LuxDL/Lux.jl

Enzyme wrappers for AD utilities in Lux

Open

#738 opened on Jun 27, 2024

View on GitHub
 (0 comments) (1 reaction) (0 assignees)Julia (700 stars) (84 forks)batch import
enzymegood first issue

Description

Functions

  • batched_jacobian
    • Enzyme.Forward
    • Enzyme.Reverse
  • vector_jacobian_product
  • jacobian_vector_product

Nested Autodiff

Enzyme needs us to use autodiff_thunk for the internal call, so we can easily replace the call inside custom rrules.

  • EnzymeRules
  • CRC.rrule

Contributor guide

Tech stack
None
Domain
machine learningai
Issue type
feature
DifficultyEstimated implementation difficulty for a new contributor, from 1 for very small changes to 5 for expert-level work.
3
Estimated timeA rough time range for an experienced contributor to investigate, implement, test, and prepare a pull request.
3-5 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
fresh
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
clear
Prerequisites
Familiarity with JuliaUnderstanding of Enzyme ADBasic knowledge of Lux.jl
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
35
Research direction
The issue requests implementing Enzyme wrappers for several AD utilities including batched jacobian, vector jacobian product, jacobian vector product, and handling nested autodiff via autodiff thunk. A contributor should study the existing Enzyme.jl package and Lux.jl's AD interface, then implement each function following the patterns in similar libraries. No linked PRs or comments are present; initial work can start by forking the repo and referencing the issue.
Enzyme wrappers for AD utilities in Lux · LuxDL/Lux.jl#738 | Good First Issue