{
|
| 195938 |
+
"epoch": 611.99,
|
| 195939 |
+
"learning_rate": 8.799983974358976e-06,
|
| 195940 |
+
"loss": 0.8997,
|
| 195941 |
+
"step": 75910
|
| 195942 |
+
},
|
| 195943 |
+
{
|
| 195944 |
+
"epoch": 612.0,
|
| 195945 |
+
"eval_loss": 0.4298833906650543,
|
| 195946 |
+
"eval_runtime": 40.7832,
|
| 195947 |
+
"eval_samples_per_second": 20.597,
|
| 195948 |
+
"eval_steps_per_second": 0.662,
|
| 195949 |
+
"eval_wer": 0.18921908413833788,
|
| 195950 |
+
"step": 75911
|
| 195951 |
+
},
|
| 195952 |
+
{
|
| 195953 |
+
"epoch": 612.03,
|
| 195954 |
+
"learning_rate": 8.79021001615509e-06,
|
| 195955 |
+
"loss": 0.3468,
|
| 195956 |
+
"step": 75915
|
| 195957 |
+
},
|
| 195958 |
+
{
|
| 195959 |
+
"epoch": 612.07,
|
| 195960 |
+
"learning_rate": 8.790129240710824e-06,
|
| 195961 |
+
"loss": 0.2899,
|
| 195962 |
+
"step": 75920
|
| 195963 |
+
},
|
| 195964 |
+
{
|
| 195965 |
+
"epoch": 612.11,
|
| 195966 |
+
"learning_rate": 8.79004846526656e-06,
|
| 195967 |
+
"loss": 0.357,
|
| 195968 |
+
"step": 75925
|
| 195969 |
+
},
|
| 195970 |
+
{
|
| 195971 |
+
"epoch": 612.15,
|
| 195972 |
+
"learning_rate": 8.789967689822296e-06,
|
| 195973 |
+
"loss": 0.4627,
|
| 195974 |
+
"step": 75930
|
| 195975 |
+
},
|
| 195976 |
+
{
|
| 195977 |
+
"epoch": 612.19,
|
| 195978 |
+
"learning_rate": 8.78988691437803e-06,
|
| 195979 |
+
"loss": 0.8461,
|
| 195980 |
+
"step": 75935
|
| 195981 |
+
},
|
| 195982 |
+
{
|
| 195983 |
+
"epoch": 612.23,
|
| 195984 |
+
"learning_rate": 8.789806138933766e-06,
|
| 195985 |
+
"loss": 0.7147,
|
| 195986 |
+
"step": 75940
|
| 195987 |
+
},
|
| 195988 |
+
{
|
| 195989 |
+
"epoch": 612.27,
|
| 195990 |
+
"learning_rate": 8.7897253634895e-06,
|
| 195991 |
+
"loss": 0.2815,
|
| 195992 |
+
"step": 75945
|
| 195993 |
+
},
|
| 195994 |
+
{
|
| 195995 |
+
"epoch": 612.31,
|
| 195996 |
+
"learning_rate": 8.789644588045236e-06,
|
| 195997 |
+
"loss": 0.3392,
|
| 195998 |
+
"step": 75950
|
| 195999 |
+
},
|
| 196000 |
+
{
|
| 196001 |
+
"epoch": 612.35,
|
| 196002 |
+
"learning_rate": 8.78956381260097e-06,
|
| 196003 |
+
"loss": 0.4352,
|
| 196004 |
+
"step": 75955
|
| 196005 |
+
},
|
| 196006 |
+
{
|
| 196007 |
+
"epoch": 612.4,
|
| 196008 |
+
"learning_rate": 8.789483037156706e-06,
|
| 196009 |
+
"loss": 0.9716,
|
| 196010 |
+
"step": 75960
|
| 196011 |
+
},
|
| 196012 |
+
{
|
| 196013 |
+
"epoch": 612.44,
|
| 196014 |
+
"learning_rate": 8.78940226171244e-06,
|
| 196015 |
+
"loss": 0.6372,
|
| 196016 |
+
"step": 75965
|
| 196017 |
+
},
|
| 196018 |
+
{
|
| 196019 |
+
"epoch": 612.48,
|
| 196020 |
+
"learning_rate": 8.789321486268176e-06,
|
| 196021 |
+
"loss": 0.3296,
|
| 196022 |
+
"step": 75970
|
| 196023 |
+
},
|
| 196024 |
+
{
|
| 196025 |
+
"epoch": 612.52,
|
| 196026 |
+
"learning_rate": 8.78924071082391e-06,
|
| 196027 |
+
"loss": 0.3629,
|
| 196028 |
+
"step": 75975
|
| 196029 |
+
},
|
| 196030 |
+
{
|
| 196031 |
+
"epoch": 612.56,
|
| 196032 |
+
"learning_rate": 8.789159935379646e-06,
|
| 196033 |
+
"loss": 0.3986,
|
| 196034 |
+
"step": 75980
|
| 196035 |
+
},
|
| 196036 |
+
{
|
| 196037 |
+
"epoch": 612.6,
|
| 196038 |
+
"learning_rate": 8.789079159935382e-06,
|
| 196039 |
+
"loss": 0.9025,
|
| 196040 |
+
"step": 75985
|
| 196041 |
+
},
|
| 196042 |
+
{
|
| 196043 |
+
"epoch": 612.64,
|
| 196044 |
+
"learning_rate": 8.788998384491116e-06,
|
| 196045 |
+
"loss": 0.888,
|
| 196046 |
+
"step": 75990
|
| 196047 |
+
},
|
| 196048 |
+
{
|
| 196049 |
+
"epoch": 612.68,
|
| 196050 |
+
"learning_rate": 8.788917609046852e-06,
|
| 196051 |
+
"loss": 0.378,
|
| 196052 |
+
"step": 75995
|
| 196053 |
+
},
|
| 196054 |
+
{
|
| 196055 |
+
"epoch": 612.72,
|
| 196056 |
+
"learning_rate": 8.788836833602586e-06,
|
| 196057 |
+
"loss": 0.3348,
|
| 196058 |
+
"step": 76000
|
| 196059 |
+
},
|
| 196060 |
+
{
|
| 196061 |
+
"epoch": 612.76,
|
| 196062 |
+
"learning_rate": 8.788756058158322e-06,
|
| 196063 |
+
"loss": 0.4165,
|
| 196064 |
+
"step": 76005
|
| 196065 |
+
},
|
| 196066 |
+
{
|
| 196067 |
+
"epoch": 612.8,
|
| 196068 |
+
"learning_rate": 8.788675282714056e-06,
|
| 196069 |
+
"loss": 0.8769,
|
| 196070 |
+
"step": 76010
|
| 196071 |
+
},
|
| 196072 |
+
{
|
| 196073 |
+
"epoch": 612.84,
|
| 196074 |
+
"learning_rate": 8.788594507269792e-06,
|
| 196075 |
+
"loss": 0.6691,
|
| 196076 |
+
"step": 76015
|
| 196077 |
+
},
|
| 196078 |
+
{
|
| 196079 |
+
"epoch": 612.88,
|
| 196080 |
+
"learning_rate": 8.788513731825526e-06,
|
| 196081 |
+
"loss": 0.3324,
|
| 196082 |
+
"step": 76020
|
| 196083 |
+
},
|
| 196084 |
+
{
|
| 196085 |
+
"epoch": 612.92,
|
| 196086 |
+
"learning_rate": 8.788432956381262e-06,
|
| 196087 |
+
"loss": 0.3127,
|
| 196088 |
+
"step": 76025
|
| 196089 |
+
},
|
| 196090 |
+
{
|
| 196091 |
+
"epoch": 612.96,
|
| 196092 |
+
"learning_rate": 8.788352180936996e-06,
|
| 196093 |
+
"loss": 0.4538,
|
| 196094 |
+
"step": 76030
|
| 196095 |
+
},
|
| 196096 |
+
{
|
| 196097 |
+
"epoch": 613.0,
|
| 196098 |
+
"learning_rate": 8.788271405492731e-06,
|
| 196099 |
+
"loss": 1.2654,
|
| 196100 |
+
"step": 76035
|
| 196101 |
+
},
|
| 196102 |
+
{
|
| 196103 |
+
"epoch": 613.0,
|
| 196104 |
+
"eval_loss": 0.42023032903671265,
|
| 196105 |
+
"eval_runtime": 45.5765,
|
| 196106 |
+
"eval_samples_per_second": 18.452,
|
| 196107 |
+
"eval_steps_per_second": 0.592,
|
| 196108 |
+
"eval_wer": 0.18862385321100916,
|
| 196109 |
+
"step": 76035
|
| 196110 |
+
},
|
| 196111 |
+
{
|
| 196112 |
+
"epoch": 613.04,
|
| 196113 |
+
"learning_rate": 8.788190630048467e-06,
|
| 196114 |
+
"loss": 0.3605,
|
| 196115 |
+
"step": 76040
|
| 196116 |
+
},
|
| 196117 |
+
{
|
| 196118 |
+
"epoch": 613.08,
|
| 196119 |
+
"learning_rate": 8.788109854604201e-06,
|
| 196120 |
+
"loss": 0.3461,
|
| 196121 |
+
"step": 76045
|
| 196122 |
+
},
|
| 196123 |
+
{
|
| 196124 |
+
"epoch": 613.12,
|
| 196125 |
+
"learning_rate": 8.788029079159937e-06,
|
| 196126 |
+
"loss": 0.3285,
|
| 196127 |
+
"step": 76050
|
| 196128 |
+
},
|
| 196129 |
+
{
|
| 196130 |
+
"epoch": 613.16,
|
| 196131 |
+
"learning_rate": 8.787948303715671e-06,
|
| 196132 |
+
"loss": 0.4423,
|
| 196133 |
+
"step": 76055
|
| 196134 |
+
},
|
| 196135 |
+
{
|
| 196136 |
+
"epoch": 613.2,
|
| 196137 |
+
"learning_rate": 8.787867528271407e-06,
|
| 196138 |
+
"loss": 1.1246,
|
| 196139 |
+
"step": 76060
|
| 196140 |
+
},
|
| 196141 |
+
{
|
| 196142 |
+
"epoch": 613.24,
|
| 196143 |
+
"learning_rate": 8.787786752827141e-06,
|
| 196144 |
+
"loss": 0.4245,
|
| 196145 |
+
"step": 76065
|
| 196146 |
+
},
|
| 196147 |
+
{
|
| 196148 |
+
"epoch": 613.28,
|
| 196149 |
+
"learning_rate": 8.787705977382877e-06,
|
| 196150 |
+
"loss": 0.3375,
|
| 196151 |
+
"step": 76070
|
| 196152 |
+
},
|
| 196153 |
+
{
|
| 196154 |
+
"epoch": 613.32,
|
| 196155 |
+
"learning_rate": 8.787625201938611e-06,
|
| 196156 |
+
"loss": 0.3201,
|
| 196157 |
+
"step": 76075
|
| 196158 |
+
},
|
| 196159 |
+
{
|
| 196160 |
+
"epoch": 613.36,
|
| 196161 |
+
"learning_rate": 8.787544426494347e-06,
|
| 196162 |
+
"loss": 0.4722,
|
| 196163 |
+
"step": 76080
|
| 196164 |
+
},
|
| 196165 |
+
{
|
| 196166 |
+
"epoch": 613.4,
|
| 196167 |
+
"learning_rate": 8.787463651050081e-06,
|
| 196168 |
+
"loss": 1.5154,
|
| 196169 |
+
"step": 76085
|
| 196170 |
+
},
|
| 196171 |
+
{
|
| 196172 |
+
"epoch": 613.44,
|
| 196173 |
+
"learning_rate": 8.787382875605817e-06,
|
| 196174 |
+
"loss": 0.3286,
|
| 196175 |
+
"step": 76090
|
| 196176 |
+
},
|
| 196177 |
+
{
|
| 196178 |
+
"epoch": 613.48,
|
| 196179 |
+
"learning_rate": 8.787302100161551e-06,
|
| 196180 |
+
"loss": 0.3029,
|
| 196181 |
+
"step": 76095
|
| 196182 |
+
},
|
| 196183 |
+
{
|
| 196184 |
+
"epoch": 613.52,
|
| 196185 |
+
"learning_rate": 8.787221324717287e-06,
|
| 196186 |
+
"loss": 0.3839,
|
| 196187 |
+
"step": 76100
|
| 196188 |
+
},
|
| 196189 |
+
{
|
| 196190 |
+
"epoch": 613.56,
|
| 196191 |
+
"learning_rate": 8.787140549273023e-06,
|
| 196192 |
+
"loss": 0.4412,
|
| 196193 |
+
"step": 76105
|
| 196194 |
+
},
|
| 196195 |
+
{
|
| 196196 |
+
"epoch": 613.6,
|
| 196197 |
+
"learning_rate": 8.787059773828757e-06,
|
| 196198 |
+
"loss": 1.21,
|
| 196199 |
+
"step": 76110
|
| 196200 |
+
},
|
| 196201 |
+
{
|
| 196202 |
+
"epoch": 613.64,
|
| 196203 |
+
"learning_rate": 8.786978998384493e-06,
|
| 196204 |
+
"loss": 0.3518,
|
| 196205 |
+
"step": 76115
|
| 196206 |
+
},
|
| 196207 |
+
{
|
| 196208 |
+
"epoch": 613.68,
|
| 196209 |
+
"learning_rate": 8.786898222940227e-06,
|
| 196210 |
+
"loss": 0.3239,
|
| 196211 |
+
"step": 76120
|
| 196212 |
+
},
|
| 196213 |
+
{
|
| 196214 |
+
"epoch": 613.72,
|
| 196215 |
+
"learning_rate": 8.786817447495963e-06,
|
| 196216 |
+
"loss": 0.36,
|
| 196217 |
+
"step": 76125
|
| 196218 |
+
},
|
| 196219 |
+
{
|
| 196220 |
+
"epoch": 613.76,
|
| 196221 |
+
"learning_rate": 8.786736672051697e-06,
|
| 196222 |
+
"loss": 0.4635,
|
| 196223 |
+
"step": 76130
|
| 196224 |
+
},
|
| 196225 |
+
{
|
| 196226 |
+
"epoch": 613.8,
|
| 196227 |
+
"learning_rate": 8.786655896607433e-06,
|
| 196228 |
+
"loss": 1.2953,
|
| 196229 |
+
"step": 76135
|
| 196230 |
+
},
|
| 196231 |
+
{
|
| 196232 |
+
"epoch": 613.84,
|
| 196233 |
+
"learning_rate": 8.786575121163167e-06,
|
| 196234 |
+
"loss": 0.3409,
|
| 196235 |
+
"step": 76140
|
| 196236 |
+
},
|
| 196237 |
+
{
|
| 196238 |
+
"epoch": 613.88,
|
| 196239 |
+
"learning_rate": 8.786494345718903e-06,
|
| 196240 |
+
"loss": 0.3064,
|
| 196241 |
+
"step": 76145
|
| 196242 |
+
},
|
| 196243 |
+
{
|
| 196244 |
+
"epoch": 613.92,
|
| 196245 |
+
"learning_rate": 8.786413570274637e-06,
|
| 196246 |
+
"loss": 0.3543,
|
| 196247 |
+
"step": 76150
|
| 196248 |
+
},
|
| 196249 |
+
{
|
| 196250 |
+
"epoch": 613.96,
|
| 196251 |
+
"learning_rate": 8.786332794830373e-06,
|
| 196252 |
+
"loss": 0.6372,
|
| 196253 |
+
"step": 76155
|
| 196254 |
+
},
|
| 196255 |
+
{
|
| 196256 |
+
"epoch": 614.0,
|
| 196257 |
+
"eval_loss": 0.35451439023017883,
|
| 196258 |
+
"eval_runtime": 43.26,
|
| 196259 |
+
"eval_samples_per_second": 19.417,
|
| 196260 |
+
"eval_steps_per_second": 0.624,
|
| 196261 |
+
"eval_wer": 0.1899717514124294,
|
| 196262 |
+
"step": 76159
|
| 196263 |
+
},
|
| 196264 |
+
{
|
| 196265 |
+
"epoch": 614.01,
|
| 196266 |
+
"learning_rate": 8.786252019386107e-06,
|
| 196267 |
+
"loss": 0.3222,
|
| 196268 |
+
"step": 76160
|
| 196269 |
+
},
|
| 196270 |
+
{
|
| 196271 |
+
"epoch": 614.05,
|
| 196272 |
+
"learning_rate": 8.786171243941843e-06,
|
| 196273 |
+
"loss": 0.2906,
|
| 196274 |
+
"step": 76165
|
| 196275 |
+
},
|
| 196276 |
+
{
|
| 196277 |
+
"epoch": 614.09,
|
| 196278 |
+
"learning_rate": 8.786090468497579e-06,
|
| 196279 |
+
"loss": 0.3531,
|
| 196280 |
+
"step": 76170
|
| 196281 |
+
},
|
| 196282 |
+
{
|
| 196283 |
+
"epoch": 614.13,
|
| 196284 |
+
"learning_rate": 8.786009693053313e-06,
|
| 196285 |
+
"loss": 0.3645,
|
| 196286 |
+
"step": 76175
|
| 196287 |
+
},
|
| 196288 |
+
{
|
| 196289 |
+
"epoch": 614.17,
|
| 196290 |
+
"learning_rate": 8.785928917609049e-06,
|
| 196291 |
+
"loss": 0.5599,
|
| 196292 |
+
"step": 76180
|
| 196293 |
+
},
|
| 196294 |
+
{
|
| 196295 |
+
"epoch": 614.21,
|
| 196296 |
+
"learning_rate": 8.785848142164783e-06,
|
| 196297 |
+
"loss": 1.053,
|
| 196298 |
+
"step": 76185
|
| 196299 |
+
},
|
| 196300 |
+
{
|
| 196301 |
+
"epoch": 614.25,
|
| 196302 |
+
"learning_rate": 8.785767366720519e-06,
|
| 196303 |
+
"loss": 0.3648,
|
| 196304 |
+
"step": 76190
|
| 196305 |
+
},
|
| 196306 |
+
{
|
| 196307 |
+
"epoch": 614.29,
|
| 196308 |
+
"learning_rate": 8.785686591276253e-06,
|
| 196309 |
+
"loss": 0.3137,
|
| 196310 |
+
"step": 76195
|
| 196311 |
+
},
|
| 196312 |
+
{
|
| 196313 |
+
"epoch": 614.33,
|
| 196314 |
+
"learning_rate": 8.785605815831989e-06,
|
| 196315 |
+
"loss": 0.3806,
|
| 196316 |
+
"step": 76200
|
| 196317 |
+
},
|
| 196318 |
+
{
|
| 196319 |
+
"epoch": 614.37,
|
| 196320 |
+
"learning_rate": 8.785525040387723e-06,
|
| 196321 |
+
"loss": 0.6722,
|
| 196322 |
+
"step": 76205
|
| 196323 |
+
},
|
| 196324 |
+
{
|
| 196325 |
+
"epoch": 614.41,
|
| 196326 |
+
"learning_rate": 8.785444264943459e-06,
|
| 196327 |
+
"loss": 1.2406,
|
| 196328 |
+
"step": 76210
|
| 196329 |
+
},
|
| 196330 |
+
{
|
| 196331 |
+
"epoch": 614.45,
|
| 196332 |
+
"learning_rate": 8.785363489499193e-06,
|
| 196333 |
+
"loss": 0.3377,
|
| 196334 |
+
"step": 76215
|
| 196335 |
+
},
|
| 196336 |
+
{
|
| 196337 |
+
"epoch": 614.49,
|
| 196338 |
+
"learning_rate": 8.785282714054929e-06,
|
| 196339 |
+
"loss": 0.3204,
|
| 196340 |
+
"step": 76220
|
| 196341 |
+
},
|
| 196342 |
+
{
|
| 196343 |
+
"epoch": 614.53,
|
| 196344 |
+
"learning_rate": 8.785201938610663e-06,
|
| 196345 |
+
"loss": 0.3461,
|
| 196346 |
+
"step": 76225
|
| 196347 |
+
},
|
| 196348 |
+
{
|
| 196349 |
+
"epoch": 614.57,
|
| 196350 |
+
"learning_rate": 8.785121163166399e-06,
|
| 196351 |
+
"loss": 0.5858,
|
| 196352 |
+
"step": 76230
|
| 196353 |
+
},
|
| 196354 |
+
{
|
| 196355 |
+
"epoch": 614.61,
|
| 196356 |
+
"learning_rate": 8.785040387722134e-06,
|
| 196357 |
+
"loss": 1.2116,
|
| 196358 |
+
"step": 76235
|
| 196359 |
+
},
|
| 196360 |
+
{
|
| 196361 |
+
"epoch": 614.65,
|
| 196362 |
+
"learning_rate": 8.784959612277869e-06,
|
| 196363 |
+
"loss": 0.3268,
|
| 196364 |
+
"step": 76240
|
| 196365 |
+
},
|
| 196366 |
+
{
|
| 196367 |
+
"epoch": 614.69,
|
| 196368 |
+
"learning_rate": 8.784878836833604e-06,
|
| 196369 |
+
"loss": 0.2913,
|
| 196370 |
+
"step": 76245
|
| 196371 |
+
},
|
| 196372 |
+
{
|
| 196373 |
+
"epoch": 614.73,
|
| 196374 |
+
"learning_rate": 8.784798061389338e-06,
|
| 196375 |
+
"loss": 0.3399,
|
| 196376 |
+
"step": 76250
|
| 196377 |
+
},
|
| 196378 |
+
{
|
| 196379 |
+
"epoch": 614.77,
|
| 196380 |
+
"learning_rate": 8.784717285945074e-06,
|
| 196381 |
+
"loss": 0.6186,
|
| 196382 |
+
"step": 76255
|
| 196383 |
+
},
|
| 196384 |
+
{
|
| 196385 |
+
"epoch": 614.81,
|
| 196386 |
+
"learning_rate": 8.784636510500808e-06,
|
| 196387 |
+
"loss": 1.1712,
|
| 196388 |
+
"step": 76260
|
| 196389 |
+
},
|
| 196390 |
+
{
|
| 196391 |
+
"epoch": 614.85,
|
| 196392 |
+
"learning_rate": 8.784555735056544e-06,
|
| 196393 |
+
"loss": 0.2874,
|
| 196394 |
+
"step": 76265
|
| 196395 |
+
},
|
| 196396 |
+
{
|
| 196397 |
+
"epoch": 614.89,
|
| 196398 |
+
"learning_rate": 8.784474959612278e-06,
|
| 196399 |
+
"loss": 0.3038,
|
| 196400 |
+
"step": 76270
|
| 196401 |
+
},
|
| 196402 |
+
{
|
| 196403 |
+
"epoch": 614.93,
|
| 196404 |
+
"learning_rate": 8.784394184168014e-06,
|
| 196405 |
+
"loss": 0.3714,
|
| 196406 |
+
"step": 76275
|
| 196407 |
+
},
|
| 196408 |
+
{
|
| 196409 |
+
"epoch": 614.97,
|
| 196410 |
+
"learning_rate": 8.784313408723748e-06,
|
| 196411 |
+
"loss": 0.6097,
|
| 196412 |
+
"step": 76280
|
| 196413 |
+
},
|
| 196414 |
+
{
|
| 196415 |
+
"epoch": 615.0,
|
| 196416 |
+
"eval_loss": 0.36748769879341125,
|
| 196417 |
+
"eval_runtime": 45.6308,
|
| 196418 |
+
"eval_samples_per_second": 18.321,
|
| 196419 |
+
"eval_steps_per_second": 0.592,
|
| 196420 |
+
"eval_wer": 0.19178379944935517,
|
| 196421 |
+
"step": 76283
|
| 196422 |
+
},
|
| 196423 |
+
{
|
| 196424 |
+
"epoch": 615.02,
|
| 196425 |
+
"learning_rate": 8.784232633279484e-06,
|
| 196426 |
+
"loss": 0.4276,
|
| 196427 |
+
"step": 76285
|
| 196428 |
+
},
|
| 196429 |
+
{
|
| 196430 |
+
"epoch": 615.06,
|
| 196431 |
+
"learning_rate": 8.784151857835218e-06,
|
| 196432 |
+
"loss": 0.3103,
|
| 196433 |
+
"step": 76290
|
| 196434 |
+
},
|
| 196435 |
+
{
|
| 196436 |
+
"epoch": 615.1,
|
| 196437 |
+
"learning_rate": 8.784071082390954e-06,
|
| 196438 |
+
"loss": 0.345,
|
| 196439 |
+
"step": 76295
|
| 196440 |
+
},
|
| 196441 |
+
{
|
| 196442 |
+
"epoch": 615.14,
|
| 196443 |
+
"learning_rate": 8.783990306946688e-06,
|
| 196444 |
+
"loss": 0.4322,
|
| 196445 |
+
"step": 76300
|
| 196446 |
+
},
|
| 196447 |
+
{
|
| 196448 |
+
"epoch": 615.18,
|
| 196449 |
+
"learning_rate": 8.783909531502424e-06,
|
| 196450 |
+
"loss": 0.5953,
|
| 196451 |
+
"step": 76305
|
| 196452 |
+
},
|
| 196453 |
+
{
|
| 196454 |
+
"epoch": 615.22,
|
| 196455 |
+
"learning_rate": 8.78382875605816e-06,
|
| 196456 |
+
"loss": 1.0507,
|
| 196457 |
+
"step": 76310
|
| 196458 |
+
},
|
| 196459 |
+
{
|
| 196460 |
+
"epoch": 615.26,
|
| 196461 |
+
"learning_rate": 8.783747980613894e-06,
|
| 196462 |
+
"loss": 0.2948,
|
| 196463 |
+
"step": 76315
|
| 196464 |
+
},
|
| 196465 |
+
{
|
| 196466 |
+
"epoch": 615.3,
|
| 196467 |
+
"learning_rate": 8.78366720516963e-06,
|
| 196468 |
+
"loss": 0.3103,
|
| 196469 |
+
"step": 76320
|
| 196470 |
+
},
|
| 196471 |
+
{
|
| 196472 |
+
"epoch": 615.34,
|
| 196473 |
+
"learning_rate": 8.783586429725364e-06,
|
| 196474 |
+
"loss": 0.3602,
|
| 196475 |
+
"step": 76325
|
| 196476 |
+
},
|
| 196477 |
+
{
|
| 196478 |
+
"epoch": 615.38,
|
| 196479 |
+
"learning_rate": 8.7835056542811e-06,
|
| 196480 |
+
"loss": 0.6729,
|
| 196481 |
+
"step": 76330
|
| 196482 |
+
},
|
| 196483 |
+
{
|
| 196484 |
+
"epoch": 615.42,
|
| 196485 |
+
"learning_rate": 8.783424878836834e-06,
|
| 196486 |
+
"loss": 1.1124,
|
| 196487 |
+
"step": 76335
|
| 196488 |
+
},
|
| 196489 |
+
{
|
| 196490 |
+
"epoch": 615.46,
|
| 196491 |
+
"learning_rate": 8.78334410339257e-06,
|
| 196492 |
+
"loss": 0.3663,
|
| 196493 |
+
"step": 76340
|
| 196494 |
+
},
|
| 196495 |
+
{
|
| 196496 |
+
"epoch": 615.5,
|
| 196497 |
+
"learning_rate": 8.783263327948304e-06,
|
| 196498 |
+
"loss": 0.2851,
|
| 196499 |
+
"step": 76345
|
| 196500 |
+
},
|
| 196501 |
+
{
|
| 196502 |
+
"epoch": 615.54,
|
| 196503 |
+
"learning_rate": 8.78318255250404e-06,
|
| 196504 |
+
"loss": 0.3856,
|
| 196505 |
+
"step": 76350
|
| 196506 |
+
},
|
| 196507 |
+
{
|
| 196508 |
+
"epoch": 615.58,
|
| 196509 |
+
"learning_rate": 8.783101777059774e-06,
|
| 196510 |
+
"loss": 0.6889,
|
| 196511 |
+
"step": 76355
|
| 196512 |
+
},
|
| 196513 |
+
{
|
| 196514 |
+
"epoch": 615.62,
|
| 196515 |
+
"learning_rate": 8.78302100161551e-06,
|
| 196516 |
+
"loss": 1.1928,
|
| 196517 |
+
"step": 76360
|
| 196518 |
+
},
|
| 196519 |
+
{
|
| 196520 |
+
"epoch": 615.66,
|
| 196521 |
+
"learning_rate": 8.782940226171244e-06,
|
| 196522 |
+
"loss": 0.336,
|
| 196523 |
+
"step": 76365
|
| 196524 |
+
},
|
| 196525 |
+
{
|
| 196526 |
+
"epoch": 615.7,
|
| 196527 |
+
"learning_rate": 8.78285945072698e-06,
|
| 196528 |
+
"loss": 0.3681,
|
| 196529 |
+
"step": 76370
|
| 196530 |
+
},
|
| 196531 |
+
{
|
| 196532 |
+
"epoch": 615.74,
|
| 196533 |
+
"learning_rate": 8.782778675282716e-06,
|
| 196534 |
+
"loss": 0.3314,
|
| 196535 |
+
"step": 76375
|
| 196536 |
+
},
|
| 196537 |
+
{
|
| 196538 |
+
"epoch": 615.78,
|
| 196539 |
+
"learning_rate": 8.78269789983845e-06,
|
| 196540 |
+
"loss": 0.6444,
|
| 196541 |
+
"step": 76380
|
| 196542 |
+
},
|
| 196543 |
+
{
|
| 196544 |
+
"epoch": 615.82,
|
| 196545 |
+
"learning_rate": 8.782617124394186e-06,
|
| 196546 |
+
"loss": 0.9598,
|
| 196547 |
+
"step": 76385
|
| 196548 |
+
},
|
| 196549 |
+
{
|
| 196550 |
+
"epoch": 615.86,
|
| 196551 |
+
"learning_rate": 8.78253634894992e-06,
|
| 196552 |
+
"loss": 0.3099,
|
| 196553 |
+
"step": 76390
|
| 196554 |
+
},
|
| 196555 |
+
{
|
| 196556 |
+
"epoch": 615.9,
|
| 196557 |
+
"learning_rate": 8.782455573505656e-06,
|
| 196558 |
+
"loss": 0.3403,
|
| 196559 |
+
"step": 76395
|
| 196560 |
+
},
|
| 196561 |
+
{
|
| 196562 |
+
"epoch": 615.94,
|
| 196563 |
+
"learning_rate": 8.78237479806139e-06,
|
| 196564 |
+
"loss": 0.3402,
|
| 196565 |
+
"step": 76400
|
| 196566 |
+
},
|
| 196567 |
+
{
|
| 196568 |
+
"epoch": 615.98,
|
| 196569 |
+
"learning_rate": 8.782294022617126e-06,
|
| 196570 |
+
"loss": 0.7314,
|
| 196571 |
+
"step": 76405
|
| 196572 |
+
},
|
| 196573 |
+
{
|
| 196574 |
+
"epoch": 616.0,
|
| 196575 |
+
"eval_loss": 0.37296634912490845,
|
| 196576 |
+
"eval_runtime": 44.7239,
|
| 196577 |
+
"eval_samples_per_second": 18.692,
|
| 196578 |
+
"eval_steps_per_second": 0.604,
|
| 196579 |
+
"eval_wer": 0.1930858806404658,
|
| 196580 |
+
"step": 76407
|
| 196581 |
+
},
|
| 196582 |
+
{
|
| 196583 |
+
"epoch": 611.02,
|
| 196584 |
+
"learning_rate": 8.78221324717286e-06,
|
| 196585 |
+
"loss": 0.4121,
|
| 196586 |
+
"step": 76410
|
| 196587 |
+
},
|
| 196588 |
+
{
|
| 196589 |
+
"epoch": 611.06,
|
| 196590 |
+
"learning_rate": 8.782132471728596e-06,
|
| 196591 |
+
"loss": 0.3243,
|
| 196592 |
+
"step": 76415
|
| 196593 |
+
},
|
| 196594 |
+
{
|
| 196595 |
+
"epoch": 611.1,
|
| 196596 |
+
"learning_rate": 8.78205169628433e-06,
|
| 196597 |
+
"loss": 0.2929,
|
| 196598 |
+
"step": 76420
|
| 196599 |
+
},
|
| 196600 |
+
{
|
| 196601 |
+
"epoch": 611.14,
|
| 196602 |
+
"learning_rate": 8.781970920840066e-06,
|
| 196603 |
+
"loss": 0.4178,
|
| 196604 |
+
"step": 76425
|
| 196605 |
+
},
|
| 196606 |
+
{
|
| 196607 |
+
"epoch": 611.18,
|
| 196608 |
+
"learning_rate": 8.7818901453958e-06,
|
| 196609 |
+
"loss": 0.7328,
|
| 196610 |
+
"step": 76430
|
| 196611 |
+
},
|
| 196612 |
+
{
|
| 196613 |
+
"epoch": 611.22,
|
| 196614 |
+
"learning_rate": 8.781809369951536e-06,
|
| 196615 |
+
"loss": 0.9123,
|
| 196616 |
+
"step": 76435
|
| 196617 |
+
},
|
| 196618 |
+
{
|
| 196619 |
+
"epoch": 611.26,
|
| 196620 |
+
"learning_rate": 8.781728594507271e-06,
|
| 196621 |
+
"loss": 0.3589,
|
| 196622 |
+
"step": 76440
|
| 196623 |
+
},
|
| 196624 |
+
{
|
| 196625 |
+
"epoch": 611.3,
|
| 196626 |
+
"learning_rate": 8.781647819063006e-06,
|
| 196627 |
+
"loss": 0.328,
|
| 196628 |
+
"step": 76445
|
| 196629 |
+
},
|
| 196630 |
+
{
|
| 196631 |
+
"epoch": 611.34,
|
| 196632 |
+
"learning_rate": 8.781567043618741e-06,
|
| 196633 |
+
"loss": 0.3831,
|
| 196634 |
+
"step": 76450
|
| 196635 |
+
},
|
| 196636 |
+
{
|
| 196637 |
+
"epoch": 611.38,
|
| 196638 |
+
"learning_rate": 8.781486268174476e-06,
|
| 196639 |
+
"loss": 0.8885,
|
| 196640 |
+
"step": 76455
|
| 196641 |
+
},
|
| 196642 |
+
{
|
| 196643 |
+
"epoch": 611.42,
|
| 196644 |
+
"learning_rate": 8.781405492730211e-06,
|
| 196645 |
+
"loss": 0.8831,
|
| 196646 |
+
"step": 76460
|
| 196647 |
+
},
|
| 196648 |
+
{
|
| 196649 |
+
"epoch": 611.46,
|
| 196650 |
+
"learning_rate": 8.781324717285945e-06,
|
| 196651 |
+
"loss": 0.3094,
|
| 196652 |
+
"step": 76465
|
| 196653 |
+
},
|
| 196654 |
+
{
|
| 196655 |
+
"epoch": 611.5,
|
| 196656 |
+
"learning_rate": 8.781243941841681e-06,
|
| 196657 |
+
"loss": 0.349,
|
| 196658 |
+
"step": 76470
|
| 196659 |
+
},
|
| 196660 |
+
{
|
| 196661 |
+
"epoch": 611.54,
|
| 196662 |
+
"learning_rate": 8.781163166397415e-06,
|
| 196663 |
+
"loss": 0.3866,
|
| 196664 |
+
"step": 76475
|
| 196665 |
+
},
|
| 196666 |
+
{
|
| 196667 |
+
"epoch": 611.58,
|
| 196668 |
+
"learning_rate": 8.781082390953151e-06,
|
| 196669 |
+
"loss": 0.7114,
|
| 196670 |
+
"step": 76480
|
| 196671 |
+
},
|
| 196672 |
+
{
|
| 196673 |
+
"epoch": 611.62,
|
| 196674 |
+
"learning_rate": 8.781001615508885e-06,
|
| 196675 |
+
"loss": 0.92,
|
| 196676 |
+
"step": 76485
|
| 196677 |
+
},
|
| 196678 |
+
{
|
| 196679 |
+
"epoch": 611.66,
|
| 196680 |
+
"learning_rate": 8.780920840064621e-06,
|
| 196681 |
+
"loss": 0.3971,
|
| 196682 |
+
"step": 76490
|
| 196683 |
+
},
|
| 196684 |
+
{
|
| 196685 |
+
"epoch": 611.7,
|
| 196686 |
+
"learning_rate": 8.780840064620355e-06,
|
| 196687 |
+
"loss": 0.3134,
|
| 196688 |
+
"step": 76495
|
| 196689 |
+
},
|
| 196690 |
+
{
|
| 196691 |
+
"epoch": 611.74,
|
| 196692 |
+
"learning_rate": 8.780759289176091e-06,
|
| 196693 |
+
"loss": 0.3991,
|
| 196694 |
+
"step": 76500
|
| 196695 |
+
},
|
| 196696 |
+
{
|
| 196697 |
+
"epoch": 611.78,
|
| 196698 |
+
"learning_rate": 8.780678513731825e-06,
|
| 196699 |
+
"loss": 0.7837,
|
| 196700 |
+
"step": 76505
|
| 196701 |
+
},
|
| 196702 |
+
{
|
| 196703 |
+
"epoch": 611.82,
|
| 196704 |
+
"learning_rate": 8.780597738287561e-06,
|
| 196705 |
+
"loss": 0.8353,
|
| 196706 |
+
"step": 76510
|
| 196707 |
+
},
|
| 196708 |
+
{
|
| 196709 |
+
"epoch": 611.86,
|
| 196710 |
+
"learning_rate": 8.780516962843297e-06,
|
| 196711 |
+
"loss": 0.3332,
|
| 196712 |
+
"step": 76515
|
| 196713 |
+
},
|
| 196714 |
+
{
|
| 196715 |
+
"epoch": 611.9,
|
| 196716 |
+
"learning_rate": 8.780436187399031e-06,
|
| 196717 |
+
"loss": 0.3274,
|
| 196718 |
+
"step": 76520
|
| 196719 |
+
},
|
| 196720 |
+
{
|
| 196721 |
+
"epoch": 611.94,
|
| 196722 |
+
"learning_rate": 8.780355411954767e-06,
|
| 196723 |
+
"loss": 0.3416,
|
| 196724 |
+
"step": 76525
|
| 196725 |
+
},
|
| 196726 |
+
{
|
| 196727 |
+
"epoch": 611.98,
|
| 196728 |
+
"learning_rate": 8.780274636510501e-06,
|
| 196729 |
+
"loss": 0.8589,
|
| 196730 |
+
"step": 76530
|
| 196731 |
+
},
|
| 196732 |
+
{
|
| 196733 |
+
"epoch": 612.0,
|
| 196734 |
+
"eval_loss": 0.39867308735847473,
|
| 196735 |
+
"eval_runtime": 43.2092,
|
| 196736 |
+
"eval_samples_per_second": 19.348,
|
| 196737 |
+
"eval_steps_per_second": 0.625,
|
| 196738 |
+
"eval_wer": 0.18634209392503231,
|
| 196739 |
+
"step": 76532
|
| 196740 |
}
|
| 196741 |
],
|
| 196742 |
"max_steps": 625000,
|
| 196743 |
"num_train_epochs": 5000,
|
| 196744 |
+
"total_flos": 2.1537393625642425e+20,
|
| 196745 |
"trial_name": null,
|
| 196746 |
"trial_params": null
|
| 196747 |
}
|
model-bin/finetune/base/{checkpoint-75787 → checkpoint-76532}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629885731.4375427/events.out.tfevents.1629885731.7e498afd5545.905.103
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:36a6da31eefd7ffcaa0e8679cde1cd51f449d40931f5cab27833d9e7f49fbbf5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629886673.7174067/events.out.tfevents.1629886673.7e498afd5545.6884.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0d74f46ca1c7fd0a37637440eb5e393c900b931ebffb6e9f79f89fe7b987f6d3
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629887166.966957/events.out.tfevents.1629887166.7e498afd5545.6884.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c73db601adf9bf691a8a648aea8ba46f4e54821cc80d4526a5a554e8065b4282
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629887754.1185539/events.out.tfevents.1629887754.7e498afd5545.6884.5
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e1e26f78794d7f6ffd2b00b6d9ef3e22821b21e2508f541e178364b277d0f026
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629888265.905986/events.out.tfevents.1629888265.7e498afd5545.6884.7
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f1bfa56c24bb168c6959bdfd406183d0478e7b3cea14e19c2fbddf4655dadf0c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629888670.3141375/events.out.tfevents.1629888670.7e498afd5545.7645.1
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8f272ccac5f572ccb9d62044928df58ab0257203e84d8b6072f7892177516e05
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629889158.7460623/events.out.tfevents.1629889158.7e498afd5545.7645.3
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:42c46008eee645fc3b59ab7eeb013ba6fe959e277d904c666ac3ad4b1d50f3ff
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629885731.7e498afd5545.905.102
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:32b48829287d7459553a281d2c7951e08c4d48da4bdde6fed52ea4c65723e7aa
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629886673.7e498afd5545.6884.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7dc414e5d7be985ac5c250402f9dbbc24776b0afb23bb9c1366a0f9c13dff95f
|
| 3 |
+
size 8630
|
model-bin/finetune/base/log/events.out.tfevents.1629887166.7e498afd5545.6884.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6b56762b9849e47c832a043579254ae91a2186a04633bcde57efd4ff10c002e1
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629887754.7e498afd5545.6884.4
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2df9a0dd68ca38106b7d0d418ffb03edfa664c29f7748d8201978ad3bf080b65
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629888265.7e498afd5545.6884.6
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cc29f1b159e9ced388223b12cf9a654002917d554eb44d9fbe9f72abd5fba169
|
| 3 |
+
size 3938
|
model-bin/finetune/base/log/events.out.tfevents.1629888670.7e498afd5545.7645.0
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:32f35091b2934d85891ecde443d198745a0801cd4984e4c157d396ab96c25bf7
|
| 3 |
+
size 8630
|
model-bin/finetune/base/log/events.out.tfevents.1629889158.7e498afd5545.7645.2
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c76070cc3d2f3aace7014ca70e3572d106002661d639b2a15e83c9110009f334
|
| 3 |
+
size 8622
|