RecursiveArrayTools v3 compatibility

[njk7062]

I am receiving a MethodError when attempting to run the most basic example provided with ReservoirComputing.jl The MethodError occurs when invoking ESN:

---

julia> esn = ESN(input_data; reservoir = RandSparseReservoir(res_size, radius = 1.2, sparsity = 6 / res_size), input_layer = WeightedLayer(), nla_type = NLAT2()) ERROR: MethodError: no method matching size(::RecursiveArrayTools.DiffEqArray{Float64, 2, Vector{Vector{Float64}}, Vector{Float64}, Vector{Float64}, ODEFunction{true, SciMLBase.AutoSpecialize,

Full details including package and version numbers and terminal I/O is provided below. I look forward to your feedback and guidance in resolving this issue!

nick@localhost-live:~$ julia () | Documentation: https://docs.julialang.org () | () () | | |_ | Type "?" for help, "]?" for Pkg help. | | | | | | |/ ` | | | | || | | | (| | | Version 1.9.4 (2023-11-14) / |_'|||_'_| | Official https://julialang.org/ release |/ |

julia> versioninfo() Julia Version 1.9.4 Commit 8e5136fa297 (2023-11-14 08:46 UTC) Build Info: Official https://julialang.org/ release Platform Info: OS: Linux (x86_64-linux-gnu) CPU: 12 × AMD Ryzen 5 1600 Six-Core Processor WORD_SIZE: 64 LIBM: libopenlibm LLVM: libLLVM-14.0.6 (ORCJIT, znver1) Threads: 1 on 12 virtual cores

(@v1.9) pkg> status Status ~/.julia/environments/v1.9/Project.toml [1dea7af3] OrdinaryDiffEq v6.63.0 [91a5bcdd] Plots v1.39.0 [731186ca] RecursiveArrayTools v3.1.0 [7c2d2b1e] ReservoirComputing v0.9.5

julia> using ReservoirComputing, OrdinaryDiffEq

julia> #lorenz system parameters u0 = [1.0, 0.0, 0.0] 3-element Vector{Float64}: 1.0 0.0 0.0

julia> tspan = (0.0, 200.0) (0.0, 200.0)

julia> p = [10.0, 28.0, 8 / 3] 3-element Vector{Float64}: 10.0 28.0 2.6666666666666665

julia> #define lorenz system function lorenz(du, u, p, t) du[1] = p[1] (u[2] - u[1]) du[2] = u[1] (p[2] - u[3]) - u[2] du[3] = u[1] u[2] - p[3] u[3] end lorenz (generic function with 1 method)

julia> #solve and take data prob = ODEProblem(lorenz, u0, tspan, p) ODEProblem with uType Vector{Float64} and tType Float64. In-place: true timespan: (0.0, 200.0) u0: 3-element Vector{Float64}: 1.0 0.0 0.0

julia> data = solve(prob, ABM54(), dt = 0.02) retcode: Success Interpolation: 3rd order Hermite t: 10001-element Vector{Float64}: 0.0 0.02 0.04 0.06 0.08 0.1 0.12000000000000001 0.14 0.16 0.18 ⋮ 199.84000000002843 199.86000000002844 199.88000000002845 199.90000000002846 199.92000000002847 199.94000000002848 199.9600000000285 199.9800000000285 200.0 u: 10001-element Vector{Vector{Float64}}: [1.0, 0.0, 0.0] [0.8680136241599999, 0.5116114289171798, 0.00464843405046073] [0.846922029853721, 0.9721624694997062, 0.016723754482887952] [0.9127463182632621, 1.4365930707725014, 0.03638911365118479] [1.0547161536343206, 1.9492406388218464, 0.06682954770912672] [1.27153139719415, 2.5500972526141688, 0.11416945073599366] [1.569344101287243, 3.2792182194088135, 0.18865896722509617] [1.9607159493604909, 4.179979661504116, 0.30675850312612357] [2.4642887133184064, 5.301306601644064, 0.4946463038435349] [3.1048014427360577, 6.698665411076851, 0.7936722317742436] ⋮ [-12.712535028050192, -7.991920612082245, 36.96745201041986] [-11.636377668987166, -5.691148385936144, 36.66624998028994] [-10.372587430370148, -3.7766460900344607, 35.77093362370175] [-9.034083394619811, -2.325994687676812, 34.4844534202631] [-7.718171209563215, -1.3249744565922135, 32.98711594710444] [-6.496311609072913, -0.7054966719917979, 31.41147109034457] [-5.412475842259617, -0.3802159215457954, 29.84101721524591] [-4.486956988649644, -0.26529770021033905, 28.320947150047015] [-3.722463627032463, -0.2914952957436106, 26.87140831430775]

julia> shift = 300 300

julia> train_len = 5000 5000

julia> predict_len = 1250 1250

julia> #one step ahead for generative prediction input_data = data[:, shift:(shift + train_len - 1)] t: 5000-element Vector{Float64}: 5.9799999999999605 5.99999999999996 6.01999999999996 6.039999999999959 6.059999999999959 6.079999999999958 6.099999999999958 6.1199999999999575 6.139999999999957 6.159999999999957 ⋮ 105.79999999999609 105.81999999999609 105.83999999999608 105.85999999999608 105.87999999999607 105.89999999999607 105.91999999999607 105.93999999999606 105.95999999999606 u: 5000-element Vector{Vector{Float64}}: [-9.903289241214775, -9.018088588651105, 29.81517777140388] [-9.691839850827593, -8.473011482849701, 29.93605690750639] [-9.420486660590333, -7.939065463718113, 29.90843259211291] [-9.104880950241373, -7.4449797108850175, 29.742036297886404] [-8.762474089642671, -7.01314735768699, 29.454019305573695] [-8.410935717615619, -6.658702746055927, 29.06608913077009] [-8.066811666191718, -6.389828805905225, 28.601842075390245] [-7.744566849246501, -6.2089080639741825, 28.084693027483976] [-7.456053443994025, -6.11406609169808, 27.53654671977279] [-7.210360712173874, -6.1007246144663565, 26.977154921025395] ⋮ [12.518922134483173, 10.108387400259105, 34.677601864758] [11.905676310941601, 8.232629715112148, 35.05977814254804] [11.071914312382939, 6.472053614446601, 34.882263966926644] [10.092017839786998, 4.960342280071891, 34.24498656635422] [9.04494324433941, 3.7654618681658447, 33.275347750522876] [8.001895218860973, 2.8967891643506336, 32.09636572944009] [7.018825886840775, 2.3239810199099424, 30.808116692367136] [6.13395766702855, 1.9972030630363637, 29.48232612752881] [5.3690197189757916, 1.862732065250123, 28.165385874059083]

julia> target_data = data[:, (shift + 1):(shift + train_len)] t: 5000-element Vector{Float64}: 5.99999999999996 6.01999999999996 6.039999999999959 6.059999999999959 6.079999999999958 6.099999999999958 6.1199999999999575 6.139999999999957 6.159999999999957 6.179999999999956 ⋮ 105.81999999999609 105.83999999999608 105.85999999999608 105.87999999999607 105.89999999999607 105.91999999999607 105.93999999999606 105.95999999999606 105.97999999999605 u: 5000-element Vector{Vector{Float64}}: [-9.691839850827593, -8.473011482849701, 29.93605690750639] [-9.420486660590333, -7.939065463718113, 29.90843259211291] [-9.104880950241373, -7.4449797108850175, 29.742036297886404] [-8.762474089642671, -7.01314735768699, 29.454019305573695] [-8.410935717615619, -6.658702746055927, 29.06608913077009] [-8.066811666191718, -6.389828805905225, 28.601842075390245] [-7.744566849246501, -6.2089080639741825, 28.084693027483976] [-7.456053443994025, -6.11406609169808, 27.53654671977279] [-7.210360712173874, -6.1007246144663565, 26.977154921025395] [-7.013953305940451, -6.162911685030047, 26.423996356199307] ⋮ [11.905676310941601, 8.232629715112148, 35.05977814254804] [11.071914312382939, 6.472053614446601, 34.882263966926644] [10.092017839786998, 4.960342280071891, 34.24498656635422] [9.04494324433941, 3.7654618681658447, 33.275347750522876] [8.001895218860973, 2.8967891643506336, 32.09636572944009] [7.018825886840775, 2.3239810199099424, 30.808116692367136] [6.13395766702855, 1.9972030630363637, 29.48232612752881] [5.3690197189757916, 1.862732065250123, 28.165385874059083] [4.732497640190652, 1.8723908301054109, 26.88478061861816]

julia> test = data[:, (shift + train_len):(shift + train_len + predict_len - 1)]

t: 1250-element Vector{Float64}: 105.97999999999605 105.99999999999605 106.01999999999605 106.03999999999604 106.05999999999604 106.07999999999603 106.09999999999603 106.11999999999603 106.13999999999602 106.15999999999602 ⋮ 130.7999999999931 130.81999999999312 130.83999999999313 130.85999999999314 130.87999999999315 130.89999999999316 130.91999999999317 130.93999999999318 130.9599999999932 u: 1250-element Vector{Vector{Float64}}: [4.732497640190652, 1.8723908301054109, 26.88478061861816] [4.22349538680031, 1.987820915806937, 25.655613923905776] [3.8353316780542905, 2.181372263125947, 24.485767387745945] [3.5584604091713943, 2.4351927068305277, 23.37943997904429] [3.3826188986546626, 2.739600649501364, 22.339370937107052] [3.2982766454142727, 3.0913593199467937, 21.368185196603726] [3.2975239355256036, 3.492142131634464, 20.46924395963976] [3.3745467566607004, 3.947276445140908, 19.647275341036046] [3.5258121339502693, 4.464737468128744, 18.908963446114747] [3.7500536807441187, 5.054297346775937, 18.263604797307796] ⋮ [5.667284334594522, 7.920044709382794, 21.837292439416736] [6.128504560099492, 8.495628927065395, 21.64604924227314] [6.616442143819489, 9.132648149983781, 21.6160067677496] [7.1357622016914535, 9.812756425953655, 21.762471295222465] [7.686353276111834, 10.510880836564114, 22.099963839944905] [8.262833251743796, 11.193794316023716, 22.639394570288623] [8.853950777066231, 11.819492971445888, 23.384189012804384] [9.442118229459183, 12.338166583539232, 24.3255347922894] [10.003421378477109, 12.695609055900354, 25.43743234339266]

julia> res_size = 300 300

julia> esn = ESN(input_data; reservoir = RandSparseReservoir(res_size, radius = 1.2, sparsity = 6 / res_size), input_layer = WeightedLayer(), nla_type = NLAT2()) ERROR: MethodError: no method matching size(::RecursiveArrayTools.DiffEqArray{Float64, 2, Vector{Vector{Float64}}, Vector{Float64}, Vector{Float64}, ODEFunction{true, SciMLBase.AutoSpecialize, FunctionWrappersWrappers.FunctionWrappersWrapper{Tuple{FunctionWrappers.FunctionWrapper{Nothing, Tuple{Vector{Float64}, Vector{Float64}, Vector{Float64}, Float64}}, FunctionWrappers.FunctionWrapper{Nothing, Tuple{Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{Float64}, Float64}}, FunctionWrappers.FunctionWrapper{Nothing, Tuple{Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{Float64}, Vector{Float64}, ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}}, FunctionWrappers.FunctionWrapper{Nothing, Tuple{Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{Float64}, ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}}}, false}, LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED), Nothing, SymbolicIndexingInterface.SymbolCache{Nothing, Nothing, Nothing}}}, ::Int64)

Closest candidates are: size(::Union{LinearAlgebra.QR, LinearAlgebra.QRCompactWY, LinearAlgebra.QRPivoted}, ::Integer) @ LinearAlgebra ~/.julia/juliaup/julia-1.9.4+0.x64.linux.gnu/share/julia/stdlib/v1.9/LinearAlgebra/src/qr.jl:581 size(::Union{LinearAlgebra.QRCompactWYQ, LinearAlgebra.QRPackedQ}, ::Integer) @ LinearAlgebra ~/.julia/juliaup/julia-1.9.4+0.x64.linux.gnu/share/julia/stdlib/v1.9/LinearAlgebra/src/qr.jl:583 size(::Union{LinearAlgebra.Cholesky, LinearAlgebra.CholeskyPivoted}, ::Integer) @ LinearAlgebra ~/.julia/juliaup/julia-1.9.4+0.x64.linux.gnu/share/julia/stdlib/v1.9/LinearAlgebra/src/cholesky.jl:513 ...

Stacktrace: [1] ESN(train_data::RecursiveArrayTools.DiffEqArray{Float64, 2, Vector{Vector{Float64}}, Vector{Float64}, Vector{Float64}, ODEFunction{true, SciMLBase.AutoSpecialize, FunctionWrappersWrappers.FunctionWrappersWrapper{Tuple{FunctionWrappers.FunctionWrapper{Nothing, Tuple{Vector{Float64}, Vector{Float64}, Vector{Float64}, Float64}}, FunctionWrappers.FunctionWrapper{Nothing, Tuple{Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{Float64}, Float64}}, FunctionWrappers.FunctionWrapper{Nothing, Tuple{Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{Float64}, Vector{Float64}, ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}}, FunctionWrappers.FunctionWrapper{Nothing, Tuple{Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}, Vector{Float64}, ForwardDiff.Dual{ForwardDiff.Tag{DiffEqBase.OrdinaryDiffEqTag, Float64}, Float64, 1}}}}, false}, LinearAlgebra.UniformScaling{Bool}, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, Nothing, typeof(SciMLBase.DEFAULT_OBSERVED), Nothing, SymbolicIndexingInterface.SymbolCache{Nothing, Nothing, Nothing}}}; variation::Default, input_layer::WeightedLayer{Float64}, reservoir::RandSparseReservoir{Float64, Float64}, bias::NullLayer, reservoir_driver::RNN{typeof(NNlib.tanh_fast), Float64}, nla_type::NLAT2, states_type::StandardStates, washout::Int64, matrix_type::Type) @ ReservoirComputing ~/.julia/packages/ReservoirComputing/ptuHU/src/esn/echostatenetwork.jl:112 [2] top-level scope @ REPL[17]:1

julia>

[ChrisRackauckas]

Try ]add RecursiveArrayTools@2

[MartinuzziFrancesco]

Chris' suggestion is the solution and works on my machine (Julia 1.9.4, RC 0.9.5 and ODE 6.63 like you). If for any reason you can't add that package a workaround like data = reduce(hcat, data.u) while keeping everything else the same also works. Let us know if any of these help you

[njk7062]

Thank you all for your prompt response! Everything now running perfectly.

[ChrisRackauckas]

@MartinuzziFrancesco let's yank all versions that were launched with v3 compatibility, downgrade, and make a new version. It sounds like we'll need to do a bit more to get v3 running in practice.

Reopening to catch this.

[MartinuzziFrancesco]

ok I'll get on it. I'm changing the title as well to highlight the issue better

This doesn't affect the internals though, it is just an issue over the examples and documentation I think. I will double check

[ChrisRackauckas]

@AayushSabharwal can you look into a solution here? I don't think it's easy to pull back from since this library didn't explicitly use RAT.

[AayushSabharwal]

The fix here is simple enough, but there's another error later when running the example.

Isn't this line redundant? Its output is not assigned anywhere

[AayushSabharwal]

https://github.com/SciML/RecursiveArrayTools.jl/pull/316

This along with removing the line mentioned above fixes this issue.

[ChrisRackauckas]

Master is fixed.