"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/trainer_state.json +798 -3
- model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630002482.2222688/events.out.tfevents.1630002482.8e89bd551565.924.281 +3 -0
- model-bin/finetune/base/log/1630002949.6500714/events.out.tfevents.1630002949.8e89bd551565.924.283 +3 -0
- model-bin/finetune/base/log/1630003391.3419247/events.out.tfevents.1630003391.8e89bd551565.924.285 +3 -0
- model-bin/finetune/base/log/1630003815.897631/events.out.tfevents.1630003816.8e89bd551565.924.287 +3 -0
- model-bin/finetune/base/log/1630004360.036647/events.out.tfevents.1630004360.8e89bd551565.924.289 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630002482.8e89bd551565.924.280 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630002948.8e89bd551565.924.282 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630003391.8e89bd551565.924.284 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630003815.8e89bd551565.924.286 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630004360.8e89bd551565.924.288 +3 -0
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/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:bd3b8419f60cc523b8679f4d0f4fa7b2952f84551c34f88d74d8fb08ee22b5e2
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/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 sha256:066691996eec347f84965d856d1ef54b47c0350dce33c505ba85facb5cc1e722
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/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 sha256:94c031a869815035c710064fb93d4483a65e88563c7147112c8644a5f33d9957
|
| 3 |
+
size 14439
|
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/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:a5340902b17e35966e7e8d13c891b4a72edbb8cbd01bf429032179aa2bf58052
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/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 sha256:8082a046cd9494cefba29cc26ccb1eb432ce0cc10aa49274b249f1c256896f99
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/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,
|
|
@@ -233952,11 +233952,806 @@
|
|
| 233952 |
"eval_steps_per_second": 0.69,
|
| 233953 |
"eval_wer": 0.18124235226372992,
|
| 233954 |
"step": 105783
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 233955 |
}
|
| 233956 |
],
|
| 233957 |
"max_steps": 625000,
|
| 233958 |
"num_train_epochs": 5000,
|
| 233959 |
-
"total_flos": 2.
|
| 233960 |
"trial_name": null,
|
| 233961 |
"trial_params": null
|
| 233962 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 851.0,
|
| 5 |
+
"global_step": 106405,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 233952 |
"eval_steps_per_second": 0.69,
|
| 233953 |
"eval_wer": 0.18124235226372992,
|
| 233954 |
"step": 105783
|
| 233955 |
+
},
|
| 233956 |
+
{
|
| 233957 |
+
"epoch": 853.02,
|
| 233958 |
+
"learning_rate": 8.321410256410257e-06,
|
| 233959 |
+
"loss": 0.3904,
|
| 233960 |
+
"step": 105785
|
| 233961 |
+
},
|
| 233962 |
+
{
|
| 233963 |
+
"epoch": 853.06,
|
| 233964 |
+
"learning_rate": 8.321330128205128e-06,
|
| 233965 |
+
"loss": 0.2852,
|
| 233966 |
+
"step": 105790
|
| 233967 |
+
},
|
| 233968 |
+
{
|
| 233969 |
+
"epoch": 853.1,
|
| 233970 |
+
"learning_rate": 8.321250000000001e-06,
|
| 233971 |
+
"loss": 0.3459,
|
| 233972 |
+
"step": 105795
|
| 233973 |
+
},
|
| 233974 |
+
{
|
| 233975 |
+
"epoch": 853.14,
|
| 233976 |
+
"learning_rate": 8.321169871794872e-06,
|
| 233977 |
+
"loss": 0.3696,
|
| 233978 |
+
"step": 105800
|
| 233979 |
+
},
|
| 233980 |
+
{
|
| 233981 |
+
"epoch": 853.18,
|
| 233982 |
+
"learning_rate": 8.321089743589744e-06,
|
| 233983 |
+
"loss": 0.6374,
|
| 233984 |
+
"step": 105805
|
| 233985 |
+
},
|
| 233986 |
+
{
|
| 233987 |
+
"epoch": 853.22,
|
| 233988 |
+
"learning_rate": 8.321009615384617e-06,
|
| 233989 |
+
"loss": 1.0277,
|
| 233990 |
+
"step": 105810
|
| 233991 |
+
},
|
| 233992 |
+
{
|
| 233993 |
+
"epoch": 853.26,
|
| 233994 |
+
"learning_rate": 8.320929487179488e-06,
|
| 233995 |
+
"loss": 0.3343,
|
| 233996 |
+
"step": 105815
|
| 233997 |
+
},
|
| 233998 |
+
{
|
| 233999 |
+
"epoch": 853.3,
|
| 234000 |
+
"learning_rate": 8.32084935897436e-06,
|
| 234001 |
+
"loss": 0.2998,
|
| 234002 |
+
"step": 105820
|
| 234003 |
+
},
|
| 234004 |
+
{
|
| 234005 |
+
"epoch": 853.34,
|
| 234006 |
+
"learning_rate": 8.320769230769231e-06,
|
| 234007 |
+
"loss": 0.3152,
|
| 234008 |
+
"step": 105825
|
| 234009 |
+
},
|
| 234010 |
+
{
|
| 234011 |
+
"epoch": 853.38,
|
| 234012 |
+
"learning_rate": 8.320689102564104e-06,
|
| 234013 |
+
"loss": 0.6423,
|
| 234014 |
+
"step": 105830
|
| 234015 |
+
},
|
| 234016 |
+
{
|
| 234017 |
+
"epoch": 853.42,
|
| 234018 |
+
"learning_rate": 8.320608974358975e-06,
|
| 234019 |
+
"loss": 0.8821,
|
| 234020 |
+
"step": 105835
|
| 234021 |
+
},
|
| 234022 |
+
{
|
| 234023 |
+
"epoch": 853.46,
|
| 234024 |
+
"learning_rate": 8.320528846153847e-06,
|
| 234025 |
+
"loss": 0.2717,
|
| 234026 |
+
"step": 105840
|
| 234027 |
+
},
|
| 234028 |
+
{
|
| 234029 |
+
"epoch": 853.5,
|
| 234030 |
+
"learning_rate": 8.320448717948718e-06,
|
| 234031 |
+
"loss": 0.314,
|
| 234032 |
+
"step": 105845
|
| 234033 |
+
},
|
| 234034 |
+
{
|
| 234035 |
+
"epoch": 853.54,
|
| 234036 |
+
"learning_rate": 8.320368589743591e-06,
|
| 234037 |
+
"loss": 0.384,
|
| 234038 |
+
"step": 105850
|
| 234039 |
+
},
|
| 234040 |
+
{
|
| 234041 |
+
"epoch": 853.58,
|
| 234042 |
+
"learning_rate": 8.320288461538462e-06,
|
| 234043 |
+
"loss": 0.5992,
|
| 234044 |
+
"step": 105855
|
| 234045 |
+
},
|
| 234046 |
+
{
|
| 234047 |
+
"epoch": 853.62,
|
| 234048 |
+
"learning_rate": 8.320208333333334e-06,
|
| 234049 |
+
"loss": 1.0864,
|
| 234050 |
+
"step": 105860
|
| 234051 |
+
},
|
| 234052 |
+
{
|
| 234053 |
+
"epoch": 853.66,
|
| 234054 |
+
"learning_rate": 8.320128205128207e-06,
|
| 234055 |
+
"loss": 0.2521,
|
| 234056 |
+
"step": 105865
|
| 234057 |
+
},
|
| 234058 |
+
{
|
| 234059 |
+
"epoch": 853.7,
|
| 234060 |
+
"learning_rate": 8.320048076923078e-06,
|
| 234061 |
+
"loss": 0.2983,
|
| 234062 |
+
"step": 105870
|
| 234063 |
+
},
|
| 234064 |
+
{
|
| 234065 |
+
"epoch": 853.74,
|
| 234066 |
+
"learning_rate": 8.31996794871795e-06,
|
| 234067 |
+
"loss": 0.3174,
|
| 234068 |
+
"step": 105875
|
| 234069 |
+
},
|
| 234070 |
+
{
|
| 234071 |
+
"epoch": 853.78,
|
| 234072 |
+
"learning_rate": 8.319887820512821e-06,
|
| 234073 |
+
"loss": 0.6625,
|
| 234074 |
+
"step": 105880
|
| 234075 |
+
},
|
| 234076 |
+
{
|
| 234077 |
+
"epoch": 853.82,
|
| 234078 |
+
"learning_rate": 8.319807692307694e-06,
|
| 234079 |
+
"loss": 1.014,
|
| 234080 |
+
"step": 105885
|
| 234081 |
+
},
|
| 234082 |
+
{
|
| 234083 |
+
"epoch": 853.86,
|
| 234084 |
+
"learning_rate": 8.319727564102564e-06,
|
| 234085 |
+
"loss": 0.3064,
|
| 234086 |
+
"step": 105890
|
| 234087 |
+
},
|
| 234088 |
+
{
|
| 234089 |
+
"epoch": 853.9,
|
| 234090 |
+
"learning_rate": 8.319647435897437e-06,
|
| 234091 |
+
"loss": 0.3064,
|
| 234092 |
+
"step": 105895
|
| 234093 |
+
},
|
| 234094 |
+
{
|
| 234095 |
+
"epoch": 853.94,
|
| 234096 |
+
"learning_rate": 8.31956730769231e-06,
|
| 234097 |
+
"loss": 0.3304,
|
| 234098 |
+
"step": 105900
|
| 234099 |
+
},
|
| 234100 |
+
{
|
| 234101 |
+
"epoch": 853.98,
|
| 234102 |
+
"learning_rate": 8.31948717948718e-06,
|
| 234103 |
+
"loss": 0.7452,
|
| 234104 |
+
"step": 105905
|
| 234105 |
+
},
|
| 234106 |
+
{
|
| 234107 |
+
"epoch": 854.0,
|
| 234108 |
+
"eval_loss": 0.40324002504348755,
|
| 234109 |
+
"eval_runtime": 40.2729,
|
| 234110 |
+
"eval_samples_per_second": 20.808,
|
| 234111 |
+
"eval_steps_per_second": 0.67,
|
| 234112 |
+
"eval_wer": 0.19164960514770402,
|
| 234113 |
+
"step": 105907
|
| 234114 |
+
},
|
| 234115 |
+
{
|
| 234116 |
+
"epoch": 847.02,
|
| 234117 |
+
"learning_rate": 8.319407051282052e-06,
|
| 234118 |
+
"loss": 0.369,
|
| 234119 |
+
"step": 105910
|
| 234120 |
+
},
|
| 234121 |
+
{
|
| 234122 |
+
"epoch": 847.06,
|
| 234123 |
+
"learning_rate": 8.319326923076924e-06,
|
| 234124 |
+
"loss": 0.2924,
|
| 234125 |
+
"step": 105915
|
| 234126 |
+
},
|
| 234127 |
+
{
|
| 234128 |
+
"epoch": 847.1,
|
| 234129 |
+
"learning_rate": 8.319246794871795e-06,
|
| 234130 |
+
"loss": 0.3036,
|
| 234131 |
+
"step": 105920
|
| 234132 |
+
},
|
| 234133 |
+
{
|
| 234134 |
+
"epoch": 847.14,
|
| 234135 |
+
"learning_rate": 8.319166666666666e-06,
|
| 234136 |
+
"loss": 0.3875,
|
| 234137 |
+
"step": 105925
|
| 234138 |
+
},
|
| 234139 |
+
{
|
| 234140 |
+
"epoch": 847.18,
|
| 234141 |
+
"learning_rate": 8.31908653846154e-06,
|
| 234142 |
+
"loss": 0.7866,
|
| 234143 |
+
"step": 105930
|
| 234144 |
+
},
|
| 234145 |
+
{
|
| 234146 |
+
"epoch": 847.22,
|
| 234147 |
+
"learning_rate": 8.319006410256411e-06,
|
| 234148 |
+
"loss": 0.807,
|
| 234149 |
+
"step": 105935
|
| 234150 |
+
},
|
| 234151 |
+
{
|
| 234152 |
+
"epoch": 847.26,
|
| 234153 |
+
"learning_rate": 8.318926282051282e-06,
|
| 234154 |
+
"loss": 0.4785,
|
| 234155 |
+
"step": 105940
|
| 234156 |
+
},
|
| 234157 |
+
{
|
| 234158 |
+
"epoch": 847.3,
|
| 234159 |
+
"learning_rate": 8.318846153846154e-06,
|
| 234160 |
+
"loss": 0.3017,
|
| 234161 |
+
"step": 105945
|
| 234162 |
+
},
|
| 234163 |
+
{
|
| 234164 |
+
"epoch": 847.34,
|
| 234165 |
+
"learning_rate": 8.318766025641027e-06,
|
| 234166 |
+
"loss": 0.3452,
|
| 234167 |
+
"step": 105950
|
| 234168 |
+
},
|
| 234169 |
+
{
|
| 234170 |
+
"epoch": 847.38,
|
| 234171 |
+
"learning_rate": 8.318685897435898e-06,
|
| 234172 |
+
"loss": 0.7524,
|
| 234173 |
+
"step": 105955
|
| 234174 |
+
},
|
| 234175 |
+
{
|
| 234176 |
+
"epoch": 847.42,
|
| 234177 |
+
"learning_rate": 8.31860576923077e-06,
|
| 234178 |
+
"loss": 0.7654,
|
| 234179 |
+
"step": 105960
|
| 234180 |
+
},
|
| 234181 |
+
{
|
| 234182 |
+
"epoch": 847.46,
|
| 234183 |
+
"learning_rate": 8.318525641025642e-06,
|
| 234184 |
+
"loss": 0.2805,
|
| 234185 |
+
"step": 105965
|
| 234186 |
+
},
|
| 234187 |
+
{
|
| 234188 |
+
"epoch": 847.5,
|
| 234189 |
+
"learning_rate": 8.318445512820514e-06,
|
| 234190 |
+
"loss": 0.3126,
|
| 234191 |
+
"step": 105970
|
| 234192 |
+
},
|
| 234193 |
+
{
|
| 234194 |
+
"epoch": 847.54,
|
| 234195 |
+
"learning_rate": 8.318365384615385e-06,
|
| 234196 |
+
"loss": 0.3958,
|
| 234197 |
+
"step": 105975
|
| 234198 |
+
},
|
| 234199 |
+
{
|
| 234200 |
+
"epoch": 847.58,
|
| 234201 |
+
"learning_rate": 8.318285256410257e-06,
|
| 234202 |
+
"loss": 0.8039,
|
| 234203 |
+
"step": 105980
|
| 234204 |
+
},
|
| 234205 |
+
{
|
| 234206 |
+
"epoch": 847.62,
|
| 234207 |
+
"learning_rate": 8.31820512820513e-06,
|
| 234208 |
+
"loss": 0.9644,
|
| 234209 |
+
"step": 105985
|
| 234210 |
+
},
|
| 234211 |
+
{
|
| 234212 |
+
"epoch": 847.66,
|
| 234213 |
+
"learning_rate": 8.318125000000001e-06,
|
| 234214 |
+
"loss": 0.2883,
|
| 234215 |
+
"step": 105990
|
| 234216 |
+
},
|
| 234217 |
+
{
|
| 234218 |
+
"epoch": 847.7,
|
| 234219 |
+
"learning_rate": 8.318044871794872e-06,
|
| 234220 |
+
"loss": 0.3089,
|
| 234221 |
+
"step": 105995
|
| 234222 |
+
},
|
| 234223 |
+
{
|
| 234224 |
+
"epoch": 847.74,
|
| 234225 |
+
"learning_rate": 8.317964743589745e-06,
|
| 234226 |
+
"loss": 0.3553,
|
| 234227 |
+
"step": 106000
|
| 234228 |
+
},
|
| 234229 |
+
{
|
| 234230 |
+
"epoch": 847.78,
|
| 234231 |
+
"learning_rate": 8.317884615384617e-06,
|
| 234232 |
+
"loss": 0.7987,
|
| 234233 |
+
"step": 106005
|
| 234234 |
+
},
|
| 234235 |
+
{
|
| 234236 |
+
"epoch": 847.82,
|
| 234237 |
+
"learning_rate": 8.317804487179488e-06,
|
| 234238 |
+
"loss": 0.917,
|
| 234239 |
+
"step": 106010
|
| 234240 |
+
},
|
| 234241 |
+
{
|
| 234242 |
+
"epoch": 847.86,
|
| 234243 |
+
"learning_rate": 8.31772435897436e-06,
|
| 234244 |
+
"loss": 0.2992,
|
| 234245 |
+
"step": 106015
|
| 234246 |
+
},
|
| 234247 |
+
{
|
| 234248 |
+
"epoch": 847.9,
|
| 234249 |
+
"learning_rate": 8.317644230769232e-06,
|
| 234250 |
+
"loss": 0.2879,
|
| 234251 |
+
"step": 106020
|
| 234252 |
+
},
|
| 234253 |
+
{
|
| 234254 |
+
"epoch": 847.94,
|
| 234255 |
+
"learning_rate": 8.317564102564102e-06,
|
| 234256 |
+
"loss": 0.4038,
|
| 234257 |
+
"step": 106025
|
| 234258 |
+
},
|
| 234259 |
+
{
|
| 234260 |
+
"epoch": 847.98,
|
| 234261 |
+
"learning_rate": 8.317483974358975e-06,
|
| 234262 |
+
"loss": 0.8434,
|
| 234263 |
+
"step": 106030
|
| 234264 |
+
},
|
| 234265 |
+
{
|
| 234266 |
+
"epoch": 848.0,
|
| 234267 |
+
"eval_loss": 0.35842400789260864,
|
| 234268 |
+
"eval_runtime": 38.2763,
|
| 234269 |
+
"eval_samples_per_second": 21.893,
|
| 234270 |
+
"eval_steps_per_second": 0.705,
|
| 234271 |
+
"eval_wer": 0.18294100552165068,
|
| 234272 |
+
"step": 106032
|
| 234273 |
+
},
|
| 234274 |
+
{
|
| 234275 |
+
"epoch": 855.02,
|
| 234276 |
+
"learning_rate": 8.317403846153847e-06,
|
| 234277 |
+
"loss": 0.3319,
|
| 234278 |
+
"step": 106035
|
| 234279 |
+
},
|
| 234280 |
+
{
|
| 234281 |
+
"epoch": 855.06,
|
| 234282 |
+
"learning_rate": 8.317323717948718e-06,
|
| 234283 |
+
"loss": 0.2666,
|
| 234284 |
+
"step": 106040
|
| 234285 |
+
},
|
| 234286 |
+
{
|
| 234287 |
+
"epoch": 855.1,
|
| 234288 |
+
"learning_rate": 8.31724358974359e-06,
|
| 234289 |
+
"loss": 0.2633,
|
| 234290 |
+
"step": 106045
|
| 234291 |
+
},
|
| 234292 |
+
{
|
| 234293 |
+
"epoch": 855.14,
|
| 234294 |
+
"learning_rate": 8.317163461538462e-06,
|
| 234295 |
+
"loss": 0.4081,
|
| 234296 |
+
"step": 106050
|
| 234297 |
+
},
|
| 234298 |
+
{
|
| 234299 |
+
"epoch": 855.18,
|
| 234300 |
+
"learning_rate": 8.317083333333334e-06,
|
| 234301 |
+
"loss": 0.7049,
|
| 234302 |
+
"step": 106055
|
| 234303 |
+
},
|
| 234304 |
+
{
|
| 234305 |
+
"epoch": 855.22,
|
| 234306 |
+
"learning_rate": 8.317003205128205e-06,
|
| 234307 |
+
"loss": 1.0077,
|
| 234308 |
+
"step": 106060
|
| 234309 |
+
},
|
| 234310 |
+
{
|
| 234311 |
+
"epoch": 855.27,
|
| 234312 |
+
"learning_rate": 8.316923076923078e-06,
|
| 234313 |
+
"loss": 0.3512,
|
| 234314 |
+
"step": 106065
|
| 234315 |
+
},
|
| 234316 |
+
{
|
| 234317 |
+
"epoch": 855.31,
|
| 234318 |
+
"learning_rate": 8.31684294871795e-06,
|
| 234319 |
+
"loss": 0.3082,
|
| 234320 |
+
"step": 106070
|
| 234321 |
+
},
|
| 234322 |
+
{
|
| 234323 |
+
"epoch": 855.35,
|
| 234324 |
+
"learning_rate": 8.31676282051282e-06,
|
| 234325 |
+
"loss": 0.3648,
|
| 234326 |
+
"step": 106075
|
| 234327 |
+
},
|
| 234328 |
+
{
|
| 234329 |
+
"epoch": 855.39,
|
| 234330 |
+
"learning_rate": 8.316682692307692e-06,
|
| 234331 |
+
"loss": 0.8656,
|
| 234332 |
+
"step": 106080
|
| 234333 |
+
},
|
| 234334 |
+
{
|
| 234335 |
+
"epoch": 855.43,
|
| 234336 |
+
"learning_rate": 8.316602564102565e-06,
|
| 234337 |
+
"loss": 0.8436,
|
| 234338 |
+
"step": 106085
|
| 234339 |
+
},
|
| 234340 |
+
{
|
| 234341 |
+
"epoch": 855.47,
|
| 234342 |
+
"learning_rate": 8.316522435897437e-06,
|
| 234343 |
+
"loss": 0.3225,
|
| 234344 |
+
"step": 106090
|
| 234345 |
+
},
|
| 234346 |
+
{
|
| 234347 |
+
"epoch": 855.51,
|
| 234348 |
+
"learning_rate": 8.316442307692308e-06,
|
| 234349 |
+
"loss": 0.4106,
|
| 234350 |
+
"step": 106095
|
| 234351 |
+
},
|
| 234352 |
+
{
|
| 234353 |
+
"epoch": 855.55,
|
| 234354 |
+
"learning_rate": 8.316362179487181e-06,
|
| 234355 |
+
"loss": 0.3349,
|
| 234356 |
+
"step": 106100
|
| 234357 |
+
},
|
| 234358 |
+
{
|
| 234359 |
+
"epoch": 855.59,
|
| 234360 |
+
"learning_rate": 8.316282051282052e-06,
|
| 234361 |
+
"loss": 0.8743,
|
| 234362 |
+
"step": 106105
|
| 234363 |
+
},
|
| 234364 |
+
{
|
| 234365 |
+
"epoch": 855.63,
|
| 234366 |
+
"learning_rate": 8.316201923076924e-06,
|
| 234367 |
+
"loss": 0.881,
|
| 234368 |
+
"step": 106110
|
| 234369 |
+
},
|
| 234370 |
+
{
|
| 234371 |
+
"epoch": 855.67,
|
| 234372 |
+
"learning_rate": 8.316121794871795e-06,
|
| 234373 |
+
"loss": 0.3457,
|
| 234374 |
+
"step": 106115
|
| 234375 |
+
},
|
| 234376 |
+
{
|
| 234377 |
+
"epoch": 855.71,
|
| 234378 |
+
"learning_rate": 8.316041666666668e-06,
|
| 234379 |
+
"loss": 0.2748,
|
| 234380 |
+
"step": 106120
|
| 234381 |
+
},
|
| 234382 |
+
{
|
| 234383 |
+
"epoch": 855.75,
|
| 234384 |
+
"learning_rate": 8.31596153846154e-06,
|
| 234385 |
+
"loss": 0.3406,
|
| 234386 |
+
"step": 106125
|
| 234387 |
+
},
|
| 234388 |
+
{
|
| 234389 |
+
"epoch": 855.79,
|
| 234390 |
+
"learning_rate": 8.31588141025641e-06,
|
| 234391 |
+
"loss": 0.6802,
|
| 234392 |
+
"step": 106130
|
| 234393 |
+
},
|
| 234394 |
+
{
|
| 234395 |
+
"epoch": 855.83,
|
| 234396 |
+
"learning_rate": 8.315801282051282e-06,
|
| 234397 |
+
"loss": 0.8238,
|
| 234398 |
+
"step": 106135
|
| 234399 |
+
},
|
| 234400 |
+
{
|
| 234401 |
+
"epoch": 855.87,
|
| 234402 |
+
"learning_rate": 8.315721153846155e-06,
|
| 234403 |
+
"loss": 0.3085,
|
| 234404 |
+
"step": 106140
|
| 234405 |
+
},
|
| 234406 |
+
{
|
| 234407 |
+
"epoch": 855.91,
|
| 234408 |
+
"learning_rate": 8.315641025641027e-06,
|
| 234409 |
+
"loss": 0.3074,
|
| 234410 |
+
"step": 106145
|
| 234411 |
+
},
|
| 234412 |
+
{
|
| 234413 |
+
"epoch": 855.95,
|
| 234414 |
+
"learning_rate": 8.315560897435898e-06,
|
| 234415 |
+
"loss": 0.3876,
|
| 234416 |
+
"step": 106150
|
| 234417 |
+
},
|
| 234418 |
+
{
|
| 234419 |
+
"epoch": 855.99,
|
| 234420 |
+
"learning_rate": 8.315480769230771e-06,
|
| 234421 |
+
"loss": 0.8758,
|
| 234422 |
+
"step": 106155
|
| 234423 |
+
},
|
| 234424 |
+
{
|
| 234425 |
+
"epoch": 856.0,
|
| 234426 |
+
"eval_loss": 0.36880195140838623,
|
| 234427 |
+
"eval_runtime": 39.3444,
|
| 234428 |
+
"eval_samples_per_second": 21.299,
|
| 234429 |
+
"eval_steps_per_second": 0.686,
|
| 234430 |
+
"eval_wer": 0.18161304411122434,
|
| 234431 |
+
"step": 106156
|
| 234432 |
+
},
|
| 234433 |
+
{
|
| 234434 |
+
"epoch": 856.03,
|
| 234435 |
+
"learning_rate": 8.315400641025642e-06,
|
| 234436 |
+
"loss": 0.3303,
|
| 234437 |
+
"step": 106160
|
| 234438 |
+
},
|
| 234439 |
+
{
|
| 234440 |
+
"epoch": 856.07,
|
| 234441 |
+
"learning_rate": 8.315320512820514e-06,
|
| 234442 |
+
"loss": 0.3295,
|
| 234443 |
+
"step": 106165
|
| 234444 |
+
},
|
| 234445 |
+
{
|
| 234446 |
+
"epoch": 856.11,
|
| 234447 |
+
"learning_rate": 8.315240384615385e-06,
|
| 234448 |
+
"loss": 0.3238,
|
| 234449 |
+
"step": 106170
|
| 234450 |
+
},
|
| 234451 |
+
{
|
| 234452 |
+
"epoch": 856.15,
|
| 234453 |
+
"learning_rate": 8.315160256410258e-06,
|
| 234454 |
+
"loss": 0.4173,
|
| 234455 |
+
"step": 106175
|
| 234456 |
+
},
|
| 234457 |
+
{
|
| 234458 |
+
"epoch": 856.19,
|
| 234459 |
+
"learning_rate": 8.315080128205128e-06,
|
| 234460 |
+
"loss": 0.8644,
|
| 234461 |
+
"step": 106180
|
| 234462 |
+
},
|
| 234463 |
+
{
|
| 234464 |
+
"epoch": 856.23,
|
| 234465 |
+
"learning_rate": 8.315000000000001e-06,
|
| 234466 |
+
"loss": 0.7193,
|
| 234467 |
+
"step": 106185
|
| 234468 |
+
},
|
| 234469 |
+
{
|
| 234470 |
+
"epoch": 856.27,
|
| 234471 |
+
"learning_rate": 8.314919871794872e-06,
|
| 234472 |
+
"loss": 0.3008,
|
| 234473 |
+
"step": 106190
|
| 234474 |
+
},
|
| 234475 |
+
{
|
| 234476 |
+
"epoch": 856.31,
|
| 234477 |
+
"learning_rate": 8.314839743589744e-06,
|
| 234478 |
+
"loss": 0.3143,
|
| 234479 |
+
"step": 106195
|
| 234480 |
+
},
|
| 234481 |
+
{
|
| 234482 |
+
"epoch": 856.35,
|
| 234483 |
+
"learning_rate": 8.314759615384615e-06,
|
| 234484 |
+
"loss": 0.3981,
|
| 234485 |
+
"step": 106200
|
| 234486 |
+
},
|
| 234487 |
+
{
|
| 234488 |
+
"epoch": 856.39,
|
| 234489 |
+
"learning_rate": 8.314679487179488e-06,
|
| 234490 |
+
"loss": 0.8131,
|
| 234491 |
+
"step": 106205
|
| 234492 |
+
},
|
| 234493 |
+
{
|
| 234494 |
+
"epoch": 856.43,
|
| 234495 |
+
"learning_rate": 8.31459935897436e-06,
|
| 234496 |
+
"loss": 0.6746,
|
| 234497 |
+
"step": 106210
|
| 234498 |
+
},
|
| 234499 |
+
{
|
| 234500 |
+
"epoch": 856.47,
|
| 234501 |
+
"learning_rate": 8.31451923076923e-06,
|
| 234502 |
+
"loss": 0.2869,
|
| 234503 |
+
"step": 106215
|
| 234504 |
+
},
|
| 234505 |
+
{
|
| 234506 |
+
"epoch": 856.51,
|
| 234507 |
+
"learning_rate": 8.314439102564104e-06,
|
| 234508 |
+
"loss": 0.2746,
|
| 234509 |
+
"step": 106220
|
| 234510 |
+
},
|
| 234511 |
+
{
|
| 234512 |
+
"epoch": 856.55,
|
| 234513 |
+
"learning_rate": 8.314358974358975e-06,
|
| 234514 |
+
"loss": 0.3885,
|
| 234515 |
+
"step": 106225
|
| 234516 |
+
},
|
| 234517 |
+
{
|
| 234518 |
+
"epoch": 856.59,
|
| 234519 |
+
"learning_rate": 8.314278846153846e-06,
|
| 234520 |
+
"loss": 0.887,
|
| 234521 |
+
"step": 106230
|
| 234522 |
+
},
|
| 234523 |
+
{
|
| 234524 |
+
"epoch": 856.63,
|
| 234525 |
+
"learning_rate": 8.314198717948718e-06,
|
| 234526 |
+
"loss": 0.6237,
|
| 234527 |
+
"step": 106235
|
| 234528 |
+
},
|
| 234529 |
+
{
|
| 234530 |
+
"epoch": 856.67,
|
| 234531 |
+
"learning_rate": 8.314118589743591e-06,
|
| 234532 |
+
"loss": 0.2606,
|
| 234533 |
+
"step": 106240
|
| 234534 |
+
},
|
| 234535 |
+
{
|
| 234536 |
+
"epoch": 856.71,
|
| 234537 |
+
"learning_rate": 8.314038461538462e-06,
|
| 234538 |
+
"loss": 0.311,
|
| 234539 |
+
"step": 106245
|
| 234540 |
+
},
|
| 234541 |
+
{
|
| 234542 |
+
"epoch": 856.76,
|
| 234543 |
+
"learning_rate": 8.313958333333334e-06,
|
| 234544 |
+
"loss": 0.368,
|
| 234545 |
+
"step": 106250
|
| 234546 |
+
},
|
| 234547 |
+
{
|
| 234548 |
+
"epoch": 856.8,
|
| 234549 |
+
"learning_rate": 8.313894230769231e-06,
|
| 234550 |
+
"loss": 1.11,
|
| 234551 |
+
"step": 106255
|
| 234552 |
+
},
|
| 234553 |
+
{
|
| 234554 |
+
"epoch": 856.84,
|
| 234555 |
+
"learning_rate": 8.313814102564102e-06,
|
| 234556 |
+
"loss": 0.6469,
|
| 234557 |
+
"step": 106260
|
| 234558 |
+
},
|
| 234559 |
+
{
|
| 234560 |
+
"epoch": 856.88,
|
| 234561 |
+
"learning_rate": 8.313733974358975e-06,
|
| 234562 |
+
"loss": 0.298,
|
| 234563 |
+
"step": 106265
|
| 234564 |
+
},
|
| 234565 |
+
{
|
| 234566 |
+
"epoch": 856.92,
|
| 234567 |
+
"learning_rate": 8.313653846153847e-06,
|
| 234568 |
+
"loss": 0.3238,
|
| 234569 |
+
"step": 106270
|
| 234570 |
+
},
|
| 234571 |
+
{
|
| 234572 |
+
"epoch": 856.96,
|
| 234573 |
+
"learning_rate": 8.313573717948718e-06,
|
| 234574 |
+
"loss": 0.4681,
|
| 234575 |
+
"step": 106275
|
| 234576 |
+
},
|
| 234577 |
+
{
|
| 234578 |
+
"epoch": 857.0,
|
| 234579 |
+
"learning_rate": 8.313493589743591e-06,
|
| 234580 |
+
"loss": 1.1579,
|
| 234581 |
+
"step": 106280
|
| 234582 |
+
},
|
| 234583 |
+
{
|
| 234584 |
+
"epoch": 857.0,
|
| 234585 |
+
"eval_loss": 0.36302605271339417,
|
| 234586 |
+
"eval_runtime": 38.705,
|
| 234587 |
+
"eval_samples_per_second": 21.651,
|
| 234588 |
+
"eval_steps_per_second": 0.698,
|
| 234589 |
+
"eval_wer": 0.1839215112022258,
|
| 234590 |
+
"step": 106280
|
| 234591 |
+
},
|
| 234592 |
+
{
|
| 234593 |
+
"epoch": 850.04,
|
| 234594 |
+
"learning_rate": 8.313413461538463e-06,
|
| 234595 |
+
"loss": 0.2904,
|
| 234596 |
+
"step": 106285
|
| 234597 |
+
},
|
| 234598 |
+
{
|
| 234599 |
+
"epoch": 850.08,
|
| 234600 |
+
"learning_rate": 8.313333333333334e-06,
|
| 234601 |
+
"loss": 0.2598,
|
| 234602 |
+
"step": 106290
|
| 234603 |
+
},
|
| 234604 |
+
{
|
| 234605 |
+
"epoch": 850.12,
|
| 234606 |
+
"learning_rate": 8.313253205128205e-06,
|
| 234607 |
+
"loss": 0.2281,
|
| 234608 |
+
"step": 106295
|
| 234609 |
+
},
|
| 234610 |
+
{
|
| 234611 |
+
"epoch": 850.16,
|
| 234612 |
+
"learning_rate": 8.313173076923078e-06,
|
| 234613 |
+
"loss": 0.3974,
|
| 234614 |
+
"step": 106300
|
| 234615 |
+
},
|
| 234616 |
+
{
|
| 234617 |
+
"epoch": 850.2,
|
| 234618 |
+
"learning_rate": 8.31309294871795e-06,
|
| 234619 |
+
"loss": 1.2344,
|
| 234620 |
+
"step": 106305
|
| 234621 |
+
},
|
| 234622 |
+
{
|
| 234623 |
+
"epoch": 850.24,
|
| 234624 |
+
"learning_rate": 8.313012820512821e-06,
|
| 234625 |
+
"loss": 0.2952,
|
| 234626 |
+
"step": 106310
|
| 234627 |
+
},
|
| 234628 |
+
{
|
| 234629 |
+
"epoch": 850.28,
|
| 234630 |
+
"learning_rate": 8.312932692307692e-06,
|
| 234631 |
+
"loss": 0.3068,
|
| 234632 |
+
"step": 106315
|
| 234633 |
+
},
|
| 234634 |
+
{
|
| 234635 |
+
"epoch": 850.32,
|
| 234636 |
+
"learning_rate": 8.312852564102565e-06,
|
| 234637 |
+
"loss": 0.2807,
|
| 234638 |
+
"step": 106320
|
| 234639 |
+
},
|
| 234640 |
+
{
|
| 234641 |
+
"epoch": 850.36,
|
| 234642 |
+
"learning_rate": 8.312772435897437e-06,
|
| 234643 |
+
"loss": 0.4926,
|
| 234644 |
+
"step": 106325
|
| 234645 |
+
},
|
| 234646 |
+
{
|
| 234647 |
+
"epoch": 850.4,
|
| 234648 |
+
"learning_rate": 8.312692307692308e-06,
|
| 234649 |
+
"loss": 1.2578,
|
| 234650 |
+
"step": 106330
|
| 234651 |
+
},
|
| 234652 |
+
{
|
| 234653 |
+
"epoch": 850.44,
|
| 234654 |
+
"learning_rate": 8.312612179487181e-06,
|
| 234655 |
+
"loss": 0.348,
|
| 234656 |
+
"step": 106335
|
| 234657 |
+
},
|
| 234658 |
+
{
|
| 234659 |
+
"epoch": 850.48,
|
| 234660 |
+
"learning_rate": 8.312532051282053e-06,
|
| 234661 |
+
"loss": 0.2576,
|
| 234662 |
+
"step": 106340
|
| 234663 |
+
},
|
| 234664 |
+
{
|
| 234665 |
+
"epoch": 850.52,
|
| 234666 |
+
"learning_rate": 8.312451923076924e-06,
|
| 234667 |
+
"loss": 0.283,
|
| 234668 |
+
"step": 106345
|
| 234669 |
+
},
|
| 234670 |
+
{
|
| 234671 |
+
"epoch": 850.56,
|
| 234672 |
+
"learning_rate": 8.312371794871795e-06,
|
| 234673 |
+
"loss": 0.5078,
|
| 234674 |
+
"step": 106350
|
| 234675 |
+
},
|
| 234676 |
+
{
|
| 234677 |
+
"epoch": 850.6,
|
| 234678 |
+
"learning_rate": 8.312291666666668e-06,
|
| 234679 |
+
"loss": 1.2305,
|
| 234680 |
+
"step": 106355
|
| 234681 |
+
},
|
| 234682 |
+
{
|
| 234683 |
+
"epoch": 850.64,
|
| 234684 |
+
"learning_rate": 8.312211538461538e-06,
|
| 234685 |
+
"loss": 0.3288,
|
| 234686 |
+
"step": 106360
|
| 234687 |
+
},
|
| 234688 |
+
{
|
| 234689 |
+
"epoch": 850.68,
|
| 234690 |
+
"learning_rate": 8.312131410256411e-06,
|
| 234691 |
+
"loss": 0.4184,
|
| 234692 |
+
"step": 106365
|
| 234693 |
+
},
|
| 234694 |
+
{
|
| 234695 |
+
"epoch": 850.72,
|
| 234696 |
+
"learning_rate": 8.312051282051282e-06,
|
| 234697 |
+
"loss": 0.3502,
|
| 234698 |
+
"step": 106370
|
| 234699 |
+
},
|
| 234700 |
+
{
|
| 234701 |
+
"epoch": 850.76,
|
| 234702 |
+
"learning_rate": 8.311971153846154e-06,
|
| 234703 |
+
"loss": 0.4781,
|
| 234704 |
+
"step": 106375
|
| 234705 |
+
},
|
| 234706 |
+
{
|
| 234707 |
+
"epoch": 850.8,
|
| 234708 |
+
"learning_rate": 8.311891025641027e-06,
|
| 234709 |
+
"loss": 1.188,
|
| 234710 |
+
"step": 106380
|
| 234711 |
+
},
|
| 234712 |
+
{
|
| 234713 |
+
"epoch": 850.84,
|
| 234714 |
+
"learning_rate": 8.311810897435898e-06,
|
| 234715 |
+
"loss": 0.2652,
|
| 234716 |
+
"step": 106385
|
| 234717 |
+
},
|
| 234718 |
+
{
|
| 234719 |
+
"epoch": 850.88,
|
| 234720 |
+
"learning_rate": 8.31173076923077e-06,
|
| 234721 |
+
"loss": 0.3013,
|
| 234722 |
+
"step": 106390
|
| 234723 |
+
},
|
| 234724 |
+
{
|
| 234725 |
+
"epoch": 850.92,
|
| 234726 |
+
"learning_rate": 8.311650641025641e-06,
|
| 234727 |
+
"loss": 0.3647,
|
| 234728 |
+
"step": 106395
|
| 234729 |
+
},
|
| 234730 |
+
{
|
| 234731 |
+
"epoch": 850.96,
|
| 234732 |
+
"learning_rate": 8.311570512820514e-06,
|
| 234733 |
+
"loss": 0.4848,
|
| 234734 |
+
"step": 106400
|
| 234735 |
+
},
|
| 234736 |
+
{
|
| 234737 |
+
"epoch": 851.0,
|
| 234738 |
+
"learning_rate": 8.311490384615385e-06,
|
| 234739 |
+
"loss": 1.3993,
|
| 234740 |
+
"step": 106405
|
| 234741 |
+
},
|
| 234742 |
+
{
|
| 234743 |
+
"epoch": 851.0,
|
| 234744 |
+
"eval_loss": 0.351011723279953,
|
| 234745 |
+
"eval_runtime": 38.478,
|
| 234746 |
+
"eval_samples_per_second": 21.753,
|
| 234747 |
+
"eval_steps_per_second": 0.702,
|
| 234748 |
+
"eval_wer": 0.18673609604397193,
|
| 234749 |
+
"step": 106405
|
| 234750 |
}
|
| 234751 |
],
|
| 234752 |
"max_steps": 625000,
|
| 234753 |
"num_train_epochs": 5000,
|
| 234754 |
+
"total_flos": 2.994431103969445e+20,
|
| 234755 |
"trial_name": null,
|
| 234756 |
"trial_params": null
|
| 234757 |
}
|
model-bin/finetune/base/{checkpoint-105783 → checkpoint-106405}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630002482.2222688/events.out.tfevents.1630002482.8e89bd551565.924.281
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8c9806712b0099ce0f347fddb419dfc968d76e4e68631621db01a95173559fa0
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630002949.6500714/events.out.tfevents.1630002949.8e89bd551565.924.283
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fa7a85b9b507bef14b7a9dad87cbdcd975e77b570a6a3a9c192c3d66f4dc5f80
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630003391.3419247/events.out.tfevents.1630003391.8e89bd551565.924.285
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fcf1097ed155d479da70f557596c5be31a940c8ee080302586c0950b0f87264f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630003815.897631/events.out.tfevents.1630003816.8e89bd551565.924.287
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:29db253b3562f8b6fa5838f4ed5adc50a6d48c2d80f3795c50ff37253333181f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630004360.036647/events.out.tfevents.1630004360.8e89bd551565.924.289
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da3761c62d2851a4589e32d391eeb4b9bbd6af43b94cd66727d62acfe9081b91
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630002482.8e89bd551565.924.280
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f07d053115242feed53d09da5f337fbeafc59871aa9dd026466508e07fefd33d
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630002948.8e89bd551565.924.282
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3737e440b0a39f73614c03fb3825781dcb8105a680c0e05930c27da7faf54df1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630003391.8e89bd551565.924.284
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:825719d37337672e588823e4c97e95b2dab166d51f461be2ed5124525a9f2730
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630003815.8e89bd551565.924.286
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7105560708997ce10038d3b8490ed6e08ff5b475299305421fd670fa9a9da8bd
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630004360.8e89bd551565.924.288
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4b08d756c2fa25a4e3b4f0eb022fdcd1840b77446e3b17199dd5053af343dbff
|
| 3 |
+
size 8622
|