Remove all dependencies except torch

[msaroufim]

1. got rid of the packaging dependency in favor of importlib
2. numpy dependency was implicit with benchmark Timer, so also removed that but needed to ignore a nested tensor warning
3. our only real requirements left is torch
4. everything else should be a dev dependency for testing

Tested locally import torchao which works fine and CI is working fine

Packaging logic also works fine

>>> from importlib.metadata import version
>>> version("torch")
'2.3.1'
>>> version("torch") >= "2.3.0.dev"
True
>>> version("torch") >= "2.4.0.dev"
False

We can also get rid of torch as a dependency, work is tracked here #371

CI failure is unrelated and will get fixed when we merge #377 so will rebase this PR before merge

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/369

• 📄 Preview Python docs built from this PR

Note: Links to docs will display an error until the docs builds have been completed.

✅ No Failures

As of commit 10cd6b0 with merge base 664f073 ():
💚 Looks good so far! There are no failures yet. 💚

This comment was automatically generated by Dr. CI and updates every 15 minutes.

[drisspg]

Is there another timer that could be used?
That being said I can't remember if numpy is a dependency of torch

[msaroufim]

numpy is not supposed to be a dependency of torch but I've seen that assumption broken multiple times - I remember chatting about this with @malfet at some point too

That said the reason why this is needed is because of pytorch/pytorch#128759 which fixes the problem but only in 2.5+ lol

And yes we could just use another timer and drop this dependency completely even from our dev dependencies - although I really need to be writing docs tomorrow so would rather do this in a future PR unless @vkuzo gets to it first

[msaroufim]

Honestly I could also just remove the numpy dependency from dev dependencies because it's only used for a benchmark for mx that doesn't run in CI