-
Notifications
You must be signed in to change notification settings - Fork 26.3k
Build magma binary tarballs for various cuda #139888
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
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/139888
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit e2987ff with merge base 3abbde9 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
4549d3f to
8518b99
Compare
a8f0c87 to
c1c4fd6
Compare
atalman
left a comment
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 looks great! Please create another PR removing these magma builds from builder repo.
Also would make sence to move the windows magma builds workflows here as well.
Thanks - I need to check what the Windows will look like without conda, once I figure it out I will make another PR like this one. I don't have a windows environment handy to test the Windows builds though, so there may be a bit more of trial and error there. |
|
@atalman would it make sense to configure the |
c1c4fd6 to
76e070e
Compare
It looks like the windows path does not use conda, so it should be only a matter of copying the files over and fixing whatever linter rules break in the new repo. |
I think so magma builds are pretty much standalone. They are used in Docker builds after being published. |
76e070e to
1a7d9c4
Compare
malfet
left a comment
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.
LGTM, will edit PR description to specify specific commit those magma scripts are picked from
Add a version of the manylinux 2.28 image with cuda 12.6. Once this is done, cuda 12.6 can be enable for the new magma non-conda distribution provided by pytorch#139888 Signed-off-by: Andrea Frittoli <[email protected]>
813278e to
11b6f8e
Compare
11b6f8e to
9376f96
Compare
da548c0 to
f7e02ad
Compare
Add a version of the manylinux 2.28 image with cuda 12.6. Once this is done, cuda 12.6 can be enable for the new magma non-conda distribution provided by #139888 Partially-fixes #139397 Pull Request resolved: #139909 Approved by: https://github.com/atalman
|
@pytorchbot merge -f "signal is green" |
Merge startedYour change will be merged immediately since you used the force (-f) flag, bypassing any CI checks (ETA: 1-5 minutes). Please use Learn more about merging in the wiki. Questions? Feedback? Please reach out to the PyTorch DevX Team |
Copy the magma for windows job and script from pytorch/builder https://github.com/pytorch/builder/blob/c9aac65e12734a16d9c285c366bcbf6cc0c67e43/.github/workflows/build-magma-windows.yml The linux version is moved here in pytorch#139888 Use the upload_aws_ossci.sh upload script used by the linux magma builds as well. Signed-off-by: Andrea Frittoli <[email protected]>
Add a version of the manylinux 2.28 image with cuda 12.6. Once this is done, cuda 12.6 can be enable for the new magma non-conda distribution provided by pytorch#139888 Partially-fixes pytorch#139397 Pull Request resolved: pytorch#139909 Approved by: https://github.com/atalman
Copy the magma for windows job and script from pytorch/builder https://github.com/pytorch/builder/blob/c9aac65e12734a16d9c285c366bcbf6cc0c67e43/.github/workflows/build-magma-windows.yml The linux version is moved here in #139888 Fixes #140001 Pull Request resolved: #139924 Approved by: https://github.com/atalman
Add a version of the manylinux 2.28 image with cuda 12.6. Once this is done, cuda 12.6 can be enable for the new magma non-conda distribution provided by pytorch#139888 Partially-fixes pytorch#139397 Pull Request resolved: pytorch#139909 Approved by: https://github.com/atalman
This is a first step towards removing builds dependency to conda. Currently we build magma as a conda package in a pytorch conda channel, implemented in https://github.com/pytorch/builder/tree/a1b372dbda2e9e3bd946cf135aa3b0137dfdf052/magma. This commit adapts the logic from pytorch/builder as follows: - use pytorch/manylinux-cuda<cuda-version> as base image - apply patches and invoke the build.sh script directly (not anymore through conda build) - stores license and build files along with the built artifact, in an info subfolder - create a tarball file which resembles that created by conda, without any conda-specific metadata A new matrix workflow is added, which runs the build for each supported cuda version, and uploads the binaries to pyorch s3 bucket. For the upload, define an upload.sh script, which will be used by the magma windows job as well, to upload to `s3://ossci-*` buckets. The build runs on PR and push, upload runs in DRY_RUN mode in case of PR. Fixes pytorch#139397 Pull Request resolved: pytorch#139888 Approved by: https://github.com/atalman, https://github.com/malfet, https://github.com/seemethere
Copy the magma for windows job and script from pytorch/builder https://github.com/pytorch/builder/blob/c9aac65e12734a16d9c285c366bcbf6cc0c67e43/.github/workflows/build-magma-windows.yml The linux version is moved here in pytorch#139888 Fixes pytorch#140001 Pull Request resolved: pytorch#139924 Approved by: https://github.com/atalman
Resolves #151707 for CUDA Nvidia Magma builds. Removes deprecated cuda 12.4 build. Using `pytorch/manylinux2_28-builder` image for magma build creates circular dependency. For a while for magma builds we used `conda-builder` image since it does not have circular dependency: https://github.com/pytorch/builder/blob/release/2.4/magma/Makefile#L13 However during migration to pytorch/pytorch: #139888 we introduced circular dependency using Manylinux 2.28 docker image. Hence using almalinux image which suppose to be general usage image Please note: Magma builds using Docker build : https://github.com/pytorch/pytorch/blob/main/.ci/magma/README.md we can look into migrating them to Docker images if required as a followup BE change if needed TODO: Make same change for rocm builds. I believe some more work for rocm is required, since maga-rocm is requires rocm dev, utils and lib to be installed : https://github.com/pytorch/pytorch/blob/main/.ci/docker/common/install_rocm.sh Pull Request resolved: #152358 Approved by: https://github.com/nWEIdia, https://github.com/huydhn
This is a first step towards removing builds dependency to conda.
Currently we build magma as a conda package in a pytorch conda channel, implemented in https://github.com/pytorch/builder/tree/a1b372dbda2e9e3bd946cf135aa3b0137dfdf052/magma.
This commit adapts the logic from pytorch/builder as follows:
A new matrix workflow is added, which runs the build for each supported cuda version, and uploads the binaries to pyorch s3 bucket.
For the upload, define an upload.sh script, which will be used by the magma windows job as well, to upload to
s3://ossci-*buckets.The build runs on PR and push, upload runs in DRY_RUN mode in case of PR.
Fixes #139397