enable cpu integer matmul

[maktukmak]

Regarding issue #64, torch now supports integer matmul (toch._int_mm) on the CPU at its master branch with this PR merged. Here, I made small changes to enable this kernel on quanto. To use the new backend, install torch nightly build after quanto.

[dacorvo]

Thank you for the pull-request ! I'd rather guard the call to int_mm CPU to a minimum pytorch version to avoid errors with current pytorch.
Something like:

from packaging import version

...
if version.parse(torch.__version__) >= version.parse(2.3.0):
    # Use CPU int_mm

[maktukmak]

Got it. I added release version control so it can work with the current nightly build.

[dacorvo]

Looks good to me, but you need to:

• fix style (make style from the root directory),
• rebase on main (no merge please: git rebase <commit-before-your-branch> --onto upstream/main).

[maktukmak]

@dacorvo, I rebased and did make style. Let me know if I missed something

[dacorvo]

Merged as #130

[dacorvo]

@dacorvo, I rebased and did make style. Let me know if I missed something

I don't know how you applied styling, but it was all wrong ... I fixed your branch and created another pull-request.