Bug for progressive_nerf_only

[zhanglonghao1992]

When setting prog_nerf_only to true, we should upsample x and img to resolution_vol:

StyleNeRF/training/stylenerf.py

Lines 1848 to 1851 in a1f6094

	if img is not None:
	if self.progressive_nerf_only and (img.size(-1) < self.resolution_vol):
	x = upsample(x, self.resolution_vol)
	img = upsample(img, self.resolution_vol)

However, there is a bug in the function upsample:

StyleNeRF/dnnlib/geometry.py

Lines 26 to 37 in a1f6094

	def upsample(img_nerf, size, filter=None):
	up = size // img_nerf.size(-1)
	if up <= 1:
	return img_nerf
	if filter is not None:
	from torch_utils.ops import upfirdn2d
	for _ in range(int(math.log2(up))):
	img_nerf = upfirdn2d.downsample2d(img_nerf, filter, up=2)
	else:
	img_nerf = F.interpolate(img_nerf, (size, size), mode='bilinear', align_corners=False)
	return img_nerf

For example, when the img_size gradually increases to 65 and the resolution_vol is 128, the upsampling operation is not performed, which will lead to subsequent size errors.

[MultiPath]

Thanks for pointing. This is an argument once I hoped to reproduce EG3D's operation. However, I could not get any good results and did not check carefully on this again. It might be incompatible now.