Convert cuda module (StyleGAN v2 Pytorch) to torchscript (pt)

Hello,

I want to convert .pth model from StyleGAN v2 (pytorch implementation) to .pt (torchscript model) for triton inference with jit trace. How I could do that ?

How I could convert cuda function to torchscript ? fused_bias_act and upfirdn_2d module ?

I have no documentation anywhere to do that.

Thanks a lot.

If you would still like a response, please consider re-posting your question on: Triton Inference Server · GitHub , the NVIDIA and other teams will be able to help you there.
Sorry for the inconvenience and thanks for your patience.