auto-commit
Browse files- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/config.json +0 -0
- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/optimizer.pt +1 -1
- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/preprocessor_config.json +0 -0
- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/pytorch_model.bin +1 -1
- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/rng_state.pth +1 -1
- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/scaler.pt +1 -1
- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/scheduler.pt +1 -1
- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/trainer_state.json +322 -4
- model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/training_args.bin +0 -0
- model-bin/finetune/base/log/1630050545.555799/events.out.tfevents.1630050545.52f5c7e305a3.886.91 +3 -0
- model-bin/finetune/base/log/1630051145.811182/events.out.tfevents.1630051145.52f5c7e305a3.886.93 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630050545.52f5c7e305a3.886.90 +3 -0
- model-bin/finetune/base/log/events.out.tfevents.1630051145.52f5c7e305a3.886.92 +3 -0
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/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 sha256:76e57d6f91737dbbfab290b5599d5b3c4aade067aae28aad47ddc55053256d5c
|
| 3 |
size 722165393
|
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/preprocessor_config.json
RENAMED
|
File without changes
|
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/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:ecd6c18707f6f175faf10b4faeebe9a418239519113272879c515e06333a64f3
|
| 3 |
size 377909911
|
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/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 sha256:7e8f08e3f19a28bcef3f6bda763058ad4ec179589a461d0c48f62c72069e50c3
|
| 3 |
size 14503
|
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/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 sha256:9423e2e940526608df5ef6924192b590e1b3251579ffbe9455b67f9f06c55439
|
| 3 |
size 559
|
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/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 sha256:01c66c6381dce1adcf2bea3883f64b5ca782f6591999977f33146f32cb6de4b0
|
| 3 |
size 623
|
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/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,
|
|
@@ -241866,11 +241866,329 @@
|
|
| 241866 |
"eval_steps_per_second": 0.625,
|
| 241867 |
"eval_wer": 0.1886928860613071,
|
| 241868 |
"step": 112002
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 241869 |
}
|
| 241870 |
],
|
| 241871 |
-
"max_steps":
|
| 241872 |
"num_train_epochs": 5000,
|
| 241873 |
-
"total_flos": 3.
|
| 241874 |
"trial_name": null,
|
| 241875 |
"trial_params": null
|
| 241876 |
}
|
|
|
|
| 1 |
{
|
| 2 |
"best_metric": 0.1743826049391605,
|
| 3 |
"best_model_checkpoint": "./model-bin/finetune/base/checkpoint-101551",
|
| 4 |
+
"epoch": 897.9960159362549,
|
| 5 |
+
"global_step": 112251,
|
| 6 |
"is_hyper_param_search": false,
|
| 7 |
"is_local_process_zero": true,
|
| 8 |
"is_world_process_zero": true,
|
|
|
|
| 241866 |
"eval_steps_per_second": 0.625,
|
| 241867 |
"eval_wer": 0.1886928860613071,
|
| 241868 |
"step": 112002
|
| 241869 |
+
},
|
| 241870 |
+
{
|
| 241871 |
+
"epoch": 903.02,
|
| 241872 |
+
"learning_rate": 8.207415185783522e-06,
|
| 241873 |
+
"loss": 0.3015,
|
| 241874 |
+
"step": 112005
|
| 241875 |
+
},
|
| 241876 |
+
{
|
| 241877 |
+
"epoch": 903.06,
|
| 241878 |
+
"learning_rate": 8.207334410339256e-06,
|
| 241879 |
+
"loss": 0.2631,
|
| 241880 |
+
"step": 112010
|
| 241881 |
+
},
|
| 241882 |
+
{
|
| 241883 |
+
"epoch": 903.1,
|
| 241884 |
+
"learning_rate": 8.207253634894992e-06,
|
| 241885 |
+
"loss": 0.2948,
|
| 241886 |
+
"step": 112015
|
| 241887 |
+
},
|
| 241888 |
+
{
|
| 241889 |
+
"epoch": 903.14,
|
| 241890 |
+
"learning_rate": 8.207172859450726e-06,
|
| 241891 |
+
"loss": 0.3397,
|
| 241892 |
+
"step": 112020
|
| 241893 |
+
},
|
| 241894 |
+
{
|
| 241895 |
+
"epoch": 903.18,
|
| 241896 |
+
"learning_rate": 8.207092084006462e-06,
|
| 241897 |
+
"loss": 0.7385,
|
| 241898 |
+
"step": 112025
|
| 241899 |
+
},
|
| 241900 |
+
{
|
| 241901 |
+
"epoch": 903.22,
|
| 241902 |
+
"learning_rate": 8.207011308562198e-06,
|
| 241903 |
+
"loss": 0.9632,
|
| 241904 |
+
"step": 112030
|
| 241905 |
+
},
|
| 241906 |
+
{
|
| 241907 |
+
"epoch": 903.27,
|
| 241908 |
+
"learning_rate": 8.206930533117932e-06,
|
| 241909 |
+
"loss": 0.3019,
|
| 241910 |
+
"step": 112035
|
| 241911 |
+
},
|
| 241912 |
+
{
|
| 241913 |
+
"epoch": 903.31,
|
| 241914 |
+
"learning_rate": 8.206849757673668e-06,
|
| 241915 |
+
"loss": 0.2995,
|
| 241916 |
+
"step": 112040
|
| 241917 |
+
},
|
| 241918 |
+
{
|
| 241919 |
+
"epoch": 903.35,
|
| 241920 |
+
"learning_rate": 8.206768982229402e-06,
|
| 241921 |
+
"loss": 0.3895,
|
| 241922 |
+
"step": 112045
|
| 241923 |
+
},
|
| 241924 |
+
{
|
| 241925 |
+
"epoch": 903.39,
|
| 241926 |
+
"learning_rate": 8.206688206785138e-06,
|
| 241927 |
+
"loss": 0.7798,
|
| 241928 |
+
"step": 112050
|
| 241929 |
+
},
|
| 241930 |
+
{
|
| 241931 |
+
"epoch": 903.43,
|
| 241932 |
+
"learning_rate": 8.206607431340872e-06,
|
| 241933 |
+
"loss": 0.801,
|
| 241934 |
+
"step": 112055
|
| 241935 |
+
},
|
| 241936 |
+
{
|
| 241937 |
+
"epoch": 903.47,
|
| 241938 |
+
"learning_rate": 8.206526655896608e-06,
|
| 241939 |
+
"loss": 0.2984,
|
| 241940 |
+
"step": 112060
|
| 241941 |
+
},
|
| 241942 |
+
{
|
| 241943 |
+
"epoch": 903.51,
|
| 241944 |
+
"learning_rate": 8.206445880452342e-06,
|
| 241945 |
+
"loss": 0.2676,
|
| 241946 |
+
"step": 112065
|
| 241947 |
+
},
|
| 241948 |
+
{
|
| 241949 |
+
"epoch": 903.55,
|
| 241950 |
+
"learning_rate": 8.206365105008078e-06,
|
| 241951 |
+
"loss": 0.3573,
|
| 241952 |
+
"step": 112070
|
| 241953 |
+
},
|
| 241954 |
+
{
|
| 241955 |
+
"epoch": 903.59,
|
| 241956 |
+
"learning_rate": 8.206284329563812e-06,
|
| 241957 |
+
"loss": 0.6617,
|
| 241958 |
+
"step": 112075
|
| 241959 |
+
},
|
| 241960 |
+
{
|
| 241961 |
+
"epoch": 903.63,
|
| 241962 |
+
"learning_rate": 8.206203554119548e-06,
|
| 241963 |
+
"loss": 0.8261,
|
| 241964 |
+
"step": 112080
|
| 241965 |
+
},
|
| 241966 |
+
{
|
| 241967 |
+
"epoch": 903.67,
|
| 241968 |
+
"learning_rate": 8.206122778675282e-06,
|
| 241969 |
+
"loss": 0.2804,
|
| 241970 |
+
"step": 112085
|
| 241971 |
+
},
|
| 241972 |
+
{
|
| 241973 |
+
"epoch": 903.71,
|
| 241974 |
+
"learning_rate": 8.206042003231018e-06,
|
| 241975 |
+
"loss": 0.3004,
|
| 241976 |
+
"step": 112090
|
| 241977 |
+
},
|
| 241978 |
+
{
|
| 241979 |
+
"epoch": 903.75,
|
| 241980 |
+
"learning_rate": 8.205961227786754e-06,
|
| 241981 |
+
"loss": 0.3085,
|
| 241982 |
+
"step": 112095
|
| 241983 |
+
},
|
| 241984 |
+
{
|
| 241985 |
+
"epoch": 903.79,
|
| 241986 |
+
"learning_rate": 8.205880452342488e-06,
|
| 241987 |
+
"loss": 0.6766,
|
| 241988 |
+
"step": 112100
|
| 241989 |
+
},
|
| 241990 |
+
{
|
| 241991 |
+
"epoch": 903.83,
|
| 241992 |
+
"learning_rate": 8.205799676898224e-06,
|
| 241993 |
+
"loss": 0.8378,
|
| 241994 |
+
"step": 112105
|
| 241995 |
+
},
|
| 241996 |
+
{
|
| 241997 |
+
"epoch": 903.87,
|
| 241998 |
+
"learning_rate": 8.205718901453958e-06,
|
| 241999 |
+
"loss": 0.3095,
|
| 242000 |
+
"step": 112110
|
| 242001 |
+
},
|
| 242002 |
+
{
|
| 242003 |
+
"epoch": 903.91,
|
| 242004 |
+
"learning_rate": 8.205638126009694e-06,
|
| 242005 |
+
"loss": 0.2864,
|
| 242006 |
+
"step": 112115
|
| 242007 |
+
},
|
| 242008 |
+
{
|
| 242009 |
+
"epoch": 903.95,
|
| 242010 |
+
"learning_rate": 8.205557350565428e-06,
|
| 242011 |
+
"loss": 0.3575,
|
| 242012 |
+
"step": 112120
|
| 242013 |
+
},
|
| 242014 |
+
{
|
| 242015 |
+
"epoch": 903.99,
|
| 242016 |
+
"learning_rate": 8.205476575121164e-06,
|
| 242017 |
+
"loss": 0.8758,
|
| 242018 |
+
"step": 112125
|
| 242019 |
+
},
|
| 242020 |
+
{
|
| 242021 |
+
"epoch": 904.0,
|
| 242022 |
+
"eval_loss": 0.4589906334877014,
|
| 242023 |
+
"eval_runtime": 43.1012,
|
| 242024 |
+
"eval_samples_per_second": 19.419,
|
| 242025 |
+
"eval_steps_per_second": 0.626,
|
| 242026 |
+
"eval_wer": 0.1792831541218638,
|
| 242027 |
+
"step": 112126
|
| 242028 |
+
},
|
| 242029 |
+
{
|
| 242030 |
+
"epoch": 897.03,
|
| 242031 |
+
"learning_rate": 8.205395799676898e-06,
|
| 242032 |
+
"loss": 0.3184,
|
| 242033 |
+
"step": 112130
|
| 242034 |
+
},
|
| 242035 |
+
{
|
| 242036 |
+
"epoch": 897.07,
|
| 242037 |
+
"learning_rate": 8.205315024232634e-06,
|
| 242038 |
+
"loss": 0.281,
|
| 242039 |
+
"step": 112135
|
| 242040 |
+
},
|
| 242041 |
+
{
|
| 242042 |
+
"epoch": 897.11,
|
| 242043 |
+
"learning_rate": 8.205234248788368e-06,
|
| 242044 |
+
"loss": 0.3728,
|
| 242045 |
+
"step": 112140
|
| 242046 |
+
},
|
| 242047 |
+
{
|
| 242048 |
+
"epoch": 897.15,
|
| 242049 |
+
"learning_rate": 8.205153473344104e-06,
|
| 242050 |
+
"loss": 0.3522,
|
| 242051 |
+
"step": 112145
|
| 242052 |
+
},
|
| 242053 |
+
{
|
| 242054 |
+
"epoch": 897.19,
|
| 242055 |
+
"learning_rate": 8.20507269789984e-06,
|
| 242056 |
+
"loss": 0.8106,
|
| 242057 |
+
"step": 112150
|
| 242058 |
+
},
|
| 242059 |
+
{
|
| 242060 |
+
"epoch": 897.23,
|
| 242061 |
+
"learning_rate": 8.204991922455574e-06,
|
| 242062 |
+
"loss": 0.7244,
|
| 242063 |
+
"step": 112155
|
| 242064 |
+
},
|
| 242065 |
+
{
|
| 242066 |
+
"epoch": 897.27,
|
| 242067 |
+
"learning_rate": 8.20491114701131e-06,
|
| 242068 |
+
"loss": 0.2733,
|
| 242069 |
+
"step": 112160
|
| 242070 |
+
},
|
| 242071 |
+
{
|
| 242072 |
+
"epoch": 897.31,
|
| 242073 |
+
"learning_rate": 8.204830371567044e-06,
|
| 242074 |
+
"loss": 0.3499,
|
| 242075 |
+
"step": 112165
|
| 242076 |
+
},
|
| 242077 |
+
{
|
| 242078 |
+
"epoch": 897.35,
|
| 242079 |
+
"learning_rate": 8.20474959612278e-06,
|
| 242080 |
+
"loss": 0.3831,
|
| 242081 |
+
"step": 112170
|
| 242082 |
+
},
|
| 242083 |
+
{
|
| 242084 |
+
"epoch": 897.39,
|
| 242085 |
+
"learning_rate": 8.204668820678514e-06,
|
| 242086 |
+
"loss": 0.8812,
|
| 242087 |
+
"step": 112175
|
| 242088 |
+
},
|
| 242089 |
+
{
|
| 242090 |
+
"epoch": 897.43,
|
| 242091 |
+
"learning_rate": 8.20458804523425e-06,
|
| 242092 |
+
"loss": 0.5804,
|
| 242093 |
+
"step": 112180
|
| 242094 |
+
},
|
| 242095 |
+
{
|
| 242096 |
+
"epoch": 897.47,
|
| 242097 |
+
"learning_rate": 8.204507269789984e-06,
|
| 242098 |
+
"loss": 0.2352,
|
| 242099 |
+
"step": 112185
|
| 242100 |
+
},
|
| 242101 |
+
{
|
| 242102 |
+
"epoch": 897.51,
|
| 242103 |
+
"learning_rate": 8.20442649434572e-06,
|
| 242104 |
+
"loss": 0.3085,
|
| 242105 |
+
"step": 112190
|
| 242106 |
+
},
|
| 242107 |
+
{
|
| 242108 |
+
"epoch": 897.55,
|
| 242109 |
+
"learning_rate": 8.204345718901454e-06,
|
| 242110 |
+
"loss": 0.4456,
|
| 242111 |
+
"step": 112195
|
| 242112 |
+
},
|
| 242113 |
+
{
|
| 242114 |
+
"epoch": 897.59,
|
| 242115 |
+
"learning_rate": 8.20426494345719e-06,
|
| 242116 |
+
"loss": 1.0291,
|
| 242117 |
+
"step": 112200
|
| 242118 |
+
},
|
| 242119 |
+
{
|
| 242120 |
+
"epoch": 897.63,
|
| 242121 |
+
"learning_rate": 8.204184168012925e-06,
|
| 242122 |
+
"loss": 0.6779,
|
| 242123 |
+
"step": 112205
|
| 242124 |
+
},
|
| 242125 |
+
{
|
| 242126 |
+
"epoch": 897.67,
|
| 242127 |
+
"learning_rate": 8.20410339256866e-06,
|
| 242128 |
+
"loss": 0.3265,
|
| 242129 |
+
"step": 112210
|
| 242130 |
+
},
|
| 242131 |
+
{
|
| 242132 |
+
"epoch": 897.71,
|
| 242133 |
+
"learning_rate": 8.204022617124395e-06,
|
| 242134 |
+
"loss": 0.3081,
|
| 242135 |
+
"step": 112215
|
| 242136 |
+
},
|
| 242137 |
+
{
|
| 242138 |
+
"epoch": 897.75,
|
| 242139 |
+
"learning_rate": 8.20394184168013e-06,
|
| 242140 |
+
"loss": 0.3493,
|
| 242141 |
+
"step": 112220
|
| 242142 |
+
},
|
| 242143 |
+
{
|
| 242144 |
+
"epoch": 897.79,
|
| 242145 |
+
"learning_rate": 8.203861066235865e-06,
|
| 242146 |
+
"loss": 1.0938,
|
| 242147 |
+
"step": 112225
|
| 242148 |
+
},
|
| 242149 |
+
{
|
| 242150 |
+
"epoch": 897.83,
|
| 242151 |
+
"learning_rate": 8.2037802907916e-06,
|
| 242152 |
+
"loss": 0.6269,
|
| 242153 |
+
"step": 112230
|
| 242154 |
+
},
|
| 242155 |
+
{
|
| 242156 |
+
"epoch": 897.87,
|
| 242157 |
+
"learning_rate": 8.203699515347335e-06,
|
| 242158 |
+
"loss": 0.2867,
|
| 242159 |
+
"step": 112235
|
| 242160 |
+
},
|
| 242161 |
+
{
|
| 242162 |
+
"epoch": 897.91,
|
| 242163 |
+
"learning_rate": 8.20361873990307e-06,
|
| 242164 |
+
"loss": 0.2649,
|
| 242165 |
+
"step": 112240
|
| 242166 |
+
},
|
| 242167 |
+
{
|
| 242168 |
+
"epoch": 897.95,
|
| 242169 |
+
"learning_rate": 8.203537964458805e-06,
|
| 242170 |
+
"loss": 0.3792,
|
| 242171 |
+
"step": 112245
|
| 242172 |
+
},
|
| 242173 |
+
{
|
| 242174 |
+
"epoch": 897.99,
|
| 242175 |
+
"learning_rate": 8.20345718901454e-06,
|
| 242176 |
+
"loss": 0.8685,
|
| 242177 |
+
"step": 112250
|
| 242178 |
+
},
|
| 242179 |
+
{
|
| 242180 |
+
"epoch": 898.0,
|
| 242181 |
+
"eval_loss": 0.3989468812942505,
|
| 242182 |
+
"eval_runtime": 41.9774,
|
| 242183 |
+
"eval_samples_per_second": 19.939,
|
| 242184 |
+
"eval_steps_per_second": 0.643,
|
| 242185 |
+
"eval_wer": 0.18449336594766882,
|
| 242186 |
+
"step": 112251
|
| 242187 |
}
|
| 242188 |
],
|
| 242189 |
+
"max_steps": 625000,
|
| 242190 |
"num_train_epochs": 5000,
|
| 242191 |
+
"total_flos": 3.1590504384111575e+20,
|
| 242192 |
"trial_name": null,
|
| 242193 |
"trial_params": null
|
| 242194 |
}
|
model-bin/finetune/base/{checkpoint-112002 → checkpoint-112251}/training_args.bin
RENAMED
|
File without changes
|
model-bin/finetune/base/log/1630050545.555799/events.out.tfevents.1630050545.52f5c7e305a3.886.91
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1b85c2968aa83ff0f693eff4476fc2b9181761fb526fed2f883a45abf46c87b1
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/1630051145.811182/events.out.tfevents.1630051145.52f5c7e305a3.886.93
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:335fda2c5e8f4d514c540425f74d9f5cbabdac3af328383abbdd4d89fc0bc2db
|
| 3 |
+
size 4194
|
model-bin/finetune/base/log/events.out.tfevents.1630050545.52f5c7e305a3.886.90
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:77bb63ff76d7de4d2c5a2692cedc707f46a50e1723e17f41e4e37445616375b1
|
| 3 |
+
size 8622
|
model-bin/finetune/base/log/events.out.tfevents.1630051145.52f5c7e305a3.886.92
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:a15adaf57895b016132c8c82eefaac064357dffb3d30c30422b3343a5fdf63d2
|
| 3 |
+
size 8622
|