NVIDIA / NVIDIA/cuvs

Accept VPQ dataset in IVF-PQ search

Open
#2,344 1 comment 0 reactions 1 assignee View on GitHub

@aamijar is already working on this.

Since Jul 22, 2026.

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

Description

Today, IVF-PQ search only accepts full-precision datasets.
We would like support for searching when the queries are VPQ quantized.

The baseline strategy to do this is to decompress the queries (based on batch_size to avoid full materialization) and reuse the existing search kernels. The current search kernels use Asymmetric Distance Computation (ADC).

We may investigate a better search kernel strategy to exploit the idea that the queries are also quantized.

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.

Assessment

This issue has not been assessed yet.

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.