Where is the actual implementation/source code for nn.PixelShuffle?

Anshumaan_Dash · January 21, 2022, 10:01am

I can find just a docstring when I go to torch.nn.functional.pixel_shuffle. Where can I find the actual implementation?

github.com

pytorch/pytorch/blob/master/torch/nn/modules/pixelshuffle.py

from .module import Module
from .. import functional as F

from torch import Tensor


class PixelShuffle(Module):
    r"""Rearranges elements in a tensor of shape :math:`(*, C \times r^2, H, W)`
    to a tensor of shape :math:`(*, C, H \times r, W \times r)`, where r is an upscale factor.

    This is useful for implementing efficient sub-pixel convolution
    with a stride of :math:`1/r`.

    See the paper:
    `Real-Time Single Image and Video Super-Resolution Using an Efficient Sub-Pixel Convolutional Neural Network`_
    by Shi et. al (2016) for more details.

    Args:
        upscale_factor (int): factor to increase spatial resolution by

This file has been truncated. show original

ptrblck · January 21, 2022, 7:53pm

The implementation can be found here.

Anshumaan_Dash · January 24, 2022, 12:41pm

@ptrblck Thanks a bunch!!

Anshumaan_Dash · January 25, 2022, 7:29am

Hey,
I cannot find the cpp file in the installed packages anywhere. I want to change the implementation locally. Where can I do that?

ptrblck · January 25, 2022, 7:32am

You won’t be able to locally change a C++ source file and see the changes without rebuilding it.
To apply changes to the PixelShuffle implementation, use this guide to build PyTorch from source, manipulate the file, and rebuild it.
This Contributing README might also be useful for incremental builds.

Anshumaan_Dash · January 25, 2022, 7:36am

Thanks a lot for helping out. Really appreciate the effort.