Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Enhance edge and vertex weights for HyPar optimizer #43

Merged
merged 8 commits into from
Nov 3, 2023
Merged

Enhance edge and vertex weights for HyPar optimizer #43

Changes from 1 commit
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
c39a7b1
Enhance edge and vertex weights for HyPar optimizer
jofrevalles Nov 2, 2023
5856a5f
Enhance syntax
jofrevalles Nov 2, 2023
6962242
Use Base.Fix1
jofrevalles Nov 3, 2023
a679316
Apply @mofeing suggestions from code review
jofrevalles Nov 3, 2023
c92f8bf
Fix code from code review
jofrevalles Nov 3, 2023
628671a
Fix KaHyPar test
jofrevalles Nov 3, 2023
72c3562
Fix kahypar test
jofrevalles Nov 3, 2023
694b3b2
Fix typo
jofrevalles Nov 3, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 4 additions & 2 deletions src/Optimizers/KaHyPar.jl
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,6 +7,8 @@ using KaHyPar
imbalance::Float32 = 0.03
stop::Function = <=(2) ∘ length ∘ Base.Fix1(getfield, :args)
configuration::Union{Nothing,Symbol,String} = nothing
edge_weight_scaling::Function = (ind_size) -> 1000 * Int(round(log2(ind_size)))
vertex_weight_scaling::Function = (prod_size) -> 1000 * Int(round(log2(prod_size)))
jofrevalles marked this conversation as resolved.
Show resolved Hide resolved
end

function EinExprs.einexpr(config::HyPar, path)
Expand All @@ -21,8 +23,8 @@ function EinExprs.einexpr(config::HyPar, path)
incidence_matrix = sparse(I, J, V)

# NOTE indices in `inds` should be in the same order as unique indices appear by iterating on `path.args` because `∪` retains order
edge_weights = map(Base.Fix1(size, path), inds)
vertex_weights = ones(Int, length(path.args))
edge_weights = map(ind -> config.edge_weight_scaling(size(path, ind)), inds)
mofeing marked this conversation as resolved.
Show resolved Hide resolved
vertex_weights = map(tensor -> config.vertex_weight_scaling(prod(size(tensor))), path.args)
mofeing marked this conversation as resolved.
Show resolved Hide resolved

hypergraph = KaHyPar.HyperGraph(incidence_matrix, vertex_weights, edge_weights)

Expand Down