Fix more bugs

Author: penelopeysm

src/logdensityfunction.jl

@@ -135,7 +135,7 @@ function LogDensityProblems.logdensity_and_gradient(
     f::LogDensityFunction, θ::AbstractVector, adtype::ADTypes.AbstractADType
 )
     # Ensure we concretise the elements of the params.
-    # θ = map(identity, θ) # TODO: Is this needed?
+    θ = map(identity, θ) # TODO: Is this needed?
     prep = DI.prepare_gradient(_flipped_logdensity, adtype, θ, DI.Constant(f))
     return DI.value_and_gradient(_flipped_logdensity, prep, adtype, θ, DI.Constant(f))
 end

test/ad.jl

@@ -1,6 +1,5 @@
 @testset "AD: ForwardDiff, ReverseDiff, and Mooncake" begin
     @testset "$(m.f)" for m in DynamicPPL.TestUtils.DEMO_MODELS
-        f = DynamicPPL.LogDensityFunction(m)
         rand_param_values = DynamicPPL.TestUtils.rand_prior_true(m)
         vns = DynamicPPL.TestUtils.varnames(m)
         varinfos = DynamicPPL.TestUtils.setup_varinfos(m, rand_param_values, vns)
@@ -11,7 +10,7 @@
             # reference: https://github.com/TuringLang/DynamicPPL.jl/pull/571#issuecomment-1924304489
             θ = convert(Vector{Float64}, varinfo[:])
             # Calculate reference logp + gradient of logp using ForwardDiff
-            default_adtype = ADTypes.AutoForwardDiff(; chunksize=0)
+            default_adtype = ADTypes.AutoForwardDiff()
             ref_logp, ref_grad = LogDensityProblems.logdensity_and_gradient(
                 f, θ, default_adtype
             )