"auto-commit"
Browse files- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39323}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39323}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39323}/trainer_state.json +476 -5
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39323}/training_args.bin +0 -0
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39572}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/rng_state.pth +2 -2
- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39572}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39572}/trainer_state.json +0 -0
- model-bin/finetune/base/{checkpoint-31352 → checkpoint-39572}/training_args.bin +1 -1
- model-bin/finetune/base/log/1629683460.0346928/events.out.tfevents.1629683460.74272264b15c.932.13 +3 -0
- model-bin/finetune/base/log/1629684116.1803105/events.out.tfevents.1629684116.74272264b15c.932.15 +3 -0
- model-bin/finetune/base/log/1629684758.8980677/events.out.tfevents.1629684758.74272264b15c.932.17 +3 -0
- model-bin/finetune/base/log/1629685405.8484218/events.out.tfevents.1629685405.74272264b15c.932.19 +3 -0
- model-bin/finetune/base/log/1629686057.09065/events.out.tfevents.1629686057.74272264b15c.932.21 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629683460.74272264b15c.932.12 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629684116.74272264b15c.932.14 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629684758.74272264b15c.932.16 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629685405.74272264b15c.932.18 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1629686057.74272264b15c.932.20 +3 -0
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d098e471715476d3b967b945f037df339ad1b0d95518095f0430182b12106bdc
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39323}/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 sha256:2c876fce063546bf4be1119f3fc2be8ec5f4f1f0cf2e13d57416c78ccc995ae2
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/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:c5110edb27f9fecca2b7a4f02fa3f21ed3db364c9c4530dbd785422622d301c5
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39323}/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 sha256:068b3f6a1364ee3a2aaaf0b72c0c5b3749099a73ddfd7fcd2f4937e239990da0
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39323}/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 sha256:20d1c53989c221e096c75373aa28c103107e7faff35d1ffbf180e7ae52a556b9
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39323}/trainer_state.json
RENAMED
|
@@ -1,8 +1,8 @@
|
|
| 1 |
{
|
| 2 |
-
"best_metric": 0.
|
| 3 |
-
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-
|
| 4 |
-
"epoch":
|
| 5 |
-
"global_step":
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
@@ -148923,11 +148923,482 @@
|
|
| 148923 |
"eval_steps_per_second": 0.685,
|
| 148924 |
"eval_wer": 0.20816059349770893,
|
| 148925 |
"step": 38950
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 148926 |
}
|
| 148927 |
],
|
| 148928 |
"max_steps": 625000,
|
| 148929 |
"num_train_epochs": 5000,
|
| 148930 |
-
"total_flos": 1.
|
| 148931 |
"trial_name": null,
|
| 148932 |
"trial_params": null
|
| 148933 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"best_metric": 0.2028974903051145,
|
| 3 |
+
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-39323",
|
| 4 |
+
"epoch": 314.0,
|
| 5 |
+
"global_step": 39323,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 148923 |
"eval_steps_per_second": 0.685,
|
| 148924 |
"eval_wer": 0.20816059349770893,
|
| 148925 |
"step": 38950
|
| 148926 |
+
},
|
| 148927 |
+
{
|
| 148928 |
+
"epoch": 314.04,
|
| 148929 |
+
"learning_rate": 9.39198717948718e-06,
|
| 148930 |
+
"loss": 0.3884,
|
| 148931 |
+
"step": 38955
|
| 148932 |
+
},
|
| 148933 |
+
{
|
| 148934 |
+
"epoch": 314.08,
|
| 148935 |
+
"learning_rate": 9.391907051282053e-06,
|
| 148936 |
+
"loss": 0.3614,
|
| 148937 |
+
"step": 38960
|
| 148938 |
+
},
|
| 148939 |
+
{
|
| 148940 |
+
"epoch": 314.12,
|
| 148941 |
+
"learning_rate": 9.391826923076924e-06,
|
| 148942 |
+
"loss": 0.3707,
|
| 148943 |
+
"step": 38965
|
| 148944 |
+
},
|
| 148945 |
+
{
|
| 148946 |
+
"epoch": 314.16,
|
| 148947 |
+
"learning_rate": 9.391746794871795e-06,
|
| 148948 |
+
"loss": 0.5056,
|
| 148949 |
+
"step": 38970
|
| 148950 |
+
},
|
| 148951 |
+
{
|
| 148952 |
+
"epoch": 314.2,
|
| 148953 |
+
"learning_rate": 9.391666666666669e-06,
|
| 148954 |
+
"loss": 1.3484,
|
| 148955 |
+
"step": 38975
|
| 148956 |
+
},
|
| 148957 |
+
{
|
| 148958 |
+
"epoch": 314.24,
|
| 148959 |
+
"learning_rate": 9.391586538461538e-06,
|
| 148960 |
+
"loss": 0.4397,
|
| 148961 |
+
"step": 38980
|
| 148962 |
+
},
|
| 148963 |
+
{
|
| 148964 |
+
"epoch": 314.28,
|
| 148965 |
+
"learning_rate": 9.391506410256411e-06,
|
| 148966 |
+
"loss": 0.3384,
|
| 148967 |
+
"step": 38985
|
| 148968 |
+
},
|
| 148969 |
+
{
|
| 148970 |
+
"epoch": 314.32,
|
| 148971 |
+
"learning_rate": 9.391426282051283e-06,
|
| 148972 |
+
"loss": 0.3773,
|
| 148973 |
+
"step": 38990
|
| 148974 |
+
},
|
| 148975 |
+
{
|
| 148976 |
+
"epoch": 314.36,
|
| 148977 |
+
"learning_rate": 9.391346153846154e-06,
|
| 148978 |
+
"loss": 0.4941,
|
| 148979 |
+
"step": 38995
|
| 148980 |
+
},
|
| 148981 |
+
{
|
| 148982 |
+
"epoch": 314.4,
|
| 148983 |
+
"learning_rate": 9.391266025641025e-06,
|
| 148984 |
+
"loss": 1.4008,
|
| 148985 |
+
"step": 39000
|
| 148986 |
+
},
|
| 148987 |
+
{
|
| 148988 |
+
"epoch": 314.44,
|
| 148989 |
+
"learning_rate": 9.391185897435898e-06,
|
| 148990 |
+
"loss": 0.4474,
|
| 148991 |
+
"step": 39005
|
| 148992 |
+
},
|
| 148993 |
+
{
|
| 148994 |
+
"epoch": 314.48,
|
| 148995 |
+
"learning_rate": 9.39110576923077e-06,
|
| 148996 |
+
"loss": 0.3198,
|
| 148997 |
+
"step": 39010
|
| 148998 |
+
},
|
| 148999 |
+
{
|
| 149000 |
+
"epoch": 314.52,
|
| 149001 |
+
"learning_rate": 9.391025641025641e-06,
|
| 149002 |
+
"loss": 0.4109,
|
| 149003 |
+
"step": 39015
|
| 149004 |
+
},
|
| 149005 |
+
{
|
| 149006 |
+
"epoch": 314.56,
|
| 149007 |
+
"learning_rate": 9.390945512820514e-06,
|
| 149008 |
+
"loss": 0.5026,
|
| 149009 |
+
"step": 39020
|
| 149010 |
+
},
|
| 149011 |
+
{
|
| 149012 |
+
"epoch": 314.6,
|
| 149013 |
+
"learning_rate": 9.390865384615385e-06,
|
| 149014 |
+
"loss": 1.5935,
|
| 149015 |
+
"step": 39025
|
| 149016 |
+
},
|
| 149017 |
+
{
|
| 149018 |
+
"epoch": 314.64,
|
| 149019 |
+
"learning_rate": 9.390785256410257e-06,
|
| 149020 |
+
"loss": 0.4271,
|
| 149021 |
+
"step": 39030
|
| 149022 |
+
},
|
| 149023 |
+
{
|
| 149024 |
+
"epoch": 314.68,
|
| 149025 |
+
"learning_rate": 9.390705128205128e-06,
|
| 149026 |
+
"loss": 0.4287,
|
| 149027 |
+
"step": 39035
|
| 149028 |
+
},
|
| 149029 |
+
{
|
| 149030 |
+
"epoch": 314.72,
|
| 149031 |
+
"learning_rate": 9.390625000000001e-06,
|
| 149032 |
+
"loss": 0.3599,
|
| 149033 |
+
"step": 39040
|
| 149034 |
+
},
|
| 149035 |
+
{
|
| 149036 |
+
"epoch": 314.76,
|
| 149037 |
+
"learning_rate": 9.390544871794873e-06,
|
| 149038 |
+
"loss": 0.5263,
|
| 149039 |
+
"step": 39045
|
| 149040 |
+
},
|
| 149041 |
+
{
|
| 149042 |
+
"epoch": 314.8,
|
| 149043 |
+
"learning_rate": 9.390464743589744e-06,
|
| 149044 |
+
"loss": 1.3151,
|
| 149045 |
+
"step": 39050
|
| 149046 |
+
},
|
| 149047 |
+
{
|
| 149048 |
+
"epoch": 314.84,
|
| 149049 |
+
"learning_rate": 9.390384615384615e-06,
|
| 149050 |
+
"loss": 0.4487,
|
| 149051 |
+
"step": 39055
|
| 149052 |
+
},
|
| 149053 |
+
{
|
| 149054 |
+
"epoch": 314.88,
|
| 149055 |
+
"learning_rate": 9.390304487179488e-06,
|
| 149056 |
+
"loss": 0.3071,
|
| 149057 |
+
"step": 39060
|
| 149058 |
+
},
|
| 149059 |
+
{
|
| 149060 |
+
"epoch": 314.92,
|
| 149061 |
+
"learning_rate": 9.39022435897436e-06,
|
| 149062 |
+
"loss": 0.4116,
|
| 149063 |
+
"step": 39065
|
| 149064 |
+
},
|
| 149065 |
+
{
|
| 149066 |
+
"epoch": 314.96,
|
| 149067 |
+
"learning_rate": 9.390144230769231e-06,
|
| 149068 |
+
"loss": 0.5645,
|
| 149069 |
+
"step": 39070
|
| 149070 |
+
},
|
| 149071 |
+
{
|
| 149072 |
+
"epoch": 315.0,
|
| 149073 |
+
"eval_loss": 0.4270016551017761,
|
| 149074 |
+
"eval_runtime": 40.3005,
|
| 149075 |
+
"eval_samples_per_second": 20.819,
|
| 149076 |
+
"eval_steps_per_second": 0.67,
|
| 149077 |
+
"eval_wer": 0.2158767253341123,
|
| 149078 |
+
"step": 39074
|
| 149079 |
+
},
|
| 149080 |
+
{
|
| 149081 |
+
"epoch": 315.01,
|
| 149082 |
+
"learning_rate": 9.390064102564104e-06,
|
| 149083 |
+
"loss": 0.4925,
|
| 149084 |
+
"step": 39075
|
| 149085 |
+
},
|
| 149086 |
+
{
|
| 149087 |
+
"epoch": 315.05,
|
| 149088 |
+
"learning_rate": 9.389983974358976e-06,
|
| 149089 |
+
"loss": 0.4834,
|
| 149090 |
+
"step": 39080
|
| 149091 |
+
},
|
| 149092 |
+
{
|
| 149093 |
+
"epoch": 315.09,
|
| 149094 |
+
"learning_rate": 9.389903846153847e-06,
|
| 149095 |
+
"loss": 0.3437,
|
| 149096 |
+
"step": 39085
|
| 149097 |
+
},
|
| 149098 |
+
{
|
| 149099 |
+
"epoch": 315.13,
|
| 149100 |
+
"learning_rate": 9.389823717948718e-06,
|
| 149101 |
+
"loss": 0.4061,
|
| 149102 |
+
"step": 39090
|
| 149103 |
+
},
|
| 149104 |
+
{
|
| 149105 |
+
"epoch": 315.17,
|
| 149106 |
+
"learning_rate": 9.389743589743591e-06,
|
| 149107 |
+
"loss": 0.5769,
|
| 149108 |
+
"step": 39095
|
| 149109 |
+
},
|
| 149110 |
+
{
|
| 149111 |
+
"epoch": 315.21,
|
| 149112 |
+
"learning_rate": 9.389663461538463e-06,
|
| 149113 |
+
"loss": 1.3031,
|
| 149114 |
+
"step": 39100
|
| 149115 |
+
},
|
| 149116 |
+
{
|
| 149117 |
+
"epoch": 315.25,
|
| 149118 |
+
"learning_rate": 9.389583333333334e-06,
|
| 149119 |
+
"loss": 0.3801,
|
| 149120 |
+
"step": 39105
|
| 149121 |
+
},
|
| 149122 |
+
{
|
| 149123 |
+
"epoch": 315.29,
|
| 149124 |
+
"learning_rate": 9.389503205128205e-06,
|
| 149125 |
+
"loss": 0.3464,
|
| 149126 |
+
"step": 39110
|
| 149127 |
+
},
|
| 149128 |
+
{
|
| 149129 |
+
"epoch": 315.33,
|
| 149130 |
+
"learning_rate": 9.389423076923078e-06,
|
| 149131 |
+
"loss": 0.468,
|
| 149132 |
+
"step": 39115
|
| 149133 |
+
},
|
| 149134 |
+
{
|
| 149135 |
+
"epoch": 315.37,
|
| 149136 |
+
"learning_rate": 9.38934294871795e-06,
|
| 149137 |
+
"loss": 0.6171,
|
| 149138 |
+
"step": 39120
|
| 149139 |
+
},
|
| 149140 |
+
{
|
| 149141 |
+
"epoch": 315.41,
|
| 149142 |
+
"learning_rate": 9.389262820512821e-06,
|
| 149143 |
+
"loss": 1.3169,
|
| 149144 |
+
"step": 39125
|
| 149145 |
+
},
|
| 149146 |
+
{
|
| 149147 |
+
"epoch": 315.45,
|
| 149148 |
+
"learning_rate": 9.389182692307694e-06,
|
| 149149 |
+
"loss": 0.3742,
|
| 149150 |
+
"step": 39130
|
| 149151 |
+
},
|
| 149152 |
+
{
|
| 149153 |
+
"epoch": 315.49,
|
| 149154 |
+
"learning_rate": 9.389102564102564e-06,
|
| 149155 |
+
"loss": 0.3478,
|
| 149156 |
+
"step": 39135
|
| 149157 |
+
},
|
| 149158 |
+
{
|
| 149159 |
+
"epoch": 315.53,
|
| 149160 |
+
"learning_rate": 9.389022435897437e-06,
|
| 149161 |
+
"loss": 0.3726,
|
| 149162 |
+
"step": 39140
|
| 149163 |
+
},
|
| 149164 |
+
{
|
| 149165 |
+
"epoch": 315.57,
|
| 149166 |
+
"learning_rate": 9.388942307692308e-06,
|
| 149167 |
+
"loss": 0.6079,
|
| 149168 |
+
"step": 39145
|
| 149169 |
+
},
|
| 149170 |
+
{
|
| 149171 |
+
"epoch": 315.61,
|
| 149172 |
+
"learning_rate": 9.38886217948718e-06,
|
| 149173 |
+
"loss": 1.2394,
|
| 149174 |
+
"step": 39150
|
| 149175 |
+
},
|
| 149176 |
+
{
|
| 149177 |
+
"epoch": 315.65,
|
| 149178 |
+
"learning_rate": 9.388782051282051e-06,
|
| 149179 |
+
"loss": 0.3411,
|
| 149180 |
+
"step": 39155
|
| 149181 |
+
},
|
| 149182 |
+
{
|
| 149183 |
+
"epoch": 315.69,
|
| 149184 |
+
"learning_rate": 9.388701923076924e-06,
|
| 149185 |
+
"loss": 0.3573,
|
| 149186 |
+
"step": 39160
|
| 149187 |
+
},
|
| 149188 |
+
{
|
| 149189 |
+
"epoch": 315.73,
|
| 149190 |
+
"learning_rate": 9.388621794871795e-06,
|
| 149191 |
+
"loss": 0.427,
|
| 149192 |
+
"step": 39165
|
| 149193 |
+
},
|
| 149194 |
+
{
|
| 149195 |
+
"epoch": 315.77,
|
| 149196 |
+
"learning_rate": 9.388541666666667e-06,
|
| 149197 |
+
"loss": 0.692,
|
| 149198 |
+
"step": 39170
|
| 149199 |
+
},
|
| 149200 |
+
{
|
| 149201 |
+
"epoch": 315.81,
|
| 149202 |
+
"learning_rate": 9.38846153846154e-06,
|
| 149203 |
+
"loss": 1.2786,
|
| 149204 |
+
"step": 39175
|
| 149205 |
+
},
|
| 149206 |
+
{
|
| 149207 |
+
"epoch": 315.85,
|
| 149208 |
+
"learning_rate": 9.388381410256411e-06,
|
| 149209 |
+
"loss": 0.3708,
|
| 149210 |
+
"step": 39180
|
| 149211 |
+
},
|
| 149212 |
+
{
|
| 149213 |
+
"epoch": 315.89,
|
| 149214 |
+
"learning_rate": 9.388301282051283e-06,
|
| 149215 |
+
"loss": 0.358,
|
| 149216 |
+
"step": 39185
|
| 149217 |
+
},
|
| 149218 |
+
{
|
| 149219 |
+
"epoch": 315.93,
|
| 149220 |
+
"learning_rate": 9.388221153846154e-06,
|
| 149221 |
+
"loss": 0.4877,
|
| 149222 |
+
"step": 39190
|
| 149223 |
+
},
|
| 149224 |
+
{
|
| 149225 |
+
"epoch": 315.97,
|
| 149226 |
+
"learning_rate": 9.388141025641027e-06,
|
| 149227 |
+
"loss": 0.6957,
|
| 149228 |
+
"step": 39195
|
| 149229 |
+
},
|
| 149230 |
+
{
|
| 149231 |
+
"epoch": 316.0,
|
| 149232 |
+
"eval_loss": 0.4085172414779663,
|
| 149233 |
+
"eval_runtime": 38.9888,
|
| 149234 |
+
"eval_samples_per_second": 21.545,
|
| 149235 |
+
"eval_steps_per_second": 0.693,
|
| 149236 |
+
"eval_wer": 0.21160508083140878,
|
| 149237 |
+
"step": 39198
|
| 149238 |
+
},
|
| 149239 |
+
{
|
| 149240 |
+
"epoch": 313.02,
|
| 149241 |
+
"learning_rate": 9.388060897435898e-06,
|
| 149242 |
+
"loss": 0.4283,
|
| 149243 |
+
"step": 39200
|
| 149244 |
+
},
|
| 149245 |
+
{
|
| 149246 |
+
"epoch": 313.06,
|
| 149247 |
+
"learning_rate": 9.38798076923077e-06,
|
| 149248 |
+
"loss": 0.4645,
|
| 149249 |
+
"step": 39205
|
| 149250 |
+
},
|
| 149251 |
+
{
|
| 149252 |
+
"epoch": 313.1,
|
| 149253 |
+
"learning_rate": 9.387900641025641e-06,
|
| 149254 |
+
"loss": 0.381,
|
| 149255 |
+
"step": 39210
|
| 149256 |
+
},
|
| 149257 |
+
{
|
| 149258 |
+
"epoch": 313.14,
|
| 149259 |
+
"learning_rate": 9.387820512820514e-06,
|
| 149260 |
+
"loss": 0.5081,
|
| 149261 |
+
"step": 39215
|
| 149262 |
+
},
|
| 149263 |
+
{
|
| 149264 |
+
"epoch": 313.18,
|
| 149265 |
+
"learning_rate": 9.387740384615385e-06,
|
| 149266 |
+
"loss": 0.7461,
|
| 149267 |
+
"step": 39220
|
| 149268 |
+
},
|
| 149269 |
+
{
|
| 149270 |
+
"epoch": 313.22,
|
| 149271 |
+
"learning_rate": 9.387660256410257e-06,
|
| 149272 |
+
"loss": 1.0202,
|
| 149273 |
+
"step": 39225
|
| 149274 |
+
},
|
| 149275 |
+
{
|
| 149276 |
+
"epoch": 313.26,
|
| 149277 |
+
"learning_rate": 9.38758012820513e-06,
|
| 149278 |
+
"loss": 0.4627,
|
| 149279 |
+
"step": 39230
|
| 149280 |
+
},
|
| 149281 |
+
{
|
| 149282 |
+
"epoch": 313.3,
|
| 149283 |
+
"learning_rate": 9.387500000000001e-06,
|
| 149284 |
+
"loss": 0.4061,
|
| 149285 |
+
"step": 39235
|
| 149286 |
+
},
|
| 149287 |
+
{
|
| 149288 |
+
"epoch": 313.34,
|
| 149289 |
+
"learning_rate": 9.387419871794873e-06,
|
| 149290 |
+
"loss": 0.3766,
|
| 149291 |
+
"step": 39240
|
| 149292 |
+
},
|
| 149293 |
+
{
|
| 149294 |
+
"epoch": 313.38,
|
| 149295 |
+
"learning_rate": 9.387339743589744e-06,
|
| 149296 |
+
"loss": 0.8813,
|
| 149297 |
+
"step": 39245
|
| 149298 |
+
},
|
| 149299 |
+
{
|
| 149300 |
+
"epoch": 313.42,
|
| 149301 |
+
"learning_rate": 9.387259615384617e-06,
|
| 149302 |
+
"loss": 1.1809,
|
| 149303 |
+
"step": 39250
|
| 149304 |
+
},
|
| 149305 |
+
{
|
| 149306 |
+
"epoch": 313.46,
|
| 149307 |
+
"learning_rate": 9.387179487179488e-06,
|
| 149308 |
+
"loss": 0.3465,
|
| 149309 |
+
"step": 39255
|
| 149310 |
+
},
|
| 149311 |
+
{
|
| 149312 |
+
"epoch": 313.5,
|
| 149313 |
+
"learning_rate": 9.38709935897436e-06,
|
| 149314 |
+
"loss": 0.3262,
|
| 149315 |
+
"step": 39260
|
| 149316 |
+
},
|
| 149317 |
+
{
|
| 149318 |
+
"epoch": 313.54,
|
| 149319 |
+
"learning_rate": 9.387019230769233e-06,
|
| 149320 |
+
"loss": 0.4011,
|
| 149321 |
+
"step": 39265
|
| 149322 |
+
},
|
| 149323 |
+
{
|
| 149324 |
+
"epoch": 313.58,
|
| 149325 |
+
"learning_rate": 9.386939102564104e-06,
|
| 149326 |
+
"loss": 0.6261,
|
| 149327 |
+
"step": 39270
|
| 149328 |
+
},
|
| 149329 |
+
{
|
| 149330 |
+
"epoch": 313.62,
|
| 149331 |
+
"learning_rate": 9.386858974358975e-06,
|
| 149332 |
+
"loss": 1.103,
|
| 149333 |
+
"step": 39275
|
| 149334 |
+
},
|
| 149335 |
+
{
|
| 149336 |
+
"epoch": 313.66,
|
| 149337 |
+
"learning_rate": 9.386778846153847e-06,
|
| 149338 |
+
"loss": 0.3716,
|
| 149339 |
+
"step": 39280
|
| 149340 |
+
},
|
| 149341 |
+
{
|
| 149342 |
+
"epoch": 313.7,
|
| 149343 |
+
"learning_rate": 9.38669871794872e-06,
|
| 149344 |
+
"loss": 0.3951,
|
| 149345 |
+
"step": 39285
|
| 149346 |
+
},
|
| 149347 |
+
{
|
| 149348 |
+
"epoch": 313.74,
|
| 149349 |
+
"learning_rate": 9.38661858974359e-06,
|
| 149350 |
+
"loss": 0.4319,
|
| 149351 |
+
"step": 39290
|
| 149352 |
+
},
|
| 149353 |
+
{
|
| 149354 |
+
"epoch": 313.78,
|
| 149355 |
+
"learning_rate": 9.386538461538463e-06,
|
| 149356 |
+
"loss": 0.8423,
|
| 149357 |
+
"step": 39295
|
| 149358 |
+
},
|
| 149359 |
+
{
|
| 149360 |
+
"epoch": 313.82,
|
| 149361 |
+
"learning_rate": 9.386458333333334e-06,
|
| 149362 |
+
"loss": 1.1871,
|
| 149363 |
+
"step": 39300
|
| 149364 |
+
},
|
| 149365 |
+
{
|
| 149366 |
+
"epoch": 313.86,
|
| 149367 |
+
"learning_rate": 9.386378205128205e-06,
|
| 149368 |
+
"loss": 0.2941,
|
| 149369 |
+
"step": 39305
|
| 149370 |
+
},
|
| 149371 |
+
{
|
| 149372 |
+
"epoch": 313.9,
|
| 149373 |
+
"learning_rate": 9.386298076923077e-06,
|
| 149374 |
+
"loss": 0.429,
|
| 149375 |
+
"step": 39310
|
| 149376 |
+
},
|
| 149377 |
+
{
|
| 149378 |
+
"epoch": 313.94,
|
| 149379 |
+
"learning_rate": 9.38621794871795e-06,
|
| 149380 |
+
"loss": 0.5199,
|
| 149381 |
+
"step": 39315
|
| 149382 |
+
},
|
| 149383 |
+
{
|
| 149384 |
+
"epoch": 313.98,
|
| 149385 |
+
"learning_rate": 9.386137820512821e-06,
|
| 149386 |
+
"loss": 0.8307,
|
| 149387 |
+
"step": 39320
|
| 149388 |
+
},
|
| 149389 |
+
{
|
| 149390 |
+
"epoch": 314.0,
|
| 149391 |
+
"eval_loss": 0.4118651747703552,
|
| 149392 |
+
"eval_runtime": 39.5212,
|
| 149393 |
+
"eval_samples_per_second": 21.254,
|
| 149394 |
+
"eval_steps_per_second": 0.683,
|
| 149395 |
+
"eval_wer": 0.2028974903051145,
|
| 149396 |
+
"step": 39323
|
| 149397 |
}
|
| 149398 |
],
|
| 149399 |
"max_steps": 625000,
|
| 149400 |
"num_train_epochs": 5000,
|
| 149401 |
+
"total_flos": 1.1065541388694205e+20,
|
| 149402 |
"trial_name": null,
|
| 149403 |
"trial_params": null
|
| 149404 |
}
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39323}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/optimizer.pt
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 722165009
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:107a92561a6cff82fbd33d0aecf3fa48f3bfab520f60a41584ebfe49ab000dfc
|
| 3 |
size 722165009
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39572}/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 sha256:9b6994fb2d07eaff2e19b5cab3d2eaf7d456da0818f11b0b9895a8490de1eecf
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/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 sha256:8efbf88b9aa3bdcb5c19f1ddd0e36363cbee840aff0d77a1b777eaf904ee5222
|
| 3 |
+
size 14503
|
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39572}/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 sha256:51d3e5c75e0c15ca92b8071b56ff282c0e7dd9cc17d1a6e0cfe3f42c5a83ea4f
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-38950 → checkpoint-39572}/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:f5963bef90550fc81663e6d8faeb8a82fce4b0a862818b270f0dbd8ac7c3bea7
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39572}/trainer_state.json
RENAMED
|
The diff for this file is too large to render.
See raw diff
|
|
|
model-bin/finetune/base/{checkpoint-31352 → checkpoint-39572}/training_args.bin
RENAMED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 2671
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4ecdcb08af58e83c186b71f5e90d56c28975bcef64ca557cbbb3c92f56fdf4ea
|
| 3 |
size 2671
|
model-bin/finetune/base/log/1629683460.0346928/events.out.tfevents.1629683460.74272264b15c.932.13
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2ea42a352819103d4232511ae33f096df1628cf83001106ab12fffd75ba67334
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629684116.1803105/events.out.tfevents.1629684116.74272264b15c.932.15
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:df9c4751e195bf9f1f33e23e0770f6220b400326b5870b68a47d01a1e4393b1f
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629684758.8980677/events.out.tfevents.1629684758.74272264b15c.932.17
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6adacd73066f6238b4a6699e57539e3330ffcb97cfd96760a792ee7f645f0eed
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629685405.8484218/events.out.tfevents.1629685405.74272264b15c.932.19
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b414c11fbbadcd61b8554ac264cdd25febd7ecd5f0b4a9e8433d6d5bdc5f91f5
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1629686057.09065/events.out.tfevents.1629686057.74272264b15c.932.21
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e925ba97681a28533e3cde93aa1d0d2144ff41daa645452f9951b2f9e8309a03
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1629683460.74272264b15c.932.12
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9464dfa359c49e66ef514365c8bdcf44180e2f6fea04166c28068ffb24cc09cb
|
| 3 |
+
size 8462
|
model-bin/finetune/base/log/events.out.tfevents.1629684116.74272264b15c.932.14
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:24cf400b110721507df8897c102261cca46bd94d19f3e04b8cdd5320547f6be8
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629684758.74272264b15c.932.16
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a2a57e46870d6557165c843a42104f920fec8f91df898c11687d9c655f284b69
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629685405.74272264b15c.932.18
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f7e0573b35d4108da9b629cb0b91676ff6dfedc1e9e3e6b22cb2ec6417d35df6
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1629686057.74272264b15c.932.20
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1433dabac2af35eb2c8ba73336a8636dc85e300e501bad8b60ccb4e2ab99314
|
| 3 |
+
size 8622
|