Moving MG functions into unified API + `raft::device_resources_snmg` as device resource type for MG functions by viclafargue · Pull Request #454 · NVIDIA/cuvs

viclafargue · 2024-11-08T17:25:54Z

Moving MG functions into unified API
Using raft::device_resources_snmg as device resource type for MG functions
Removing instantiation of raft::comms::build_comms_nccl_only solving [FEA] Remove dependency of build_comms_nccl_only on UCP raft#2465

raft::device_resources_snmg is introduced in this PR : NVIDIA/raft#2549

cjnolet · 2025-01-07T22:03:11Z

 * @return the constructed IVF-Flat MG index
 */
-auto build(const raft::device_resources& handle,
+auto build(const raft::device_resources_snmg& clique,


One of the main goals with having the "snmg" resources object match the single-gpu object is that we wanted to be able to remove this additional MG. We should now be able to accept device_resources and then check to see if a nccl clique has been set on it (which would imply that it's a multi-gpu resources object and not a single-gpu object). The whole goal with doing this was to consolidate code paths.

The NCCL clique is not set as a resource anymore, but we should still be able to implement the dispatching by checking the dynamic type of the device_resources. The real question then is, do we truly want dispatching on both the regular API (cuvs::neighbors::build) and the mg namespace (cuvs::neighbors::mg::build)? It kind of make sense that a user providing a device_resources_snmg instance to the regular API (cuvs::neighbors::build) would want things to be deployed on multiple GPUs. However, the reverse is not necessarily true. A user who explicitly chose the mg namespace, but did not provide the adequate device_resources_snmg would fallback to single GPU, potentially unintentionally. Is this what we want?

I propose that we implement the dispatching mechanism solely on the regular API (cuvs::neighbors::build) in a dedicated follow-up PR? This also allows the MG doc to explicitly avert users that they should use an adequate device_resources_snmg to use the MG API. What do you think?

cjnolet · 2025-01-17T00:45:18Z

Not sure why CI is still failing. The RAFT PR was merged quite awhile ago but I'm still seeing errors like this:

/home/coder/cuvs/cpp/include/cuvs/neighbors/mg.hpp:24:10: fatal error: raft/core/device_resources_snmg.hpp: No such file or directory
     24 | #include <raft/core/device_resources_snmg.hpp>
        |          ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

bdice · 2025-01-17T00:46:57Z

This fell back to old 24.12 RAFT packages for unclear reasons:

    libraft-headers:             24.12.00-cuda12_241211_geaf9cc72_0 rapidsai   
    libraft-headers-only:        24.12.00-cuda12_241211_geaf9cc72_0 rapidsai

https://github.com/rapidsai/cuvs/actions/runs/12818869804/job/35749392606?pr=454#step:8:521

cjnolet · 2025-01-17T00:48:17Z

This fell back to old 24.12 RAFT packages for unclear reasons:

Oh fooey! Thanks for noticing that @bdice, I just realized the darn target branch was never updated!

copy-pr-bot · 2025-01-20T17:56:45Z

This pull request requires additional validation before any workflows can run on NVIDIA's runners.

Pull request vetters can view their responsibilities here.

Contributors can view more details about this message here.

copy-pr-bot · 2025-04-22T22:27:10Z

/ok to test

@cjnolet, there was an error processing your request: E1

See the following link for more information: https://docs.gha-runners.nvidia.com/cpr/e/1/

cjnolet · 2025-04-22T22:35:23Z

/ok to test

copy-pr-bot · 2025-04-22T22:35:26Z

/ok to test

@cjnolet, there was an error processing your request: E1

See the following link for more information: https://docs.gha-runners.nvidia.com/cpr/e/1/

cjnolet · 2025-04-22T22:37:16Z

/ok to test da48376

cjnolet · 2025-04-22T22:37:40Z

/merge

cjnolet · 2025-04-23T03:58:19Z

/ok to test da48

cjnolet · 2025-04-23T03:59:02Z

/ok to test da48376

copy-pr-bot · 2025-04-23T03:59:05Z

/ok to test da48376

@cjnolet, there was an error processing your request: E2

See the following link for more information: https://docs.gha-runners.nvidia.com/cpr/e/2/

divyegala · 2025-04-23T04:00:42Z

/ok to test d2e2be9

divyegala · 2025-04-23T04:32:19Z

/merge

viclafargue · 2025-04-23T11:39:32Z

/ok to test

copy-pr-bot · 2025-04-23T11:39:35Z

/ok to test

@viclafargue, there was an error processing your request: E1

See the following link for more information: https://docs.gha-runners.nvidia.com/cpr/e/1/

achirkin · 2025-04-23T11:42:31Z

/ok to test

copy-pr-bot · 2025-04-23T11:42:34Z

/ok to test

@achirkin, there was an error processing your request: E1

See the following link for more information: https://docs.gha-runners.nvidia.com/cpr/e/1/

viclafargue · 2025-04-23T11:44:29Z

/ok to test 1e1d97e

achirkin · 2025-04-23T11:49:58Z

/ok to test 1e1d97e

Co-authored-by: Artem M. Chirkin <9253178+achirkin@users.noreply.github.com>

cjnolet · 2025-04-23T15:34:28Z

/ok to test 34d67ee

cjnolet · 2025-04-23T20:53:38Z

/merge

Account for RAFT update

ee98593

viclafargue requested review from a team as code owners November 8, 2024 17:25

github-actions Bot added cpp CMake labels Nov 8, 2024

viclafargue added 6 commits November 15, 2024 14:25

use new device_resources_snmg

3a99b40

improved device_resources_snmg

e16b68e

Merge branch 'branch-24.12' into account-for-raft-update

cdd5cfb

switch from RAFT_LOG_INFO to RAFT_LOG_DEBUG for mg logs

96e69fc

clique as device_resource

657bf9e

updating MG tests

1fdccd4

cjnolet changed the title ~~Account for RAFT update~~ Account for RAFT update to SNMG APIs Dec 18, 2024

cjnolet assigned viclafargue Dec 18, 2024

cjnolet reviewed Jan 7, 2025

View reviewed changes

Merge branch 'branch-25.02' into account-for-raft-update

d7fff4c

cjnolet mentioned this pull request Jan 17, 2025

Introduction of the raft::device_resources_snmg type NVIDIA/raft#2487

Merged

cjnolet changed the base branch from branch-24.12 to branch-25.02 January 17, 2025 00:47

cjnolet added improvement Improves an existing functionality non-breaking Introduces a non-breaking change labels Jan 17, 2025

This was referenced Jan 17, 2025

add libraft wheels rapidsai/devcontainers#438

Merged

Use GCC 13 in CUDA 12+ builds rapidsai/build-planning#129

Closed

viclafargue added 2 commits January 20, 2025 17:53

API unification (removal of mg namespace)

45a41fa

Merge branch 'branch-25.02' into account-for-raft-update

b0c7ab9

viclafargue force-pushed the account-for-raft-update branch from 6ae1c70 to b0c7ab9 Compare January 20, 2025 17:56

Adding CUVS_BUILD_MG_ALGOS macro back

db12ab9

viclafargue changed the title ~~Account for RAFT update to SNMG APIs~~ Moving MG functions into unified API + raft::device_resources_snmg as device resource type for MG functions Jan 22, 2025

cjnolet approved these changes Apr 22, 2025

View reviewed changes

KyleFromNVIDIA approved these changes Apr 22, 2025

View reviewed changes

cjnolet removed the request for review from msarahan April 22, 2025 22:35

Merge branch 'branch-25.06' into account-for-raft-update

d2e2be9

temporary pin to RAFT compilation fix

1e1d97e

viclafargue force-pushed the account-for-raft-update branch from 92ced86 to 1e1d97e Compare April 23, 2025 11:27

achirkin reviewed Apr 23, 2025

View reviewed changes

Comment thread cpp/cmake/thirdparty/get_raft.cmake Outdated

Update cpp/cmake/thirdparty/get_raft.cmake

34d67ee

Co-authored-by: Artem M. Chirkin <9253178+achirkin@users.noreply.github.com>

rapids-bot Bot merged commit d2adc6f into NVIDIA:branch-25.06 Apr 23, 2025

github-project-automation Bot moved this from In Progress to Done in Unstructured Data Processing Apr 23, 2025

josephine-wolf-oberholtzer added this to Unstructured Data Processing Jul 1, 2026

josephine-wolf-oberholtzer moved this to Done in Unstructured Data Processing Jul 1, 2026

Uh oh!

Conversation

viclafargue commented Nov 8, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

cjnolet Jan 7, 2025

Choose a reason for hiding this comment

Uh oh!

viclafargue Jan 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

cjnolet commented Jan 17, 2025

Uh oh!

bdice commented Jan 17, 2025

Uh oh!

cjnolet commented Jan 17, 2025

Uh oh!

copy-pr-bot Bot commented Jan 20, 2025

Uh oh!

copy-pr-bot Bot commented Apr 22, 2025

Uh oh!

cjnolet commented Apr 22, 2025

Uh oh!

copy-pr-bot Bot commented Apr 22, 2025

Uh oh!

cjnolet commented Apr 22, 2025

Uh oh!

cjnolet commented Apr 22, 2025

Uh oh!

cjnolet commented Apr 23, 2025

Uh oh!

cjnolet commented Apr 23, 2025

Uh oh!

copy-pr-bot Bot commented Apr 23, 2025

Uh oh!

divyegala commented Apr 23, 2025

Uh oh!

divyegala commented Apr 23, 2025

Uh oh!

viclafargue commented Apr 23, 2025

Uh oh!

copy-pr-bot Bot commented Apr 23, 2025

Uh oh!

achirkin commented Apr 23, 2025

Uh oh!

copy-pr-bot Bot commented Apr 23, 2025

Uh oh!

viclafargue commented Apr 23, 2025

Uh oh!

achirkin commented Apr 23, 2025

Uh oh!

Uh oh!

cjnolet commented Apr 23, 2025

Uh oh!

cjnolet commented Apr 23, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

7 participants

viclafargue commented Nov 8, 2024 •

edited

Loading

viclafargue Jan 8, 2025 •

edited

Loading