JuliaDiff / JuliaDiff/ChainRules.jl

Make `OneElement` more GPU friendly

Open
#767 0 comments 1 reaction 0 assignees View on GitHub

Nobody has claimed this yet.

GPU help wanted
Dominant language
Julia
Stars
475
Forks
98
PR merge metrics
No merged PRs in 30d

Description

Ref. https://github.com/FluxML/Flux.jl/pull/2368. I see a couple of possibly complementary ways to go about this. Easiest would be to define an Adapt rule for OneElement so it's materialized or substituted with some GPU-friendly equivalent when run through CUDA.cu. The other would be defining overloads for certain functions such as mul! which can take advantage of the sparsity.

Contributor guide

No contributing guide indexed for this repository

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 reading the referenced FluxML/Flux.jl pull request 2368 and the existing OneElement implementation. Compare the two suggested directions—an Adapt rule used by CUDA.cu or overloads such as mul!—and establish which GPU behavior and sparse operations should be supported before defining done.

Written by the indexing model from the issue text.

Assessment

Tech stack
julia
Domain
performance
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.