RuntimeError: Sizes of tensors must match except in dimension 0. Expected size 4 but got size 3 for tensor number 1 in the list

vision
1

Hi ya,

I got this error while training my NeRF on llff synthetic image dataset.

Error:

RuntimeError: Sizes of tensors must match except in dimension 0. Expected size 4 but got size 3 for tensor number 1 in the list.

Code:

dummy_num = ((len(poses) - 1) // args.num_gpu + 1) * args.num_gpu - len(poses)
dummy_poses = torch.eye(3, 4).unsqueeze(0).expand(dummy_num, 3, 4).type_as(render_poses)
print(f"Append {dummy_num} # of poses to fill all the GPUs")
with torch.no_grad():
  nerf.eval()
   rgbs,_ = nerf(H, W, K, args.chunk, poses=torch.cat([poses, dummy_poses], dim=0).cuda(),
                               render_kwargs=render_kwargs_test)
   rgbs = rgbs[:len(rgbs) - dummy_num]
   rgbs_save = rgbs  # (rgbs - rgbs.min()) / (rgbs.max() - rgbs.min())

The error is in this line

rgbs,_ = nerf(H, W, K, args.chunk, poses=torch.cat([poses, dummy_poses], dim=0).cuda(),
                               render_kwargs=render_kwargs_test)

Could you help, @ptrblck