Use OrderedSet in _functorch/partitioners #146102
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
Summary: Test Plan: [ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/146102
Note: Links to docs will display an error until the docs builds have been completed. ❗ 1 Active SEVsThere are 1 currently active SEVs. If your PR is affected, please view them below: ✅ No FailuresAs of commit fd75193 with merge base 354fe48 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
masnesral
changed the title
masnesral
changed the title
Copied from #145024, but made _all_ sets OrderedSets and fixed mypy complaints. Note that this change does not fix the non-determinism we're seeing in the internal model. But let's at least eliminate this potential source of non-determinism before investigating any changes to the mincut approach? [ghstack-poisoned]
Copied from #145024, but made _all_ sets OrderedSets and fixed mypy complaints. Note that this change does not fix the non-determinism we're seeing in the internal model. But let's at least eliminate this potential source of non-determinism before investigating any changes to the mincut approach? [ghstack-poisoned]
Copied from #145024, but made _all_ sets OrderedSets and fixed mypy complaints. Note that this change does not fix the non-determinism we're seeing in the internal model. But let's at least eliminate this potential source of non-determinism before investigating any changes to the mincut approach? [ghstack-poisoned]
oulgen
approved these changes
Jan 31, 2025
torch/_functorch/partitioners.py
Outdated
Comment on lines
1405
to
1409
| OrderedSet(fusible_ops), | ||
| OrderedSet(compute_intensive_ops), | ||
| OrderedSet(random_ops), | ||
| OrderedSet(view_ops), | ||
| OrderedSet(recomputable_ops), |
Contributor
There was a problem hiding this comment.
arent some of these already orderedset?
Copied from #145024, but made _all_ sets OrderedSets and fixed mypy complaints. Note that this change does not fix the non-determinism we're seeing in the internal model. But let's at least eliminate this potential source of non-determinism before investigating any changes to the mincut approach? [ghstack-poisoned]
Skylion007
reviewed
Feb 1, 2025
torch/_functorch/partitioners.py
Outdated
| node.name for node in bw_module.graph.nodes if node.op == "call_function" | ||
| } | ||
| fw_module_nodes = OrderedSet( | ||
| [node.name for node in fw_module.graph.nodes if node.op == "call_function"] |
Collaborator
There was a problem hiding this comment.
These list comprehensions can be generators if you do not want an extra copy (input only needs to be an Iterable)
Copied from #145024, but made _all_ sets OrderedSets and fixed mypy complaints. Note that this change does not fix the non-determinism we're seeing in the internal model. But let's at least eliminate this potential source of non-determinism before investigating any changes to the mincut approach? [ghstack-poisoned]
Copied from #145024, but made _all_ sets OrderedSets and fixed mypy complaints. Note that this change does not fix the non-determinism we're seeing in the internal model. But let's at least eliminate this potential source of non-determinism before investigating any changes to the mincut approach? [ghstack-poisoned]
masnesral
changed the title
Contributor
Author
|
@pytorchbot merge |
Collaborator
Merge startedYour 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 |
masnesral
added a commit
that referenced
this pull request
Feb 5, 2025
In an attempt to make partitioning more deterministic, change all sets in partitioners.py to OrderedSets. Note that this change does not fix the non-determinism we're seeing in the internal model. But let's at least eliminate this potential source of non-determinism before investigating any changes to the mincut approach? Pull Request resolved: #146102 Approved by: https://github.com/oulgen ghstack-source-id: 3cc1c96
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
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):
In an attempt to make partitioning more deterministic, change all sets in partitioners.py to OrderedSets. Note that this change does not fix the non-determinism we're seeing in the internal model. But let's at least eliminate this potential source of non-determinism before investigating any changes to the mincut approach?