fULL support of sparse AD for NLS models

src/ADNLPModels.jl

@@ -4,7 +4,11 @@ module ADNLPModels
 using LinearAlgebra, SparseArrays
 
 # external
+<<<<<<< HEAD
 using ADTypes: ADTypes, AbstractSparsityDetector
+=======
+using ADTypes: ADTypes
+>>>>>>> 5cacfb0 (fULL support of sparse AD for NLS models)
 using SparseConnectivityTracer, ColPack, ForwardDiff, ReverseDiff
 
 # JSO

src/ad_api.jl

@@ -289,6 +289,25 @@ function NLPModels.hess_coord!(
   return vals
 end
 
+function NLPModels.hess_structure_residuals!(
+  b::ADBackend,
+  nls::AbstractADNLSModel,
+  rows::AbstractVector{<:Integer},
+  cols::AbstractVector{<:Integer},
+)
+  nothing
+end
+
+function NLPModels.hess_coord_residuals!(
+  b::ADBackend,
+  nls::AbstractADNLSModel,
+  x::AbstractVector,
+  v::AbstractVector,
+  vals::AbstractVector,
+)
+  nothing
+end
+
 function NLPModels.hprod!(
   b::ADBackend,
   nlp::ADModel,

src/sparse_hessian.jl

@@ -294,3 +294,44 @@ function NLPModels.hess_coord!(
   sparse_hess_coord!(ℓ, b, x, obj_weight, b.y, vals)
   return vals
 end
+
+function NLPModels.hess_structure_residuals!(
+  b::Union{SparseADHessian, SparseReverseADHessian},
+  nls::AbstractADNLSModel,
+  rows::AbstractVector{<:Integer},
+  cols::AbstractVector{<:Integer},
+)
+  function objective(x)
+    F = get_F(nls, b)
+    Fx = F(x)
+    return dot(Fx, Fx) / 2
+  end
+
+  H = compute_hessian_sparsity(objective, nls.meta.nvar, nothing, 0)
+  trilH = tril(H)
+  rowval = trilH.rowval
+  colptr = trilH.colptr
+  rows .= rowval
+  for i = 1:(nls.meta.nvar)
+    for j = colptr[i]:(colptr[i + 1] - 1)
+      cols[j] = i
+    end
+  end
+  return rows, cols
+end
+
+function NLPModels.hess_coord_residuals!(
+  b::Union{SparseADHessian, SparseReverseADHessian},
+  nls::AbstractADNLSModel,
+  x::AbstractVector,
+  v::AbstractVector,
+  vals::AbstractVector,
+)
+  function objective(x)
+    F = get_F(nls, b)
+    Fx = F(x)
+    return dot(Fx, Fx) / 2
+  end
+
+  sparse_hess_coord!(objective, b, x, 1.0, v, vals)
+end

test/runtests.jl

@@ -29,6 +29,7 @@ end
 
 @testset "Basic Hessian derivative test" begin
   include("sparse_hessian.jl")
+  include("sparse_hessian_nls.jl")
 end
 
 for problem in NLPModelsTest.nlp_problems ∪ ["GENROSE"]

test/sparse_hessian_nls.jl

@@ -0,0 +1,45 @@
+list_sparse_hess_backend = (
+  (ADNLPModels.SparseADHessian, Dict()),
+  (ADNLPModels.ForwardDiffADHessian, Dict()),
+)
+
+dt = (Float32, Float64)
+
+@testset "Basic Hessian of residual derivative with backend=$(backend) and T=$(T)" for T in dt,
+  (backend, kw) in list_sparse_hess_backend
+
+  F!(Fx, x) = begin
+    Fx[1] = x[1] - 1
+    Fx[2] = 10 * (x[2] - x[1]^2)
+    Fx[3] = x[2] + 1
+    Fx
+  end
+  x0 = T[-1.2; 1.0]
+  nvar = 2
+  nequ = 3
+  nls = ADNLPModels.ADNLSModel!(F!, x0, 3, hessian_residual_backend = backend; kw...)
+
+  x = rand(T, nvar)
+  v = rand(T, nequ)
+  rows, cols = zeros(Int, nls.nls_meta.nnzh), zeros(Int, nls.nls_meta.nnzh)
+  vals = zeros(T, nls.nls_meta.nnzh)
+  hess_structure_residual!(nls, rows, cols)
+  hess_coord_residual!(nls, x, v, vals)
+  @test eltype(vals) == T
+  H = sparse(rows, cols, vals, nvar, nvar)
+  # @test H == []
+
+  # Test also the implementation of the backends
+  b = nls.adbackend.hessian_residual_backend
+  obj_weight = 0.5
+  @test nls.nls_meta.nnzh == ADNLPModels.get_nln_nnzh(b, nvar)
+  ADNLPModels.hess_structure_residual!(b, nls, rows, cols)
+  ADNLPModels.hess_coord_residual!(b, nls, x, y, obj_weight, vals)
+  @test eltype(vals) == T
+  H = sparse(rows, cols, vals, nvar, nvar)
+  # @test H == []
+
+  nls = ADNLPModels.ADNLSModel!(F!, x0, 3, matrix_free = true; kw...)
+  @test nls.adbackend.hessian_backend isa ADNLPModels.EmptyADbackend
+  @test nls.adbackend.hessian_residual_backend isa ADNLPModels.EmptyADbackend
+end

test/sparse_jacobian.jl

@@ -1,8 +1,10 @@
 list_sparse_jac_backend = (
-  (ADNLPModels.SparseADJacobian, Dict()), # default
+  (ADNLPModels.SparseADJacobian, Dict()),
   (ADNLPModels.ForwardDiffADJacobian, Dict()),
 )
+
 dt = (Float32, Float64)
+
 @testset "Basic Jacobian derivative with backend=$(backend) and T=$(T)" for T in dt,
   (backend, kw) in list_sparse_jac_backend
 

test/sparse_jacobian_nls.jl

@@ -1,8 +1,10 @@
 list_sparse_jac_backend = (
-  (ADNLPModels.SparseADJacobian, Dict()), # default
+  (ADNLPModels.SparseADJacobian, Dict()),
   (ADNLPModels.ForwardDiffADJacobian, Dict()),
 )
+
 dt = (Float32, Float64)
+
 @testset "Basic Jacobian of residual derivative with backend=$(backend) and T=$(T)" for T in dt,
   (backend, kw) in list_sparse_jac_backend