Why does the generalization ability of ffhq dataset training feel very low?

NVlabs / stylegan3

Official PyTorch implementation of StyleGAN3

Other

6.38k stars 1.12k forks source link

Why does the generalization ability of ffhq dataset training feel very low? #176

Open huixingaidiqiu opened 2 years ago

huixingaidiqiu commented 2 years ago

This is the parameter I have trained for half a month Is my parameter wrong?

python train.py --outdir=training-runs --cfg=stylegan3-t --data=datasets/dent.zip --gpus=4 --batch=16 --gamma=10 --mirror=1 --kimg=5000 --snap=10 --workers=10 --metrics=none

Can give a stylegan3-t-ffhq-1024x1024.pkl Training parameters?