Skip to content
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 NCCL pins in build and test environments #341

Merged
merged 2 commits into from
Sep 25, 2024

Conversation

jameslamb
Copy link
Member

@jameslamb jameslamb commented Sep 20, 2024

Description

Contributes to rapidsai/build-planning#102

Some RAPIDS libraries are using ncclCommSplit(), which was introduced in nccl==2.18.1.1. This is part of a series of PRs across RAPIDS updating libraries' pins to nccl>=2.18.1.1 to ensure they get a new-enough version that supports that.

cuvs doesn't have any direct uses of NCCL... it only uses it via raft. This PR proposes removing cuvs's dependency pinnings on NCCL, in favor of just using whatever it gets transitively via raft.

@jameslamb jameslamb added improvement Improves an existing functionality non-breaking Introduces a non-breaking change labels Sep 20, 2024
@jameslamb jameslamb changed the title WIP: bump nccl floor to 2.18.1.1 WIP: bump NCCL floor to 2.18.1.1 Sep 20, 2024
@jameslamb jameslamb changed the title WIP: bump NCCL floor to 2.18.1.1 bump NCCL floor to 2.18.1.1 Sep 20, 2024
@jameslamb jameslamb marked this pull request as ready for review September 20, 2024 21:58
@jameslamb jameslamb requested a review from a team as a code owner September 20, 2024 21:58
@jameslamb jameslamb requested a review from AyodeAwe September 20, 2024 21:58
Copy link
Contributor

@vyasr vyasr left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Does cuvs use nccl directly, or only transitively via raft? I notice that it's not specified in the nccl recipes, and if I search the repo I don't see other references which makes me think it's just via raft. If so, we should just drop it here.

@jameslamb
Copy link
Member Author

Does cuvs use nccl directly, or only transitively via raft? I notice that it's not specified in the nccl recipes, and if I search the repo I don't see other references which makes me think it's just via raft. If so, we should just drop it here.

I searched around and it seems like it's only used transitively through raft.

git grep -i nccl

Looked through the diff of #231 too, and I only see stuff like this:

const raft::comms::nccl_clique& clique = handle.get_nccl_clique_handle();

@vyasr
Copy link
Contributor

vyasr commented Sep 24, 2024

Yeah that matches my expectation. Let's see if tests pass. If they do, please update the title and description accordingly and I can approve.

@jameslamb jameslamb changed the title bump NCCL floor to 2.18.1.1 remove NCCL pins in build and test environments Sep 24, 2024
@jameslamb
Copy link
Member Author

Tests look happy! I've updated the title and description here.

Thanks for the nudge, removing these pins will make the next such update a little easier 😁

@jameslamb jameslamb requested a review from vyasr September 24, 2024 19:58
@jakirkham jakirkham requested a review from cjnolet September 25, 2024 02:22
@jakirkham
Copy link
Member

/merge

@rapids-bot rapids-bot bot merged commit 8e02150 into rapidsai:branch-24.10 Sep 25, 2024
54 checks passed
@jakirkham
Copy link
Member

Thanks all! 🙏

@jameslamb jameslamb deleted the update-nccl branch September 25, 2024 19:39
@jameslamb jameslamb mentioned this pull request Oct 2, 2024
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
improvement Improves an existing functionality non-breaking Introduces a non-breaking change
Projects
None yet
Development

Successfully merging this pull request may close these issues.

4 participants