feat: addons for FP8 attention bmm and linear in FMS

[andrea-fasoli]

Description of the change

Implementation of addons FP8 attention BMMs for AIU. Similarly to GPTQ and INT8 addons, the functions build on FMS and expand its functionalities.

Original implementation by @ani300

Components:

• scaled BMM operation registered for torch.compile
• custom scaled attention BMM operation registered with FMS

Related issues or PRs

n/a

How to verify the PR

Code review (including math) is required.
cc: @ani300 @nwang-ibm @chichun-charlie-liu

Was the PR tested

• I have added >=1 unit test(s) for every new method I have added (if that coverage is difficult, please briefly explain the reason)
• I have ensured all unit tests pass

Checklist for passing CI/CD:

• All commits are signed showing "Signed-off-by: Name <email@domain.com>" with git commit -signoff or equivalent
• PR title and commit messages adhere to Conventional Commits
• Contribution is formatted with tox -e fix
• Contribution passes linting with tox -e lint
• Contribution passes spellcheck with tox -e spellcheck
• Contribution passes all unit tests with tox -e unit

Note: CI/CD performs unit tests on multiple versions of Python from a fresh install. There may be differences with your local environment and the test environment.

[andrea-fasoli]

Bringing in FP8 attention and linear modules into FMS-MO.

CI/CD only fails lint at this time. All errors current errors are import errors. They come in two kinds:

• torchao import error -> need to add this to our installed packages
• import of some fms functions -> I'm investigating this, same functions are loaded fine from other addons...

cc: @chichun-charlie-liu @BrandonGroth @tharapalanivel @ani300

[andrea-fasoli]

tox -e lint run locally does not report fms import errors, only torchao errors