-
Notifications
You must be signed in to change notification settings - Fork 250
Migrate extension-cpp to stable API/ABI #126
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
base: master
Are you sure you want to change the base?
Conversation
98ea688 to
75df317
Compare
|
|
||
| echo '::group::Install extension-cpp' | ||
| python setup.py develop | ||
| pip install -e . --no-build-isolation |
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.
This change was needed to get CI to run properly.
Otherwise, CI was installing the nightly but using torch 2.7 to build the extension
| runner: linux.g5.4xlarge.nvidia.gpu | ||
| gpu-arch-type: cuda | ||
| gpu-arch-version: "12.4" | ||
| gpu-arch-version: "12.9" |
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.
we no longer have cu nightlies for 12.4
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.
lol CI for this repo is so outdated
07ef7e7 to
6f2afea
Compare
6f2afea to
ffab4c0
Compare
| # C++/CUDA Extensions in PyTorch with LibTorch Stable ABI | ||
|
|
||
| An example of writing a C++/CUDA extension for PyTorch using the [LibTorch Stable ABI](https://pytorch.org/docs/main/notes/libtorch_stable_abi.html). | ||
| See [here](https://pytorch.org/tutorials/advanced/cpp_custom_ops.html) for the accompanying tutorial. |
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.
are we pushing for everyone to use libtorch stable abi? Or do we expect people will want to use the regular api? How do people select between the two?
IMO if we're not ready for literally everyone to move to libtorch stable abi right now, we should provide multiple examples. One for libtorch stable, and one without.
Accompanying tutorials change pytorch/tutorials#3701