"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/trainer_state.json +792 -3
- model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/training_args.bin +0 -0
- model-bin/finetune/base/log/1629948643.035603/events.out.tfevents.1629948643.8e89bd551565.924.41 +3 -0
- model-bin/finetune/base/log/1629949081.6864243/events.out.tfevents.1629949081.8e89bd551565.924.43 +3 -0
- model-bin/finetune/base/log/1629949522.6644921/events.out.tfevents.1629949522.8e89bd551565.924.45 +3 -0
- model-bin/finetune/base/log/1629949956.9957566/events.out.tfevents.1629949956.8e89bd551565.924.47 +3 -0
- model-bin/finetune/base/log/1629950393.869526/events.out.tfevents.1629950393.8e89bd551565.924.49 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629948643.8e89bd551565.924.40 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629949081.8e89bd551565.924.42 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629949521.8e89bd551565.924.44 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629949955.8e89bd551565.924.46 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629950393.8e89bd551565.924.48 +3 -0
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/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:fccedbb272de74ab25b2035aeade8d1e49eff475d4b868bd879b45181e3c7ad3
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/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:a18df0deb63865cc8efa82c77473fae3876daf550dfbd245e80481e240d527ef
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/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:eb079d7ed4b71c04d28c14c590fcc5ba695809dc09ef0cde234ad0b53c0629ce
|
| 3 |
+
size 14567
|
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/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 sha256:1e1991f32f5f69855e1bd3da9a5830a4f1aaef8433c71f35e8c344ce4867ad9e
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/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 sha256:50381413d8b427dd961e0ecfc10029215d94b0125a94f79ea725055ba283c20d
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -214950,11 +214950,800 @@
|
|
| 214950 |
"eval_steps_per_second": 0.676,
|
| 214951 |
"eval_wer": 0.17860725032239577,
|
| 214952 |
"step": 90847
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 214953 |
}
|
| 214954 |
],
|
| 214955 |
"max_steps": 620000,
|
| 214956 |
"num_train_epochs": 5000,
|
| 214957 |
-
"total_flos": 2.
|
| 214958 |
"trial_name": null,
|
| 214959 |
"trial_params": null
|
| 214960 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.17731766728757703,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-88857",
|
| 4 |
+
"epoch": 736.995983935743,
|
| 5 |
+
"global_step": 91468,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 214950 |
"eval_steps_per_second": 0.676,
|
| 214951 |
"eval_wer": 0.17860725032239577,
|
| 214952 |
"step": 90847
|
| 214953 |
+
},
|
| 214954 |
+
{
|
| 214955 |
+
"epoch": 726.02,
|
| 214956 |
+
"learning_rate": 8.560673076923077e-06,
|
| 214957 |
+
"loss": 0.3015,
|
| 214958 |
+
"step": 90850
|
| 214959 |
+
},
|
| 214960 |
+
{
|
| 214961 |
+
"epoch": 726.06,
|
| 214962 |
+
"learning_rate": 8.56059294871795e-06,
|
| 214963 |
+
"loss": 0.3127,
|
| 214964 |
+
"step": 90855
|
| 214965 |
+
},
|
| 214966 |
+
{
|
| 214967 |
+
"epoch": 726.1,
|
| 214968 |
+
"learning_rate": 8.56051282051282e-06,
|
| 214969 |
+
"loss": 0.2894,
|
| 214970 |
+
"step": 90860
|
| 214971 |
+
},
|
| 214972 |
+
{
|
| 214973 |
+
"epoch": 726.14,
|
| 214974 |
+
"learning_rate": 8.560432692307693e-06,
|
| 214975 |
+
"loss": 0.4391,
|
| 214976 |
+
"step": 90865
|
| 214977 |
+
},
|
| 214978 |
+
{
|
| 214979 |
+
"epoch": 726.18,
|
| 214980 |
+
"learning_rate": 8.560352564102566e-06,
|
| 214981 |
+
"loss": 0.6959,
|
| 214982 |
+
"step": 90870
|
| 214983 |
+
},
|
| 214984 |
+
{
|
| 214985 |
+
"epoch": 726.22,
|
| 214986 |
+
"learning_rate": 8.560272435897436e-06,
|
| 214987 |
+
"loss": 0.8727,
|
| 214988 |
+
"step": 90875
|
| 214989 |
+
},
|
| 214990 |
+
{
|
| 214991 |
+
"epoch": 726.26,
|
| 214992 |
+
"learning_rate": 8.560192307692309e-06,
|
| 214993 |
+
"loss": 0.2816,
|
| 214994 |
+
"step": 90880
|
| 214995 |
+
},
|
| 214996 |
+
{
|
| 214997 |
+
"epoch": 726.3,
|
| 214998 |
+
"learning_rate": 8.56011217948718e-06,
|
| 214999 |
+
"loss": 0.3963,
|
| 215000 |
+
"step": 90885
|
| 215001 |
+
},
|
| 215002 |
+
{
|
| 215003 |
+
"epoch": 726.34,
|
| 215004 |
+
"learning_rate": 8.560032051282051e-06,
|
| 215005 |
+
"loss": 0.42,
|
| 215006 |
+
"step": 90890
|
| 215007 |
+
},
|
| 215008 |
+
{
|
| 215009 |
+
"epoch": 726.38,
|
| 215010 |
+
"learning_rate": 8.559951923076923e-06,
|
| 215011 |
+
"loss": 0.6985,
|
| 215012 |
+
"step": 90895
|
| 215013 |
+
},
|
| 215014 |
+
{
|
| 215015 |
+
"epoch": 726.42,
|
| 215016 |
+
"learning_rate": 8.559871794871796e-06,
|
| 215017 |
+
"loss": 0.8407,
|
| 215018 |
+
"step": 90900
|
| 215019 |
+
},
|
| 215020 |
+
{
|
| 215021 |
+
"epoch": 726.46,
|
| 215022 |
+
"learning_rate": 8.559791666666667e-06,
|
| 215023 |
+
"loss": 0.2827,
|
| 215024 |
+
"step": 90905
|
| 215025 |
+
},
|
| 215026 |
+
{
|
| 215027 |
+
"epoch": 726.5,
|
| 215028 |
+
"learning_rate": 8.559711538461539e-06,
|
| 215029 |
+
"loss": 0.3499,
|
| 215030 |
+
"step": 90910
|
| 215031 |
+
},
|
| 215032 |
+
{
|
| 215033 |
+
"epoch": 726.54,
|
| 215034 |
+
"learning_rate": 8.55963141025641e-06,
|
| 215035 |
+
"loss": 0.6144,
|
| 215036 |
+
"step": 90915
|
| 215037 |
+
},
|
| 215038 |
+
{
|
| 215039 |
+
"epoch": 726.58,
|
| 215040 |
+
"learning_rate": 8.559551282051283e-06,
|
| 215041 |
+
"loss": 0.7495,
|
| 215042 |
+
"step": 90920
|
| 215043 |
+
},
|
| 215044 |
+
{
|
| 215045 |
+
"epoch": 726.62,
|
| 215046 |
+
"learning_rate": 8.559471153846154e-06,
|
| 215047 |
+
"loss": 0.8357,
|
| 215048 |
+
"step": 90925
|
| 215049 |
+
},
|
| 215050 |
+
{
|
| 215051 |
+
"epoch": 726.66,
|
| 215052 |
+
"learning_rate": 8.559391025641026e-06,
|
| 215053 |
+
"loss": 0.3003,
|
| 215054 |
+
"step": 90930
|
| 215055 |
+
},
|
| 215056 |
+
{
|
| 215057 |
+
"epoch": 726.7,
|
| 215058 |
+
"learning_rate": 8.559310897435899e-06,
|
| 215059 |
+
"loss": 0.3175,
|
| 215060 |
+
"step": 90935
|
| 215061 |
+
},
|
| 215062 |
+
{
|
| 215063 |
+
"epoch": 726.74,
|
| 215064 |
+
"learning_rate": 8.55923076923077e-06,
|
| 215065 |
+
"loss": 0.3643,
|
| 215066 |
+
"step": 90940
|
| 215067 |
+
},
|
| 215068 |
+
{
|
| 215069 |
+
"epoch": 726.78,
|
| 215070 |
+
"learning_rate": 8.559150641025642e-06,
|
| 215071 |
+
"loss": 0.8117,
|
| 215072 |
+
"step": 90945
|
| 215073 |
+
},
|
| 215074 |
+
{
|
| 215075 |
+
"epoch": 726.82,
|
| 215076 |
+
"learning_rate": 8.559070512820513e-06,
|
| 215077 |
+
"loss": 0.7839,
|
| 215078 |
+
"step": 90950
|
| 215079 |
+
},
|
| 215080 |
+
{
|
| 215081 |
+
"epoch": 726.86,
|
| 215082 |
+
"learning_rate": 8.558990384615386e-06,
|
| 215083 |
+
"loss": 0.3361,
|
| 215084 |
+
"step": 90955
|
| 215085 |
+
},
|
| 215086 |
+
{
|
| 215087 |
+
"epoch": 726.9,
|
| 215088 |
+
"learning_rate": 8.558910256410257e-06,
|
| 215089 |
+
"loss": 0.2919,
|
| 215090 |
+
"step": 90960
|
| 215091 |
+
},
|
| 215092 |
+
{
|
| 215093 |
+
"epoch": 726.94,
|
| 215094 |
+
"learning_rate": 8.558830128205129e-06,
|
| 215095 |
+
"loss": 0.3633,
|
| 215096 |
+
"step": 90965
|
| 215097 |
+
},
|
| 215098 |
+
{
|
| 215099 |
+
"epoch": 726.98,
|
| 215100 |
+
"learning_rate": 8.558750000000002e-06,
|
| 215101 |
+
"loss": 0.8311,
|
| 215102 |
+
"step": 90970
|
| 215103 |
+
},
|
| 215104 |
+
{
|
| 215105 |
+
"epoch": 727.0,
|
| 215106 |
+
"eval_loss": 0.326074481010437,
|
| 215107 |
+
"eval_runtime": 40.1779,
|
| 215108 |
+
"eval_samples_per_second": 20.932,
|
| 215109 |
+
"eval_steps_per_second": 0.672,
|
| 215110 |
+
"eval_wer": 0.18255121313380793,
|
| 215111 |
+
"step": 90972
|
| 215112 |
+
},
|
| 215113 |
+
{
|
| 215114 |
+
"epoch": 733.02,
|
| 215115 |
+
"learning_rate": 8.558669871794873e-06,
|
| 215116 |
+
"loss": 0.4064,
|
| 215117 |
+
"step": 90975
|
| 215118 |
+
},
|
| 215119 |
+
{
|
| 215120 |
+
"epoch": 733.06,
|
| 215121 |
+
"learning_rate": 8.558589743589744e-06,
|
| 215122 |
+
"loss": 0.2823,
|
| 215123 |
+
"step": 90980
|
| 215124 |
+
},
|
| 215125 |
+
{
|
| 215126 |
+
"epoch": 733.1,
|
| 215127 |
+
"learning_rate": 8.558509615384616e-06,
|
| 215128 |
+
"loss": 0.3044,
|
| 215129 |
+
"step": 90985
|
| 215130 |
+
},
|
| 215131 |
+
{
|
| 215132 |
+
"epoch": 733.14,
|
| 215133 |
+
"learning_rate": 8.558429487179489e-06,
|
| 215134 |
+
"loss": 0.3886,
|
| 215135 |
+
"step": 90990
|
| 215136 |
+
},
|
| 215137 |
+
{
|
| 215138 |
+
"epoch": 733.18,
|
| 215139 |
+
"learning_rate": 8.55834935897436e-06,
|
| 215140 |
+
"loss": 0.776,
|
| 215141 |
+
"step": 90995
|
| 215142 |
+
},
|
| 215143 |
+
{
|
| 215144 |
+
"epoch": 733.22,
|
| 215145 |
+
"learning_rate": 8.558269230769232e-06,
|
| 215146 |
+
"loss": 0.9261,
|
| 215147 |
+
"step": 91000
|
| 215148 |
+
},
|
| 215149 |
+
{
|
| 215150 |
+
"epoch": 733.27,
|
| 215151 |
+
"learning_rate": 8.558189102564103e-06,
|
| 215152 |
+
"loss": 0.3216,
|
| 215153 |
+
"step": 91005
|
| 215154 |
+
},
|
| 215155 |
+
{
|
| 215156 |
+
"epoch": 733.31,
|
| 215157 |
+
"learning_rate": 8.558108974358976e-06,
|
| 215158 |
+
"loss": 0.3408,
|
| 215159 |
+
"step": 91010
|
| 215160 |
+
},
|
| 215161 |
+
{
|
| 215162 |
+
"epoch": 733.35,
|
| 215163 |
+
"learning_rate": 8.558028846153846e-06,
|
| 215164 |
+
"loss": 0.3725,
|
| 215165 |
+
"step": 91015
|
| 215166 |
+
},
|
| 215167 |
+
{
|
| 215168 |
+
"epoch": 733.39,
|
| 215169 |
+
"learning_rate": 8.557948717948719e-06,
|
| 215170 |
+
"loss": 0.7755,
|
| 215171 |
+
"step": 91020
|
| 215172 |
+
},
|
| 215173 |
+
{
|
| 215174 |
+
"epoch": 733.43,
|
| 215175 |
+
"learning_rate": 8.557868589743592e-06,
|
| 215176 |
+
"loss": 0.8765,
|
| 215177 |
+
"step": 91025
|
| 215178 |
+
},
|
| 215179 |
+
{
|
| 215180 |
+
"epoch": 733.47,
|
| 215181 |
+
"learning_rate": 8.557788461538461e-06,
|
| 215182 |
+
"loss": 0.3093,
|
| 215183 |
+
"step": 91030
|
| 215184 |
+
},
|
| 215185 |
+
{
|
| 215186 |
+
"epoch": 733.51,
|
| 215187 |
+
"learning_rate": 8.557708333333334e-06,
|
| 215188 |
+
"loss": 0.283,
|
| 215189 |
+
"step": 91035
|
| 215190 |
+
},
|
| 215191 |
+
{
|
| 215192 |
+
"epoch": 733.55,
|
| 215193 |
+
"learning_rate": 8.557628205128206e-06,
|
| 215194 |
+
"loss": 0.3365,
|
| 215195 |
+
"step": 91040
|
| 215196 |
+
},
|
| 215197 |
+
{
|
| 215198 |
+
"epoch": 733.59,
|
| 215199 |
+
"learning_rate": 8.557548076923077e-06,
|
| 215200 |
+
"loss": 0.6963,
|
| 215201 |
+
"step": 91045
|
| 215202 |
+
},
|
| 215203 |
+
{
|
| 215204 |
+
"epoch": 733.63,
|
| 215205 |
+
"learning_rate": 8.557467948717949e-06,
|
| 215206 |
+
"loss": 0.8104,
|
| 215207 |
+
"step": 91050
|
| 215208 |
+
},
|
| 215209 |
+
{
|
| 215210 |
+
"epoch": 733.67,
|
| 215211 |
+
"learning_rate": 8.557387820512822e-06,
|
| 215212 |
+
"loss": 0.2954,
|
| 215213 |
+
"step": 91055
|
| 215214 |
+
},
|
| 215215 |
+
{
|
| 215216 |
+
"epoch": 733.71,
|
| 215217 |
+
"learning_rate": 8.557307692307693e-06,
|
| 215218 |
+
"loss": 0.2938,
|
| 215219 |
+
"step": 91060
|
| 215220 |
+
},
|
| 215221 |
+
{
|
| 215222 |
+
"epoch": 733.75,
|
| 215223 |
+
"learning_rate": 8.557227564102564e-06,
|
| 215224 |
+
"loss": 0.336,
|
| 215225 |
+
"step": 91065
|
| 215226 |
+
},
|
| 215227 |
+
{
|
| 215228 |
+
"epoch": 733.79,
|
| 215229 |
+
"learning_rate": 8.557147435897437e-06,
|
| 215230 |
+
"loss": 0.7033,
|
| 215231 |
+
"step": 91070
|
| 215232 |
+
},
|
| 215233 |
+
{
|
| 215234 |
+
"epoch": 733.83,
|
| 215235 |
+
"learning_rate": 8.557067307692309e-06,
|
| 215236 |
+
"loss": 0.8494,
|
| 215237 |
+
"step": 91075
|
| 215238 |
+
},
|
| 215239 |
+
{
|
| 215240 |
+
"epoch": 733.87,
|
| 215241 |
+
"learning_rate": 8.55698717948718e-06,
|
| 215242 |
+
"loss": 0.2807,
|
| 215243 |
+
"step": 91080
|
| 215244 |
+
},
|
| 215245 |
+
{
|
| 215246 |
+
"epoch": 733.91,
|
| 215247 |
+
"learning_rate": 8.556907051282051e-06,
|
| 215248 |
+
"loss": 0.3524,
|
| 215249 |
+
"step": 91085
|
| 215250 |
+
},
|
| 215251 |
+
{
|
| 215252 |
+
"epoch": 733.95,
|
| 215253 |
+
"learning_rate": 8.556826923076924e-06,
|
| 215254 |
+
"loss": 0.4139,
|
| 215255 |
+
"step": 91090
|
| 215256 |
+
},
|
| 215257 |
+
{
|
| 215258 |
+
"epoch": 733.99,
|
| 215259 |
+
"learning_rate": 8.556746794871796e-06,
|
| 215260 |
+
"loss": 0.8845,
|
| 215261 |
+
"step": 91095
|
| 215262 |
+
},
|
| 215263 |
+
{
|
| 215264 |
+
"epoch": 734.0,
|
| 215265 |
+
"eval_loss": 0.39281827211380005,
|
| 215266 |
+
"eval_runtime": 39.2375,
|
| 215267 |
+
"eval_samples_per_second": 21.434,
|
| 215268 |
+
"eval_steps_per_second": 0.688,
|
| 215269 |
+
"eval_wer": 0.19532831935203934,
|
| 215270 |
+
"step": 91096
|
| 215271 |
+
},
|
| 215272 |
+
{
|
| 215273 |
+
"epoch": 734.03,
|
| 215274 |
+
"learning_rate": 8.556666666666667e-06,
|
| 215275 |
+
"loss": 0.3269,
|
| 215276 |
+
"step": 91100
|
| 215277 |
+
},
|
| 215278 |
+
{
|
| 215279 |
+
"epoch": 734.07,
|
| 215280 |
+
"learning_rate": 8.556586538461539e-06,
|
| 215281 |
+
"loss": 0.2877,
|
| 215282 |
+
"step": 91105
|
| 215283 |
+
},
|
| 215284 |
+
{
|
| 215285 |
+
"epoch": 734.11,
|
| 215286 |
+
"learning_rate": 8.556506410256412e-06,
|
| 215287 |
+
"loss": 0.2866,
|
| 215288 |
+
"step": 91110
|
| 215289 |
+
},
|
| 215290 |
+
{
|
| 215291 |
+
"epoch": 734.15,
|
| 215292 |
+
"learning_rate": 8.556426282051283e-06,
|
| 215293 |
+
"loss": 0.4321,
|
| 215294 |
+
"step": 91115
|
| 215295 |
+
},
|
| 215296 |
+
{
|
| 215297 |
+
"epoch": 734.19,
|
| 215298 |
+
"learning_rate": 8.556346153846154e-06,
|
| 215299 |
+
"loss": 1.0139,
|
| 215300 |
+
"step": 91120
|
| 215301 |
+
},
|
| 215302 |
+
{
|
| 215303 |
+
"epoch": 734.23,
|
| 215304 |
+
"learning_rate": 8.556266025641027e-06,
|
| 215305 |
+
"loss": 0.6581,
|
| 215306 |
+
"step": 91125
|
| 215307 |
+
},
|
| 215308 |
+
{
|
| 215309 |
+
"epoch": 734.27,
|
| 215310 |
+
"learning_rate": 8.556185897435899e-06,
|
| 215311 |
+
"loss": 0.2806,
|
| 215312 |
+
"step": 91130
|
| 215313 |
+
},
|
| 215314 |
+
{
|
| 215315 |
+
"epoch": 734.31,
|
| 215316 |
+
"learning_rate": 8.55610576923077e-06,
|
| 215317 |
+
"loss": 0.2541,
|
| 215318 |
+
"step": 91135
|
| 215319 |
+
},
|
| 215320 |
+
{
|
| 215321 |
+
"epoch": 734.35,
|
| 215322 |
+
"learning_rate": 8.556025641025641e-06,
|
| 215323 |
+
"loss": 0.3847,
|
| 215324 |
+
"step": 91140
|
| 215325 |
+
},
|
| 215326 |
+
{
|
| 215327 |
+
"epoch": 734.4,
|
| 215328 |
+
"learning_rate": 8.555945512820514e-06,
|
| 215329 |
+
"loss": 0.8591,
|
| 215330 |
+
"step": 91145
|
| 215331 |
+
},
|
| 215332 |
+
{
|
| 215333 |
+
"epoch": 734.44,
|
| 215334 |
+
"learning_rate": 8.555865384615384e-06,
|
| 215335 |
+
"loss": 0.7456,
|
| 215336 |
+
"step": 91150
|
| 215337 |
+
},
|
| 215338 |
+
{
|
| 215339 |
+
"epoch": 734.48,
|
| 215340 |
+
"learning_rate": 8.555785256410257e-06,
|
| 215341 |
+
"loss": 0.2987,
|
| 215342 |
+
"step": 91155
|
| 215343 |
+
},
|
| 215344 |
+
{
|
| 215345 |
+
"epoch": 734.52,
|
| 215346 |
+
"learning_rate": 8.555705128205129e-06,
|
| 215347 |
+
"loss": 0.3111,
|
| 215348 |
+
"step": 91160
|
| 215349 |
+
},
|
| 215350 |
+
{
|
| 215351 |
+
"epoch": 734.56,
|
| 215352 |
+
"learning_rate": 8.555625e-06,
|
| 215353 |
+
"loss": 0.4555,
|
| 215354 |
+
"step": 91165
|
| 215355 |
+
},
|
| 215356 |
+
{
|
| 215357 |
+
"epoch": 734.6,
|
| 215358 |
+
"learning_rate": 8.555544871794873e-06,
|
| 215359 |
+
"loss": 0.8859,
|
| 215360 |
+
"step": 91170
|
| 215361 |
+
},
|
| 215362 |
+
{
|
| 215363 |
+
"epoch": 734.64,
|
| 215364 |
+
"learning_rate": 8.555464743589744e-06,
|
| 215365 |
+
"loss": 0.6728,
|
| 215366 |
+
"step": 91175
|
| 215367 |
+
},
|
| 215368 |
+
{
|
| 215369 |
+
"epoch": 734.68,
|
| 215370 |
+
"learning_rate": 8.555384615384616e-06,
|
| 215371 |
+
"loss": 0.2605,
|
| 215372 |
+
"step": 91180
|
| 215373 |
+
},
|
| 215374 |
+
{
|
| 215375 |
+
"epoch": 734.72,
|
| 215376 |
+
"learning_rate": 8.555304487179487e-06,
|
| 215377 |
+
"loss": 0.3072,
|
| 215378 |
+
"step": 91185
|
| 215379 |
+
},
|
| 215380 |
+
{
|
| 215381 |
+
"epoch": 734.76,
|
| 215382 |
+
"learning_rate": 8.55522435897436e-06,
|
| 215383 |
+
"loss": 0.3961,
|
| 215384 |
+
"step": 91190
|
| 215385 |
+
},
|
| 215386 |
+
{
|
| 215387 |
+
"epoch": 734.8,
|
| 215388 |
+
"learning_rate": 8.555144230769231e-06,
|
| 215389 |
+
"loss": 0.8639,
|
| 215390 |
+
"step": 91195
|
| 215391 |
+
},
|
| 215392 |
+
{
|
| 215393 |
+
"epoch": 734.84,
|
| 215394 |
+
"learning_rate": 8.555064102564103e-06,
|
| 215395 |
+
"loss": 0.6852,
|
| 215396 |
+
"step": 91200
|
| 215397 |
+
},
|
| 215398 |
+
{
|
| 215399 |
+
"epoch": 734.88,
|
| 215400 |
+
"learning_rate": 8.554983974358974e-06,
|
| 215401 |
+
"loss": 0.3206,
|
| 215402 |
+
"step": 91205
|
| 215403 |
+
},
|
| 215404 |
+
{
|
| 215405 |
+
"epoch": 734.92,
|
| 215406 |
+
"learning_rate": 8.554903846153847e-06,
|
| 215407 |
+
"loss": 0.3195,
|
| 215408 |
+
"step": 91210
|
| 215409 |
+
},
|
| 215410 |
+
{
|
| 215411 |
+
"epoch": 734.96,
|
| 215412 |
+
"learning_rate": 8.554823717948719e-06,
|
| 215413 |
+
"loss": 0.4959,
|
| 215414 |
+
"step": 91215
|
| 215415 |
+
},
|
| 215416 |
+
{
|
| 215417 |
+
"epoch": 735.0,
|
| 215418 |
+
"learning_rate": 8.55474358974359e-06,
|
| 215419 |
+
"loss": 1.2116,
|
| 215420 |
+
"step": 91220
|
| 215421 |
+
},
|
| 215422 |
+
{
|
| 215423 |
+
"epoch": 735.0,
|
| 215424 |
+
"eval_loss": 0.448379248380661,
|
| 215425 |
+
"eval_runtime": 38.2149,
|
| 215426 |
+
"eval_samples_per_second": 22.007,
|
| 215427 |
+
"eval_steps_per_second": 0.707,
|
| 215428 |
+
"eval_wer": 0.1929014572211802,
|
| 215429 |
+
"step": 91220
|
| 215430 |
+
},
|
| 215431 |
+
{
|
| 215432 |
+
"epoch": 735.04,
|
| 215433 |
+
"learning_rate": 8.554663461538463e-06,
|
| 215434 |
+
"loss": 0.3294,
|
| 215435 |
+
"step": 91225
|
| 215436 |
+
},
|
| 215437 |
+
{
|
| 215438 |
+
"epoch": 735.08,
|
| 215439 |
+
"learning_rate": 8.554583333333334e-06,
|
| 215440 |
+
"loss": 0.2684,
|
| 215441 |
+
"step": 91230
|
| 215442 |
+
},
|
| 215443 |
+
{
|
| 215444 |
+
"epoch": 735.12,
|
| 215445 |
+
"learning_rate": 8.554503205128206e-06,
|
| 215446 |
+
"loss": 0.3815,
|
| 215447 |
+
"step": 91235
|
| 215448 |
+
},
|
| 215449 |
+
{
|
| 215450 |
+
"epoch": 735.16,
|
| 215451 |
+
"learning_rate": 8.554423076923077e-06,
|
| 215452 |
+
"loss": 0.5039,
|
| 215453 |
+
"step": 91240
|
| 215454 |
+
},
|
| 215455 |
+
{
|
| 215456 |
+
"epoch": 735.2,
|
| 215457 |
+
"learning_rate": 8.55434294871795e-06,
|
| 215458 |
+
"loss": 1.2851,
|
| 215459 |
+
"step": 91245
|
| 215460 |
+
},
|
| 215461 |
+
{
|
| 215462 |
+
"epoch": 735.24,
|
| 215463 |
+
"learning_rate": 8.554262820512821e-06,
|
| 215464 |
+
"loss": 0.2845,
|
| 215465 |
+
"step": 91250
|
| 215466 |
+
},
|
| 215467 |
+
{
|
| 215468 |
+
"epoch": 735.28,
|
| 215469 |
+
"learning_rate": 8.554182692307693e-06,
|
| 215470 |
+
"loss": 0.346,
|
| 215471 |
+
"step": 91255
|
| 215472 |
+
},
|
| 215473 |
+
{
|
| 215474 |
+
"epoch": 735.32,
|
| 215475 |
+
"learning_rate": 8.554102564102564e-06,
|
| 215476 |
+
"loss": 0.3353,
|
| 215477 |
+
"step": 91260
|
| 215478 |
+
},
|
| 215479 |
+
{
|
| 215480 |
+
"epoch": 735.36,
|
| 215481 |
+
"learning_rate": 8.554022435897437e-06,
|
| 215482 |
+
"loss": 0.4677,
|
| 215483 |
+
"step": 91265
|
| 215484 |
+
},
|
| 215485 |
+
{
|
| 215486 |
+
"epoch": 735.4,
|
| 215487 |
+
"learning_rate": 8.553942307692309e-06,
|
| 215488 |
+
"loss": 1.0487,
|
| 215489 |
+
"step": 91270
|
| 215490 |
+
},
|
| 215491 |
+
{
|
| 215492 |
+
"epoch": 735.44,
|
| 215493 |
+
"learning_rate": 8.55386217948718e-06,
|
| 215494 |
+
"loss": 0.339,
|
| 215495 |
+
"step": 91275
|
| 215496 |
+
},
|
| 215497 |
+
{
|
| 215498 |
+
"epoch": 735.48,
|
| 215499 |
+
"learning_rate": 8.553782051282053e-06,
|
| 215500 |
+
"loss": 0.2874,
|
| 215501 |
+
"step": 91280
|
| 215502 |
+
},
|
| 215503 |
+
{
|
| 215504 |
+
"epoch": 735.52,
|
| 215505 |
+
"learning_rate": 8.553701923076924e-06,
|
| 215506 |
+
"loss": 0.3102,
|
| 215507 |
+
"step": 91285
|
| 215508 |
+
},
|
| 215509 |
+
{
|
| 215510 |
+
"epoch": 735.56,
|
| 215511 |
+
"learning_rate": 8.553621794871796e-06,
|
| 215512 |
+
"loss": 0.4229,
|
| 215513 |
+
"step": 91290
|
| 215514 |
+
},
|
| 215515 |
+
{
|
| 215516 |
+
"epoch": 735.6,
|
| 215517 |
+
"learning_rate": 8.553541666666667e-06,
|
| 215518 |
+
"loss": 1.0708,
|
| 215519 |
+
"step": 91295
|
| 215520 |
+
},
|
| 215521 |
+
{
|
| 215522 |
+
"epoch": 735.64,
|
| 215523 |
+
"learning_rate": 8.55346153846154e-06,
|
| 215524 |
+
"loss": 0.3132,
|
| 215525 |
+
"step": 91300
|
| 215526 |
+
},
|
| 215527 |
+
{
|
| 215528 |
+
"epoch": 735.68,
|
| 215529 |
+
"learning_rate": 8.55338141025641e-06,
|
| 215530 |
+
"loss": 0.2999,
|
| 215531 |
+
"step": 91305
|
| 215532 |
+
},
|
| 215533 |
+
{
|
| 215534 |
+
"epoch": 735.72,
|
| 215535 |
+
"learning_rate": 8.553301282051283e-06,
|
| 215536 |
+
"loss": 0.3107,
|
| 215537 |
+
"step": 91310
|
| 215538 |
+
},
|
| 215539 |
+
{
|
| 215540 |
+
"epoch": 735.76,
|
| 215541 |
+
"learning_rate": 8.553221153846154e-06,
|
| 215542 |
+
"loss": 0.5146,
|
| 215543 |
+
"step": 91315
|
| 215544 |
+
},
|
| 215545 |
+
{
|
| 215546 |
+
"epoch": 735.8,
|
| 215547 |
+
"learning_rate": 8.553141025641026e-06,
|
| 215548 |
+
"loss": 1.2286,
|
| 215549 |
+
"step": 91320
|
| 215550 |
+
},
|
| 215551 |
+
{
|
| 215552 |
+
"epoch": 735.84,
|
| 215553 |
+
"learning_rate": 8.553060897435899e-06,
|
| 215554 |
+
"loss": 0.3302,
|
| 215555 |
+
"step": 91325
|
| 215556 |
+
},
|
| 215557 |
+
{
|
| 215558 |
+
"epoch": 735.88,
|
| 215559 |
+
"learning_rate": 8.55298076923077e-06,
|
| 215560 |
+
"loss": 0.4933,
|
| 215561 |
+
"step": 91330
|
| 215562 |
+
},
|
| 215563 |
+
{
|
| 215564 |
+
"epoch": 735.92,
|
| 215565 |
+
"learning_rate": 8.552900641025641e-06,
|
| 215566 |
+
"loss": 0.4702,
|
| 215567 |
+
"step": 91335
|
| 215568 |
+
},
|
| 215569 |
+
{
|
| 215570 |
+
"epoch": 735.96,
|
| 215571 |
+
"learning_rate": 8.552820512820513e-06,
|
| 215572 |
+
"loss": 0.4743,
|
| 215573 |
+
"step": 91340
|
| 215574 |
+
},
|
| 215575 |
+
{
|
| 215576 |
+
"epoch": 736.0,
|
| 215577 |
+
"eval_loss": 0.3884158432483673,
|
| 215578 |
+
"eval_runtime": 37.6587,
|
| 215579 |
+
"eval_samples_per_second": 22.332,
|
| 215580 |
+
"eval_steps_per_second": 0.717,
|
| 215581 |
+
"eval_wer": 0.1877105610077633,
|
| 215582 |
+
"step": 91344
|
| 215583 |
+
},
|
| 215584 |
+
{
|
| 215585 |
+
"epoch": 736.01,
|
| 215586 |
+
"learning_rate": 8.552740384615386e-06,
|
| 215587 |
+
"loss": 0.4299,
|
| 215588 |
+
"step": 91345
|
| 215589 |
+
},
|
| 215590 |
+
{
|
| 215591 |
+
"epoch": 736.05,
|
| 215592 |
+
"learning_rate": 8.552660256410257e-06,
|
| 215593 |
+
"loss": 0.3343,
|
| 215594 |
+
"step": 91350
|
| 215595 |
+
},
|
| 215596 |
+
{
|
| 215597 |
+
"epoch": 736.09,
|
| 215598 |
+
"learning_rate": 8.552580128205128e-06,
|
| 215599 |
+
"loss": 0.2703,
|
| 215600 |
+
"step": 91355
|
| 215601 |
+
},
|
| 215602 |
+
{
|
| 215603 |
+
"epoch": 736.13,
|
| 215604 |
+
"learning_rate": 8.5525e-06,
|
| 215605 |
+
"loss": 0.3447,
|
| 215606 |
+
"step": 91360
|
| 215607 |
+
},
|
| 215608 |
+
{
|
| 215609 |
+
"epoch": 736.17,
|
| 215610 |
+
"learning_rate": 8.552419871794873e-06,
|
| 215611 |
+
"loss": 0.5808,
|
| 215612 |
+
"step": 91365
|
| 215613 |
+
},
|
| 215614 |
+
{
|
| 215615 |
+
"epoch": 736.21,
|
| 215616 |
+
"learning_rate": 8.552339743589744e-06,
|
| 215617 |
+
"loss": 1.3182,
|
| 215618 |
+
"step": 91370
|
| 215619 |
+
},
|
| 215620 |
+
{
|
| 215621 |
+
"epoch": 736.25,
|
| 215622 |
+
"learning_rate": 8.552259615384616e-06,
|
| 215623 |
+
"loss": 0.3096,
|
| 215624 |
+
"step": 91375
|
| 215625 |
+
},
|
| 215626 |
+
{
|
| 215627 |
+
"epoch": 736.29,
|
| 215628 |
+
"learning_rate": 8.552179487179489e-06,
|
| 215629 |
+
"loss": 0.3141,
|
| 215630 |
+
"step": 91380
|
| 215631 |
+
},
|
| 215632 |
+
{
|
| 215633 |
+
"epoch": 736.33,
|
| 215634 |
+
"learning_rate": 8.55209935897436e-06,
|
| 215635 |
+
"loss": 0.2899,
|
| 215636 |
+
"step": 91385
|
| 215637 |
+
},
|
| 215638 |
+
{
|
| 215639 |
+
"epoch": 736.37,
|
| 215640 |
+
"learning_rate": 8.552019230769231e-06,
|
| 215641 |
+
"loss": 0.5201,
|
| 215642 |
+
"step": 91390
|
| 215643 |
+
},
|
| 215644 |
+
{
|
| 215645 |
+
"epoch": 736.41,
|
| 215646 |
+
"learning_rate": 8.551939102564103e-06,
|
| 215647 |
+
"loss": 1.0685,
|
| 215648 |
+
"step": 91395
|
| 215649 |
+
},
|
| 215650 |
+
{
|
| 215651 |
+
"epoch": 736.45,
|
| 215652 |
+
"learning_rate": 8.551858974358976e-06,
|
| 215653 |
+
"loss": 0.2799,
|
| 215654 |
+
"step": 91400
|
| 215655 |
+
},
|
| 215656 |
+
{
|
| 215657 |
+
"epoch": 736.49,
|
| 215658 |
+
"learning_rate": 8.551778846153847e-06,
|
| 215659 |
+
"loss": 0.3173,
|
| 215660 |
+
"step": 91405
|
| 215661 |
+
},
|
| 215662 |
+
{
|
| 215663 |
+
"epoch": 736.53,
|
| 215664 |
+
"learning_rate": 8.551698717948719e-06,
|
| 215665 |
+
"loss": 0.4231,
|
| 215666 |
+
"step": 91410
|
| 215667 |
+
},
|
| 215668 |
+
{
|
| 215669 |
+
"epoch": 736.57,
|
| 215670 |
+
"learning_rate": 8.55161858974359e-06,
|
| 215671 |
+
"loss": 0.4921,
|
| 215672 |
+
"step": 91415
|
| 215673 |
+
},
|
| 215674 |
+
{
|
| 215675 |
+
"epoch": 736.61,
|
| 215676 |
+
"learning_rate": 8.551538461538463e-06,
|
| 215677 |
+
"loss": 1.0186,
|
| 215678 |
+
"step": 91420
|
| 215679 |
+
},
|
| 215680 |
+
{
|
| 215681 |
+
"epoch": 736.65,
|
| 215682 |
+
"learning_rate": 8.551458333333334e-06,
|
| 215683 |
+
"loss": 0.3173,
|
| 215684 |
+
"step": 91425
|
| 215685 |
+
},
|
| 215686 |
+
{
|
| 215687 |
+
"epoch": 736.69,
|
| 215688 |
+
"learning_rate": 8.551378205128206e-06,
|
| 215689 |
+
"loss": 0.2881,
|
| 215690 |
+
"step": 91430
|
| 215691 |
+
},
|
| 215692 |
+
{
|
| 215693 |
+
"epoch": 736.73,
|
| 215694 |
+
"learning_rate": 8.551298076923079e-06,
|
| 215695 |
+
"loss": 0.2937,
|
| 215696 |
+
"step": 91435
|
| 215697 |
+
},
|
| 215698 |
+
{
|
| 215699 |
+
"epoch": 736.77,
|
| 215700 |
+
"learning_rate": 8.551217948717948e-06,
|
| 215701 |
+
"loss": 0.5759,
|
| 215702 |
+
"step": 91440
|
| 215703 |
+
},
|
| 215704 |
+
{
|
| 215705 |
+
"epoch": 736.81,
|
| 215706 |
+
"learning_rate": 8.551137820512821e-06,
|
| 215707 |
+
"loss": 1.1206,
|
| 215708 |
+
"step": 91445
|
| 215709 |
+
},
|
| 215710 |
+
{
|
| 215711 |
+
"epoch": 736.85,
|
| 215712 |
+
"learning_rate": 8.551057692307693e-06,
|
| 215713 |
+
"loss": 0.2569,
|
| 215714 |
+
"step": 91450
|
| 215715 |
+
},
|
| 215716 |
+
{
|
| 215717 |
+
"epoch": 736.89,
|
| 215718 |
+
"learning_rate": 8.550977564102564e-06,
|
| 215719 |
+
"loss": 0.3088,
|
| 215720 |
+
"step": 91455
|
| 215721 |
+
},
|
| 215722 |
+
{
|
| 215723 |
+
"epoch": 736.93,
|
| 215724 |
+
"learning_rate": 8.550897435897435e-06,
|
| 215725 |
+
"loss": 0.3157,
|
| 215726 |
+
"step": 91460
|
| 215727 |
+
},
|
| 215728 |
+
{
|
| 215729 |
+
"epoch": 736.97,
|
| 215730 |
+
"learning_rate": 8.550817307692309e-06,
|
| 215731 |
+
"loss": 0.5843,
|
| 215732 |
+
"step": 91465
|
| 215733 |
+
},
|
| 215734 |
+
{
|
| 215735 |
+
"epoch": 737.0,
|
| 215736 |
+
"eval_loss": 0.39843711256980896,
|
| 215737 |
+
"eval_runtime": 39.5694,
|
| 215738 |
+
"eval_samples_per_second": 21.254,
|
| 215739 |
+
"eval_steps_per_second": 0.682,
|
| 215740 |
+
"eval_wer": 0.18542114230683754,
|
| 215741 |
+
"step": 91468
|
| 215742 |
}
|
| 215743 |
],
|
| 215744 |
"max_steps": 620000,
|
| 215745 |
"num_train_epochs": 5000,
|
| 215746 |
+
"total_flos": 2.5740082672739317e+20,
|
| 215747 |
"trial_name": null,
|
| 215748 |
"trial_params": null
|
| 215749 |
}
|
model-bin/finetune/base/{checkpoint-90847 → checkpoint-91468}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1629948643.035603/events.out.tfevents.1629948643.8e89bd551565.924.41
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c10b837c215b3881a88bbe1e65db93660304880c10fe785f3c6fdb4d5baa180f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629949081.6864243/events.out.tfevents.1629949081.8e89bd551565.924.43
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f90e32a19713236eec52dd7eb7e78e82257c08591aed45070958046eab8c8ee9
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629949522.6644921/events.out.tfevents.1629949522.8e89bd551565.924.45
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:818ab456b5923720df646c860a3147aa0edcd8d70d352c8b5c0f874b6c20a8d1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629949956.9957566/events.out.tfevents.1629949956.8e89bd551565.924.47
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f52ca1bb5cb4ca44922ed7b1e444f34715844eccdfcb6a63a06e9baa008a5e44
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629950393.869526/events.out.tfevents.1629950393.8e89bd551565.924.49
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:acf46d137ee19dbe958b4f79ca11df54ecb26cca95a43078a24da168e3b8629c
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629948643.8e89bd551565.924.40
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4297b531fbd899ddab11f37530e357aaf403c68be268ed14aa67530465cc57f0
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629949081.8e89bd551565.924.42
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f60b1588d7d2ced08cacfca1a51608ef53fbdd915dab82f3614a389389e0f441
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629949521.8e89bd551565.924.44
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a48fe200c805dab368c956d23897e19a48f4b2d73b452f141067be5300f56572
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629949955.8e89bd551565.924.46
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9035381bd655c0c12795012eb0f323ba4c4ecba7c5701702aa379fe45d11a860
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629950393.8e89bd551565.924.48
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6ccbab421519432e5d9df4f8b1529971136d8f5cd758feaba5a761feb4dfa9d1
|
| 3 |
+
size 8622
|