This is the source code for pytorch, and what does the Tensor stride size mean?

As far as I understand it, the memory format of pytorch tensors is basically the same as that of numpy ndarrays, so I shall give some basic pointers and let the experts fill in any gaps.

I assume you know how strides work in numpy, if not here are some links.

Basic docs https://docs.scipy.org/doc/numpy/reference/generated/numpy.ndarray.strides.html

A more wordy explanation https://www.datacamp.com/community/tutorials/python-numpy-tutorial

A lengthy explanation of numpy C code https://docs.scipy.org/doc/numpy-1.13.0/reference/internals.code-explanations.html

Do you have a more specific question?