The code is freely available and on Youtube you find videos on „How to use Stylegan“.

But:

• As already mentioned you‘ll need more data. 5-10k are already at the low end. You need more like 50-100k I think.
• You need some serious hardware to train something like this. For Stylegan they used 8 high end GPUs for a week. I guess for Stylegan2/3 even more. You might be happy with less than NVIDIA, but it‘s no quick thing to do.
• And even then there‘s no guarantee that it will turn out perfectly. It might need some fine-tuning, it‘s not a simple algorithm.