About the function to get point_features

[silence-tang]

Thanks for your awesome work! I am new to 3D-aware image synthesis and this work inspired me a lot.
I notice that you use your own function grid_sample_3d() to get point_features instead of using F.grid_sample, and I wonder if there is any idea or insight behind it. Thanks again!
# point_features = F.grid_sample(volume, grid_coords, padding_mode='zeros', align_corners=True)
point_features = grid_sample_3d(volume, grid_coords)