NVIDIA / NVIDIA/cuvs

Lucene: Improve CAGRA_HNSW search latency via better graph quality

Open
#2,458 2 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Lucene
Dominant language
Cuda
Stars
854
Forks
236
Avg merge
3d 3h
Merged PRs (30d)
62

Description

Current logic only allows 1 to 3 graph layers when converting from CAGRA to HNSW which results in less "navigable" HNSW graphs that take longer to search.

PR: https://github.com/rapidsai/cuvs-lucene/pull/140

Image

In the attached plot, we can see that some datasets (such as the one seen here) appear to demand higher hnsw-layers in order to achieve the natural search-latency as seen in the natural HNSW benchmark runs

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start with PR #140 and the CAGRA-to-HNSW conversion logic described in the issue, then review the attached latency plot and compare the affected search behavior with the natural HNSW benchmarks. Done means the converted graphs provide better navigability and improved search latency on datasets that require more HNSW layers.

Written by the indexing model from the issue text.

Assessment

Domain
performance, search
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.