Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

训练的loss不收敛 #259

Open
kingrt666666 opened this issue Apr 26, 2024 · 1 comment
Open

训练的loss不收敛 #259

kingrt666666 opened this issue Apr 26, 2024 · 1 comment

Comments

@kingrt666666
Copy link

老师您好!为什么使用VIG_S训练自己的数据的时候loss一直在震荡而没办法收敛呢。
微信截图_20240426224743
下面是输出的summary.xls文件:

<style> </style>
epoch train_loss eval_loss eval_top1 eval_top5
         
0 1.467884 1.461614 22.4 100
         
1 1.395868 1.445047 23.475 100
         
2 1.380566 1.396237 26.65 100
         
3 1.38526 1.339708 27.55 100
         
4 1.388591 1.308261 26.55 100
         
5 1.397199 1.286827 28.075 100
         
6 1.398853 1.285091 35.15 100
         
7 1.403012 1.296711 43.675 100
         
8 1.407558 1.323002 46.9 100
         
9 1.410531 1.359494 29.75 100
         
10 1.405997 1.372851 25 100
         
11 1.406798 1.385787 25 100
         
12 1.402455 1.389221 25 100
         
13 1.4007 1.38869 25 100
         
14 1.397808 1.387725 21 100
训练的配置是:python train.py ./dataset/ --model vig_s_224_gelu --sched cosine --epochs 300 --opt adamw -j 8 --warmup-lr 1e-6 --mixup .8 --cutmix 1.0 --model-ema --model-ema-decay 0.99996 --aa rand-m9-mstd0.5-inc1 --color-jitter 0.4 --warmup-epochs 20 --opt-eps 1e-8 --repeated-aug --remode pixel --reprob 0.25 --amp --lr 2e-3 --weight-decay .05 --drop 0 --drop-path .1 -b 2 --output H:\\VIG-Pytorch\\vig_pytorch\\output 您能帮我看看吗?麻烦了
@iamhankai
Copy link
Member

数据没有错的话,调调学习率

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

No branches or pull requests

2 participants