There is the following issue on this page: https://docs.pytorch.org/docs/stable/generated/torch.nn.Conv2d.html#torch.nn.Conv2d

There are a number of variables used in this documentation, and after considerable effort, one can relate those variables to the parameters of the Conv2d function, but the mapping is not made obvious. A tiny section that denotes C_in --> in_channels, C_out --> out_channels, H_in --> kernel_size or kernel_size[0], etc. would be helpful.

There is a small attempt to do something like this in the following paragraph from the documentation:

"where ⋆ is the valid 2D cross-correlation operator, N is a batch size, C denotes a number of channels, H is a height of input planes in pixels, and W is width in pixels."

Yet I don't think it goes far enough. A reader has to figure out on their own that H_in (or H depending on context) is either the parameter kernel_size (if specified as an integer) or the first element of kernel_size if specified as a tuple. The information is eventually made available when describing the specific parameters directly, but it's not obvious when first looking at the initial summation formula.

In addition, the link for the cross-correlation operator is rather misleading. It links to the Wikipedia article, which discusses continuous convolution operators rather than the discrete operators expected by pytorch.

cc @svekars @sekyondaMeta @AlannaBurke @albanD @mruberry @jbschlosser @walterddr @mikaylagawarecki