torch currently only supports 2d pixel shuffle from https://arxiv.org/abs/1609.05158. This is the 1d version.

torch_pixel_shuffle1d.py

def pixel_shuffle_1d(x, upscale_factor):
    batch_size, channels, steps = x.size()
    channels //= upscale_factor
    input_view = x.contiguous().view(batch_size, channels, upscale_factor, steps)
    shuffle_out = input_view.permute(0, 1, 3, 2).contiguous()
    return shuffle_out.view(batch_size, channels, steps * upscale_factor)

Thanks, this version works ok

Thanks for noticing the bug in the PR implementation! Would probably have just used that code otherwise.