Reverse Mode AD doesn't work if NamedTuples are used as parameters

[mattborghi]

Lately, I have been trying to use NamedTuples for passing parameters to a DE problem in order to compute derivatives but this method seems to fail.

For comparison, I make the same calculation using a Vector of parameters instead.

I am using the latest stable versions for the presented packages and Julia v1.5.0.

# load packages
using DifferentialEquations
using DiffEqSensitivity
using Statistics: mean
using Parameters: @unpack
using ForwardDiff
using ReverseDiff
using Zygote

# initialize parameters 
tspan = (0., 1.)
u₀ = 1.
r = 0.05
σ = 0.4

# receives p = [ r, σ] as an array
function fs(u, p, t)
    r = p[1]
    r .* u
end

function gs(u, p, t)
    σ = p[2]
    σ .* u
end

# receives p = (; r, σ) as a NamedTuple
function f(u, p, t)
    @unpack r = p
    r .* u
end

function g(u, p, t)
    @unpack σ = p
    σ .* u
end

# Define problems
prob = SDEProblem{false}(f, g, u₀, tspan)
oop = EnsembleProblem(prob)

probs = SDEProblem{false}(fs, gs, u₀, tspan)
oops = EnsembleProblem(probs)

# define parameters
p = [u₀, r, σ]

the solve method:

1. receives a Vector as input for p when simple function is used.
2. receives a NamedTuple as input for p when calculate function is used.

# Define functions to be derivated
simple = function (up, prob = probs, sensealg = TrackerAdjoint(); seed=111)
    Array(solve(
    prob,
    SRIW1();
    u0=[up[1]],
    p=up[2:end],
    saveat=0.1,
    sensealg,
    seed,
    trajectories=1_000,
  ))[:,end,:] |> mean
end

calculate = function (up, prob = prob, sensealg = TrackerAdjoint(); seed=111)
    p = (;r=up[2], σ=up[3])
    Array(solve(
    prob,
    SRIW1();
    u0=[up[1]],
    p,
    saveat=0.1,
    sensealg,
    seed,
    trajectories=1_000,
  ))[:,end,:] |> mean
end

# evaluate the functions
simple(p, oops) # 1.4020033221163284 
calculate(p, oop) # 1.4020033221163284

Just for comparison we calculate Forward Mode AD without problems.

# using ForwardDiff.jl
ForwardDiff.gradient(x -> calculate(x, oop), p)
# 3-element Array{Float64,1}:
# 1.5799924814996311
#  1.581962649205976
#  1.2930442221318288
ForwardDiff.gradient(x -> simple(x, oops), p)
# 3-element Array{Float64,1}:
# 1.5799924814996311
#  1.581962649205976
#  1.2930442221318288

Reverse mode AD using ReverseDiff.jl

Using (1) with different sensealgs gives different results but works. Using (2) fails.

# using ReverseDiff.jl
ReverseDiff.gradient(x -> simple(x, oops, TrackerAdjoint()), p)
# 3-element Array{Float64,1}:
# 0.0
#  1.4032274627896102
#  0.7290267892368372
ReverseDiff.gradient(x -> simple(x, oops, QuadratureAdjoint()), p)
# 3-element Array{Float64,1}:
# 0.0
#  1.2517556141290906
#  0.0
ReverseDiff.gradient(x -> calculate(x, oop, TrackerAdjoint()), p)

ERROR: type Float64 has no field partials

ERROR: type Float64 has no field partials
Stacktrace: 2. getproperty(::Float64, ::Symbol) at ./Base.jl:33 3. _deriv(::Function, ::Float64, ::Val{1}, ::Float64, ::Float64, ::Float64, ::Float64, ::Float64, ::Float64) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/derivatives/broadcast.jl:224 4. _broadcast_getindex_evalf at ./broadcast.jl:648 5. _broadcast_getindex at ./broadcast.jl:621 6. getindex at ./broadcast.jl:575 7. copy at ./broadcast.jl:876 8. materialize at ./broadcast.jl:837 9. macro expansion at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/derivatives/broadcast.jl:0 10. _derivs(::Base.Broadcast.var"#2#4"{Base.Broadcast.var"#8#10"{Base.Broadcast.var"#8#10"{Base.Broadcast.var"#1#3",Base.Broadcast.var"#8#10"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"},Base.Broadcast.var"#5#6"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"}},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(+)},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(DiffEqBase.ODE_DEFAULT_NORM)},Base.Broadcast.var"#8#10"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"},Base.Broadcast.var"#5#6"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"}},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(-)},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(DiffEqBase.ODE_DEFAULT_NORM)},typeof(max)}, ::Array{Float64,1}, ::Array{Float64,1}, ::Array{Float64,1}, ::Float64, ::Array{Float64,1}, ::Array{Float64,1}, ::Float64) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/derivatives/broadcast.jl:226 11. special_reverse_exec!(::ReverseDiff.SpecialInstruction{typeof(ReverseDiff.tracker_∇broadcast),Tuple{ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},Float64,ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},Float64},ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},Tuple{Base.Broadcast.var"#2#4"{Base.Broadcast.var"#8#10"{Base.Broadcast.var"#8#10"{Base.Broadcast.var"#1#3",Base.Broadcast.var"#8#10"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"},Base.Broadcast.var"#5#6"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"}},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(+)},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(DiffEqBase.ODE_DEFAULT_NORM)},Base.Broadcast.var"#8#10"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"},Base.Broadcast.var"#5#6"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"}},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(-)},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(DiffEqBase.ODE_DEFAULT_NORM)},typeof(max)}}}) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/derivatives/broadcast.jl:255 12. reverse_exec!(::ReverseDiff.SpecialInstruction{typeof(ReverseDiff.tracker_∇broadcast),Tuple{ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},Float64,ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},Float64},ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},Tuple{Base.Broadcast.var"#2#4"{Base.Broadcast.var"#8#10"{Base.Broadcast.var"#8#10"{Base.Broadcast.var"#1#3",Base.Broadcast.var"#8#10"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"},Base.Broadcast.var"#5#6"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"}},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(+)},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(DiffEqBase.ODE_DEFAULT_NORM)},Base.Broadcast.var"#8#10"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"},Base.Broadcast.var"#5#6"{Base.Broadcast.var"#5#6"{Base.Broadcast.var"#7#9"}},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(-)},Base.Broadcast.var"#11#12"{Base.Broadcast.var"#11#12"{Base.Broadcast.var"#13#14"}},Base.Broadcast.var"#15#16"{Base.Broadcast.var"#15#16"{Base.Broadcast.var"#17#18"}},typeof(DiffEqBase.ODE_DEFAULT_NORM)},typeof(max)}}}) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/tape.jl:93 13. reverse_pass!(::Array{ReverseDiff.AbstractInstruction,1}) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/tape.jl:87 14. reverse_pass! at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/api/tape.jl:36 15. seeded_reverse_pass!(::Array{Float64,1}, ::ReverseDiff.TrackedReal{Float64,Float64,Nothing}, ::ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}}, ::ReverseDiff.GradientTape{var"#19#20",ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},ReverseDiff.TrackedReal{Float64,Float64,Nothing}}) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/api/utils.jl:31 16. seeded_reverse_pass!(::Array{Float64,1}, ::ReverseDiff.GradientTape{var"#19#20",ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}},ReverseDiff.TrackedReal{Float64,Float64,Nothing}}) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/api/tape.jl:47 17. gradient(::Function, ::Array{Float64,1}, ::ReverseDiff.GradientConfig{ReverseDiff.TrackedArray{Float64,Float64,1,Array{Float64,1},Array{Float64,1}}}) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/api/gradients.jl:24 18. gradient(::Function, ::Array{Float64,1}) at /home/borghi/.julia/packages/ReverseDiff/jFRo1/src/api/gradients.jl:22 19. top-level scope at /home/borghi/Desktop/UniversalMonteCarlo.jl/examples/greeks/fix.jl:99 20. include_string(::Function, ::Module, ::String, ::String) at ./loading.jl:1091 21. invokelatest(::Any, ::Any, ::Vararg{Any,N} where N; kwargs::Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}}) at ./essentials.jl:710 22. invokelatest(::Any, ::Any, ::Vararg{Any,N} where N) at ./essentials.jl:709 23. inlineeval(::Module, ::String, ::Int64, ::Int64, ::String; softscope::Bool) at /home/borghi/.vscode/extensions/julialang.language-julia-1.0.8/scripts/packages/VSCodeServer/src/eval.jl:83 24. (::VSCodeServer.var"#43#45"{VSCodeServer.ReplRunCodeRequestParams,String,Int64,Int64,String,Module,Bool})() at /home/borghi/.vscode/extensions/julialang.language-julia-1.0.8/scripts/packages/VSCodeServer/src/eval.jl:45 25. withpath(::VSCodeServer.var"#43#45"{VSCodeServer.ReplRunCodeRequestParams,String,Int64,Int64,String,Module,Bool}, ::String) at /home/borghi/.vscode/extensions/julialang.language-julia-1.0.8/scripts/packages/VSCodeServer/src/repl.jl:118 26. (::VSCodeServer.var"#42#44"{VSCodeServer.ReplRunCodeRequestParams,String,Int64,Int64,String,Module,Bool,Bool})() at /home/borghi/.vscode/extensions/julialang.language-julia-1.0.8/scripts/packages/VSCodeServer/src/eval.jl:43 27. hideprompt(::VSCodeServer.var"#42#44"{VSCodeServer.ReplRunCodeRequestParams,String,Int64,Int64,String,Module,Bool,Bool}) at /home/borghi/.vscode/extensions/julialang.language-julia-1.0.8/scripts/packages/VSCodeServer/src/repl.jl:36 28. repl_runcode_request(::VSCodeServer.JSONRPC.JSONRPCEndpoint{Base.PipeEndpoint,Base.PipeEndpoint}, ::VSCodeServer.ReplRunCodeRequestParams) at /home/borghi/.vscode/extensions/julialang.language-julia-1.0.8/scripts/packages/VSCodeServer/src/eval.jl:23 29. dispatch_msg(::VSCodeServer.JSONRPC.JSONRPCEndpoint{Base.PipeEndpoint,Base.PipeEndpoint}, ::VSCodeServer.JSONRPC.MsgDispatcher, ::Dict{String,Any}) at /home/borghi/.vscode/extensions/julialang.language-julia-1.0.8/scripts/packages/JSONRPC/src/typed.jl:66 30. macro expansion at /home/borghi/.vscode/extensions/julialang.language-julia-1.0.8/scripts/packages/VSCodeServer/src/VSCodeServer.jl:95 31. (::VSCodeServer.var"#61#63"{Bool,String})() at ./task.jl:356

Closes call of ReverseDiff.gradient

Reverse mode AD using Zygote.jl

Using (1) works. Using (2) fails.

# using Zygote.jl
Zygote.gradient(x -> simple(x, oops, QuadratureAdjoint()), p)[1]
# 3-element Array{Float64,1}:
# 1.0512710963760281
#  1.2517556141290906
#  0.0
Zygote.gradient(x -> calculate(x, oops, QuadratureAdjoint()), p)[1]

ERROR:MethodError: no method matching getproperty(::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Int64)

ERROR: MethodError: no method matching getproperty(::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Int64)
Closest candidates are:
  getproperty(!Matched::Tuple, ::Int64) at Base.jl:30
  getproperty(!Matched::ChainRulesCore.Composite, ::Int64) at /home/borghi/.julia/packages/ChainRulesCore/tBJg6/src/differentials/composite.jl:63
  getproperty(!Matched::Union{Tables.AbstractColumns, Tables.AbstractRow}, ::Int64) at /home/borghi/.julia/packages/Tables/xHhzi/src/Tables.jl:164

Stacktrace: 2. adjoint at /home/borghi/.julia/packages/Zygote/chgvX/src/lib/lib.jl:197 3. _pullback at /home/borghi/.julia/packages/ZygoteRules/6nssF/src/adjoint.jl:47 4. _pullback(::Zygote.Context, ::typeof(Zygote.literal_getindex), ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Val{1}) at /home/borghi/.julia/packages/Zygote/chgvX/src/lib/lib.jl:217 5. fs at /home/borghi/Desktop/UniversalMonteCarlo.jl/examples/greeks/fix.jl:16 6. _pullback(::Zygote.Context, ::typeof(fs), ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Float64) at /home/borghi/.julia/packages/Zygote/chgvX/src/compiler/interface2.jl:0 7. adjoint at /home/borghi/.julia/packages/Zygote/chgvX/src/lib/lib.jl:172 8. _pullback at /home/borghi/.julia/packages/ZygoteRules/6nssF/src/adjoint.jl:47 9. SDEFunction at /home/borghi/.julia/packages/DiffEqBase/gLFRA/src/diffeqfunction.jl:286 10. _pullback(::Zygote.Context, ::SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing}, ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Float64) at /home/borghi/.julia/packages/Zygote/chgvX/src/compiler/interface2.jl:0 11. #30 at /home/borghi/.julia/dev/DiffEqSensitivity/src/local_sensitivity/derivative_wrappers.jl:290 12. _pullback(::Zygote.Context, ::DiffEqSensitivity.var"#30#32"{Float64,SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing}}, ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}) at /home/borghi/.julia/packages/Zygote/chgvX/src/compiler/interface2.jl:0 13. _pullback(::Function, ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}) at /home/borghi/.julia/packages/Zygote/chgvX/src/compiler/interface.jl:38 14. pullback(::Function, ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}) at /home/borghi/.julia/packages/Zygote/chgvX/src/compiler/interface.jl:44 15. _vecjacobian!(::Array{Float64,1}, ::Array{Float64,1}, ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Float64, ::DiffEqSensitivity.ODEQuadratureAdjointSensitivityFunction{DiffEqSensitivity.AdjointDiffCache{Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Base.OneTo{Int64},UnitRange{Int64},LinearAlgebra.UniformScaling{Bool}},QuadratureAdjoint{0,true,Val{:central},Bool},Array{Float64,1},RODESolution{Float64,2,Array{Array{Float64,1},1},Nothing,Nothing,Array{Float64,1},NoiseProcess{Float64,2,Float64,Array{Float64,1},Array{Float64,1},Array{Array{Float64,1},1},typeof(DiffEqNoiseProcess.WHITE_NOISE_DIST),typeof(DiffEqNoiseProcess.WHITE_NOISE_BRIDGE),false,ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},RSWM{Float64},Nothing,RandomNumbers.Xorshifts.Xoroshiro128Plus},SDEProblem{Array{Float64,1},Tuple{Float64,Float64},false,NamedTuple{(:r, :σ),Tuple{Float64,Float64}},Nothing,SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing},typeof(gs),Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}},Nothing},SRIW1,StochasticDiffEq.LinearInterpolationData{Array{Array{Float64,1},1},Array{Float64,1}},DiffEqBase.DEStats},SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing}}, ::ZygoteVJP, ::Nothing, ::Nothing) at /home/borghi/.julia/dev/DiffEqSensitivity/src/local_sensitivity/derivative_wrappers.jl:289 16. _vecjacobian!(::Array{Float64,1}, ::Array{Float64,1}, ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Float64, ::DiffEqSensitivity.ODEQuadratureAdjointSensitivityFunction{DiffEqSensitivity.AdjointDiffCache{Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Base.OneTo{Int64},UnitRange{Int64},LinearAlgebra.UniformScaling{Bool}},QuadratureAdjoint{0,true,Val{:central},Bool},Array{Float64,1},RODESolution{Float64,2,Array{Array{Float64,1},1},Nothing,Nothing,Array{Float64,1},NoiseProcess{Float64,2,Float64,Array{Float64,1},Array{Float64,1},Array{Array{Float64,1},1},typeof(DiffEqNoiseProcess.WHITE_NOISE_DIST),typeof(DiffEqNoiseProcess.WHITE_NOISE_BRIDGE),false,ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},RSWM{Float64},Nothing,RandomNumbers.Xorshifts.Xoroshiro128Plus},SDEProblem{Array{Float64,1},Tuple{Float64,Float64},false,NamedTuple{(:r, :σ),Tuple{Float64,Float64}},Nothing,SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing},typeof(gs),Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}},Nothing},SRIW1,StochasticDiffEq.LinearInterpolationData{Array{Array{Float64,1},1},Array{Float64,1}},DiffEqBase.DEStats},SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing}}, ::Bool, ::Nothing, ::Nothing) at /home/borghi/.julia/dev/DiffEqSensitivity/src/local_sensitivity/derivative_wrappers.jl:193 17. vecjacobian!(::Array{Float64,1}, ::Array{Float64,1}, ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Float64, ::DiffEqSensitivity.ODEQuadratureAdjointSensitivityFunction{DiffEqSensitivity.AdjointDiffCache{Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Base.OneTo{Int64},UnitRange{Int64},LinearAlgebra.UniformScaling{Bool}},QuadratureAdjoint{0,true,Val{:central},Bool},Array{Float64,1},RODESolution{Float64,2,Array{Array{Float64,1},1},Nothing,Nothing,Array{Float64,1},NoiseProcess{Float64,2,Float64,Array{Float64,1},Array{Float64,1},Array{Array{Float64,1},1},typeof(DiffEqNoiseProcess.WHITE_NOISE_DIST),typeof(DiffEqNoiseProcess.WHITE_NOISE_BRIDGE),false,ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},RSWM{Float64},Nothing,RandomNumbers.Xorshifts.Xoroshiro128Plus},SDEProblem{Array{Float64,1},Tuple{Float64,Float64},false,NamedTuple{(:r, :σ),Tuple{Float64,Float64}},Nothing,SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing},typeof(gs),Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}},Nothing},SRIW1,StochasticDiffEq.LinearInterpolationData{Array{Array{Float64,1},1},Array{Float64,1}},DiffEqBase.DEStats},SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing}}; dgrad::Nothing, dy::Nothing) at /home/borghi/.julia/dev/DiffEqSensitivity/src/local_sensitivity/derivative_wrappers.jl:147 18. vecjacobian!(::Array{Float64,1}, ::Array{Float64,1}, ::Array{Float64,1}, ::NamedTuple{(:r, :σ),Tuple{Float64,Float64}}, ::Float64, ::DiffEqSensitivity.ODEQuadratureAdjointSensitivityFunction{DiffEqSensitivity.AdjointDiffCache{Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Base.OneTo{Int64},UnitRange{Int64},LinearAlgebra.UniformScaling{Bool}},QuadratureAdjoint{0,true,Val{:central},Bool},Array{Float64,1},RODESolution{Float64,2,Array{Array{Float64,1},1},Nothing,Nothing,Array{Float64,1},NoiseProcess{Float64,2,Float64,Array{Float64,1},Array{Float64,1},Array{Array{Float64,1},1},typeof(DiffEqNoiseProcess.WHITE_NOISE_DIST),typeof(DiffEqNoiseProcess.WHITE_NOISE_BRIDGE),false,ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},RSWM{Float64},Nothing,RandomNumbers.Xorshifts.Xoroshiro128Plus},SDEProblem{Array{Float64,1},Tuple{Float64,Float64},false,NamedTuple{(:r, :σ),Tuple{Float64,Float64}},Nothing,SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing},typeof(gs),Base.Iterators.Pairs{Union{},Union{},Tuple{},NamedTuple{(),Tuple{}}},Nothing},SRIW1,StochasticDiffEq.LinearInterpolationData{Array{Array{Float64,1},1},Array{Float64,1}},DiffEqBase.DEStats},SDEFunction{false,typeof(fs),typeof(gs),LinearAlgebra.UniformScaling{Bool},Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing}}) at /home/borghi/.julia/dev/DiffEqSensitivity/src/local_sensitivity/derivative_wrappers.jl:147 [18] (::DiffEqSensitivity.ODEQuadratureAdjointSensitivityFunction{DiffEqSensitivity.AdjointDiffCache{Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Nothing,Base.OneTo{Int64},UnitRange{Int64},LinearAlgebra.UniformScaling{Bool}},QuadratureAdjoint{0,true,Val{:central},Bool},Array{Float64,1},RODESolution{Float64,2,Array{Array{Float64,1},1},Nothing,Nothing,Array{Float64,1},NoiseProcess{Float64,2,Float64,Array{Float64,1},Array{Float64,1},Array{Array{Float64,1},1},typeof(DiffEqNoiseProcess.WHITE_NOISE_DIST),typeof(DiffEqNoiseProcess.WHITE_NOISE_BRIDGE),false,ResettableStacks.ResettableStack{Tuple{Float64,Array{Float64,1},Array{Float64,1}},false},ResettableStacks.ResettableStack{Tuple{Floa

Thanks in advance.

[ChrisRackauckas]

Yes, the object for the parameters needs to be some kind of abstract array, otherwise math is not defined on it. This is something I'm hoping to fix, but for now, you need a type that is suitable as a differential equation array (such as any <:AbstractArray)

[mattborghi]

Thank you Chris! Is there any issue open right now in order to track the progress?

[ChrisRackauckas]

I guess https://github.com/SciML/DiffEqFlux.jl/issues/178 was it, but this is a better issue to track it from so I'll leave this open. DiffEqFlux's issue here is essentially a DiffEqSensitivity issue in that it needs to build a new differential equation.