Use a custom Symbol class for performance #135651
Closed
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/135651
Note: Links to docs will display an error until the docs builds have been completed. ❌ 24 New FailuresAs of commit bf891a1 with merge base d2207c5 ( NEW FAILURES - The following jobs have failed:
This comment was automatically generated by Dr. CI and updates every 15 minutes. |
pytorch-bot
bot
added
ciflow/inductor
module: cpu
ezyang
reviewed
Sep 11, 2024
| def make_symbol(prefix: SymT, idx: int, **kwargs) -> sympy.Symbol: | ||
| # TODO: maybe put the assumptions here directly | ||
| return sympy.Symbol(f"{prefix_str[prefix]}{idx}", **kwargs) | ||
| class PrefixedSymbol(sympy.Symbol): |
ezyang
reviewed
Sep 11, 2024
| else: | ||
| return name_str.startswith(tuple(prefix_str[p] for p in prefix)) | ||
| assert isinstance(sym, sympy.Symbol) | ||
| # TODO: remove all these Symbols and use PrefixedSymbol |
Contributor
There was a problem hiding this comment.
lol, I think I did think about storing the prefix explicitly and then gave up when trying to figure out how to update everyone.
ezyang
reviewed
Sep 11, 2024
| return name_str.startswith(tuple(prefix_str[p] for p in prefix)) | ||
| assert isinstance(sym, sympy.Symbol) | ||
| # TODO: remove all these Symbols and use PrefixedSymbol | ||
| sym_name = sym.name.lower() |
Contributor
There was a problem hiding this comment.
I hate this so frickin much lol. @shunting314, @jansel let's not take changes like this please lol (c.f. #127342 )
ezyang
reviewed
Sep 11, 2024
torch/utils/_sympy/symbol.py
Outdated
| def symbol_is_type(sym: sympy.Basic, prefix: Union[SymT, Tuple[SymT, ...]]) -> bool: | ||
| if isinstance(sym, PrefixedSymbol): | ||
| # This function is called a *lot* of times, so it needs to be fast | ||
| if type(prefix) == tuple: |
Contributor
There was a problem hiding this comment.
maybe is here is slightly faster?
Contributor
There was a problem hiding this comment.
Hmm, do you think we can afford an assert type(prefix) == SymT in the else case? Because otherwise I'm afraid of someone feeding us a list and not have typechecking on lol
Collaborator
Author
There was a problem hiding this comment.
Didn't see any measurable difference. Fixed now
ezyang
approved these changes
Sep 11, 2024
isuruf
added a commit
to isuruf/pytorch
that referenced
this pull request
Sep 23, 2024
ghstack-source-id: 08df448 Pull Request resolved: pytorch#135651
Contributor
|
So is this hopeless or is there still hope haha |
Collaborator
Author
|
I don't think this is easy to do this across the codebase. I'll close this for now and open it later if the performance becomes too much of a bottleneck. |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Stack from ghstack (oldest at bottom):
cc @jgong5 @mingfeima @XiaobingSuper @sanchitintel @ashokei @jingxu10 @voznesenskym @penguinwu @EikanWang @Guobing-Chen @zhuhaozhe @blzheng @wenzhe-nrv @jiayisunx @ipiszy @yf225 @chenyang78 @kadeng @muchulee8 @ColinPeppler @amjames @desertfire @chauhang