"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630045620.5655968/events.out.tfevents.1630045620.52f5c7e305a3.886.71 +3 -0
- model-bin/finetune/base/log/1630046118.6803894/events.out.tfevents.1630046118.52f5c7e305a3.886.73 +3 -0
- model-bin/finetune/base/log/1630046681.8092928/events.out.tfevents.1630046681.52f5c7e305a3.886.75 +3 -0
- model-bin/finetune/base/log/1630047172.2078328/events.out.tfevents.1630047172.52f5c7e305a3.886.77 +3 -0
- model-bin/finetune/base/log/1630047639.1210005/events.out.tfevents.1630047639.52f5c7e305a3.886.79 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630045620.52f5c7e305a3.886.70 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630046118.52f5c7e305a3.886.72 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630046681.52f5c7e305a3.886.74 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630047172.52f5c7e305a3.886.76 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630047639.52f5c7e305a3.886.78 +3 -0
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/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:a1b8197993eb37d9ee36acbb12a1f9605ec9f468a36950f0b19cfe128de2b5f0
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/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 sha256:8232bc1150a76590269afa478f75b4236d9220bb172b83e409418781d8efeebd
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/rng_state.pth
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 14503
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a14f61f7dddaa53dbcb08f44b28b663a43e59c31dc4d069de7af3154f39d5215
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/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:f60fd12782841de6f909cf814170025073e913f168ae702ff0cdd93272700668
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/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 sha256:7e9a559e52c3b4b08e9232cdf54f10d4fa2d1e564b84998252aef70d42eaf820
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/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,
|
|
@@ -240282,11 +240282,800 @@
|
|
| 240282 |
"eval_steps_per_second": 0.631,
|
| 240283 |
"eval_wer": 0.19120699881376038,
|
| 240284 |
"step": 110758
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 240285 |
}
|
| 240286 |
],
|
| 240287 |
"max_steps": 620000,
|
| 240288 |
"num_train_epochs": 5000,
|
| 240289 |
-
"total_flos": 3.
|
| 240290 |
"trial_name": null,
|
| 240291 |
"trial_params": null
|
| 240292 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 898.0,
|
| 5 |
+
"global_step": 111379,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 240282 |
"eval_steps_per_second": 0.631,
|
| 240283 |
"eval_wer": 0.19120699881376038,
|
| 240284 |
"step": 110758
|
| 240285 |
+
},
|
| 240286 |
+
{
|
| 240287 |
+
"epoch": 900.02,
|
| 240288 |
+
"learning_rate": 8.227528271405494e-06,
|
| 240289 |
+
"loss": 0.3431,
|
| 240290 |
+
"step": 110760
|
| 240291 |
+
},
|
| 240292 |
+
{
|
| 240293 |
+
"epoch": 900.06,
|
| 240294 |
+
"learning_rate": 8.227447495961228e-06,
|
| 240295 |
+
"loss": 0.2264,
|
| 240296 |
+
"step": 110765
|
| 240297 |
+
},
|
| 240298 |
+
{
|
| 240299 |
+
"epoch": 900.1,
|
| 240300 |
+
"learning_rate": 8.227366720516964e-06,
|
| 240301 |
+
"loss": 0.3079,
|
| 240302 |
+
"step": 110770
|
| 240303 |
+
},
|
| 240304 |
+
{
|
| 240305 |
+
"epoch": 900.14,
|
| 240306 |
+
"learning_rate": 8.227285945072698e-06,
|
| 240307 |
+
"loss": 0.4818,
|
| 240308 |
+
"step": 110775
|
| 240309 |
+
},
|
| 240310 |
+
{
|
| 240311 |
+
"epoch": 900.18,
|
| 240312 |
+
"learning_rate": 8.227205169628434e-06,
|
| 240313 |
+
"loss": 0.7211,
|
| 240314 |
+
"step": 110780
|
| 240315 |
+
},
|
| 240316 |
+
{
|
| 240317 |
+
"epoch": 900.22,
|
| 240318 |
+
"learning_rate": 8.227124394184168e-06,
|
| 240319 |
+
"loss": 1.0533,
|
| 240320 |
+
"step": 110785
|
| 240321 |
+
},
|
| 240322 |
+
{
|
| 240323 |
+
"epoch": 900.26,
|
| 240324 |
+
"learning_rate": 8.227043618739904e-06,
|
| 240325 |
+
"loss": 0.3337,
|
| 240326 |
+
"step": 110790
|
| 240327 |
+
},
|
| 240328 |
+
{
|
| 240329 |
+
"epoch": 900.3,
|
| 240330 |
+
"learning_rate": 8.226962843295638e-06,
|
| 240331 |
+
"loss": 0.3322,
|
| 240332 |
+
"step": 110795
|
| 240333 |
+
},
|
| 240334 |
+
{
|
| 240335 |
+
"epoch": 900.34,
|
| 240336 |
+
"learning_rate": 8.226882067851374e-06,
|
| 240337 |
+
"loss": 0.3738,
|
| 240338 |
+
"step": 110800
|
| 240339 |
+
},
|
| 240340 |
+
{
|
| 240341 |
+
"epoch": 900.38,
|
| 240342 |
+
"learning_rate": 8.22680129240711e-06,
|
| 240343 |
+
"loss": 0.6992,
|
| 240344 |
+
"step": 110805
|
| 240345 |
+
},
|
| 240346 |
+
{
|
| 240347 |
+
"epoch": 900.42,
|
| 240348 |
+
"learning_rate": 8.226720516962844e-06,
|
| 240349 |
+
"loss": 0.9494,
|
| 240350 |
+
"step": 110810
|
| 240351 |
+
},
|
| 240352 |
+
{
|
| 240353 |
+
"epoch": 900.46,
|
| 240354 |
+
"learning_rate": 8.22663974151858e-06,
|
| 240355 |
+
"loss": 0.2978,
|
| 240356 |
+
"step": 110815
|
| 240357 |
+
},
|
| 240358 |
+
{
|
| 240359 |
+
"epoch": 900.5,
|
| 240360 |
+
"learning_rate": 8.226558966074314e-06,
|
| 240361 |
+
"loss": 0.3104,
|
| 240362 |
+
"step": 110820
|
| 240363 |
+
},
|
| 240364 |
+
{
|
| 240365 |
+
"epoch": 900.54,
|
| 240366 |
+
"learning_rate": 8.22647819063005e-06,
|
| 240367 |
+
"loss": 0.3612,
|
| 240368 |
+
"step": 110825
|
| 240369 |
+
},
|
| 240370 |
+
{
|
| 240371 |
+
"epoch": 900.58,
|
| 240372 |
+
"learning_rate": 8.226397415185784e-06,
|
| 240373 |
+
"loss": 0.6377,
|
| 240374 |
+
"step": 110830
|
| 240375 |
+
},
|
| 240376 |
+
{
|
| 240377 |
+
"epoch": 900.62,
|
| 240378 |
+
"learning_rate": 8.22631663974152e-06,
|
| 240379 |
+
"loss": 0.8597,
|
| 240380 |
+
"step": 110835
|
| 240381 |
+
},
|
| 240382 |
+
{
|
| 240383 |
+
"epoch": 900.66,
|
| 240384 |
+
"learning_rate": 8.226235864297254e-06,
|
| 240385 |
+
"loss": 0.3028,
|
| 240386 |
+
"step": 110840
|
| 240387 |
+
},
|
| 240388 |
+
{
|
| 240389 |
+
"epoch": 900.7,
|
| 240390 |
+
"learning_rate": 8.22615508885299e-06,
|
| 240391 |
+
"loss": 0.3288,
|
| 240392 |
+
"step": 110845
|
| 240393 |
+
},
|
| 240394 |
+
{
|
| 240395 |
+
"epoch": 900.74,
|
| 240396 |
+
"learning_rate": 8.226074313408724e-06,
|
| 240397 |
+
"loss": 0.3319,
|
| 240398 |
+
"step": 110850
|
| 240399 |
+
},
|
| 240400 |
+
{
|
| 240401 |
+
"epoch": 900.79,
|
| 240402 |
+
"learning_rate": 8.22599353796446e-06,
|
| 240403 |
+
"loss": 0.5342,
|
| 240404 |
+
"step": 110855
|
| 240405 |
+
},
|
| 240406 |
+
{
|
| 240407 |
+
"epoch": 900.83,
|
| 240408 |
+
"learning_rate": 8.225912762520194e-06,
|
| 240409 |
+
"loss": 0.994,
|
| 240410 |
+
"step": 110860
|
| 240411 |
+
},
|
| 240412 |
+
{
|
| 240413 |
+
"epoch": 900.87,
|
| 240414 |
+
"learning_rate": 8.22583198707593e-06,
|
| 240415 |
+
"loss": 0.3007,
|
| 240416 |
+
"step": 110865
|
| 240417 |
+
},
|
| 240418 |
+
{
|
| 240419 |
+
"epoch": 900.91,
|
| 240420 |
+
"learning_rate": 8.225751211631666e-06,
|
| 240421 |
+
"loss": 0.3091,
|
| 240422 |
+
"step": 110870
|
| 240423 |
+
},
|
| 240424 |
+
{
|
| 240425 |
+
"epoch": 900.95,
|
| 240426 |
+
"learning_rate": 8.2256704361874e-06,
|
| 240427 |
+
"loss": 0.3254,
|
| 240428 |
+
"step": 110875
|
| 240429 |
+
},
|
| 240430 |
+
{
|
| 240431 |
+
"epoch": 900.99,
|
| 240432 |
+
"learning_rate": 8.225589660743136e-06,
|
| 240433 |
+
"loss": 0.7219,
|
| 240434 |
+
"step": 110880
|
| 240435 |
+
},
|
| 240436 |
+
{
|
| 240437 |
+
"epoch": 901.0,
|
| 240438 |
+
"eval_loss": 0.36234134435653687,
|
| 240439 |
+
"eval_runtime": 43.3119,
|
| 240440 |
+
"eval_samples_per_second": 19.325,
|
| 240441 |
+
"eval_steps_per_second": 0.623,
|
| 240442 |
+
"eval_wer": 0.18013960819635216,
|
| 240443 |
+
"step": 110881
|
| 240444 |
+
},
|
| 240445 |
+
{
|
| 240446 |
+
"epoch": 894.03,
|
| 240447 |
+
"learning_rate": 8.22550888529887e-06,
|
| 240448 |
+
"loss": 0.4134,
|
| 240449 |
+
"step": 110885
|
| 240450 |
+
},
|
| 240451 |
+
{
|
| 240452 |
+
"epoch": 894.07,
|
| 240453 |
+
"learning_rate": 8.225428109854606e-06,
|
| 240454 |
+
"loss": 0.2736,
|
| 240455 |
+
"step": 110890
|
| 240456 |
+
},
|
| 240457 |
+
{
|
| 240458 |
+
"epoch": 894.11,
|
| 240459 |
+
"learning_rate": 8.22534733441034e-06,
|
| 240460 |
+
"loss": 0.3003,
|
| 240461 |
+
"step": 110895
|
| 240462 |
+
},
|
| 240463 |
+
{
|
| 240464 |
+
"epoch": 894.15,
|
| 240465 |
+
"learning_rate": 8.225266558966076e-06,
|
| 240466 |
+
"loss": 0.457,
|
| 240467 |
+
"step": 110900
|
| 240468 |
+
},
|
| 240469 |
+
{
|
| 240470 |
+
"epoch": 894.19,
|
| 240471 |
+
"learning_rate": 8.22518578352181e-06,
|
| 240472 |
+
"loss": 0.8826,
|
| 240473 |
+
"step": 110905
|
| 240474 |
+
},
|
| 240475 |
+
{
|
| 240476 |
+
"epoch": 894.23,
|
| 240477 |
+
"learning_rate": 8.225105008077546e-06,
|
| 240478 |
+
"loss": 0.6164,
|
| 240479 |
+
"step": 110910
|
| 240480 |
+
},
|
| 240481 |
+
{
|
| 240482 |
+
"epoch": 894.27,
|
| 240483 |
+
"learning_rate": 8.22502423263328e-06,
|
| 240484 |
+
"loss": 0.2989,
|
| 240485 |
+
"step": 110915
|
| 240486 |
+
},
|
| 240487 |
+
{
|
| 240488 |
+
"epoch": 894.31,
|
| 240489 |
+
"learning_rate": 8.224943457189016e-06,
|
| 240490 |
+
"loss": 0.3474,
|
| 240491 |
+
"step": 110920
|
| 240492 |
+
},
|
| 240493 |
+
{
|
| 240494 |
+
"epoch": 894.35,
|
| 240495 |
+
"learning_rate": 8.224862681744751e-06,
|
| 240496 |
+
"loss": 0.3896,
|
| 240497 |
+
"step": 110925
|
| 240498 |
+
},
|
| 240499 |
+
{
|
| 240500 |
+
"epoch": 894.4,
|
| 240501 |
+
"learning_rate": 8.224781906300486e-06,
|
| 240502 |
+
"loss": 0.8943,
|
| 240503 |
+
"step": 110930
|
| 240504 |
+
},
|
| 240505 |
+
{
|
| 240506 |
+
"epoch": 894.44,
|
| 240507 |
+
"learning_rate": 8.224701130856221e-06,
|
| 240508 |
+
"loss": 0.6775,
|
| 240509 |
+
"step": 110935
|
| 240510 |
+
},
|
| 240511 |
+
{
|
| 240512 |
+
"epoch": 894.48,
|
| 240513 |
+
"learning_rate": 8.224620355411956e-06,
|
| 240514 |
+
"loss": 0.3509,
|
| 240515 |
+
"step": 110940
|
| 240516 |
+
},
|
| 240517 |
+
{
|
| 240518 |
+
"epoch": 894.52,
|
| 240519 |
+
"learning_rate": 8.224539579967691e-06,
|
| 240520 |
+
"loss": 0.2667,
|
| 240521 |
+
"step": 110945
|
| 240522 |
+
},
|
| 240523 |
+
{
|
| 240524 |
+
"epoch": 894.56,
|
| 240525 |
+
"learning_rate": 8.224458804523426e-06,
|
| 240526 |
+
"loss": 0.3828,
|
| 240527 |
+
"step": 110950
|
| 240528 |
+
},
|
| 240529 |
+
{
|
| 240530 |
+
"epoch": 894.6,
|
| 240531 |
+
"learning_rate": 8.224378029079161e-06,
|
| 240532 |
+
"loss": 0.8308,
|
| 240533 |
+
"step": 110955
|
| 240534 |
+
},
|
| 240535 |
+
{
|
| 240536 |
+
"epoch": 894.64,
|
| 240537 |
+
"learning_rate": 8.224297253634896e-06,
|
| 240538 |
+
"loss": 0.6964,
|
| 240539 |
+
"step": 110960
|
| 240540 |
+
},
|
| 240541 |
+
{
|
| 240542 |
+
"epoch": 894.68,
|
| 240543 |
+
"learning_rate": 8.224216478190631e-06,
|
| 240544 |
+
"loss": 0.3014,
|
| 240545 |
+
"step": 110965
|
| 240546 |
+
},
|
| 240547 |
+
{
|
| 240548 |
+
"epoch": 894.72,
|
| 240549 |
+
"learning_rate": 8.224135702746365e-06,
|
| 240550 |
+
"loss": 0.3124,
|
| 240551 |
+
"step": 110970
|
| 240552 |
+
},
|
| 240553 |
+
{
|
| 240554 |
+
"epoch": 894.76,
|
| 240555 |
+
"learning_rate": 8.224054927302101e-06,
|
| 240556 |
+
"loss": 0.3854,
|
| 240557 |
+
"step": 110975
|
| 240558 |
+
},
|
| 240559 |
+
{
|
| 240560 |
+
"epoch": 894.8,
|
| 240561 |
+
"learning_rate": 8.223974151857837e-06,
|
| 240562 |
+
"loss": 0.9649,
|
| 240563 |
+
"step": 110980
|
| 240564 |
+
},
|
| 240565 |
+
{
|
| 240566 |
+
"epoch": 894.84,
|
| 240567 |
+
"learning_rate": 8.223893376413571e-06,
|
| 240568 |
+
"loss": 0.7675,
|
| 240569 |
+
"step": 110985
|
| 240570 |
+
},
|
| 240571 |
+
{
|
| 240572 |
+
"epoch": 894.88,
|
| 240573 |
+
"learning_rate": 8.223812600969307e-06,
|
| 240574 |
+
"loss": 0.3051,
|
| 240575 |
+
"step": 110990
|
| 240576 |
+
},
|
| 240577 |
+
{
|
| 240578 |
+
"epoch": 894.92,
|
| 240579 |
+
"learning_rate": 8.223731825525041e-06,
|
| 240580 |
+
"loss": 0.2799,
|
| 240581 |
+
"step": 110995
|
| 240582 |
+
},
|
| 240583 |
+
{
|
| 240584 |
+
"epoch": 894.96,
|
| 240585 |
+
"learning_rate": 8.223651050080777e-06,
|
| 240586 |
+
"loss": 0.411,
|
| 240587 |
+
"step": 111000
|
| 240588 |
+
},
|
| 240589 |
+
{
|
| 240590 |
+
"epoch": 895.0,
|
| 240591 |
+
"learning_rate": 8.223570274636511e-06,
|
| 240592 |
+
"loss": 1.2268,
|
| 240593 |
+
"step": 111005
|
| 240594 |
+
},
|
| 240595 |
+
{
|
| 240596 |
+
"epoch": 895.0,
|
| 240597 |
+
"eval_loss": 0.4037691652774811,
|
| 240598 |
+
"eval_runtime": 44.1709,
|
| 240599 |
+
"eval_samples_per_second": 18.949,
|
| 240600 |
+
"eval_steps_per_second": 0.611,
|
| 240601 |
+
"eval_wer": 0.1897884018198888,
|
| 240602 |
+
"step": 111005
|
| 240603 |
+
},
|
| 240604 |
+
{
|
| 240605 |
+
"epoch": 888.04,
|
| 240606 |
+
"learning_rate": 8.223489499192247e-06,
|
| 240607 |
+
"loss": 0.2999,
|
| 240608 |
+
"step": 111010
|
| 240609 |
+
},
|
| 240610 |
+
{
|
| 240611 |
+
"epoch": 888.08,
|
| 240612 |
+
"learning_rate": 8.223408723747981e-06,
|
| 240613 |
+
"loss": 0.3209,
|
| 240614 |
+
"step": 111015
|
| 240615 |
+
},
|
| 240616 |
+
{
|
| 240617 |
+
"epoch": 888.12,
|
| 240618 |
+
"learning_rate": 8.223327948303717e-06,
|
| 240619 |
+
"loss": 0.398,
|
| 240620 |
+
"step": 111020
|
| 240621 |
+
},
|
| 240622 |
+
{
|
| 240623 |
+
"epoch": 888.16,
|
| 240624 |
+
"learning_rate": 8.223247172859451e-06,
|
| 240625 |
+
"loss": 0.4157,
|
| 240626 |
+
"step": 111025
|
| 240627 |
+
},
|
| 240628 |
+
{
|
| 240629 |
+
"epoch": 888.2,
|
| 240630 |
+
"learning_rate": 8.223166397415187e-06,
|
| 240631 |
+
"loss": 1.1022,
|
| 240632 |
+
"step": 111030
|
| 240633 |
+
},
|
| 240634 |
+
{
|
| 240635 |
+
"epoch": 888.24,
|
| 240636 |
+
"learning_rate": 8.223085621970921e-06,
|
| 240637 |
+
"loss": 0.4123,
|
| 240638 |
+
"step": 111035
|
| 240639 |
+
},
|
| 240640 |
+
{
|
| 240641 |
+
"epoch": 888.28,
|
| 240642 |
+
"learning_rate": 8.223004846526657e-06,
|
| 240643 |
+
"loss": 0.2387,
|
| 240644 |
+
"step": 111040
|
| 240645 |
+
},
|
| 240646 |
+
{
|
| 240647 |
+
"epoch": 888.32,
|
| 240648 |
+
"learning_rate": 8.222924071082393e-06,
|
| 240649 |
+
"loss": 0.3189,
|
| 240650 |
+
"step": 111045
|
| 240651 |
+
},
|
| 240652 |
+
{
|
| 240653 |
+
"epoch": 888.36,
|
| 240654 |
+
"learning_rate": 8.222843295638127e-06,
|
| 240655 |
+
"loss": 0.5308,
|
| 240656 |
+
"step": 111050
|
| 240657 |
+
},
|
| 240658 |
+
{
|
| 240659 |
+
"epoch": 888.4,
|
| 240660 |
+
"learning_rate": 8.222762520193863e-06,
|
| 240661 |
+
"loss": 1.1537,
|
| 240662 |
+
"step": 111055
|
| 240663 |
+
},
|
| 240664 |
+
{
|
| 240665 |
+
"epoch": 888.44,
|
| 240666 |
+
"learning_rate": 8.222681744749597e-06,
|
| 240667 |
+
"loss": 0.3102,
|
| 240668 |
+
"step": 111060
|
| 240669 |
+
},
|
| 240670 |
+
{
|
| 240671 |
+
"epoch": 888.48,
|
| 240672 |
+
"learning_rate": 8.222600969305333e-06,
|
| 240673 |
+
"loss": 0.2456,
|
| 240674 |
+
"step": 111065
|
| 240675 |
+
},
|
| 240676 |
+
{
|
| 240677 |
+
"epoch": 888.52,
|
| 240678 |
+
"learning_rate": 8.222520193861067e-06,
|
| 240679 |
+
"loss": 0.2704,
|
| 240680 |
+
"step": 111070
|
| 240681 |
+
},
|
| 240682 |
+
{
|
| 240683 |
+
"epoch": 888.56,
|
| 240684 |
+
"learning_rate": 8.222439418416803e-06,
|
| 240685 |
+
"loss": 0.3987,
|
| 240686 |
+
"step": 111075
|
| 240687 |
+
},
|
| 240688 |
+
{
|
| 240689 |
+
"epoch": 888.6,
|
| 240690 |
+
"learning_rate": 8.222358642972537e-06,
|
| 240691 |
+
"loss": 1.1706,
|
| 240692 |
+
"step": 111080
|
| 240693 |
+
},
|
| 240694 |
+
{
|
| 240695 |
+
"epoch": 888.64,
|
| 240696 |
+
"learning_rate": 8.222277867528273e-06,
|
| 240697 |
+
"loss": 0.3518,
|
| 240698 |
+
"step": 111085
|
| 240699 |
+
},
|
| 240700 |
+
{
|
| 240701 |
+
"epoch": 888.68,
|
| 240702 |
+
"learning_rate": 8.222197092084007e-06,
|
| 240703 |
+
"loss": 0.2963,
|
| 240704 |
+
"step": 111090
|
| 240705 |
+
},
|
| 240706 |
+
{
|
| 240707 |
+
"epoch": 888.72,
|
| 240708 |
+
"learning_rate": 8.222116316639743e-06,
|
| 240709 |
+
"loss": 0.3069,
|
| 240710 |
+
"step": 111095
|
| 240711 |
+
},
|
| 240712 |
+
{
|
| 240713 |
+
"epoch": 888.76,
|
| 240714 |
+
"learning_rate": 8.222035541195479e-06,
|
| 240715 |
+
"loss": 0.5053,
|
| 240716 |
+
"step": 111100
|
| 240717 |
+
},
|
| 240718 |
+
{
|
| 240719 |
+
"epoch": 888.8,
|
| 240720 |
+
"learning_rate": 8.221954765751213e-06,
|
| 240721 |
+
"loss": 1.1867,
|
| 240722 |
+
"step": 111105
|
| 240723 |
+
},
|
| 240724 |
+
{
|
| 240725 |
+
"epoch": 888.84,
|
| 240726 |
+
"learning_rate": 8.221873990306949e-06,
|
| 240727 |
+
"loss": 0.3344,
|
| 240728 |
+
"step": 111110
|
| 240729 |
+
},
|
| 240730 |
+
{
|
| 240731 |
+
"epoch": 888.88,
|
| 240732 |
+
"learning_rate": 8.221793214862683e-06,
|
| 240733 |
+
"loss": 0.2685,
|
| 240734 |
+
"step": 111115
|
| 240735 |
+
},
|
| 240736 |
+
{
|
| 240737 |
+
"epoch": 888.92,
|
| 240738 |
+
"learning_rate": 8.221712439418419e-06,
|
| 240739 |
+
"loss": 0.346,
|
| 240740 |
+
"step": 111120
|
| 240741 |
+
},
|
| 240742 |
+
{
|
| 240743 |
+
"epoch": 888.96,
|
| 240744 |
+
"learning_rate": 8.221631663974153e-06,
|
| 240745 |
+
"loss": 0.413,
|
| 240746 |
+
"step": 111125
|
| 240747 |
+
},
|
| 240748 |
+
{
|
| 240749 |
+
"epoch": 889.0,
|
| 240750 |
+
"learning_rate": 8.221550888529888e-06,
|
| 240751 |
+
"loss": 1.43,
|
| 240752 |
+
"step": 111130
|
| 240753 |
+
},
|
| 240754 |
+
{
|
| 240755 |
+
"epoch": 889.0,
|
| 240756 |
+
"eval_loss": 0.4563208818435669,
|
| 240757 |
+
"eval_runtime": 45.6982,
|
| 240758 |
+
"eval_samples_per_second": 18.316,
|
| 240759 |
+
"eval_steps_per_second": 0.591,
|
| 240760 |
+
"eval_wer": 0.17906810035842294,
|
| 240761 |
+
"step": 111130
|
| 240762 |
+
},
|
| 240763 |
+
{
|
| 240764 |
+
"epoch": 889.04,
|
| 240765 |
+
"learning_rate": 8.221470113085623e-06,
|
| 240766 |
+
"loss": 0.3338,
|
| 240767 |
+
"step": 111135
|
| 240768 |
+
},
|
| 240769 |
+
{
|
| 240770 |
+
"epoch": 889.08,
|
| 240771 |
+
"learning_rate": 8.221389337641358e-06,
|
| 240772 |
+
"loss": 0.2645,
|
| 240773 |
+
"step": 111140
|
| 240774 |
+
},
|
| 240775 |
+
{
|
| 240776 |
+
"epoch": 889.12,
|
| 240777 |
+
"learning_rate": 8.221308562197093e-06,
|
| 240778 |
+
"loss": 0.3281,
|
| 240779 |
+
"step": 111145
|
| 240780 |
+
},
|
| 240781 |
+
{
|
| 240782 |
+
"epoch": 889.16,
|
| 240783 |
+
"learning_rate": 8.221227786752828e-06,
|
| 240784 |
+
"loss": 0.4546,
|
| 240785 |
+
"step": 111150
|
| 240786 |
+
},
|
| 240787 |
+
{
|
| 240788 |
+
"epoch": 889.2,
|
| 240789 |
+
"learning_rate": 8.221147011308563e-06,
|
| 240790 |
+
"loss": 1.1137,
|
| 240791 |
+
"step": 111155
|
| 240792 |
+
},
|
| 240793 |
+
{
|
| 240794 |
+
"epoch": 889.24,
|
| 240795 |
+
"learning_rate": 8.221066235864298e-06,
|
| 240796 |
+
"loss": 0.3458,
|
| 240797 |
+
"step": 111160
|
| 240798 |
+
},
|
| 240799 |
+
{
|
| 240800 |
+
"epoch": 889.28,
|
| 240801 |
+
"learning_rate": 8.220985460420034e-06,
|
| 240802 |
+
"loss": 0.2908,
|
| 240803 |
+
"step": 111165
|
| 240804 |
+
},
|
| 240805 |
+
{
|
| 240806 |
+
"epoch": 889.32,
|
| 240807 |
+
"learning_rate": 8.220904684975768e-06,
|
| 240808 |
+
"loss": 0.3467,
|
| 240809 |
+
"step": 111170
|
| 240810 |
+
},
|
| 240811 |
+
{
|
| 240812 |
+
"epoch": 889.36,
|
| 240813 |
+
"learning_rate": 8.220823909531504e-06,
|
| 240814 |
+
"loss": 0.4733,
|
| 240815 |
+
"step": 111175
|
| 240816 |
+
},
|
| 240817 |
+
{
|
| 240818 |
+
"epoch": 889.4,
|
| 240819 |
+
"learning_rate": 8.220743134087238e-06,
|
| 240820 |
+
"loss": 1.119,
|
| 240821 |
+
"step": 111180
|
| 240822 |
+
},
|
| 240823 |
+
{
|
| 240824 |
+
"epoch": 889.44,
|
| 240825 |
+
"learning_rate": 8.220662358642974e-06,
|
| 240826 |
+
"loss": 0.3293,
|
| 240827 |
+
"step": 111185
|
| 240828 |
+
},
|
| 240829 |
+
{
|
| 240830 |
+
"epoch": 889.48,
|
| 240831 |
+
"learning_rate": 8.220581583198708e-06,
|
| 240832 |
+
"loss": 0.2844,
|
| 240833 |
+
"step": 111190
|
| 240834 |
+
},
|
| 240835 |
+
{
|
| 240836 |
+
"epoch": 889.52,
|
| 240837 |
+
"learning_rate": 8.220500807754444e-06,
|
| 240838 |
+
"loss": 0.294,
|
| 240839 |
+
"step": 111195
|
| 240840 |
+
},
|
| 240841 |
+
{
|
| 240842 |
+
"epoch": 889.56,
|
| 240843 |
+
"learning_rate": 8.220420032310178e-06,
|
| 240844 |
+
"loss": 0.4952,
|
| 240845 |
+
"step": 111200
|
| 240846 |
+
},
|
| 240847 |
+
{
|
| 240848 |
+
"epoch": 889.6,
|
| 240849 |
+
"learning_rate": 8.220339256865914e-06,
|
| 240850 |
+
"loss": 1.2199,
|
| 240851 |
+
"step": 111205
|
| 240852 |
+
},
|
| 240853 |
+
{
|
| 240854 |
+
"epoch": 889.64,
|
| 240855 |
+
"learning_rate": 8.220258481421648e-06,
|
| 240856 |
+
"loss": 0.329,
|
| 240857 |
+
"step": 111210
|
| 240858 |
+
},
|
| 240859 |
+
{
|
| 240860 |
+
"epoch": 889.68,
|
| 240861 |
+
"learning_rate": 8.220177705977384e-06,
|
| 240862 |
+
"loss": 0.3095,
|
| 240863 |
+
"step": 111215
|
| 240864 |
+
},
|
| 240865 |
+
{
|
| 240866 |
+
"epoch": 889.72,
|
| 240867 |
+
"learning_rate": 8.220096930533118e-06,
|
| 240868 |
+
"loss": 0.3582,
|
| 240869 |
+
"step": 111220
|
| 240870 |
+
},
|
| 240871 |
+
{
|
| 240872 |
+
"epoch": 889.76,
|
| 240873 |
+
"learning_rate": 8.220016155088854e-06,
|
| 240874 |
+
"loss": 0.4658,
|
| 240875 |
+
"step": 111225
|
| 240876 |
+
},
|
| 240877 |
+
{
|
| 240878 |
+
"epoch": 889.8,
|
| 240879 |
+
"learning_rate": 8.21993537964459e-06,
|
| 240880 |
+
"loss": 1.2943,
|
| 240881 |
+
"step": 111230
|
| 240882 |
+
},
|
| 240883 |
+
{
|
| 240884 |
+
"epoch": 889.84,
|
| 240885 |
+
"learning_rate": 8.219854604200324e-06,
|
| 240886 |
+
"loss": 0.3334,
|
| 240887 |
+
"step": 111235
|
| 240888 |
+
},
|
| 240889 |
+
{
|
| 240890 |
+
"epoch": 889.88,
|
| 240891 |
+
"learning_rate": 8.21977382875606e-06,
|
| 240892 |
+
"loss": 0.2945,
|
| 240893 |
+
"step": 111240
|
| 240894 |
+
},
|
| 240895 |
+
{
|
| 240896 |
+
"epoch": 889.92,
|
| 240897 |
+
"learning_rate": 8.219693053311794e-06,
|
| 240898 |
+
"loss": 0.2881,
|
| 240899 |
+
"step": 111245
|
| 240900 |
+
},
|
| 240901 |
+
{
|
| 240902 |
+
"epoch": 889.96,
|
| 240903 |
+
"learning_rate": 8.21961227786753e-06,
|
| 240904 |
+
"loss": 0.4068,
|
| 240905 |
+
"step": 111250
|
| 240906 |
+
},
|
| 240907 |
+
{
|
| 240908 |
+
"epoch": 890.0,
|
| 240909 |
+
"learning_rate": 8.219531502423264e-06,
|
| 240910 |
+
"loss": 1.2953,
|
| 240911 |
+
"step": 111255
|
| 240912 |
+
},
|
| 240913 |
+
{
|
| 240914 |
+
"epoch": 890.0,
|
| 240915 |
+
"eval_loss": 0.39579546451568604,
|
| 240916 |
+
"eval_runtime": 41.6212,
|
| 240917 |
+
"eval_samples_per_second": 20.11,
|
| 240918 |
+
"eval_steps_per_second": 0.649,
|
| 240919 |
+
"eval_wer": 0.18293677266325697,
|
| 240920 |
+
"step": 111255
|
| 240921 |
+
},
|
| 240922 |
+
{
|
| 240923 |
+
"epoch": 897.04,
|
| 240924 |
+
"learning_rate": 8.219450726979e-06,
|
| 240925 |
+
"loss": 0.3568,
|
| 240926 |
+
"step": 111260
|
| 240927 |
+
},
|
| 240928 |
+
{
|
| 240929 |
+
"epoch": 897.08,
|
| 240930 |
+
"learning_rate": 8.219369951534734e-06,
|
| 240931 |
+
"loss": 0.2539,
|
| 240932 |
+
"step": 111265
|
| 240933 |
+
},
|
| 240934 |
+
{
|
| 240935 |
+
"epoch": 897.12,
|
| 240936 |
+
"learning_rate": 8.21928917609047e-06,
|
| 240937 |
+
"loss": 0.3165,
|
| 240938 |
+
"step": 111270
|
| 240939 |
+
},
|
| 240940 |
+
{
|
| 240941 |
+
"epoch": 897.16,
|
| 240942 |
+
"learning_rate": 8.219208400646204e-06,
|
| 240943 |
+
"loss": 0.4533,
|
| 240944 |
+
"step": 111275
|
| 240945 |
+
},
|
| 240946 |
+
{
|
| 240947 |
+
"epoch": 897.2,
|
| 240948 |
+
"learning_rate": 8.21912762520194e-06,
|
| 240949 |
+
"loss": 1.0925,
|
| 240950 |
+
"step": 111280
|
| 240951 |
+
},
|
| 240952 |
+
{
|
| 240953 |
+
"epoch": 897.24,
|
| 240954 |
+
"learning_rate": 8.219046849757674e-06,
|
| 240955 |
+
"loss": 0.3357,
|
| 240956 |
+
"step": 111285
|
| 240957 |
+
},
|
| 240958 |
+
{
|
| 240959 |
+
"epoch": 897.28,
|
| 240960 |
+
"learning_rate": 8.21896607431341e-06,
|
| 240961 |
+
"loss": 0.2692,
|
| 240962 |
+
"step": 111290
|
| 240963 |
+
},
|
| 240964 |
+
{
|
| 240965 |
+
"epoch": 897.32,
|
| 240966 |
+
"learning_rate": 8.218885298869146e-06,
|
| 240967 |
+
"loss": 0.3194,
|
| 240968 |
+
"step": 111295
|
| 240969 |
+
},
|
| 240970 |
+
{
|
| 240971 |
+
"epoch": 897.36,
|
| 240972 |
+
"learning_rate": 8.21880452342488e-06,
|
| 240973 |
+
"loss": 0.555,
|
| 240974 |
+
"step": 111300
|
| 240975 |
+
},
|
| 240976 |
+
{
|
| 240977 |
+
"epoch": 897.4,
|
| 240978 |
+
"learning_rate": 8.218723747980616e-06,
|
| 240979 |
+
"loss": 1.1612,
|
| 240980 |
+
"step": 111305
|
| 240981 |
+
},
|
| 240982 |
+
{
|
| 240983 |
+
"epoch": 897.44,
|
| 240984 |
+
"learning_rate": 8.21864297253635e-06,
|
| 240985 |
+
"loss": 0.3347,
|
| 240986 |
+
"step": 111310
|
| 240987 |
+
},
|
| 240988 |
+
{
|
| 240989 |
+
"epoch": 897.48,
|
| 240990 |
+
"learning_rate": 8.218562197092086e-06,
|
| 240991 |
+
"loss": 0.2972,
|
| 240992 |
+
"step": 111315
|
| 240993 |
+
},
|
| 240994 |
+
{
|
| 240995 |
+
"epoch": 897.52,
|
| 240996 |
+
"learning_rate": 8.21848142164782e-06,
|
| 240997 |
+
"loss": 0.3082,
|
| 240998 |
+
"step": 111320
|
| 240999 |
+
},
|
| 241000 |
+
{
|
| 241001 |
+
"epoch": 897.56,
|
| 241002 |
+
"learning_rate": 8.218400646203556e-06,
|
| 241003 |
+
"loss": 0.4182,
|
| 241004 |
+
"step": 111325
|
| 241005 |
+
},
|
| 241006 |
+
{
|
| 241007 |
+
"epoch": 897.6,
|
| 241008 |
+
"learning_rate": 8.21831987075929e-06,
|
| 241009 |
+
"loss": 1.192,
|
| 241010 |
+
"step": 111330
|
| 241011 |
+
},
|
| 241012 |
+
{
|
| 241013 |
+
"epoch": 897.65,
|
| 241014 |
+
"learning_rate": 8.218239095315026e-06,
|
| 241015 |
+
"loss": 0.3286,
|
| 241016 |
+
"step": 111335
|
| 241017 |
+
},
|
| 241018 |
+
{
|
| 241019 |
+
"epoch": 897.69,
|
| 241020 |
+
"learning_rate": 8.21815831987076e-06,
|
| 241021 |
+
"loss": 0.2918,
|
| 241022 |
+
"step": 111340
|
| 241023 |
+
},
|
| 241024 |
+
{
|
| 241025 |
+
"epoch": 897.73,
|
| 241026 |
+
"learning_rate": 8.218077544426495e-06,
|
| 241027 |
+
"loss": 0.3403,
|
| 241028 |
+
"step": 111345
|
| 241029 |
+
},
|
| 241030 |
+
{
|
| 241031 |
+
"epoch": 897.77,
|
| 241032 |
+
"learning_rate": 8.21799676898223e-06,
|
| 241033 |
+
"loss": 0.3825,
|
| 241034 |
+
"step": 111350
|
| 241035 |
+
},
|
| 241036 |
+
{
|
| 241037 |
+
"epoch": 897.81,
|
| 241038 |
+
"learning_rate": 8.217915993537965e-06,
|
| 241039 |
+
"loss": 1.0865,
|
| 241040 |
+
"step": 111355
|
| 241041 |
+
},
|
| 241042 |
+
{
|
| 241043 |
+
"epoch": 897.85,
|
| 241044 |
+
"learning_rate": 8.2178352180937e-06,
|
| 241045 |
+
"loss": 0.3358,
|
| 241046 |
+
"step": 111360
|
| 241047 |
+
},
|
| 241048 |
+
{
|
| 241049 |
+
"epoch": 897.89,
|
| 241050 |
+
"learning_rate": 8.217754442649435e-06,
|
| 241051 |
+
"loss": 0.3336,
|
| 241052 |
+
"step": 111365
|
| 241053 |
+
},
|
| 241054 |
+
{
|
| 241055 |
+
"epoch": 897.93,
|
| 241056 |
+
"learning_rate": 8.217673667205171e-06,
|
| 241057 |
+
"loss": 0.3484,
|
| 241058 |
+
"step": 111370
|
| 241059 |
+
},
|
| 241060 |
+
{
|
| 241061 |
+
"epoch": 897.97,
|
| 241062 |
+
"learning_rate": 8.217592891760905e-06,
|
| 241063 |
+
"loss": 0.4983,
|
| 241064 |
+
"step": 111375
|
| 241065 |
+
},
|
| 241066 |
+
{
|
| 241067 |
+
"epoch": 898.0,
|
| 241068 |
+
"eval_loss": 0.3549356162548065,
|
| 241069 |
+
"eval_runtime": 42.074,
|
| 241070 |
+
"eval_samples_per_second": 19.894,
|
| 241071 |
+
"eval_steps_per_second": 0.642,
|
| 241072 |
+
"eval_wer": 0.18447376051029285,
|
| 241073 |
+
"step": 111379
|
| 241074 |
}
|
| 241075 |
],
|
| 241076 |
"max_steps": 620000,
|
| 241077 |
"num_train_epochs": 5000,
|
| 241078 |
+
"total_flos": 3.134467323546747e+20,
|
| 241079 |
"trial_name": null,
|
| 241080 |
"trial_params": null
|
| 241081 |
}
|
model-bin/finetune/base/{checkpoint-110758 → checkpoint-111379}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630045620.5655968/events.out.tfevents.1630045620.52f5c7e305a3.886.71
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:789077e8a431041de909256226a12d9fd795a843e189968fcbd73a7d940c1bb0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630046118.6803894/events.out.tfevents.1630046118.52f5c7e305a3.886.73
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2d4196b10fc377834886dc0d987e6817a8ea8f93225248f6d0c19c43600641b2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630046681.8092928/events.out.tfevents.1630046681.52f5c7e305a3.886.75
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:abdd13375aec82b2ef9f339c6eb574e37141831c3243c001d07c32e4ab0cf46f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630047172.2078328/events.out.tfevents.1630047172.52f5c7e305a3.886.77
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:232390e2831e3b474ff01088741dca913da5c1a45a34e06d55f1e7af7f05c211
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630047639.1210005/events.out.tfevents.1630047639.52f5c7e305a3.886.79
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:63970938dab9c44f0940f5ad235a4c8e66651ae8540d2f66744893ea940f0ac2
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630045620.52f5c7e305a3.886.70
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b8314966b5c7b8b998ed76cc6f32d38a8a2fd310b0f566a0ba90b2151c0e7b1f
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630046118.52f5c7e305a3.886.72
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f2fca3e324b891d951182c535e721c6721fa0e5b2649c8e079560f44cf322a59
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630046681.52f5c7e305a3.886.74
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f79dfe9c06a70a4a2dc57702c07e9552251702809224aa774e01719c63bd929a
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630047172.52f5c7e305a3.886.76
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:86cc4daafbb29ca0d18efe9c264f9c5feb271b6fcd28a84322fbf266ffb68542
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630047639.52f5c7e305a3.886.78
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2f6abc8f2020bcd704cade369621eba3e292108e363a1184c6f8b5a5f47ca085
|
| 3 |
+
size 8462
|