Do not import sympy within torch._prims_common

[lezcano]

Stack from ghstack (oldest at bottom):

• -> Do not import sympy within torch._prims_common #112034

This is the first of a few PRs that avoid importing SymPy at import time.
The pitch here is that we (almost!) do not have SymPy on our API, so
this should be feasible.

This should speed-up torch imports by a good 15% as per
https://dev-discuss.pytorch.org/t/delving-into-what-happens-when-you-import-torch/1589

In this PR we just move a few global imports into local imports.

[pytorch-bot]

🔗 Helpful Links

🧪 See artifacts and rendered test results at hud.pytorch.org/pr/112034

• 📄 Preview Python docs built from this PR
• 📄 Preview C++ docs built from this PR
• ❓ Need help or want to give feedback on the CI? Visit the bot commands wiki or our office hours

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

✅ No Failures

As of commit 260cee0 with merge base edafe2d ():
💚 Looks good so far! There are no failures yet. 💚

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

[lezcano]

@malfet @ezyang This may may need some minor tweaking, but as it stands, it works and we get a hefty 15% speed-up on import times.
What do you guys think about this? All of it was done via a few local imports, a TYPE_CHECK and two PRs that do some code movement.

[ezyang]

I'm not fundamentally opposed to this, but I don't see how we're going to manage to keep enforcing it, it's very easy to mess this up.

[lezcano]

I add a test in the last PR of this stack, have a look

[mthrok]

I'm not fundamentally opposed to this, but I don't see how we're going to manage to keep enforcing it, it's very easy to mess this up.

I found TensorFlow's LazyLoader an elegant implementation of lazy import. It uses types.ModuleType and defining __getattr__, then updating self.__dict__. Perhaps something similar can be used as a standard way to access the module.

https://github.com/tensorflow/tensorflow/blob/51b601fa6bb7e801c0b6ae73c25580e40a8b5745/tensorflow/python/util/lazy_loader.py#L24

[lezcano]

@pytorchbot merge -r

[pytorchmergebot]

@pytorchbot started a rebase job onto refs/remotes/origin/viable/strict. Check the current status here

[pytorchmergebot]

Successfully rebased gh/Lezcano/243/orig onto refs/remotes/origin/viable/strict, please pull locally before adding more changes (for example, via ghstack checkout https://github.com/pytorch/pytorch/pull/112034)

[pytorchmergebot]

Merge started

Your change will be merged once all checks pass (ETA 0-4 Hours).

Learn more about merging in the wiki.

Questions? Feedback? Please reach out to the PyTorch DevX Team

Advanced Debugging

Check the merge workflow status
here