EnzymeAD/Enzyme.jl

API for Hessian

Open

#1,495 opened on Jun 6, 2024

View on GitHub
 (1 comment) (1 reaction) (0 assignees)Julia (96 forks)batch import
good first issue

Repository metrics

Stars
 (563 stars)
PR merge metrics
 (Avg merge 5d 1h) (76 merged PRs in 30d)

Description

A hessian / hessian! function in Enzyme would be cool!

Right now the method described in https://enzymead.github.io/Enzyme.jl/stable/generated/autodiff/#Forward-over-reverse uses BatchDuplicated. Am I correct in assuming that, like ForwardDiff's "chunks", Enzyme's "batches" are limited in size by memory bandwitch, typically to the order of 10 elements? If so, should there be a Hessian mode that computes HVPs in batches smaller than the size n?

Contributor guide

Research direction
Examine the feasibility of implementing a Hessian function in Enzyme.jl that computes Hessian vector products in batches to handle large Hessians, building on existing forward over reverse mode.
Tech stack
None
Domain
machine learning
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.
1-2 days
Activity statusHow available the issue appears right now: fresh, active, stale, blocked, or waiting on maintainer input.
Active
ClarityHow clearly the issue explains the expected change, acceptance criteria, and next step.
Clear
Prerequisites
JuliaAutomatic Differentiation
Newbie friendlinessA 1-100 score estimating how approachable this issue is for first-time contributors.
50