# PolyesterForwardDiff - backend choice is not available

**URL:** <https://discourse.julialang.org/t/polyesterforwarddiff-backend-choice-is-not-available/110985>\
**Category:** Modelling & Simulations\
**Created:** [March 1, 2024, 12:08am UTC](https://discourse.julialang.org/t/polyesterforwarddiff-backend-choice-is-not-available/110985 "2024-03-01T00:08:17Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![sob](https://avatars.discourse-cdn.com/v4/letter/s/65b543/32.png) [@sob](https://discourse.julialang.org/u/sob)\
**Post date:** [March 1, 2024, 12:08am UTC](https://discourse.julialang.org/t/polyesterforwarddiff-backend-choice-is-not-available/110985/1 "2024-03-01T00:08:17Z")

</div>

Hi everybody,

I’m trying to use PolyesterForwardDiff in my code but stumbling upon an error for which the MWE is [inspired by the doc](https://docs.sciml.ai/Optimization/stable/examples/rosenbrock/):

```julia
julia> using Optimization, OptimizationOptimJL, ForwardDiff, PolyesterForwardDiff

julia> rosenbrock(x, p) = (p[1] - x[1])^2 + p[2] * (x[2] - x[1]^2)^2
rosenbrock (generic function with 1 method)

julia> x0 = zeros(2)
2-element Vector{Float64}:
 0.0
 0.0

julia> _p = [1.0, 100.0]
2-element Vector{Float64}:
   1.0
 100.0

julia> optf = OptimizationFunction(rosenbrock, Optimization.AutoForwardDiff())
(::OptimizationFunction{true, AutoForwardDiff{nothing, Nothing}, typeof(rosenbrock), Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED_NO_TIME), Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}) (generic function with 1 method)

julia> prob = OptimizationProblem(optf, x0, _p)
OptimizationProblem. In-place: true
u0: 2-element Vector{Float64}:
 0.0
 0.0

julia> sol = solve(prob, BFGS())
retcode: Success
u: 2-element Vector{Float64}:
 0.9999999999373603
 0.9999999998686199

julia> optf = OptimizationFunction(rosenbrock, Optimization.AutoPolyesterForwardDiff())
(::OptimizationFunction{true, AutoPolyesterForwardDiff{nothing}, typeof(rosenbrock), Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED_NO_TIME), Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing}) (generic function with 1 method)

julia> prob = OptimizationProblem(optf, x0, _p)
OptimizationProblem. In-place: true
u0: 2-element Vector{Float64}:
 0.0
 0.0

julia> sol = solve(prob, BFGS())
ERROR: ArgumentError: The passed automatic differentiation backend choice is not available. Please load the corresponding AD package PolyesterForwardDiff.
Stacktrace:
  [1] instantiate_function(f::Function, x::Optimization.ReInitCache{…}, adtype::AutoPolyesterForwardDiff{…}, p::Int64, num_cons::Int64)
    @ Optimization ~/.julia/packages/Optimization/xbuNJ/src/function.jl:114
  [2] instantiate_function(f::Function, x::Optimization.ReInitCache{…}, adtype::AutoPolyesterForwardDiff{…}, p::Int64)
    @ Optimization ~/.julia/packages/Optimization/xbuNJ/src/function.jl:106
  [3] OptimizationCache(prob::OptimizationProblem{…}, opt::BFGS{…}, data::Base.Iterators.Cycle{…}; callback::Function, maxiters::Nothing, maxtime::Nothing, abstol::Nothing, reltol::Nothing, progress::Bool, kwargs::@Kwargs{})
    @ Optimization ~/.julia/packages/Optimization/xbuNJ/src/cache.jl:27
  [4] __init(prob::OptimizationProblem{…}, opt::BFGS{…}, data::Base.Iterators.Cycle{…}; callback::Function, maxiters::Nothing, maxtime::Nothing, abstol::Nothing, reltol::Nothing, progress::Bool, kwargs::@Kwargs{})
    @ OptimizationOptimJL ~/.julia/packages/OptimizationOptimJL/yMF3E/src/OptimizationOptimJL.jl:101
  [5] __init(prob::OptimizationProblem{…}, opt::BFGS{…}, data::Base.Iterators.Cycle{…})
    @ OptimizationOptimJL ~/.julia/packages/OptimizationOptimJL/yMF3E/src/OptimizationOptimJL.jl:68
  [6] init(::OptimizationProblem{…}, ::BFGS{…}; kwargs::@Kwargs{})
    @ SciMLBase ~/.julia/packages/SciMLBase/Dwomw/src/solve.jl:166
  [7] init(::OptimizationProblem{…}, ::BFGS{…})
    @ SciMLBase ~/.julia/packages/SciMLBase/Dwomw/src/solve.jl:164
  [8] solve(::OptimizationProblem{…}, ::BFGS{…}; kwargs::@Kwargs{})
    @ SciMLBase ~/.julia/packages/SciMLBase/Dwomw/src/solve.jl:96
  [9] solve(::OptimizationProblem{…}, ::BFGS{…})
    @ SciMLBase ~/.julia/packages/SciMLBase/Dwomw/src/solve.jl:93
 [10] top-level scope
    @ REPL[17]:1
Some type information was truncated. Use `show(err)` to see complete types.

```

I looked at the code for `OptimizationCache` but couldn’t figure it out. Has somebody encountered this?

My package versions:

```julia
(Test) pkg> st
Project Test v0.1.0
Status `~/workspace/julia/1.10/Test/Project.toml`
  [052768ef] CUDA v5.2.0
  [5ae59095] Colors v0.12.10
  [a93c6f00] DataFrames v1.6.1
  [aae7a2af] DiffEqFlux v3.3.1
  [071ae1c0] DiffEqGPU v3.4.1
  [0c46a032] DifferentialEquations v7.12.0
  [31c24e10] Distributions v0.25.107
  [587475ba] Flux v0.14.12
  [f6369f11] ForwardDiff v0.10.36
  [e9467ef8] GLMakie v0.9.9
  [cd3eb016] HTTP v1.10.2
  [7f7a1694] Optimization v3.23.0
  [36348300] OptimizationOptimJL v0.2.2
  [42dfb2eb] OptimizationOptimisers v0.2.1
  [1dea7af3] OrdinaryDiffEq v6.73.1
  [f0f68f2c] PlotlyJS v0.18.13
  [91a5bcdd] Plots v1.40.1
  [98d1487c] PolyesterForwardDiff v0.1.1
  [92933f4c] ProgressMeter v1.9.0
  [ce6b1742] RDatasets v0.7.7
  [1ed8b502] SciMLSensitivity v7.56.0
  [789caeaf] StochasticDiffEq v6.65.1
  [e88e6eb3] Zygote v0.6.69
  [02a925ec] cuDNN v1.3.0
  [ea8e919c] SHA v0.7.0
  [10745b16] Statistics v1.10.0

(Test) pkg> 

```

Note that I’m able to successfully call `PolyesterForwardDiff.threaded_gradient!(f, dx, x, ForwardDiff.Chunk(8));` for MWE style values of `f`, `dx` and `x`

---

<div class="post-metadata">

**Author:** ![sob](https://avatars.discourse-cdn.com/v4/letter/s/65b543/32.png) [@sob](https://discourse.julialang.org/u/sob)\
**Post date:** [March 1, 2024, 12:55am UTC](https://discourse.julialang.org/t/polyesterforwarddiff-backend-choice-is-not-available/110985/2 "2024-03-01T00:55:48Z")

</div>

As I was trying to recreate a smaller cleaner env, I got this, which I suspect could be related, but I’m not sure how to interpret it

 ![image](https://global.discourse-cdn.com/julialang/original/3X/8/3/8396cdf063db12c721a1d9e08fcdf7fb0e7b1dca.png)

---

<div class="post-metadata">

**Author:** ![Vaibhavdixit02](https://sea2.discourse-cdn.com/julialang/user_avatar/discourse.julialang.org/vaibhavdixit02/32/2916_2.png) [@Vaibhavdixit02](https://discourse.julialang.org/u/Vaibhavdixit02)\
**Post date:** [March 1, 2024, 1:00am UTC](https://discourse.julialang.org/t/polyesterforwarddiff-backend-choice-is-not-available/110985/3 "2024-03-01T01:00:27Z")

</div>

PolyesterForwardDiff is not a supported AD backend for Optimization.jl yet, only the ones listed in the docs [OptimizationFunction · Optimization.jl](https://docs.sciml.ai/Optimization/stable/API/optimization_function/#Automatic-Differentiation-Construction-Choice-Recommendations) are supported. Create an issue in [GitHub - SciML/OptimizationBase.jl: The base package for Optimization.jl, containing the structs and basic functions for it.](https://github.com/SciML/OptimizationBase.jl) if you’d like it supported.

---

<div class="post-metadata">

**Author:** ![sob](https://avatars.discourse-cdn.com/v4/letter/s/65b543/32.png) [@sob](https://discourse.julialang.org/u/sob)\
**Post date:** [March 1, 2024, 1:23am UTC](https://discourse.julialang.org/t/polyesterforwarddiff-backend-choice-is-not-available/110985/4 "2024-03-01T01:23:42Z")

</div>

That would explain it. Thank you sir. I’ve created [this issue](https://github.com/SciML/OptimizationBase.jl/issues/21). I’m unfortunately not knowledgeable enough to prove more details, but happy to try to help if you think it needs to be altered.

---

<div class="post-metadata">

**Author:** ![sob](https://avatars.discourse-cdn.com/v4/letter/s/65b543/32.png) [@sob](https://discourse.julialang.org/u/sob)\
**Post date:** [March 1, 2024, 1:25am UTC](https://discourse.julialang.org/t/polyesterforwarddiff-backend-choice-is-not-available/110985/5 "2024-03-01T01:25:57Z")

</div>

For context, I’m trying to minimize a loss function that is decently well behaved but unfortunately has a couple dozen dimensions and for which speedy convergence is necessary. Hoping to leverage every core I have.