Enable CPP/CUDAExtension with py_limited_api for python agnosticism#138088
Closed
janeyx99 wants to merge 12 commits into
Closed
Enable CPP/CUDAExtension with py_limited_api for python agnosticism#138088janeyx99 wants to merge 12 commits into
janeyx99 wants to merge 12 commits into
Conversation
[ghstack-poisoned]
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/138088
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit f00a87b with merge base c30dd35 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
[ghstack-poisoned]
…nosticism" [ghstack-poisoned]
…nosticism" Getting tested with ao [ghstack-poisoned]
janeyx99
commented
Nov 14, 2024
janeyx99
commented
Nov 14, 2024
…nosticism" Getting tested with ao, but now there is a real test i added. [ghstack-poisoned]
…nosticism" Getting tested with ao, but now there is a real test i added. [ghstack-poisoned]
albanD
reviewed
Dec 10, 2024
| constructor. Full list arguments can be found at | ||
| https://setuptools.pypa.io/en/latest/userguide/ext_modules.html#extension-api-reference | ||
|
|
||
| Note! |
Collaborator
There was a problem hiding this comment.
Shouldn't this be ..note: or something like that?
Contributor
Author
There was a problem hiding this comment.
I wasn't swayed either way, the cudaextension one had a big chunk of text so I stuck with it. but yea, can change
Comment on lines
+1033
to
+1040
| If passing in the ``py_limited_api`` flag signifying that the extension uses | ||
| only the Python limited API, CUDAExtension will respect the flag and SKIP | ||
| linking torch_python with your custom extension!!!! This means that your | ||
| extension CANNOT rely on any APIs provided in torch_python!!!! Notably, your | ||
| extension cannot call PYBIND11_MODULE, which is commonly used to create a | ||
| Python module associated with the extension. The APIs in torch_python are | ||
| blatantly NOT within the Python limited API, which is why we avoid linking | ||
| torch_python when the ``py_limited_api`` is specified. |
Collaborator
There was a problem hiding this comment.
I think we could make this shorter and to the point? I feel like there are only 2 ideas here:
- PyTorch python API (as provided in libtorch_python) cannot be built with py_limited_api
- When this flag is passed, it is the user responsibility in their library to not use APIs from libtorch_python (in particular pytorch/python bindings) and only use APIs from libtorch (aten objects, operators and registering operations through the dispatcher for access from python).
Contributor
Author
There was a problem hiding this comment.
oki let me try again!
…nosticism" Getting tested with ao, but now there is a real test i added. ## What does this PR do? We want to allow custom PyTorch extensions to be able to build one wheel for multiple Python versions, in other words, achieve python agnosticism. It turns out that there is such a way that setuptools/Python provides already! Namely, if the user promises to use only the Python limited API in their extension, they can pass in `py_limited_api` to their Extension class and to the bdist_wheel command (with a min python version) in order to build 1 wheel that will suffice across multiple Python versions. Sounds lovely! Why don't people do that already with PyTorch? Well 2 things. This workflow is hardly documented (even searching for python agnostic specifically does not reveal many answers) so I'd expect that people simply don't know about it. But even if they did, _PyTorch_ custom Extensions would still not work because we always link torch_python, which does not abide by py_limited_api rules. So this is where this PR comes in! We respect when the user specifies py_limited_api and skip linking torch_python under that condition, allowing users to enroll in the provided functionality I just described. ## How do I know this PR works? I manually tested my silly little ultra_norm locally (with `import python_agnostic`) and wrote a test case for the extension showing that - torch_python doesn't show up in the ldd tree - no Py- symbols show up It may be a little confusing that our test case is actually python-free (more clean than python-agnostic) but it is sufficient (and not necessary) towards showing that this change works. [ghstack-poisoned]
…nosticism" Getting tested with ao, but now there is a real test i added. ## What does this PR do? We want to allow custom PyTorch extensions to be able to build one wheel for multiple Python versions, in other words, achieve python agnosticism. It turns out that there is such a way that setuptools/Python provides already! Namely, if the user promises to use only the Python limited API in their extension, they can pass in `py_limited_api` to their Extension class and to the bdist_wheel command (with a min python version) in order to build 1 wheel that will suffice across multiple Python versions. Sounds lovely! Why don't people do that already with PyTorch? Well 2 things. This workflow is hardly documented (even searching for python agnostic specifically does not reveal many answers) so I'd expect that people simply don't know about it. But even if they did, _PyTorch_ custom Extensions would still not work because we always link torch_python, which does not abide by py_limited_api rules. So this is where this PR comes in! We respect when the user specifies py_limited_api and skip linking torch_python under that condition, allowing users to enroll in the provided functionality I just described. ## How do I know this PR works? I manually tested my silly little ultra_norm locally (with `import python_agnostic`) and wrote a test case for the extension showing that - torch_python doesn't show up in the ldd tree - no Py- symbols show up It may be a little confusing that our test case is actually python-free (more clean than python-agnostic) but it is sufficient (and not necessary) towards showing that this change works. [ghstack-poisoned]
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 |
Collaborator
Merge failedReason: 1 mandatory check(s) failed. The first few are: Dig deeper by viewing the failures on hud |
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 |
Contributor
Author
|
@pytorchbot cherry-pick --onto release/2.6 -c fixnewfeature This fixes py_limited_api usage for torchao to be able to be python agnostic in their releases, as their releases depend on stable torch to build. |
pytorchbot
pushed a commit
that referenced
this pull request
Dec 17, 2024
…138088) Getting tested with ao, but now there is a real test i added. ## What does this PR do? We want to allow custom PyTorch extensions to be able to build one wheel for multiple Python versions, in other words, achieve python agnosticism. It turns out that there is such a way that setuptools/Python provides already! Namely, if the user promises to use only the Python limited API in their extension, they can pass in `py_limited_api` to their Extension class and to the bdist_wheel command (with a min python version) in order to build 1 wheel that will suffice across multiple Python versions. Sounds lovely! Why don't people do that already with PyTorch? Well 2 things. This workflow is hardly documented (even searching for python agnostic specifically does not reveal many answers) so I'd expect that people simply don't know about it. But even if they did, _PyTorch_ custom Extensions would still not work because we always link torch_python, which does not abide by py_limited_api rules. So this is where this PR comes in! We respect when the user specifies py_limited_api and skip linking torch_python under that condition, allowing users to enroll in the provided functionality I just described. ## How do I know this PR works? I manually tested my silly little ultra_norm locally (with `import python_agnostic`) and wrote a test case for the extension showing that - torch_python doesn't show up in the ldd tree - no Py- symbols show up It may be a little confusing that our test case is actually python-free (more clean than python-agnostic) but it is sufficient (and not necessary) towards showing that this change works. Pull Request resolved: #138088 Approved by: https://github.com/ezyang, https://github.com/albanD (cherry picked from commit be27dbf)
Collaborator
Cherry picking #138088The cherry pick PR is at #143448 and it is recommended to link a fixnewfeature cherry pick PR with an issue. The following tracker issues are updated: Details for Dev Infra teamRaised by workflow job |
malfet
pushed a commit
that referenced
this pull request
Dec 17, 2024
…143448) Enable CPP/CUDAExtension with py_limited_api for python agnosticism (#138088) Getting tested with ao, but now there is a real test i added. ## What does this PR do? We want to allow custom PyTorch extensions to be able to build one wheel for multiple Python versions, in other words, achieve python agnosticism. It turns out that there is such a way that setuptools/Python provides already! Namely, if the user promises to use only the Python limited API in their extension, they can pass in `py_limited_api` to their Extension class and to the bdist_wheel command (with a min python version) in order to build 1 wheel that will suffice across multiple Python versions. Sounds lovely! Why don't people do that already with PyTorch? Well 2 things. This workflow is hardly documented (even searching for python agnostic specifically does not reveal many answers) so I'd expect that people simply don't know about it. But even if they did, _PyTorch_ custom Extensions would still not work because we always link torch_python, which does not abide by py_limited_api rules. So this is where this PR comes in! We respect when the user specifies py_limited_api and skip linking torch_python under that condition, allowing users to enroll in the provided functionality I just described. ## How do I know this PR works? I manually tested my silly little ultra_norm locally (with `import python_agnostic`) and wrote a test case for the extension showing that - torch_python doesn't show up in the ldd tree - no Py- symbols show up It may be a little confusing that our test case is actually python-free (more clean than python-agnostic) but it is sufficient (and not necessary) towards showing that this change works. Pull Request resolved: #138088 Approved by: https://github.com/ezyang, https://github.com/albanD (cherry picked from commit be27dbf) Co-authored-by: Jane Xu <janeyx@meta.com>
This was referenced Jan 28, 2025
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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
5 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.
Getting tested with ao, but now there is a real test i added.
What does this PR do?
We want to allow custom PyTorch extensions to be able to build one wheel for multiple Python versions, in other words, achieve python agnosticism. It turns out that there is such a way that setuptools/Python provides already! Namely, if the user promises to use only the Python limited API in their extension, they can pass in
py_limited_apito their Extension class and to the bdist_wheel command (with a min python version) in order to build 1 wheel that will suffice across multiple Python versions.Sounds lovely! Why don't people do that already with PyTorch? Well 2 things. This workflow is hardly documented (even searching for python agnostic specifically does not reveal many answers) so I'd expect that people simply don't know about it. But even if they did, PyTorch custom Extensions would still not work because we always link torch_python, which does not abide by py_limited_api rules.
So this is where this PR comes in! We respect when the user specifies py_limited_api and skip linking torch_python under that condition, allowing users to enroll in the provided functionality I just described.
How do I know this PR works?
I manually tested my silly little ultra_norm locally (with
import python_agnostic) and wrote a test case for the extension showing thatIt may be a little confusing that our test case is actually python-free (more clean than python-agnostic) but it is sufficient (and not necessary) towards showing that this change works.
Stack from ghstack (oldest at bottom):