search

autonomousvision / stylegan-xl

[SIGGRAPH'22] StyleGAN-XL: Scaling StyleGAN to Large Diverse Datasets
MIT License
961 stars 113 forks source link

about cifar10 setting #79

Open Youzebin opened 2 years ago

Youzebin commented 2 years ago

hello, can you release your cifar10 setting? I have tried stylegan-xl in cifar10, by training the 16 stem 2wkimg, and 2wking 16->32. cbase 16734, cmax 256 head-layer 4 syn-layer 7

can you release your cifar10 setting?? such as how many kimg train 16 stem and 16->32 and Hyperparameters such as cbase cmax head layer and syn-layer

thank you very much!

Youzebin commented 2 years ago

and my try can get the result 16 stem fid 1.50 and 32 resolution 2.3 It looks a bit worse than in your paper, so I'd like to consult you about the above

mingqiJ commented 1 year ago

Hello, could I get your command of training? because when I use this command to train, the fid is very big:python3 train.py --outdir=./out --cfg=stylegan3-r --cond=True --data=./dataset/cifar10.zip --gpus=4 --batch=128 --mirror=True --batch-gpu 32 --cbase 16734 --cmax 256 --kimg 25000 --syn_layers 7 --head_layers 4