Error finding gradient of poislogpdf

[dtjohnson]

I'm getting an error when trying to find the gradient of the poislogpdf for a MAP analysis.

using StatsFuns
using ForwardDiff
ForwardDiff.gradient(x -> poislogpdf(x[1], 2), [0.75])

MethodError: no method matching Float64(::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1})
Closest candidates are:
Float64(::Real, !Matched::RoundingMode) where T<:AbstractFloat at rounding.jl:185
Float64(::T<:Number) where T<:Number at boot.jl:725
Float64(!Matched::Int8) at float.jl:60
...

Stacktrace:
[1] convert(::Type{Float64}, ::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1}) at ./number.jl:7
[2] cconvert(::Type, ::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1}) at ./essentials.jl:344
[3] poislogpdf(::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1}, ::Int64) at /Users/djohnson/.julia/packages/StatsFuns/0W2sM/src/rmath.jl:58
[4] (::getfield(Main, Symbol("##31#32")))(::Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1},1}) at ./In[40]:2
[5] vector_mode_dual_eval(::getfield(Main, Symbol("##31#32")), ::Array{Float64,1}, ::ForwardDiff.GradientConfig{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1,Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1},1}}) at /Users/djohnson/.julia/packages/ForwardDiff/hnKaN/src/apiutils.jl:35
[6] vector_mode_gradient(::getfield(Main, Symbol("##31#32")), ::Array{Float64,1}, ::ForwardDiff.GradientConfig{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1,Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1},1}}) at /Users/djohnson/.julia/packages/ForwardDiff/hnKaN/src/gradient.jl:96
[7] gradient(::Function, ::Array{Float64,1}, ::ForwardDiff.GradientConfig{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1,Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1},1}}, ::Val{true}) at /Users/djohnson/.julia/packages/ForwardDiff/hnKaN/src/gradient.jl:17
[8] gradient(::Function, ::Array{Float64,1}, ::ForwardDiff.GradientConfig{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1,Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##31#32")),Float64},Float64,1},1}}) at /Users/djohnson/.julia/packages/ForwardDiff/hnKaN/src/gradient.jl:15 (repeats 2 times)
[9] top-level scope at In[40]:2

Looks like poislogpdf is fine accepting duals but something fails farther in.

[dtjohnson]

I should add that the gradient works for normlogpdf:

ForwardDiff.gradient(x -> normlogpdf(x[1], x[2], 2), [0.75, 1.0])

2-element Array{Float64,1}:
1.25
0.5625

However, I get the same failure for gammalogpdf:

ForwardDiff.gradient(x -> gammalogpdf(x[1], x[2], 2), [0.75, 1.0])

MethodError: no method matching Float64(::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2})
Closest candidates are:
Float64(::Real, !Matched::RoundingMode) where T<:AbstractFloat at rounding.jl:185
Float64(::T<:Number) where T<:Number at boot.jl:725
Float64(!Matched::Int8) at float.jl:60
...

Stacktrace:
[1] convert(::Type{Float64}, ::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2}) at ./number.jl:7
[2] cconvert(::Type, ::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2}) at ./essentials.jl:344
[3] gammalogpdf(::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2}, ::ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2}, ::Int64) at /Users/djohnson/.julia/packages/StatsFuns/0W2sM/src/rmath.jl:58
[4] (::getfield(Main, Symbol("##77#78")))(::Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2},1}) at ./In[93]:3
[5] vector_mode_dual_eval(::getfield(Main, Symbol("##77#78")), ::Array{Float64,1}, ::ForwardDiff.GradientConfig{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2,Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2},1}}) at /Users/djohnson/.julia/packages/ForwardDiff/hnKaN/src/apiutils.jl:35
[6] vector_mode_gradient(::getfield(Main, Symbol("##77#78")), ::Array{Float64,1}, ::ForwardDiff.GradientConfig{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2,Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2},1}}) at /Users/djohnson/.julia/packages/ForwardDiff/hnKaN/src/gradient.jl:96
[7] gradient(::Function, ::Array{Float64,1}, ::ForwardDiff.GradientConfig{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2,Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2},1}}, ::Val{true}) at /Users/djohnson/.julia/packages/ForwardDiff/hnKaN/src/gradient.jl:17
[8] gradient(::Function, ::Array{Float64,1}, ::ForwardDiff.GradientConfig{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2,Array{ForwardDiff.Dual{ForwardDiff.Tag{getfield(Main, Symbol("##77#78")),Float64},Float64,2},1}}) at /Users/djohnson/.julia/packages/ForwardDiff/hnKaN/src/gradient.jl:15 (repeats 2 times)
[9] top-level scope at In[93]:3

[ngiann]

Same problem occurs with the Hessian too.

[devmotion]

Do you still have these issues? All examples mentioned here (including Hessians) work for me on Julia 1.1.1 with StatsFuns v0.8.0 and ForwardDiff v0.10.3. Maybe the problems were fixed by #65?

[devmotion]

As mentioned above, I can't reproduce the errors anymore. Both examples work on recent versions of StatsFuns, therefore I'll close this issue.