feat(cugraph): upgrade to RAPIDS 25.12 / CUDA 13.1 with comprehensive e2e tests

[mattkjames7]

MAGE PR #710 transferred to this repo. Closes #3564.

Summary

This PR upgrades MAGE's cuGraph integration from the legacy RAPIDS 22.02 / CUDA 11.5 stack to modern RAPIDS 25.12 / CUDA 13.1, bringing GPU-accelerated graph algorithms up to date with current NVIDIA tooling.

Key Changes

Infrastructure Upgrade:

• CUDA 11.5.2 → 13.1.0
• RAPIDS/cuGraph 22.02 → 25.12
• Ubuntu 20.04 → 24.04
• Python 3.8 → 3.12

API Migration:
All 9 cuGraph algorithms updated to use the modern pylibcugraph API:

• cugraph::pagerank → cugraph::pagerank() with explicit graph view
• cugraph::betweenness_centrality → normalized output handling
• cugraph::hits → proper hub/authority vector management
• cugraph::katz_centrality → updated alpha/beta parameter handling
• cugraph::louvain / cugraph::leiden → new clustering return types
• cugraph::personalized_pagerank → vertex list handling

Legacy API Preserved:
Two algorithms remain on cugraph::ext_raft:: API as they haven't been migrated in RAPIDS 25.x:

• balanced_cut_clustering
• spectral_clustering

E2E Tests Added

Comprehensive end-to-end tests for all 9 algorithms following MAGE's existing test framework:

e2e/pagerank_test/test_cugraph_networkx_validation/
e2e/betweenness_centrality_test/test_cugraph_networkx_validation/
e2e/hits_test/test_cugraph_networkx_validation/
e2e/katz_test/test_cugraph_networkx_validation/
e2e/louvain_test/test_cugraph_networkx_validation/
e2e/leiden_cugraph_test/test_cugraph_networkx_validation/
e2e/personalized_pagerank_test/test_cugraph_networkx_validation/
e2e/balanced_cut_clustering_test/test_cugraph_networkx_validation/
e2e/spectral_clustering_test/test_cugraph_networkx_validation/

Each test uses a 9-node two-community graph topology with expected values validated against NetworkX ground truth (5% tolerance for GPU floating-point variance).

Validation Script

Added scripts/validate_cugraph_algorithms.py - a standalone debugging tool that:

1. Builds identical graph in NetworkX (ground truth)
2. Spins up Memgraph container with cuGraph modules
3. Runs each algorithm and compares against NetworkX
4. Reports pass/fail with detailed value comparisons

This is for developer debugging, not CI.

Test Plan

• All 9 cuGraph algorithms pass validation against NetworkX ground truth
• Docker image builds successfully with Dockerfile.cugraph
• E2E tests follow existing MAGE test conventions
• CI pipeline runs (pending merge)

Breaking Changes

None. All algorithm signatures and return types preserved.

[mattkjames7]

Tracking

• [Link to Epic/Issue]

Standard development

• Update unit/E2E tests
• Compare the benchmarking results between the master branch this branch

CI Testing Labels

• Select the appropriate CI test labels (CI -build=build-name -test=test-suite)

Documentation checklist

• Add the documentation label
• Add the bug / feature label
• Add the milestone for which this feature is intended
  • If not known, set for a later milestone
• Write a release note, including added/changed clauses
  • This PR upgrades MAGE's cuGraph integration from the legacy RAPIDS 22.02 / CUDA 11.5 stack to modern RAPIDS 25.12 / CUDA 13.1, bringing GPU-accelerated graph algorithms up to date with current NVIDIA tooling. #3709
• cuGraph API changes documentation#1517
  • Is back linked to this development PR

[mattkjames7]

TODO: update docs with changes to module inputs

[antejavor]

This is a huge PR that mostly contains the changes in the cuGraph API and updates from legacy.

Since there is validate_cugraph_algorithms.py script there, I assume that the changes here are correct.

The only missing step is to update the docs API, since some args have changed in this PR.

[antejavor]

Two questions @mattkjames7 now that Mage is in the main repo, are we running the Mage test anywhere? Did you try to use this on Nvidia hardware? I assume you did.

[mattkjames7]

This is a huge PR that mostly contains the changes in the cuGraph API and updates from legacy.

Since there is validate_cugraph_algorithms.py script there, I assume that the changes here are correct.

The only missing step is to update the docs API, since some args have changed in this PR.

I haven't written the docs for this just yet, but I will link them here and ping once I have. I think the API changes are fairly minimal.

Two questions @mattkjames7 now that Mage is in the main repo, are we running the Mage test anywhere? Did you try to use this on Nvidia hardware? I assume you did.

We run the MAGE tests in this repo now as part of the diff workflow + daily build and RC build. We do not run any cuGraph tests in CI - the image built with this code will only run if a Nvidia GPU is present, though I did build it and run the tests locally. There will hopefully soon be a cuGraph image that is built regularly too, with this WIP PR: #3723

[sonarqubecloud]

Quality Gate failed

Failed conditions
1 Security Hotspot

See analysis details on SonarQube Cloud