Hi I am trying to quantize StyleGAN3 model and I found almost all the layers are custom layers thus could not be quantized directly.
I have a few questions about quantization
would quantization be supported in the future?
if I rewrite a StyleGAN3 with native Pytorch layers, would there be any pitfalls or parameters missing which would result in significant performance drop?
Hi I am trying to quantize StyleGAN3 model and I found almost all the layers are custom layers thus could not be quantized directly.
I have a few questions about quantization