FluxML / FluxML/Tracker.jl

Support for tracked SharedArrays?

Open
#7 3 comments 0 reactions 0 assignees View on GitHub
Dominant language
Julia
Stars
54
Forks
37
Avg merge
5h 49m
Merged PRs (30d)
1

Description

Currently, a SharedArray becomes non-shared when passed to the function param.

```julia
julia> W = SharedArray(rand(4, 3))
b = SharedArray(rand(4))
l = Dense(param(W), param(b), relu)
l.W.data
4×3 Array{Float64,2}:
0.97669 0.26793 0.893204
0.775216 0.679546 0.176578
0.857517 0.289327 0.0719966
0.638023 0.0768014 0.669786
```
Would it be easily possible to keep it shared?

Contributor guide

No contributing guide indexed for this repository

Research direction

Reproduce the Julia example using SharedArray, param, Dense, and relu, then trace where the parameter is converted before it appears as l.W.data. Determine how shared storage should be preserved while tracking remains correct. Done means SharedArray inputs remain shared in the resulting tracked parameters without breaking existing behavior.

Written by the indexing model from the issue text.

Assessment

Tech stack
julia
Domain
machine-learning
Issue type
Feature
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.