Fix deprecations

Project.toml

@@ -1,6 +1,6 @@
 name = "DistributionsAD"
 uuid = "ced4e74d-a319-5a8a-b0ac-84af2272839c"
-version = "0.6.42"
+version = "0.6.43"
 
 [deps]
 Adapt = "79e6a3ab-5dfb-504d-930d-738a2a938a0e"

README.md

@@ -24,6 +24,7 @@ AD of `logpdf` is fully supported and tested for the following distributions wrt
     - `Chi`
     - `Chisq`
     - `Cosine`
+    - `Distributions.AffineDistribution`
     - `Epanechnikov`
     - `Erlang`
     - `Exponential`
@@ -38,7 +39,6 @@ AD of `logpdf` is fully supported and tested for the following distributions wrt
     - `Kolmogorov`
     - `Laplace`
     - `Levy`
-    - `LocationScale`
     - `Logistic`
     - `LogitNormal`
     - `LogNormal`

src/common.jl

@@ -1,11 +1,13 @@
 ## Linear Algebra ##
 
+const CHOLESKY_NoPivot = VERSION >= v"1.8.0-rc1" ? LinearAlgebra.NoPivot() : Val(false)
+
 function turing_chol(A::AbstractMatrix, check)
     chol = cholesky(A, check=check)
     (chol.factors, chol.info)
 end
 function turing_chol_back(A::AbstractMatrix, check)
-    C, chol_pullback = rrule(cholesky, A, Val(false), check=check)
+    C, chol_pullback = rrule(cholesky, A, CHOLESKY_NoPivot; check=check)
     function back(Δ)
         Ȳ = Tangent{typeof(C)}(; factors=Δ[1])
         ∂C = chol_pullback(Ȳ)[2]
@@ -19,7 +21,7 @@ function symm_turing_chol(A::AbstractMatrix, check, uplo)
     (chol.factors, chol.info)
 end
 function symm_turing_chol_back(A::AbstractMatrix, check, uplo)
-    C, chol_pullback = rrule(cholesky, Symmetric(A,uplo), Val(false), check=check)
+    C, chol_pullback = rrule(cholesky, Symmetric(A,uplo), CHOLESKY_NoPivot; check=check)
     function back(Δ)
         Ȳ = Tangent{typeof(C)}(; factors=Δ[1])
         ∂C = chol_pullback(Ȳ)[2]

src/flatten.jl

@@ -46,7 +46,7 @@ const flattened_dists = [   Bernoulli,
                             Kolmogorov,
                             Laplace,
                             Levy,
-                            LocationScale,
+                            Distributions.AffineDistribution,
                             Logistic,
                             LogitNormal,
                             LogNormal,

src/reversediff.jl

@@ -148,14 +148,14 @@ function Gamma(α::T, θ::T; check_args=true) where {T <: TrackedReal}
 end
 
 # Work around to stop TrackedReal of Inf and -Inf from producing NaN in the derivative
-function Base.minimum(d::LocationScale{T}) where {T <: TrackedReal}
+function Base.minimum(d::Distributions.AffineDistribution{T}) where {T <: TrackedReal}
     if isfinite(minimum(d.ρ))
         return d.μ + d.σ * minimum(d.ρ)
     else
         return convert(T, ReverseDiff.@skip(minimum)(d.ρ))
     end
 end
-function Base.maximum(d::LocationScale{T}) where {T <: TrackedReal}
+function Base.maximum(d::Distributions.AffineDistribution{T}) where {T <: TrackedReal}
     if isfinite(minimum(d.ρ))
         return d.μ + d.σ * maximum(d.ρ)
     else

test/ad/distributions.jl

@@ -143,7 +143,8 @@
         DistSpec(Levy, (0.0,), 0.5),
         DistSpec(Levy, (0.0, 2.0), 0.5),
 
-        DistSpec((a, b) -> LocationScale(a, b, Normal()), (1.0, 2.0), 0.5),
+        # Test AffineDistribution
+        DistSpec((a, b) -> a + b * Beta(), (1.0, 2.0), 2.0),
 
         DistSpec(Logistic, (), 0.5),
         DistSpec(Logistic, (1.0,), 0.5),