Backward for rasterization_inria_wrapper

[ZhouCX117]

Thanks for your awesome work! I want to replace the gslat with inria for some purpose. So I use the rasterization_inria_wrapper function in the rendering.py file. However, I don't know how to get the 'mean2d' grad. Could you please give me some hints? Thanks for your help!

[ZhouCX117]

I use the code
screenspace_points = torch.zeros_like(means, dtype=means_crop.dtype, requires_grad=True, device="cuda") + 0 try: screenspace_points.retain_grad() except: pass
and use the grad of screenspace_points, but the obtained results are weird.