StyleGAN3 Setup

Paper: https://nvlabs.github.io/stylegan3

Source: https://github.com/NVlabs/stylegan3

Install Visual Studio Community Edition (2017):

https://visualstudio.microsoft.com/vs/community

Add 64-bit compiler into system environment PATH variable:

C:\Program Files (x86)\Microsoft Visual Studio\2017\Community\VC\Tools\MSVC\14.16.27023\bin\Hostx64\x64

Install CUDA (11.6):

https://developer.nvidia.com/cuda-downloads

Install Anaconda3 (2021.11.0):

https://www.anaconda.com/products/individual#windows

From Anaconda Prompt:

conda create --name StyleGAN3 python=3.81

conda activate StyleGAN3

conda install click (4.10.3)

conda install numpy (1.21.5)

conda install requests (2.27.1)

conda install pillow (8.4.0)

conda install scipy (1.7.3)

conda install ninja (1.10.2)

conda install imageio (2.9.0)

conda install tqdm (4.62.3)

conda install colorama (0.4.4)

pip install imageio-ffmpeg (0.4.5)

pip install psutil (5.9.0)

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

Test (pre-trained model):

python gen_images.py --outdir=out --trunc=1 --seeds=2 --network=https://api.ngc.nvidia.com/v2/models/nvidia/research/stylegan3/versions/1/files/stylegan3-r-afhqv2-512x512.pkl

Training (example):

python train.py --outdir=training-runs --cfg=stylegan3-r --data=metfaces.zip --gpus=1 --batch=32 --batch-gpu=4 --mirror=1 --gamma=2 --snap=5 --kimg=5000 --metrics=none

Transfer Learning (example):

python train.py --outdir=training-runs --cfg=stylegan3-r --data=metfaces.zip --gpus=1 --batch=32 --batch-gpu=4 --mirror=1 --gamma=2 --snap=5 --kimg=5000 --metrics=none --resume=training-runs/00000-stylegan3-r-metfaces-gpus1-batch32-gamma2/network-snapshot-000060.pkl