build(cudf): Pin cuDF dependencies and add script for updates#15992
Closed
bdice wants to merge 9 commits intofacebookincubator:mainfrom
Closed
build(cudf): Pin cuDF dependencies and add script for updates#15992bdice wants to merge 9 commits intofacebookincubator:mainfrom
bdice wants to merge 9 commits intofacebookincubator:mainfrom
Conversation
✅ Deploy Preview for meta-velox canceled.
|
meta-cla
bot
added
the
CLA Signed
Update cudf.cmake to pin rapids-cmake, rmm, and kvikio to specific commits with SHA256 checksums and refactor commit hashes into variables. Add --commit mode to update-cudf-deps.sh that automatically finds compatible dependency versions by selecting the most recent main branch commit before the specified cuDF commit date, ensuring all dependencies are temporally compatible.
karthikeyann
approved these changes
Jan 22, 2026
bdice
commented
Jan 27, 2026
Collaborator
Author
There was a problem hiding this comment.
This is optional -- if Velox maintainers have reservations about including this script, we don't have to include it here. We can move it somewhere else.
Its purpose is to pin a compatible cudf dependency tree (rapids-cmake, rmm, kvikio) based on the desired cuDF branch/PR/commit hash.
karthikeyann
approved these changes
Jan 27, 2026
karthikeyann
added
cudf
cuDF PR #20937 changed partitioning APIs to return num_partitions + 1 offsets where offsets[num_partitions] is the total row count. Update the offset handling to expect the new format and exclude the trailing total row count when calling cudf::split.
shrshi
approved these changes
Jan 27, 2026
|
@kagamiori has imported this pull request. If you are a Meta employee, you can view this in D91639452. |
|
@kagamiori merged this pull request in 6280634. |
paul-aiyedun
added a commit
to paul-aiyedun/velox
that referenced
this pull request
Jan 30, 2026
Fix a regression that is caused by commit 6280634 ("build(cudf): Pin cuDF dependencies and add script for updates (facebookincubator#15992)"), which updates cudf to version 26.04. This cudf version changed the partition offset API to return numPartitions + 1 elements instead of numPartitions, but the CudfPartitionedOutput.cpp file in the cudf-exchange directory was not updated to handle this change.
paul-aiyedun
added a commit
to paul-aiyedun/velox
that referenced
this pull request
Jan 31, 2026
Fix a regression that is caused by commit 6280634 ("build(cudf): Pin cuDF dependencies and add script for updates (facebookincubator#15992)"), which updates cudf to version 26.04. This cudf version changed the partition offset API to return numPartitions + 1 elements instead of numPartitions, but the CudfPartitionedOutput.cpp file in the cudf-exchange directory was not updated to handle this change.
paul-aiyedun
added a commit
to paul-aiyedun/velox
that referenced
this pull request
Feb 3, 2026
Fix a regression that is caused by commit 6280634 ("build(cudf): Pin cuDF dependencies and add script for updates (facebookincubator#15992)"), which updates cudf to version 26.04. This cudf version changed the partition offset API to return numPartitions + 1 elements instead of numPartitions, but the CudfPartitionedOutput.cpp file in the cudf-exchange directory was not updated to handle this change.
paul-aiyedun
added a commit
to paul-aiyedun/velox
that referenced
this pull request
Feb 3, 2026
Fix a regression that is caused by commit 6280634 ("build(cudf): Pin cuDF dependencies and add script for updates (facebookincubator#15992)"), which updates cudf to version 26.04. This cudf version changed the partition offset API to return numPartitions + 1 elements instead of numPartitions, but the CudfPartitionedOutput.cpp file in the cudf-exchange directory was not updated to handle this change.
paul-aiyedun
added a commit
to paul-aiyedun/velox
that referenced
this pull request
Feb 5, 2026
Fix a regression that is caused by commit 6280634 ("build(cudf): Pin cuDF dependencies and add script for updates (facebookincubator#15992)"), which updates cudf to version 26.04. This cudf version changed the partition offset API to return numPartitions + 1 elements instead of numPartitions, but the CudfPartitionedOutput.cpp file in the cudf-exchange directory was not updated to handle this change.
paul-aiyedun
added a commit
to paul-aiyedun/velox
that referenced
this pull request
Feb 6, 2026
Fix a regression that is caused by commit 6280634 ("build(cudf): Pin cuDF dependencies and add script for updates (facebookincubator#15992)"), which updates cudf to version 26.04. This cudf version changed the partition offset API to return numPartitions + 1 elements instead of numPartitions, but the CudfPartitionedOutput.cpp file in the cudf-exchange directory was not updated to handle this change.
paul-aiyedun
added a commit
to paul-aiyedun/velox
that referenced
this pull request
Feb 7, 2026
Fix a regression that is caused by commit 6280634 ("build(cudf): Pin cuDF dependencies and add script for updates (facebookincubator#15992)"), which updates cudf to version 26.04. This cudf version changed the partition offset API to return numPartitions + 1 elements instead of numPartitions, but the CudfPartitionedOutput.cpp file in the cudf-exchange directory was not updated to handle this change.
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
4 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.
This PR pins cuDF's dependencies: rapids-cmake, rmm, and kvikio.
This is a follow-up to #15937, which updates the cuDF dependency to 26.02. It pulls a recent cuDF commit from
main. However, cuDF's core RAPIDS dependencies (rapids-cmake, RMM, KvikIO) are currently unpinned, meaning that they will continue to pull the latest frommainwhile cuDF remains pinned. We would like to ensure that builds are stable and reproducible, which means we should pin these core cuDF dependencies.To make future updates simpler, I added a script
scripts/update-cudf-deps.sh. To update to a recent branch, developers can run:A pain point that cuDF developers have faced is testing Velox with cuDF pull requests in progress. To support that use case, developers can run:
To use a specific cuDF commit with compatible dependency versions, developers can run:
This automatically finds compatible rapids-cmake, rmm, and kvikio versions by selecting the most recent main branch commit before the specified cuDF commit date, ensuring all dependencies are temporally compatible.