Grayscale images

[povolann]

Hi, I am trying to use EG3D for grayscale images.
When I was using the original
image = image[:, :, np.newaxis] # HW => HWC
it didn't work for me. It gave me this error:

Exception has occurred: AssertionError
Wrong size for dimension 1: got 6, expected 2
  File "/home/anya/Programs/eg3d/eg3d/torch_utils/misc.py", line 97, in assert_shape
    raise AssertionError(f'Wrong size for dimension {idx}: got {size}, expected {ref_size}')
  File "/home/anya/Programs/eg3d/eg3d/training/networks_stylegan2.py", line 621, in forward
    misc.assert_shape(img, [None, self.img_channels, self.resolution, self.resolution]) # where does the img come from?
  File "/home/anya/Programs/eg3d/eg3d/training/dual_discriminator.py", line 164, in forward
    x, img = block(x, img, **block_kwargs)
  File "/home/anya/Programs/eg3d/eg3d/torch_utils/misc.py", line 218, in print_module_summary
    outputs = module(*inputs)
  File "/home/anya/Programs/eg3d/eg3d/training/training_loop.py", line 178, in training_loop
    misc.print_module_summary(D, [img, c])
  File "/home/anya/Programs/eg3d/eg3d/train.py", line 52, in subprocess_fn
    training_loop.training_loop(rank=rank, **c)
  File "/home/anya/Programs/eg3d/eg3d/train.py", line 101, in launch_training
    subprocess_fn(rank=0, c=c, temp_dir=temp_dir)
  File "/home/anya/Programs/eg3d/eg3d/train.py", line 402, in main
    launch_training(c=c, desc=desc, outdir=opts.outdir, dry_run=opts.dry_run)
  File "/home/anya/Programs/eg3d/eg3d/train.py", line 407, in <module>
    main() # pylint: disable=no-value-for-parameter
AssertionError: Wrong size for dimension 1: got 6, expected 2 

Therefore, I changed it to
image = np.stack((image,)*3, axis=-1) # HW => HWC, but C is not 1, but 3
so I treat the images as RGB, but I was hoping to change the code to work with the images like grayscale and I got confused about the used dimensions/tensor shapes.
I have tried to change the img_channels here to 32 * 1 (assuming that the original 32 * 3 had 3 for RGB), but then it doesn't fit with other shapes later in the code which seems to be connected to the triplane size, not the RGB colour. I guess I need to keep the triplane representation (which I guess stands for the 3D geometry, not colour)...
If anyone could explain me a bit more I would be glad. Thank you :)