ERROR: LoadError: MethodError: no method matching Float32(::ReverseDiff.TrackedReal{Float64, Float32, Nothing})

[prbzrg]

When I run this code:

using DiffEqFlux, DifferentialEquations, GalacticOptim, Distributions

nn = Chain(
    Dense(1, 3, tanh),
    Dense(3, 1, tanh),
) |> f32
tspan = (0.0f0, 10.0f0)
ffjord_mdl = FFJORD(nn, tspan, Tsit5())

data_dist = Normal(6.0f0, 0.7f0)
train_data = rand(data_dist, 1, 100)

function loss(θ)
    logpx, λ₁, λ₂ = ffjord_mdl(train_data, θ)
    -mean(logpx)
end

adtype = GalacticOptim.AutoReverseDiff()
res1 = DiffEqFlux.sciml_train(loss, ffjord_mdl.p, ADAM(0.1), adtype; maxiters=100)

I get this error:

ERROR: LoadError: MethodError: no method matching Float32(::ReverseDiff.TrackedReal{Float64, Float32, Nothing})
Closest candidates are:
  (::Type{T})(::Real, ::RoundingMode) where T<:AbstractFloat at rounding.jl:200
  (::Type{T})(::T) where T<:Number at boot.jl:760
  (::Type{T})(::AbstractChar) where T<:Union{AbstractChar, Number} at char.jl:50
  ...
Stacktrace:
  [1] _broadcast_getindex_evalf
    @ .\broadcast.jl:648 [inlined]
  [2] _broadcast_getindex
    @ .\broadcast.jl:621 [inlined]
  [3] getindex
    @ .\broadcast.jl:575 [inlined]
  [4] copy
    @ .\broadcast.jl:922 [inlined]
  [5] materialize(bc::Base.Broadcast.Broadcasted{Base.Broadcast.DefaultArrayStyle{2}, Nothing, Type{Float32}, Tuple{Matrix{ReverseDiff.TrackedReal{Float64, Float32, Nothing}}}})
    @ Base.Broadcast .\broadcast.jl:883
  [6] WARNING: both Flux and Iterators export "flatten"; uses of it in module DiffEqFlux must be qualified
WARNING: both Flux and Distributions export "params"; uses of it in module DiffEqFlux must be qualified
(::FFJORD{Chain{Tuple{Dense{typeof(tanh), Matrix{Float32}, Vector{Float32}}, Dense{typeof(tanh), Matrix{Float32}, Vector{Float32}}}}, Vector{Float32}, Flux.var"#60#62"{Chain{Tuple{Dense{typeof(tanh), Matrix{Float32}, Vector{Float32}}, Dense{typeof(tanh), Matrix{Float32}, Vector{Float32}}}}}, FullNormal, Tuple{Float32, Float32}, Tuple{Tsit5}, Base.Iterators.Pairs{Union{}, Union{}, Tuple{}, NamedTuple{(), Tuple{}}}})(x::Matrix{Float32}, p::ReverseDiff.TrackedArray{Float32, Float32, 1, Vector{Float32}, Vector{Float32}}, e::Matrix{Float32}; regularize::Bool, monte_carlo::Bool)
    @ DiffEqFlux C:\Users\Hossein Pourbozorg\.julia\packages\DiffEqFlux\N7blG\src\ffjord.jl:215
  [7] FFJORD (repeats 2 times)
    @ C:\Users\Hossein Pourbozorg\.julia\packages\DiffEqFlux\N7blG\src\ffjord.jl:192 [inlined]
  [8] loss(θ::ReverseDiff.TrackedArray{Float32, Float32, 1, Vector{Float32}, Vector{Float32}})
    @ Main C:\Users\Hossein Pourbozorg\Code Projects\Mine\ffjord-report-issues\iss-4.jl:14
  [9] #74
    @ C:\Users\Hossein Pourbozorg\.julia\packages\DiffEqFlux\N7blG\src\train.jl:84 [inlined]
 [10] (::GalacticOptim.var"#229#238"{OptimizationFunction{true, GalacticOptim.AutoReverseDiff, DiffEqFlux.var"#74#79"{typeof(loss)}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, Nothing})(::ReverseDiff.TrackedArray{Float32, Float32, 1, Vector{Float32}, Vector{Float32}})
    @ GalacticOptim C:\Users\Hossein Pourbozorg\.julia\packages\GalacticOptim\bEh06\src\function\reversediff.jl:6
 [11] #231
    @ C:\Users\Hossein Pourbozorg\.julia\packages\GalacticOptim\bEh06\src\function\reversediff.jl:9 [inlined]
 [12] ReverseDiff.GradientTape(f::GalacticOptim.var"#231#240"{Tuple{}, GalacticOptim.var"#229#238"{OptimizationFunction{true, GalacticOptim.AutoReverseDiff, DiffEqFlux.var"#74#79"{typeof(loss)}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, Nothing}}, input::Vector{Float32}, cfg::ReverseDiff.GradientConfig{ReverseDiff.TrackedArray{Float32, Float32, 1, Vector{Float32}, Vector{Float32}}})
    @ ReverseDiff C:\Users\Hossein Pourbozorg\.julia\packages\ReverseDiff\E4Tzn\src\api\tape.jl:199
 [13] gradient!(result::Vector{Float32}, f::Function, input::Vector{Float32}, cfg::ReverseDiff.GradientConfig{ReverseDiff.TrackedArray{Float32, Float32, 1, Vector{Float32}, Vector{Float32}}})
    @ ReverseDiff C:\Users\Hossein Pourbozorg\.julia\packages\ReverseDiff\E4Tzn\src\api\gradients.jl:41
 [14] (::GalacticOptim.var"#230#239"{GalacticOptim.var"#229#238"{OptimizationFunction{true, GalacticOptim.AutoReverseDiff, DiffEqFlux.var"#74#79"{typeof(loss)}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, Nothing}})(::Vector{Float32}, ::Vector{Float32})
    @ GalacticOptim C:\Users\Hossein Pourbozorg\.julia\packages\GalacticOptim\bEh06\src\function\reversediff.jl:9
 [15] macro expansion
    @ C:\Users\Hossein Pourbozorg\.julia\packages\GalacticOptim\bEh06\src\solve\flux.jl:43 [inlined]
 [16] macro expansion
    @ C:\Users\Hossein Pourbozorg\.julia\packages\GalacticOptim\bEh06\src\solve\solve.jl:35 [inlined]
 [17] __solve(prob::OptimizationProblem{false, OptimizationFunction{false, GalacticOptim.AutoReverseDiff, OptimizationFunction{true, GalacticOptim.AutoReverseDiff, DiffEqFlux.var"#74#79"{typeof(loss)}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, GalacticOptim.var"#230#239"{GalacticOptim.var"#229#238"{OptimizationFunction{true, GalacticOptim.AutoReverseDiff, DiffEqFlux.var"#74#79"{typeof(loss)}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, Nothing}}, GalacticOptim.var"#232#241"{GalacticOptim.var"#229#238"{OptimizationFunction{true, GalacticOptim.AutoReverseDiff, DiffEqFlux.var"#74#79"{typeof(loss)}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}, Nothing}}, GalacticOptim.var"#237#246", Nothing, Nothing, Nothing}, Vector{Float32}, SciMLBase.NullParameters, Nothing, Nothing, Nothing, Nothing, Base.Iterators.Pairs{Union{}, Union{}, Tuple{}, NamedTuple{(), Tuple{}}}}, opt::ADAM, data::Base.Iterators.Cycle{Tuple{GalacticOptim.NullData}}; maxiters::Int64, cb::Function, progress::Bool, save_best::Bool, kwargs::Base.Iterators.Pairs{Union{}, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
    @ GalacticOptim C:\Users\Hossein Pourbozorg\.julia\packages\GalacticOptim\bEh06\src\solve\flux.jl:41
 [18] #solve#474
    @ C:\Users\Hossein Pourbozorg\.julia\packages\SciMLBase\UIp7W\src\solve.jl:3 [inlined]
 [19] sciml_train(::typeof(loss), ::Vector{Float32}, ::ADAM, ::GalacticOptim.AutoReverseDiff; lower_bounds::Nothing, upper_bounds::Nothing, maxiters::Int64,
 kwargs::Base.Iterators.Pairs{Union{}, Union{}, Tuple{}, NamedTuple{(), Tuple{}}})
    @ DiffEqFlux C:\Users\Hossein Pourbozorg\.julia\packages\DiffEqFlux\N7blG\src\train.jl:89
 [20] top-level scope
    @ C:\Users\Hossein Pourbozorg\Code Projects\Mine\ffjord-report-issues\iss-4.jl:19
in expression starting at C:\Users\Hossein Pourbozorg\Code Projects\Mine\ffjord-report-issues\iss-4.jl:19

In this environment:

(ffjord-report-issues) pkg> status
      Status `C:\Users\Hossein Pourbozorg\Code Projects\Mine\ffjord-report-issues\Project.toml`
  [aae7a2af] DiffEqFlux v1.43.0
  [0c46a032] DifferentialEquations v6.19.0
  [31c24e10] Distributions v0.25.14
  [a75be94c] GalacticOptim v2.0.3
  [c3e4b0f8] Pluto v0.15.1

julia> versioninfo()
Julia Version 1.6.1
Commit 6aaedecc44 (2021-04-23 05:59 UTC)
Platform Info:
  OS: Windows (x86_64-w64-mingw32)
  CPU: Intel(R) Core(TM) i7-10750H CPU @ 2.60GHz
  WORD_SIZE: 64
  LIBM: libopenlibm
  LLVM: libLLVM-11.0.1 (ORCJIT, skylake)

[prbzrg]

I think it's related to type conversion in FFJORD at this line:

DiffEqFlux.jl/src/ffjord.jl

Line 215 in 4bcfc23

logpz = eltype(x).(reshape(logpdf(pz, z), 1, size(x, 2)))

[ChrisRackauckas]

This got fixed.