Open WindMirror1108 opened 3 years ago
Hello, I also have this problem. How did you solve it? Thank you for your reply
Probably pretty late but if anyone needs a quick and dirty fix, editing is_custom_kernel_supported(): in util.util.py to always return False seems to work for me, as far as I understand it basically forces the use of inbuilt torch modules for stuff like activation functions instead of the optimized ones based on StyleGAN2 and written in cuda and cpp, this specific error is caused by torch's inability to find the appropriate compiler for those.
how to solve it? thanks