NVIDIA / NVIDIA/cuvs

[FEA] Option to train Vamana OPQ codebooks and rotation matrix

Open
#905 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

Is your feature request related to a problem? Please describe.
Current (initial) version of Vamana disk index build (with quantization) requires pre-trained codebooks. This is for initial use cases, but future users may want to create codebooks within cuVS.

Describe the solution you'd like
A separate API call to train OPQ codebooks and rotation matrix that can easily be used with vamana::build. Also need the ability to serialize these so that they can be used with search through the CPU DiskANN library.

Describe alternatives you've considered
Another option is to have a flag that creates the codebooks during vamana::build, but this gives much less flexibility and complicates the build call.

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 by locating the vamana::build entry point and reviewing how pre-trained OPQ codebooks are currently supplied. Define the separate training API and serialization path needed for codebooks and the rotation matrix, then verify that the serialized output can be consumed by the CPU DiskANN library.

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
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.