lmcinnes / lmcinnes/pynndescent

Very high memory usage

Open
#220 7 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
970
Forks
107
PR merge metrics
No merged PRs in 30d

Description

Hi, I am experimenting excessive memory utilization of the library:

import numpy as np
X = np.random.rand(20_000_000, 2) # 20MM 2D vectors

index = pynndescent.NNDescent(X)
index.prepare()

The following code exceeds 30GB of memory in my machine. Surprisingly, other libraries like hnswlib are capable of building the same index in approximately 5GB of memory.

Is this a possible bug/memory leak in the library or is it the expected memory usage?
Testing with version 0.5.10

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

Reproduce the reported Python snippet with pynndescent 0.5.10, focusing on memory use during NNDescent(X) and index.prepare() for the 20-million-vector input. Compare the observed usage with the report and determine whether it is expected behavior or a memory leak; done means the cause and expected usage are established, or a concrete bug is identified.

Written by the indexing model from the issue text.

Assessment

Tech stack
numpy, python
Domain
performance
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.