-
Notifications
You must be signed in to change notification settings - Fork 185
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
remove torch dependency #449
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/pytorch/ao/449
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit 94169bc with merge base 2e3ec26 (): This comment was automatically generated by Dr. CI and updates every 15 minutes. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
IMO you need to keep runtime requirement on torch, as torchao indeed needs one
And if you want to warn users about the pytorch version it was build against, you can do it in __init__.py
or something like that
* remove torch dependency * push * push * push * push * push * push * Update pre_build_script.sh * Update version.txt * push
* remove torch dependency * push * push * push * push * push * push * Update pre_build_script.sh * Update version.txt * push
When users are running
This is causing a force downgrade to
torch==2.3.1
Instead we should tell users that if they want CPP extensions working with an official release they should install
torch==2.3.1
in documentation but we should never force downgradeSo with this PR we remove the pytorch dependency completely and instead rely on building
torchao
without build isolation to make suretorch
is available as a build time dependencyThis is an unfortunate workaround because
pyproject.toml
is not expressive enough to pick a dynamic version fortorch
Todo: figure out the right test strategy for this PR