reactivate AD tests: mean functions (#313)

* reactivate mean function AD tests
* extend mean function tests to ColVecs/RowVecs
* unify testcases
* remove rrules and ChainRulesCore
Co-authored-by: David Widmann <[email protected]>
Co-authored-by: willtebbutt <[email protected]>

Author: 3 people

Project.toml

@@ -1,10 +1,9 @@
 name = "AbstractGPs"
 uuid = "99985d1d-32ba-4be9-9821-2ec096f28918"
 authors = ["JuliaGaussianProcesses Team"]
-version = "0.5.11"
+version = "0.5.12"
 
 [deps]
-ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 FillArrays = "1a297f60-69ca-5386-bcde-b61e274b549b"
 IrrationalConstants = "92d709cd-6900-40b7-9082-c6be49f344b6"
@@ -19,7 +18,6 @@ StatsBase = "2913bbd2-ae8a-5f71-8c99-4fb6c76f3a91"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [compat]
-ChainRulesCore = "1"
 Distributions = "0.19, 0.20, 0.21, 0.22, 0.23, 0.24, 0.25"
 FillArrays = "0.7, 0.8, 0.9, 0.10, 0.11, 0.12, 0.13"
 IrrationalConstants = "0.1"

src/AbstractGPs.jl

@@ -1,6 +1,5 @@
 module AbstractGPs
 
-using ChainRulesCore
 using Distributions
 using FillArrays
 using LinearAlgebra

src/mean_function.jl

@@ -12,11 +12,6 @@ This is an AbstractGPs-internal workaround for AD issues; ideally we would just
 """
 _map_meanfunction(::ZeroMean{T}, x::AbstractVector) where {T} = Zeros{T}(length(x))
 
-function ChainRulesCore.rrule(::typeof(_map_meanfunction), m::ZeroMean, x::AbstractVector)
-    map_ZeroMean_pullback(Δ) = (NoTangent(), NoTangent(), ZeroTangent())
-    return _map_meanfunction(m, x), map_ZeroMean_pullback
-end
-
 ZeroMean() = ZeroMean{Float64}()
 
 """
@@ -40,4 +35,4 @@ struct CustomMean{Tf} <: MeanFunction
     f::Tf
 end
 
-_map_meanfunction(f::CustomMean, x::AbstractVector) = map(f.f, x)
+_map_meanfunction(m::CustomMean, x::AbstractVector) = map(m.f, x)

test/Project.toml

@@ -1,5 +1,4 @@
 [deps]
-ChainRulesCore = "d360d2e6-b24c-11e9-a2a3-2a2ae2dbcce4"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 Documenter = "e30172f5-a6a5-5a46-863b-614d45cd2de4"
 FillArrays = "1a297f60-69ca-5386-bcde-b61e274b549b"
@@ -14,7 +13,6 @@ Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [compat]
-ChainRulesCore = "1"
 Distributions = "0.19, 0.20, 0.21, 0.22, 0.23, 0.24, 0.25"
 Documenter = "0.24, 0.25, 0.26, 0.27"
 FillArrays = "0.11, 0.12, 0.13"

test/mean_function.jl

@@ -5,42 +5,23 @@
     xD_colvecs = ColVecs(randn(rng, D, N))
     xD_rowvecs = RowVecs(randn(rng, N, D))
 
-    @testset "ZeroMean" begin
-        m = ZeroMean{Float64}()
+    zero_mean_testcase = (; mean_function=ZeroMean(), calc_expected=_ -> zeros(N))
 
-        for x in [x1, xD_colvecs, xD_rowvecs]
-            @test AbstractGPs._map_meanfunction(m, x) == zeros(N)
-            #differentiable_mean_function_tests(m, randn(rng, N), x)
-
-            # Manually verify the ChainRule. Really, this should employ FiniteDifferences, but
-            # currently ChainRulesTestUtils isn't up to handling this, so this will have to do
-            # for now.
-            y, pb = rrule(AbstractGPs._map_meanfunction, m, x)
-            @test y == AbstractGPs._map_meanfunction(m, x)
-            Δmap, Δf, Δx = pb(randn(rng, N))
-            @test iszero(Δmap)
-            @test iszero(Δf)
-            @test iszero(Δx)
-        end
-    end
-
-    @testset "ConstMean" begin
-        c = randn(rng)
-        m = ConstMean(c)
-
-        for x in [x1, xD_colvecs, xD_rowvecs]
-            @test AbstractGPs._map_meanfunction(m, x) == fill(c, N)
-            #differentiable_mean_function_tests(m, randn(rng, N), x)
-        end
-    end
+    c = randn(rng)
+    const_mean_testcase = (; mean_function=ConstMean(c), calc_expected=_ -> fill(c, N))
 
-    @testset "CustomMean" begin
-        foo_mean = x -> sum(abs2, x)
-        m = CustomMean(foo_mean)
+    foo_mean = x -> sum(abs2, x)
+    custom_mean_testcase = (;
+        mean_function=CustomMean(foo_mean), calc_expected=x -> map(foo_mean, x)
+    )
 
+    @testset "$(typeof(testcase.mean_function))" for testcase in [
+        zero_mean_testcase, const_mean_testcase, custom_mean_testcase
+    ]
         for x in [x1, xD_colvecs, xD_rowvecs]
-            @test AbstractGPs._map_meanfunction(m, x) == map(foo_mean, x)
-            #differentiable_mean_function_tests(m, randn(rng, N), x)
+            m = testcase.mean_function
+            @test AbstractGPs._map_meanfunction(m, x) == testcase.calc_expected(x)
+            differentiable_mean_function_tests(rng, m, x)
         end
     end
 end

test/runtests.jl

@@ -23,7 +23,6 @@ using AbstractGPs:
     TestUtils
 
 using Documenter
-using ChainRulesCore
 using Distributions: MvNormal, PDMat, loglikelihood, Distributions
 using FillArrays
 using FiniteDifferences

test/test_util.jl

@@ -73,8 +73,9 @@ end
 Test _very_ basic consistency properties of the mean function `m`.
 """
 function mean_function_tests(m::MeanFunction, x::AbstractVector)
-    @test AbstractGPs._map_meanfunction(m, x) isa AbstractVector
-    @test length(ew(m, x)) == length(x)
+    mean = AbstractGPs._map_meanfunction(m, x)
+    @test mean isa AbstractVector
+    @test length(mean) == length(x)
 end
 
 """
@@ -87,34 +88,19 @@ end
 Ensure that the gradient w.r.t. the inputs of `MeanFunction` `m` are approximately correct.
 """
 function differentiable_mean_function_tests(
-    m::MeanFunction,
-    ȳ::AbstractVector{<:Real},
-    x::AbstractVector{<:Real};
-    rtol=_rtol,
-    atol=_atol,
+    m::MeanFunction, ȳ::AbstractVector, x::AbstractVector; rtol=_rtol, atol=_atol
 )
     # Run forward tests.
     mean_function_tests(m, x)
 
     # Check adjoint.
     @assert length(ȳ) == length(x)
-    return adjoint_test(x -> ew(m, x), ȳ, x; rtol=rtol, atol=atol)
+    adjoint_test(
+        x -> collect(AbstractGPs._map_meanfunction(m, x)), ȳ, x; rtol=rtol, atol=atol
+    )
+    return nothing
 end
 
-# function differentiable_mean_function_tests(
-#     m::MeanFunction,
-#     ȳ::AbstractVector{<:Real},
-#     x::ColVecs{<:Real};
-#     rtol=_rtol,
-#     atol=_atol,
-# )
-#     # Run forward tests.
-#     mean_function_tests(m, x)
-
-#     @assert length(ȳ) == length(x)
-#     adjoint_test(X->ew(m, ColVecs(X)), ȳ, x.X; rtol=rtol, atol=atol)  
-# end
-
 function differentiable_mean_function_tests(
     rng::AbstractRNG, m::MeanFunction, x::AbstractVector; rtol=_rtol, atol=_atol
 )