search

JuliaSmoothOptimizers / ADNLPModels.jl

Other
29 stars 12 forks source link

Let users configure sparsity detector and coloring algorithm #242

Closed gdalle closed 1 week ago

gdalle commented 3 weeks ago

See the discussion in https://github.com/JuliaSmoothOptimizers/ADNLPModels.jl/pull/230#issuecomment-2151461371

amontoison commented 3 weeks ago

It's already the case. It's a keyword augment of SparseADHessian and SparseADJacobian. https://github.com/JuliaSmoothOptimizers/ADNLPModels.jl/blob/main/src/sparse_hessian.jl#L24

gdalle commented 3 weeks ago

Where is it documented? Can't find those in the API reference

gdalle commented 3 weeks ago

If it's not documented, users can't know where to change it

gdalle commented 2 weeks ago

Same goes for the coloring algorithm since #244

amontoison commented 2 weeks ago

Let's update everything after that we switch to DifferentiationInterface.jl.

gdalle commented 2 weeks ago

Roadmap:

  1. Track down where we inserted the sparsity detector and coloring algorithms in PRs #230 and #244
  2. Tell users how they can change it
amontoison commented 2 weeks ago

@gdalle I'm wondering if we could not use a symmetric_coloring before rely we rely on DI.jl. We probably just need to update https://github.com/JuliaSmoothOptimizers/ADNLPModels.jl/blob/main/src/sparse_hessian.jl#L208-L216.

gdalle commented 2 weeks ago

Yes you can, I outlined how to do it in #246.

amontoison commented 1 week ago

Done in #249