Is `to_opengl` a must for sampling env map?

    
           self.to_opengl = torch.tensor([[1, 0, 0], [0, 0, 1], [0, -1, 0]], dtype=torch.float32, device="cuda")

Hi! The ray direction is already in the world coordinate, and I am confused about why we need to do this conversion. Could you explain it?

Looking forward to your help and reply

Hi, I've noticed that nvdiffrast uses the y-axis upward, while we use the z-axis, just to align with nvdiffrast. I think it's also okay without this conversion.

Thanks!