ci: Remove manylinux builds for triton, except for XPU #148129
Conversation
🔗 Helpful Links🧪 See artifacts and rendered test results at hud.pytorch.org/pr/148129
Note: Links to docs will display an error until the docs builds have been completed. ✅ No FailuresAs of commit d445a4f with merge base 915b9c8 ( This comment was automatically generated by Dr. CI and updates every 15 minutes. |
| docker-image: "pytorch/manylinux-builder:cpu" | ||
| - device: "xpu" | ||
| docker-image: "pytorch/manylinux2_28-builder:cpu" | ||
| docker-image: ["pytorch/manylinux2_28-builder:cpu"] |
There was a problem hiding this comment.
I believe need to keep:
exclude:
- device: "xpu"
docker-image: "pytorch/manylinux2_28-builder:cpu"
There was a problem hiding this comment.
I'm curious why it's necessary for them to be on the older version? Looks like later on in the script they install a more modern compiler so wouldn't it make sense for them to just use the compiler in the newer image?
There was a problem hiding this comment.
I went ahead and removed the lines that install the newer compiler since the manylinux_2_28 image already includes a modern compiler
There was a problem hiding this comment.
Yes looks like the XPU wheels are not Manylinux 2.28 yet: https://download.pytorch.org/whl/nightly/xpu/torch
We should transition XPU builds to Manylinux 2.28 as well.
cc @EikanWang @chuanqi129
There was a problem hiding this comment.
Maybe @chuanqi129 has an opinion here?
There was a problem hiding this comment.
Looks like the new image uses the correct devtoolset that they want:
[251/385] /opt/rh/gcc-toolset-11/root/usr/bin/c++ -DGTEST_HAS_RTTI=0 -Ithird_party/intel/lib/TritonIntelGPUTransforms
From https://github.com/pytorch/pytorch/actions/runs/13577591113/job/37957297977?pr=148129#step:7:979
There was a problem hiding this comment.
Looks like failures are more related to glibc version:
NVVMDialect.cpp:(.text._ZN4mlir4NVVM9BarrierOp5buildERNS_9OpBuilderERNS_14OperationStateENS_9TypeRangeENS_10ValueRangeEN4llvm8ArrayRefINS_14NamedAttributeEEE+0xfa):
undefined reference to `std::__glibcxx_assert_fail(char const*, int, char const*, char const*)'
There was a problem hiding this comment.
Okay I re-added xpu builds back for manylinux
| - device: "rocm" | ||
| docker-image: "pytorch/manylinux-builder:cpu" | ||
| # TODO: Remove this for cuda when manylinux2_28 migration for xpu is done |
There was a problem hiding this comment.
Not a fan of this btw, there has to be a better way of doing this
|
@pytorchbot merge |
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 |
|
The merge job was canceled or timed out. This most often happen if two merge requests were issued for the same PR, or if merge job was waiting for more than 6 hours for tests to finish. In later case, please do not hesitate to reissue the merge command |
|
@pytorchbot merge |
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 |
Follow PR #148129 to remove manylinux builds for triton xpu Pull Request resolved: #148195 Approved by: https://github.com/seemethere
We're dropping regular old manylinux so let's drop it here too Relates to pytorch#123649 Signed-off-by: Eli Uriegas <eliuriegas@meta.com> Pull Request resolved: pytorch#148129 Approved by: https://github.com/Camyll, https://github.com/huydhn, https://github.com/malfet, https://github.com/atalman ghstack dependencies: pytorch#148126
Follow PR pytorch#148129 to remove manylinux builds for triton xpu Pull Request resolved: pytorch#148195 Approved by: https://github.com/seemethere
Stack from ghstack (oldest at bottom):
We're dropping regular old manylinux so let's drop it here too
Relates to #123649
Signed-off-by: Eli Uriegas eliuriegas@meta.com