NVIDIA / NVIDIA/cuvs

[FEA] Support chamfer distance / matching

Open
#1,190 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

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

Description

Chamfer distance is becoming popular for multi-vector matching with models like ColBert. We have been planning on support for multi-vector for awhile, but we need to prioritize it.

Chamfer distance might be computationally intensive in practice, and might provide a good value prop for using GPUs in the multi-vector world. We need to evaluate whether chamfer distance is required to provide a good performance with multi-vectors and support it if that's the case. If not, we need to find a another way to support multi-vectors.

Muvera is another mechanism we might be able to leverage, though this involves an additional encoding to convert the multiple vectors into a single fixed-length vector.

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

The issue provides no file or test entry point. Start by reviewing the Chamfer-distance and Muvera references in the description and evaluating multi-vector retrieval needs; completion requires deciding whether Chamfer is necessary and supporting the selected mechanism.

Written by the indexing model from the issue text.

Assessment

Domain
machine-learning, search
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
20/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.