lmcinnes / lmcinnes/umap

Run out of memory on a 1.6 m dataset

Open
#511 1 comment 2 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Dominant language
Python
Stars
8.3k
Forks
871
Avg merge
1d 13h
Merged PRs (30d)
5

Description

Hi, great work for UMAP. I got oom when I run UMAP on a 1.6 m point dataset with 300 dimensions for each point. UMAP eats all my 32G memory in 2 minutes. I enabled verbose and it shows building RP forest with 69 trees. I also run it with low_memory=True, init='random', and the same oom.

Is there any equation to pre-calculate how many memory the dataset gonna use for a 1.6m X 300 dataset?

Thanks a lot!

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 reproducing the out-of-memory run with 1.6 million points, 300 dimensions, and the reported UMAP settings. Inspect the RP forest construction and low_memory behavior using the repository's existing entry points; done would mean establishing the memory cause or producing a defined, tested change that prevents or clearly reports the exhaustion.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
machine-learning, performance
Issue type
Bug
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.