"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629980056.6178455/events.out.tfevents.1629980056.8e89bd551565.924.181 +3 -0
- model-bin/finetune/base/log/1629980506.0580995/events.out.tfevents.1629980506.8e89bd551565.924.183 +3 -0
- model-bin/finetune/base/log/1629980941.2291443/events.out.tfevents.1629980941.8e89bd551565.924.185 +3 -0
- model-bin/finetune/base/log/1629981376.296084/events.out.tfevents.1629981376.8e89bd551565.924.187 +3 -0
- model-bin/finetune/base/log/1629981805.4906385/events.out.tfevents.1629981805.8e89bd551565.924.189 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629980056.8e89bd551565.924.180 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629980506.8e89bd551565.924.182 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629980940.8e89bd551565.924.184 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629981375.8e89bd551565.924.186 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629981805.8e89bd551565.924.188 +3 -0
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/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:c5ca2134f4b3ba0da8edcb8e6cd1388b203548ff4a96352c981561cd13fb011f
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/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 sha256:0466fb5a9f526bdd62d1df1d0510a451ed10596dd03df9cda2a39d5cc3c2d193
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/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 sha256:93aa4f1f1b23c8a0be7433fdd4a15ed2e3afdb6362c025c63304d955b95cc4e1
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/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 sha256:0704821450ad2e472fe6ec90dc8448a1296ca59e4ae62277041ac58c687cc511
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/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 sha256:6d929aa920368a347fd351edf6478d569d94686bcf1b97c344d99188ee6c8acb
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -226038,11 +226038,800 @@
|
|
| 226038 |
"eval_steps_per_second": 0.688,
|
| 226039 |
"eval_wer": 0.17951169289601412,
|
| 226040 |
"step": 99560
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 226041 |
}
|
| 226042 |
],
|
| 226043 |
"max_steps": 620000,
|
| 226044 |
"num_train_epochs": 5000,
|
| 226045 |
-
"total_flos": 2.
|
| 226046 |
"trial_name": null,
|
| 226047 |
"trial_params": null
|
| 226048 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17565318086415285,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-94333",
|
| 4 |
+
"epoch": 807.0,
|
| 5 |
+
"global_step": 100181,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 226038 |
"eval_steps_per_second": 0.688,
|
| 226039 |
"eval_wer": 0.17951169289601412,
|
| 226040 |
"step": 99560
|
| 226041 |
+
},
|
| 226042 |
+
{
|
| 226043 |
+
"epoch": 802.04,
|
| 226044 |
+
"learning_rate": 8.421073717948718e-06,
|
| 226045 |
+
"loss": 0.3382,
|
| 226046 |
+
"step": 99565
|
| 226047 |
+
},
|
| 226048 |
+
{
|
| 226049 |
+
"epoch": 802.08,
|
| 226050 |
+
"learning_rate": 8.42099358974359e-06,
|
| 226051 |
+
"loss": 0.286,
|
| 226052 |
+
"step": 99570
|
| 226053 |
+
},
|
| 226054 |
+
{
|
| 226055 |
+
"epoch": 802.12,
|
| 226056 |
+
"learning_rate": 8.420913461538463e-06,
|
| 226057 |
+
"loss": 0.2989,
|
| 226058 |
+
"step": 99575
|
| 226059 |
+
},
|
| 226060 |
+
{
|
| 226061 |
+
"epoch": 802.16,
|
| 226062 |
+
"learning_rate": 8.420833333333334e-06,
|
| 226063 |
+
"loss": 0.4523,
|
| 226064 |
+
"step": 99580
|
| 226065 |
+
},
|
| 226066 |
+
{
|
| 226067 |
+
"epoch": 802.2,
|
| 226068 |
+
"learning_rate": 8.420753205128206e-06,
|
| 226069 |
+
"loss": 1.4162,
|
| 226070 |
+
"step": 99585
|
| 226071 |
+
},
|
| 226072 |
+
{
|
| 226073 |
+
"epoch": 802.24,
|
| 226074 |
+
"learning_rate": 8.420673076923077e-06,
|
| 226075 |
+
"loss": 0.33,
|
| 226076 |
+
"step": 99590
|
| 226077 |
+
},
|
| 226078 |
+
{
|
| 226079 |
+
"epoch": 802.28,
|
| 226080 |
+
"learning_rate": 8.42059294871795e-06,
|
| 226081 |
+
"loss": 0.3307,
|
| 226082 |
+
"step": 99595
|
| 226083 |
+
},
|
| 226084 |
+
{
|
| 226085 |
+
"epoch": 802.32,
|
| 226086 |
+
"learning_rate": 8.420512820512821e-06,
|
| 226087 |
+
"loss": 0.3105,
|
| 226088 |
+
"step": 99600
|
| 226089 |
+
},
|
| 226090 |
+
{
|
| 226091 |
+
"epoch": 802.36,
|
| 226092 |
+
"learning_rate": 8.420432692307693e-06,
|
| 226093 |
+
"loss": 0.4323,
|
| 226094 |
+
"step": 99605
|
| 226095 |
+
},
|
| 226096 |
+
{
|
| 226097 |
+
"epoch": 802.4,
|
| 226098 |
+
"learning_rate": 8.420352564102566e-06,
|
| 226099 |
+
"loss": 1.2248,
|
| 226100 |
+
"step": 99610
|
| 226101 |
+
},
|
| 226102 |
+
{
|
| 226103 |
+
"epoch": 802.44,
|
| 226104 |
+
"learning_rate": 8.420272435897437e-06,
|
| 226105 |
+
"loss": 0.3468,
|
| 226106 |
+
"step": 99615
|
| 226107 |
+
},
|
| 226108 |
+
{
|
| 226109 |
+
"epoch": 802.48,
|
| 226110 |
+
"learning_rate": 8.420192307692308e-06,
|
| 226111 |
+
"loss": 0.3016,
|
| 226112 |
+
"step": 99620
|
| 226113 |
+
},
|
| 226114 |
+
{
|
| 226115 |
+
"epoch": 802.52,
|
| 226116 |
+
"learning_rate": 8.42011217948718e-06,
|
| 226117 |
+
"loss": 0.3349,
|
| 226118 |
+
"step": 99625
|
| 226119 |
+
},
|
| 226120 |
+
{
|
| 226121 |
+
"epoch": 802.56,
|
| 226122 |
+
"learning_rate": 8.420032051282053e-06,
|
| 226123 |
+
"loss": 0.4675,
|
| 226124 |
+
"step": 99630
|
| 226125 |
+
},
|
| 226126 |
+
{
|
| 226127 |
+
"epoch": 802.6,
|
| 226128 |
+
"learning_rate": 8.419951923076922e-06,
|
| 226129 |
+
"loss": 1.047,
|
| 226130 |
+
"step": 99635
|
| 226131 |
+
},
|
| 226132 |
+
{
|
| 226133 |
+
"epoch": 802.64,
|
| 226134 |
+
"learning_rate": 8.419871794871796e-06,
|
| 226135 |
+
"loss": 0.3183,
|
| 226136 |
+
"step": 99640
|
| 226137 |
+
},
|
| 226138 |
+
{
|
| 226139 |
+
"epoch": 802.68,
|
| 226140 |
+
"learning_rate": 8.419791666666667e-06,
|
| 226141 |
+
"loss": 0.2828,
|
| 226142 |
+
"step": 99645
|
| 226143 |
+
},
|
| 226144 |
+
{
|
| 226145 |
+
"epoch": 802.72,
|
| 226146 |
+
"learning_rate": 8.419711538461538e-06,
|
| 226147 |
+
"loss": 0.327,
|
| 226148 |
+
"step": 99650
|
| 226149 |
+
},
|
| 226150 |
+
{
|
| 226151 |
+
"epoch": 802.76,
|
| 226152 |
+
"learning_rate": 8.419631410256411e-06,
|
| 226153 |
+
"loss": 0.473,
|
| 226154 |
+
"step": 99655
|
| 226155 |
+
},
|
| 226156 |
+
{
|
| 226157 |
+
"epoch": 802.8,
|
| 226158 |
+
"learning_rate": 8.419551282051283e-06,
|
| 226159 |
+
"loss": 1.1045,
|
| 226160 |
+
"step": 99660
|
| 226161 |
+
},
|
| 226162 |
+
{
|
| 226163 |
+
"epoch": 802.84,
|
| 226164 |
+
"learning_rate": 8.419471153846154e-06,
|
| 226165 |
+
"loss": 0.3454,
|
| 226166 |
+
"step": 99665
|
| 226167 |
+
},
|
| 226168 |
+
{
|
| 226169 |
+
"epoch": 802.88,
|
| 226170 |
+
"learning_rate": 8.419391025641025e-06,
|
| 226171 |
+
"loss": 0.2748,
|
| 226172 |
+
"step": 99670
|
| 226173 |
+
},
|
| 226174 |
+
{
|
| 226175 |
+
"epoch": 802.92,
|
| 226176 |
+
"learning_rate": 8.419310897435898e-06,
|
| 226177 |
+
"loss": 0.5234,
|
| 226178 |
+
"step": 99675
|
| 226179 |
+
},
|
| 226180 |
+
{
|
| 226181 |
+
"epoch": 802.96,
|
| 226182 |
+
"learning_rate": 8.41923076923077e-06,
|
| 226183 |
+
"loss": 0.477,
|
| 226184 |
+
"step": 99680
|
| 226185 |
+
},
|
| 226186 |
+
{
|
| 226187 |
+
"epoch": 803.0,
|
| 226188 |
+
"eval_loss": 0.47172078490257263,
|
| 226189 |
+
"eval_runtime": 40.19,
|
| 226190 |
+
"eval_samples_per_second": 20.727,
|
| 226191 |
+
"eval_steps_per_second": 0.672,
|
| 226192 |
+
"eval_wer": 0.18185773330432892,
|
| 226193 |
+
"step": 99684
|
| 226194 |
+
},
|
| 226195 |
+
{
|
| 226196 |
+
"epoch": 797.01,
|
| 226197 |
+
"learning_rate": 8.419150641025641e-06,
|
| 226198 |
+
"loss": 0.4009,
|
| 226199 |
+
"step": 99685
|
| 226200 |
+
},
|
| 226201 |
+
{
|
| 226202 |
+
"epoch": 797.05,
|
| 226203 |
+
"learning_rate": 8.419070512820513e-06,
|
| 226204 |
+
"loss": 0.2643,
|
| 226205 |
+
"step": 99690
|
| 226206 |
+
},
|
| 226207 |
+
{
|
| 226208 |
+
"epoch": 797.09,
|
| 226209 |
+
"learning_rate": 8.418990384615386e-06,
|
| 226210 |
+
"loss": 0.4993,
|
| 226211 |
+
"step": 99695
|
| 226212 |
+
},
|
| 226213 |
+
{
|
| 226214 |
+
"epoch": 797.13,
|
| 226215 |
+
"learning_rate": 8.418910256410257e-06,
|
| 226216 |
+
"loss": 0.4462,
|
| 226217 |
+
"step": 99700
|
| 226218 |
+
},
|
| 226219 |
+
{
|
| 226220 |
+
"epoch": 797.17,
|
| 226221 |
+
"learning_rate": 8.418830128205128e-06,
|
| 226222 |
+
"loss": 0.5344,
|
| 226223 |
+
"step": 99705
|
| 226224 |
+
},
|
| 226225 |
+
{
|
| 226226 |
+
"epoch": 797.21,
|
| 226227 |
+
"learning_rate": 8.418750000000001e-06,
|
| 226228 |
+
"loss": 1.177,
|
| 226229 |
+
"step": 99710
|
| 226230 |
+
},
|
| 226231 |
+
{
|
| 226232 |
+
"epoch": 797.25,
|
| 226233 |
+
"learning_rate": 8.418669871794873e-06,
|
| 226234 |
+
"loss": 0.3522,
|
| 226235 |
+
"step": 99715
|
| 226236 |
+
},
|
| 226237 |
+
{
|
| 226238 |
+
"epoch": 797.29,
|
| 226239 |
+
"learning_rate": 8.418589743589744e-06,
|
| 226240 |
+
"loss": 0.2821,
|
| 226241 |
+
"step": 99720
|
| 226242 |
+
},
|
| 226243 |
+
{
|
| 226244 |
+
"epoch": 797.33,
|
| 226245 |
+
"learning_rate": 8.418509615384615e-06,
|
| 226246 |
+
"loss": 0.2643,
|
| 226247 |
+
"step": 99725
|
| 226248 |
+
},
|
| 226249 |
+
{
|
| 226250 |
+
"epoch": 797.37,
|
| 226251 |
+
"learning_rate": 8.418429487179488e-06,
|
| 226252 |
+
"loss": 0.5044,
|
| 226253 |
+
"step": 99730
|
| 226254 |
+
},
|
| 226255 |
+
{
|
| 226256 |
+
"epoch": 797.41,
|
| 226257 |
+
"learning_rate": 8.41834935897436e-06,
|
| 226258 |
+
"loss": 1.3069,
|
| 226259 |
+
"step": 99735
|
| 226260 |
+
},
|
| 226261 |
+
{
|
| 226262 |
+
"epoch": 797.45,
|
| 226263 |
+
"learning_rate": 8.418269230769231e-06,
|
| 226264 |
+
"loss": 0.3163,
|
| 226265 |
+
"step": 99740
|
| 226266 |
+
},
|
| 226267 |
+
{
|
| 226268 |
+
"epoch": 797.49,
|
| 226269 |
+
"learning_rate": 8.418189102564103e-06,
|
| 226270 |
+
"loss": 0.2883,
|
| 226271 |
+
"step": 99745
|
| 226272 |
+
},
|
| 226273 |
+
{
|
| 226274 |
+
"epoch": 797.53,
|
| 226275 |
+
"learning_rate": 8.418108974358976e-06,
|
| 226276 |
+
"loss": 0.3393,
|
| 226277 |
+
"step": 99750
|
| 226278 |
+
},
|
| 226279 |
+
{
|
| 226280 |
+
"epoch": 797.57,
|
| 226281 |
+
"learning_rate": 8.418028846153847e-06,
|
| 226282 |
+
"loss": 0.5629,
|
| 226283 |
+
"step": 99755
|
| 226284 |
+
},
|
| 226285 |
+
{
|
| 226286 |
+
"epoch": 797.61,
|
| 226287 |
+
"learning_rate": 8.417948717948718e-06,
|
| 226288 |
+
"loss": 1.0497,
|
| 226289 |
+
"step": 99760
|
| 226290 |
+
},
|
| 226291 |
+
{
|
| 226292 |
+
"epoch": 797.65,
|
| 226293 |
+
"learning_rate": 8.417868589743591e-06,
|
| 226294 |
+
"loss": 0.3956,
|
| 226295 |
+
"step": 99765
|
| 226296 |
+
},
|
| 226297 |
+
{
|
| 226298 |
+
"epoch": 797.69,
|
| 226299 |
+
"learning_rate": 8.417788461538463e-06,
|
| 226300 |
+
"loss": 0.2644,
|
| 226301 |
+
"step": 99770
|
| 226302 |
+
},
|
| 226303 |
+
{
|
| 226304 |
+
"epoch": 797.73,
|
| 226305 |
+
"learning_rate": 8.417708333333334e-06,
|
| 226306 |
+
"loss": 0.3281,
|
| 226307 |
+
"step": 99775
|
| 226308 |
+
},
|
| 226309 |
+
{
|
| 226310 |
+
"epoch": 797.76,
|
| 226311 |
+
"learning_rate": 8.417628205128205e-06,
|
| 226312 |
+
"loss": 0.5832,
|
| 226313 |
+
"step": 99780
|
| 226314 |
+
},
|
| 226315 |
+
{
|
| 226316 |
+
"epoch": 797.8,
|
| 226317 |
+
"learning_rate": 8.417548076923079e-06,
|
| 226318 |
+
"loss": 1.1425,
|
| 226319 |
+
"step": 99785
|
| 226320 |
+
},
|
| 226321 |
+
{
|
| 226322 |
+
"epoch": 797.84,
|
| 226323 |
+
"learning_rate": 8.417467948717948e-06,
|
| 226324 |
+
"loss": 0.2996,
|
| 226325 |
+
"step": 99790
|
| 226326 |
+
},
|
| 226327 |
+
{
|
| 226328 |
+
"epoch": 797.88,
|
| 226329 |
+
"learning_rate": 8.417387820512821e-06,
|
| 226330 |
+
"loss": 0.3637,
|
| 226331 |
+
"step": 99795
|
| 226332 |
+
},
|
| 226333 |
+
{
|
| 226334 |
+
"epoch": 797.92,
|
| 226335 |
+
"learning_rate": 8.417307692307694e-06,
|
| 226336 |
+
"loss": 0.4234,
|
| 226337 |
+
"step": 99800
|
| 226338 |
+
},
|
| 226339 |
+
{
|
| 226340 |
+
"epoch": 797.96,
|
| 226341 |
+
"learning_rate": 8.417227564102564e-06,
|
| 226342 |
+
"loss": 0.5232,
|
| 226343 |
+
"step": 99805
|
| 226344 |
+
},
|
| 226345 |
+
{
|
| 226346 |
+
"epoch": 798.0,
|
| 226347 |
+
"eval_loss": 0.3478337228298187,
|
| 226348 |
+
"eval_runtime": 37.8125,
|
| 226349 |
+
"eval_samples_per_second": 22.03,
|
| 226350 |
+
"eval_steps_per_second": 0.714,
|
| 226351 |
+
"eval_wer": 0.17775073376762832,
|
| 226352 |
+
"step": 99809
|
| 226353 |
+
},
|
| 226354 |
+
{
|
| 226355 |
+
"epoch": 804.01,
|
| 226356 |
+
"learning_rate": 8.417147435897437e-06,
|
| 226357 |
+
"loss": 4.0085,
|
| 226358 |
+
"step": 99810
|
| 226359 |
+
},
|
| 226360 |
+
{
|
| 226361 |
+
"epoch": 804.05,
|
| 226362 |
+
"learning_rate": 8.417067307692308e-06,
|
| 226363 |
+
"loss": 0.3344,
|
| 226364 |
+
"step": 99815
|
| 226365 |
+
},
|
| 226366 |
+
{
|
| 226367 |
+
"epoch": 804.09,
|
| 226368 |
+
"learning_rate": 8.41698717948718e-06,
|
| 226369 |
+
"loss": 0.413,
|
| 226370 |
+
"step": 99820
|
| 226371 |
+
},
|
| 226372 |
+
{
|
| 226373 |
+
"epoch": 804.13,
|
| 226374 |
+
"learning_rate": 8.416907051282051e-06,
|
| 226375 |
+
"loss": 0.3468,
|
| 226376 |
+
"step": 99825
|
| 226377 |
+
},
|
| 226378 |
+
{
|
| 226379 |
+
"epoch": 804.17,
|
| 226380 |
+
"learning_rate": 8.416826923076924e-06,
|
| 226381 |
+
"loss": 0.4792,
|
| 226382 |
+
"step": 99830
|
| 226383 |
+
},
|
| 226384 |
+
{
|
| 226385 |
+
"epoch": 804.21,
|
| 226386 |
+
"learning_rate": 8.416746794871795e-06,
|
| 226387 |
+
"loss": 1.1734,
|
| 226388 |
+
"step": 99835
|
| 226389 |
+
},
|
| 226390 |
+
{
|
| 226391 |
+
"epoch": 804.25,
|
| 226392 |
+
"learning_rate": 8.416666666666667e-06,
|
| 226393 |
+
"loss": 0.319,
|
| 226394 |
+
"step": 99840
|
| 226395 |
+
},
|
| 226396 |
+
{
|
| 226397 |
+
"epoch": 804.29,
|
| 226398 |
+
"learning_rate": 8.416586538461538e-06,
|
| 226399 |
+
"loss": 0.3238,
|
| 226400 |
+
"step": 99845
|
| 226401 |
+
},
|
| 226402 |
+
{
|
| 226403 |
+
"epoch": 804.33,
|
| 226404 |
+
"learning_rate": 8.416506410256411e-06,
|
| 226405 |
+
"loss": 0.3412,
|
| 226406 |
+
"step": 99850
|
| 226407 |
+
},
|
| 226408 |
+
{
|
| 226409 |
+
"epoch": 804.37,
|
| 226410 |
+
"learning_rate": 8.416426282051283e-06,
|
| 226411 |
+
"loss": 0.5476,
|
| 226412 |
+
"step": 99855
|
| 226413 |
+
},
|
| 226414 |
+
{
|
| 226415 |
+
"epoch": 804.41,
|
| 226416 |
+
"learning_rate": 8.416346153846154e-06,
|
| 226417 |
+
"loss": 1.0643,
|
| 226418 |
+
"step": 99860
|
| 226419 |
+
},
|
| 226420 |
+
{
|
| 226421 |
+
"epoch": 804.45,
|
| 226422 |
+
"learning_rate": 8.416266025641027e-06,
|
| 226423 |
+
"loss": 1.1687,
|
| 226424 |
+
"step": 99865
|
| 226425 |
+
},
|
| 226426 |
+
{
|
| 226427 |
+
"epoch": 804.49,
|
| 226428 |
+
"learning_rate": 8.416185897435898e-06,
|
| 226429 |
+
"loss": 0.3379,
|
| 226430 |
+
"step": 99870
|
| 226431 |
+
},
|
| 226432 |
+
{
|
| 226433 |
+
"epoch": 804.53,
|
| 226434 |
+
"learning_rate": 8.41610576923077e-06,
|
| 226435 |
+
"loss": 0.3514,
|
| 226436 |
+
"step": 99875
|
| 226437 |
+
},
|
| 226438 |
+
{
|
| 226439 |
+
"epoch": 804.57,
|
| 226440 |
+
"learning_rate": 8.416025641025641e-06,
|
| 226441 |
+
"loss": 0.6156,
|
| 226442 |
+
"step": 99880
|
| 226443 |
+
},
|
| 226444 |
+
{
|
| 226445 |
+
"epoch": 804.61,
|
| 226446 |
+
"learning_rate": 8.415945512820514e-06,
|
| 226447 |
+
"loss": 1.1068,
|
| 226448 |
+
"step": 99885
|
| 226449 |
+
},
|
| 226450 |
+
{
|
| 226451 |
+
"epoch": 804.65,
|
| 226452 |
+
"learning_rate": 8.415865384615386e-06,
|
| 226453 |
+
"loss": 0.3151,
|
| 226454 |
+
"step": 99890
|
| 226455 |
+
},
|
| 226456 |
+
{
|
| 226457 |
+
"epoch": 804.69,
|
| 226458 |
+
"learning_rate": 8.415785256410257e-06,
|
| 226459 |
+
"loss": 0.2707,
|
| 226460 |
+
"step": 99895
|
| 226461 |
+
},
|
| 226462 |
+
{
|
| 226463 |
+
"epoch": 804.73,
|
| 226464 |
+
"learning_rate": 8.41570512820513e-06,
|
| 226465 |
+
"loss": 0.3389,
|
| 226466 |
+
"step": 99900
|
| 226467 |
+
},
|
| 226468 |
+
{
|
| 226469 |
+
"epoch": 804.77,
|
| 226470 |
+
"learning_rate": 8.415625000000001e-06,
|
| 226471 |
+
"loss": 0.5223,
|
| 226472 |
+
"step": 99905
|
| 226473 |
+
},
|
| 226474 |
+
{
|
| 226475 |
+
"epoch": 804.81,
|
| 226476 |
+
"learning_rate": 8.415544871794873e-06,
|
| 226477 |
+
"loss": 1.1362,
|
| 226478 |
+
"step": 99910
|
| 226479 |
+
},
|
| 226480 |
+
{
|
| 226481 |
+
"epoch": 804.85,
|
| 226482 |
+
"learning_rate": 8.415464743589744e-06,
|
| 226483 |
+
"loss": 0.2897,
|
| 226484 |
+
"step": 99915
|
| 226485 |
+
},
|
| 226486 |
+
{
|
| 226487 |
+
"epoch": 804.89,
|
| 226488 |
+
"learning_rate": 8.415384615384617e-06,
|
| 226489 |
+
"loss": 0.2626,
|
| 226490 |
+
"step": 99920
|
| 226491 |
+
},
|
| 226492 |
+
{
|
| 226493 |
+
"epoch": 804.93,
|
| 226494 |
+
"learning_rate": 8.415304487179488e-06,
|
| 226495 |
+
"loss": 0.3401,
|
| 226496 |
+
"step": 99925
|
| 226497 |
+
},
|
| 226498 |
+
{
|
| 226499 |
+
"epoch": 804.97,
|
| 226500 |
+
"learning_rate": 8.41522435897436e-06,
|
| 226501 |
+
"loss": 0.6187,
|
| 226502 |
+
"step": 99930
|
| 226503 |
+
},
|
| 226504 |
+
{
|
| 226505 |
+
"epoch": 805.0,
|
| 226506 |
+
"eval_loss": 0.37858957052230835,
|
| 226507 |
+
"eval_runtime": 38.0372,
|
| 226508 |
+
"eval_samples_per_second": 21.9,
|
| 226509 |
+
"eval_steps_per_second": 0.71,
|
| 226510 |
+
"eval_wer": 0.18095796963250935,
|
| 226511 |
+
"step": 99933
|
| 226512 |
+
},
|
| 226513 |
+
{
|
| 226514 |
+
"epoch": 805.02,
|
| 226515 |
+
"learning_rate": 8.415144230769231e-06,
|
| 226516 |
+
"loss": 0.3085,
|
| 226517 |
+
"step": 99935
|
| 226518 |
+
},
|
| 226519 |
+
{
|
| 226520 |
+
"epoch": 805.06,
|
| 226521 |
+
"learning_rate": 8.415064102564104e-06,
|
| 226522 |
+
"loss": 0.2646,
|
| 226523 |
+
"step": 99940
|
| 226524 |
+
},
|
| 226525 |
+
{
|
| 226526 |
+
"epoch": 805.1,
|
| 226527 |
+
"learning_rate": 8.414983974358974e-06,
|
| 226528 |
+
"loss": 0.3165,
|
| 226529 |
+
"step": 99945
|
| 226530 |
+
},
|
| 226531 |
+
{
|
| 226532 |
+
"epoch": 805.14,
|
| 226533 |
+
"learning_rate": 8.414903846153847e-06,
|
| 226534 |
+
"loss": 0.3132,
|
| 226535 |
+
"step": 99950
|
| 226536 |
+
},
|
| 226537 |
+
{
|
| 226538 |
+
"epoch": 805.18,
|
| 226539 |
+
"learning_rate": 8.41482371794872e-06,
|
| 226540 |
+
"loss": 0.6313,
|
| 226541 |
+
"step": 99955
|
| 226542 |
+
},
|
| 226543 |
+
{
|
| 226544 |
+
"epoch": 805.22,
|
| 226545 |
+
"learning_rate": 8.41474358974359e-06,
|
| 226546 |
+
"loss": 1.0675,
|
| 226547 |
+
"step": 99960
|
| 226548 |
+
},
|
| 226549 |
+
{
|
| 226550 |
+
"epoch": 805.26,
|
| 226551 |
+
"learning_rate": 8.414663461538463e-06,
|
| 226552 |
+
"loss": 0.286,
|
| 226553 |
+
"step": 99965
|
| 226554 |
+
},
|
| 226555 |
+
{
|
| 226556 |
+
"epoch": 805.3,
|
| 226557 |
+
"learning_rate": 8.414583333333334e-06,
|
| 226558 |
+
"loss": 0.2669,
|
| 226559 |
+
"step": 99970
|
| 226560 |
+
},
|
| 226561 |
+
{
|
| 226562 |
+
"epoch": 805.34,
|
| 226563 |
+
"learning_rate": 8.414503205128205e-06,
|
| 226564 |
+
"loss": 0.3894,
|
| 226565 |
+
"step": 99975
|
| 226566 |
+
},
|
| 226567 |
+
{
|
| 226568 |
+
"epoch": 805.38,
|
| 226569 |
+
"learning_rate": 8.414423076923077e-06,
|
| 226570 |
+
"loss": 0.628,
|
| 226571 |
+
"step": 99980
|
| 226572 |
+
},
|
| 226573 |
+
{
|
| 226574 |
+
"epoch": 805.42,
|
| 226575 |
+
"learning_rate": 8.41434294871795e-06,
|
| 226576 |
+
"loss": 0.9502,
|
| 226577 |
+
"step": 99985
|
| 226578 |
+
},
|
| 226579 |
+
{
|
| 226580 |
+
"epoch": 805.46,
|
| 226581 |
+
"learning_rate": 8.414262820512821e-06,
|
| 226582 |
+
"loss": 0.3369,
|
| 226583 |
+
"step": 99990
|
| 226584 |
+
},
|
| 226585 |
+
{
|
| 226586 |
+
"epoch": 805.5,
|
| 226587 |
+
"learning_rate": 8.414182692307693e-06,
|
| 226588 |
+
"loss": 0.267,
|
| 226589 |
+
"step": 99995
|
| 226590 |
+
},
|
| 226591 |
+
{
|
| 226592 |
+
"epoch": 805.54,
|
| 226593 |
+
"learning_rate": 8.414102564102566e-06,
|
| 226594 |
+
"loss": 0.3449,
|
| 226595 |
+
"step": 100000
|
| 226596 |
+
},
|
| 226597 |
+
{
|
| 226598 |
+
"epoch": 805.58,
|
| 226599 |
+
"learning_rate": 8.414022435897437e-06,
|
| 226600 |
+
"loss": 0.6615,
|
| 226601 |
+
"step": 100005
|
| 226602 |
+
},
|
| 226603 |
+
{
|
| 226604 |
+
"epoch": 805.62,
|
| 226605 |
+
"learning_rate": 8.413942307692308e-06,
|
| 226606 |
+
"loss": 1.0243,
|
| 226607 |
+
"step": 100010
|
| 226608 |
+
},
|
| 226609 |
+
{
|
| 226610 |
+
"epoch": 805.66,
|
| 226611 |
+
"learning_rate": 8.41386217948718e-06,
|
| 226612 |
+
"loss": 0.2643,
|
| 226613 |
+
"step": 100015
|
| 226614 |
+
},
|
| 226615 |
+
{
|
| 226616 |
+
"epoch": 805.7,
|
| 226617 |
+
"learning_rate": 8.413782051282053e-06,
|
| 226618 |
+
"loss": 0.3039,
|
| 226619 |
+
"step": 100020
|
| 226620 |
+
},
|
| 226621 |
+
{
|
| 226622 |
+
"epoch": 805.74,
|
| 226623 |
+
"learning_rate": 8.413701923076924e-06,
|
| 226624 |
+
"loss": 0.3028,
|
| 226625 |
+
"step": 100025
|
| 226626 |
+
},
|
| 226627 |
+
{
|
| 226628 |
+
"epoch": 805.78,
|
| 226629 |
+
"learning_rate": 8.413621794871795e-06,
|
| 226630 |
+
"loss": 0.5747,
|
| 226631 |
+
"step": 100030
|
| 226632 |
+
},
|
| 226633 |
+
{
|
| 226634 |
+
"epoch": 805.82,
|
| 226635 |
+
"learning_rate": 8.413541666666667e-06,
|
| 226636 |
+
"loss": 1.0076,
|
| 226637 |
+
"step": 100035
|
| 226638 |
+
},
|
| 226639 |
+
{
|
| 226640 |
+
"epoch": 805.86,
|
| 226641 |
+
"learning_rate": 8.41346153846154e-06,
|
| 226642 |
+
"loss": 0.3063,
|
| 226643 |
+
"step": 100040
|
| 226644 |
+
},
|
| 226645 |
+
{
|
| 226646 |
+
"epoch": 805.9,
|
| 226647 |
+
"learning_rate": 8.413381410256411e-06,
|
| 226648 |
+
"loss": 0.311,
|
| 226649 |
+
"step": 100045
|
| 226650 |
+
},
|
| 226651 |
+
{
|
| 226652 |
+
"epoch": 805.94,
|
| 226653 |
+
"learning_rate": 8.413301282051283e-06,
|
| 226654 |
+
"loss": 0.3488,
|
| 226655 |
+
"step": 100050
|
| 226656 |
+
},
|
| 226657 |
+
{
|
| 226658 |
+
"epoch": 805.98,
|
| 226659 |
+
"learning_rate": 8.413221153846156e-06,
|
| 226660 |
+
"loss": 0.6481,
|
| 226661 |
+
"step": 100055
|
| 226662 |
+
},
|
| 226663 |
+
{
|
| 226664 |
+
"epoch": 806.0,
|
| 226665 |
+
"eval_loss": 0.35248780250549316,
|
| 226666 |
+
"eval_runtime": 40.524,
|
| 226667 |
+
"eval_samples_per_second": 20.556,
|
| 226668 |
+
"eval_steps_per_second": 0.666,
|
| 226669 |
+
"eval_wer": 0.18215773595747736,
|
| 226670 |
+
"step": 100057
|
| 226671 |
+
},
|
| 226672 |
+
{
|
| 226673 |
+
"epoch": 806.02,
|
| 226674 |
+
"learning_rate": 8.413141025641027e-06,
|
| 226675 |
+
"loss": 0.2858,
|
| 226676 |
+
"step": 100060
|
| 226677 |
+
},
|
| 226678 |
+
{
|
| 226679 |
+
"epoch": 806.06,
|
| 226680 |
+
"learning_rate": 8.413060897435898e-06,
|
| 226681 |
+
"loss": 0.275,
|
| 226682 |
+
"step": 100065
|
| 226683 |
+
},
|
| 226684 |
+
{
|
| 226685 |
+
"epoch": 806.1,
|
| 226686 |
+
"learning_rate": 8.41298076923077e-06,
|
| 226687 |
+
"loss": 0.2892,
|
| 226688 |
+
"step": 100070
|
| 226689 |
+
},
|
| 226690 |
+
{
|
| 226691 |
+
"epoch": 806.15,
|
| 226692 |
+
"learning_rate": 8.412900641025643e-06,
|
| 226693 |
+
"loss": 0.4414,
|
| 226694 |
+
"step": 100075
|
| 226695 |
+
},
|
| 226696 |
+
{
|
| 226697 |
+
"epoch": 806.19,
|
| 226698 |
+
"learning_rate": 8.412820512820512e-06,
|
| 226699 |
+
"loss": 0.7002,
|
| 226700 |
+
"step": 100080
|
| 226701 |
+
},
|
| 226702 |
+
{
|
| 226703 |
+
"epoch": 806.23,
|
| 226704 |
+
"learning_rate": 8.412740384615385e-06,
|
| 226705 |
+
"loss": 0.7833,
|
| 226706 |
+
"step": 100085
|
| 226707 |
+
},
|
| 226708 |
+
{
|
| 226709 |
+
"epoch": 806.27,
|
| 226710 |
+
"learning_rate": 8.412660256410257e-06,
|
| 226711 |
+
"loss": 0.3272,
|
| 226712 |
+
"step": 100090
|
| 226713 |
+
},
|
| 226714 |
+
{
|
| 226715 |
+
"epoch": 806.31,
|
| 226716 |
+
"learning_rate": 8.412580128205128e-06,
|
| 226717 |
+
"loss": 0.3191,
|
| 226718 |
+
"step": 100095
|
| 226719 |
+
},
|
| 226720 |
+
{
|
| 226721 |
+
"epoch": 806.35,
|
| 226722 |
+
"learning_rate": 8.412500000000001e-06,
|
| 226723 |
+
"loss": 0.3924,
|
| 226724 |
+
"step": 100100
|
| 226725 |
+
},
|
| 226726 |
+
{
|
| 226727 |
+
"epoch": 806.39,
|
| 226728 |
+
"learning_rate": 8.412419871794873e-06,
|
| 226729 |
+
"loss": 0.8189,
|
| 226730 |
+
"step": 100105
|
| 226731 |
+
},
|
| 226732 |
+
{
|
| 226733 |
+
"epoch": 806.43,
|
| 226734 |
+
"learning_rate": 8.412339743589744e-06,
|
| 226735 |
+
"loss": 0.9466,
|
| 226736 |
+
"step": 100110
|
| 226737 |
+
},
|
| 226738 |
+
{
|
| 226739 |
+
"epoch": 806.47,
|
| 226740 |
+
"learning_rate": 8.412259615384615e-06,
|
| 226741 |
+
"loss": 0.2812,
|
| 226742 |
+
"step": 100115
|
| 226743 |
+
},
|
| 226744 |
+
{
|
| 226745 |
+
"epoch": 806.51,
|
| 226746 |
+
"learning_rate": 8.412179487179488e-06,
|
| 226747 |
+
"loss": 0.3069,
|
| 226748 |
+
"step": 100120
|
| 226749 |
+
},
|
| 226750 |
+
{
|
| 226751 |
+
"epoch": 806.55,
|
| 226752 |
+
"learning_rate": 8.41209935897436e-06,
|
| 226753 |
+
"loss": 0.3893,
|
| 226754 |
+
"step": 100125
|
| 226755 |
+
},
|
| 226756 |
+
{
|
| 226757 |
+
"epoch": 806.59,
|
| 226758 |
+
"learning_rate": 8.412019230769231e-06,
|
| 226759 |
+
"loss": 0.6543,
|
| 226760 |
+
"step": 100130
|
| 226761 |
+
},
|
| 226762 |
+
{
|
| 226763 |
+
"epoch": 806.63,
|
| 226764 |
+
"learning_rate": 8.411939102564102e-06,
|
| 226765 |
+
"loss": 0.8736,
|
| 226766 |
+
"step": 100135
|
| 226767 |
+
},
|
| 226768 |
+
{
|
| 226769 |
+
"epoch": 806.67,
|
| 226770 |
+
"learning_rate": 8.411858974358975e-06,
|
| 226771 |
+
"loss": 0.2459,
|
| 226772 |
+
"step": 100140
|
| 226773 |
+
},
|
| 226774 |
+
{
|
| 226775 |
+
"epoch": 806.71,
|
| 226776 |
+
"learning_rate": 8.411778846153847e-06,
|
| 226777 |
+
"loss": 0.3367,
|
| 226778 |
+
"step": 100145
|
| 226779 |
+
},
|
| 226780 |
+
{
|
| 226781 |
+
"epoch": 806.75,
|
| 226782 |
+
"learning_rate": 8.411698717948718e-06,
|
| 226783 |
+
"loss": 0.3835,
|
| 226784 |
+
"step": 100150
|
| 226785 |
+
},
|
| 226786 |
+
{
|
| 226787 |
+
"epoch": 806.79,
|
| 226788 |
+
"learning_rate": 8.411618589743591e-06,
|
| 226789 |
+
"loss": 0.7286,
|
| 226790 |
+
"step": 100155
|
| 226791 |
+
},
|
| 226792 |
+
{
|
| 226793 |
+
"epoch": 806.83,
|
| 226794 |
+
"learning_rate": 8.411538461538463e-06,
|
| 226795 |
+
"loss": 0.8448,
|
| 226796 |
+
"step": 100160
|
| 226797 |
+
},
|
| 226798 |
+
{
|
| 226799 |
+
"epoch": 806.87,
|
| 226800 |
+
"learning_rate": 8.411458333333334e-06,
|
| 226801 |
+
"loss": 0.2768,
|
| 226802 |
+
"step": 100165
|
| 226803 |
+
},
|
| 226804 |
+
{
|
| 226805 |
+
"epoch": 806.91,
|
| 226806 |
+
"learning_rate": 8.411378205128205e-06,
|
| 226807 |
+
"loss": 0.3267,
|
| 226808 |
+
"step": 100170
|
| 226809 |
+
},
|
| 226810 |
+
{
|
| 226811 |
+
"epoch": 806.95,
|
| 226812 |
+
"learning_rate": 8.411298076923078e-06,
|
| 226813 |
+
"loss": 0.3316,
|
| 226814 |
+
"step": 100175
|
| 226815 |
+
},
|
| 226816 |
+
{
|
| 226817 |
+
"epoch": 806.99,
|
| 226818 |
+
"learning_rate": 8.41121794871795e-06,
|
| 226819 |
+
"loss": 0.9675,
|
| 226820 |
+
"step": 100180
|
| 226821 |
+
},
|
| 226822 |
+
{
|
| 226823 |
+
"epoch": 807.0,
|
| 226824 |
+
"eval_loss": 0.4434155821800232,
|
| 226825 |
+
"eval_runtime": 38.2932,
|
| 226826 |
+
"eval_samples_per_second": 21.753,
|
| 226827 |
+
"eval_steps_per_second": 0.705,
|
| 226828 |
+
"eval_wer": 0.18565555718690355,
|
| 226829 |
+
"step": 100181
|
| 226830 |
}
|
| 226831 |
],
|
| 226832 |
"max_steps": 620000,
|
| 226833 |
"num_train_epochs": 5000,
|
| 226834 |
+
"total_flos": 2.819371019425974e+20,
|
| 226835 |
"trial_name": null,
|
| 226836 |
"trial_params": null
|
| 226837 |
}
|
model-bin/finetune/base/{checkpoint-99560 → checkpoint-100181}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629980056.6178455/events.out.tfevents.1629980056.8e89bd551565.924.181
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c22752d5e9d46587ee5483c6167738121534b5d2bdde66f150ffdafb66afdcd8
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629980506.0580995/events.out.tfevents.1629980506.8e89bd551565.924.183
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7be4ae668a669990af1b64f28be2b644cbe322156e930ceef12167821d59bc82
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629980941.2291443/events.out.tfevents.1629980941.8e89bd551565.924.185
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:17a0b7a2e7e9a0b6b13f908ec88e35e6d4e5f3a07a9dbf19167403655e20033c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629981376.296084/events.out.tfevents.1629981376.8e89bd551565.924.187
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df5a95646d67b5d8b0098832d77b54d7c5d74b8238a6fc9c93852a8217151ec9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629981805.4906385/events.out.tfevents.1629981805.8e89bd551565.924.189
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b80bea72f6d24ec071e069b3a61125b31c158a75ee1b97395653a1ac1703572
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629980056.8e89bd551565.924.180
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:45a4bef00a6173531f71d2a4df9a89fed341e8c5edcd49f5caa61f3374908a5f
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629980506.8e89bd551565.924.182
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:414d773ac83a6e2e0049f36fbb3ec46960411818101a9e9c6f0440e9c0e12879
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629980940.8e89bd551565.924.184
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb2220d7b6f4dd7eb4a0d1d3790872248f76682fa83ef73148fb7f2f2606c1fe
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629981375.8e89bd551565.924.186
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4d126c57a4b4ffed8e1a472d38e83fdde836b589a2cf158bbd0b6134cf899201
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629981805.8e89bd551565.924.188
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e5772cba43f717e6e25dcf5b339279ce7260183939d437229aaf3ddfb23886a2
|
| 3 |
+
size 8622
|