search

SciML / SciMLSensitivity.jl

A component of the DiffEq ecosystem for enabling sensitivity analysis for scientific machine learning (SciML). Optimize-then-discretize, discretize-then-optimize, adjoint methods, and more for ODEs, SDEs, DDEs, DAEs, etc.
https://docs.sciml.ai/SciMLSensitivity/stable/
Other
333 stars 71 forks source link

Grandient calculation fails when using a parameter-dependent `SciMLOperator` #1139

Open albertomercurio opened 3 weeks ago

albertomercurio commented 3 weeks ago

Describe the bug 🐞

All the examples in the SciMLSensitivity.jl Documentation use a user-defined function for the ODEProblem. I need instead to define a parameter-dependent SciMLOperator (e.g., a MatrixOperator, AddedOperator, ...), but it fails.

Expected behavior

Returning the correct gradient without errors.

Minimal Reproducible Example 👇

using LinearAlgebra
using OrdinaryDiffEq
using SciMLOperators
using Zygote
using Enzyme
using SciMLSensitivity

const T = Float64
const N = 10
const u0 = ones(T, N)
H_tmp = rand(T, N, N)
const H = H_tmp + H_tmp'

coef(a, u, p, t) = - p[1]

function my_f(γ)
    # U = MatrixOperator(-1im * H - γ * Diagonal(H))
    U = ScalarOperator(one(γ), coef) * MatrixOperator(Diagonal(H))
    tspan = (0.0, 1.0)
    # prob = ODEProblem{true}(U, u0, tspan, [γ], sensealg = InterpolatingAdjoint(autojacvec=false))
    prob = ODEProblem{true}(U, u0, tspan, [γ])
    sol = solve(prob, Tsit5())
    return real(sol.u[end][end])
end

my_f(1.9) # 0.25621142049273665

Error & Stacktrace ⚠️

Zygote Error

Zygote.gradient(my_f, 1.9)
ERROR: MethodError: no method matching Float64(::ForwardDiff.Dual{ForwardDiff.Tag{ODEFunction{…}, Float64}, Float64, 10})
The type `Float64` exists, but no method is defined for this combination of argument types when trying to construct it.

Closest candidates are:
  (::Type{T})(::Real, ::RoundingMode) where T<:AbstractFloat
   @ Base rounding.jl:265
  (::Type{T})(::T) where T<:Number
   @ Core boot.jl:900
  Float64(::IrrationalConstants.Invπ)
   @ IrrationalConstants ~/.julia/packages/IrrationalConstants/vp5v4/src/macro.jl:112
  ...

Stacktrace:
  [1] convert(::Type{Float64}, x::ForwardDiff.Dual{ForwardDiff.Tag{ODEFunction{…}, Float64}, Float64, 10})
    @ Base ./number.jl:7
  [2] setproperty!(x::ScalarOperator{Float64, typeof(coef)}, f::Symbol, v::ForwardDiff.Dual{ForwardDiff.Tag{…}, Float64, 10})
    @ Base ./Base.jl:52
  [3] update_coefficients!(L::ScalarOperator{…}, u::Vector{…}, p::Vector{…}, t::Float64; kwargs::@Kwargs{})
    @ SciMLOperators ~/.julia/packages/SciMLOperators/Q5dkx/src/scalar.jl:193
  [4] update_coefficients!(L::ScalarOperator{…}, u::Vector{…}, p::Vector{…}, t::Float64)
    @ SciMLOperators ~/.julia/packages/SciMLOperators/Q5dkx/src/scalar.jl:192
  [5] update_coefficients!(L::SciMLOperators.ScaledOperator{…}, u::Vector{…}, p::Vector{…}, t::Float64)
    @ SciMLOperators ~/.julia/packages/SciMLOperators/Q5dkx/src/basic.jl:251
  [6] (::SciMLOperators.ScaledOperator{…})(du::Vector{…}, u::Vector{…}, p::Vector{…}, t::Float64; kwargs::@Kwargs{})
    @ SciMLOperators ~/.julia/packages/SciMLOperators/Q5dkx/src/interface.jl:116
  [7] (::SciMLOperators.ScaledOperator{…})(du::Vector{…}, u::Vector{…}, p::Vector{…}, t::Float64)
    @ SciMLOperators ~/.julia/packages/SciMLOperators/Q5dkx/src/interface.jl:115
  [8] (::ODEFunction{…})(::Vector{…}, ::Vararg{…})
    @ SciMLBase ~/.julia/packages/SciMLBase/hJh6T/src/scimlfunctions.jl:2355
  [9] (::ODEFunction{…})(::Vector{…}, ::Vararg{…})
    @ SciMLBase ~/.julia/packages/SciMLBase/hJh6T/src/scimlfunctions.jl:2355
 [10] initialize!(integrator::OrdinaryDiffEqCore.ODEIntegrator{…}, cache::OrdinaryDiffEqTsit5.Tsit5Cache{…})
    @ OrdinaryDiffEqTsit5 ~/.julia/packages/OrdinaryDiffEqTsit5/DHYtz/src/tsit_perform_step.jl:175
 [11] __init(prob::ODEProblem{…}, alg::Tsit5{…}, timeseries_init::Tuple{}, ts_init::Tuple{}, ks_init::Tuple{}, recompile::Type{…}; saveat::Vector{…}, tstops::Tuple{}, d_discontinuities::Tuple{}, save_idxs::Nothing, save_everystep::Bool, save_on::Bool, save_start::Bool, save_end::Nothing, callback::Nothing, dense::Bool, calck::Bool, dt::Float64, dtmin::Float64, dtmax::Float64, force_dtmin::Bool, adaptive::Bool, gamma::Rational{…}, abstol::Nothing, reltol::Nothing, qmin::Rational{…}, qmax::Int64, qsteady_min::Int64, qsteady_max::Int64, beta1::Nothing, beta2::Nothing, qoldinit::Rational{…}, controller::Nothing, fullnormalize::Bool, failfactor::Int64, maxiters::Int64, internalnorm::typeof(DiffEqBase.ODE_DEFAULT_NORM), internalopnorm::typeof(opnorm), isoutofdomain::typeof(DiffEqBase.ODE_DEFAULT_ISOUTOFDOMAIN), unstable_check::typeof(DiffEqBase.ODE_DEFAULT_UNSTABLE_CHECK), verbose::Bool, timeseries_errors::Bool, dense_errors::Bool, advance_to_tstop::Bool, stop_at_next_tstop::Bool, initialize_save::Bool, progress::Bool, progress_steps::Int64, progress_name::String, progress_message::typeof(DiffEqBase.ODE_DEFAULT_PROG_MESSAGE), progress_id::Symbol, userdata::Nothing, allow_extrapolation::Bool, initialize_integrator::Bool, alias_u0::Bool, alias_du0::Bool, initializealg::OrdinaryDiffEqCore.DefaultInit, kwargs::@Kwargs{})
    @ OrdinaryDiffEqCore ~/.julia/packages/OrdinaryDiffEqCore/NnA60/src/solve.jl:528
 [12] __init (repeats 5 times)
    @ ~/.julia/packages/OrdinaryDiffEqCore/NnA60/src/solve.jl:11 [inlined]
 [13] #__solve#75
    @ ~/.julia/packages/OrdinaryDiffEqCore/NnA60/src/solve.jl:6 [inlined]
 [14] __solve
    @ ~/.julia/packages/OrdinaryDiffEqCore/NnA60/src/solve.jl:1 [inlined]
 [15] solve_call(_prob::ODEProblem{…}, args::Tsit5{…}; merge_callbacks::Bool, kwargshandle::Nothing, kwargs::@Kwargs{…})
    @ DiffEqBase ~/.julia/packages/DiffEqBase/frOsk/src/solve.jl:612
 [16] solve_call
    @ ~/.julia/packages/DiffEqBase/frOsk/src/solve.jl:569 [inlined]
 [17] #solve_up#53
    @ ~/.julia/packages/DiffEqBase/frOsk/src/solve.jl:1092 [inlined]
 [18] solve_up
    @ ~/.julia/packages/DiffEqBase/frOsk/src/solve.jl:1078 [inlined]
 [19] #solve#51
    @ ~/.julia/packages/DiffEqBase/frOsk/src/solve.jl:1015 [inlined]
 [20] (::SciMLSensitivity.var"#333#342"{…})()
    @ SciMLSensitivity ~/.julia/packages/SciMLSensitivity/XCu1T/src/concrete_solve.jl:1073
 [21] unthunk
    @ ~/.julia/packages/ChainRulesCore/6Pucz/src/tangent_types/thunks.jl:205 [inlined]
 [22] wrap_chainrules_output
    @ ~/.julia/packages/Zygote/NRp5C/src/compiler/chainrules.jl:110 [inlined]
 [23] map
    @ ./tuple.jl:357 [inlined]
 [24] map (repeats 3 times)
    @ ./tuple.jl:358 [inlined]
 [25] wrap_chainrules_output
    @ ~/.julia/packages/Zygote/NRp5C/src/compiler/chainrules.jl:111 [inlined]
 [26] ZBack
    @ ~/.julia/packages/Zygote/NRp5C/src/compiler/chainrules.jl:212 [inlined]
 [27] (::Zygote.var"#294#295"{…})(Δ::ODESolution{…})
    @ Zygote ~/.julia/packages/Zygote/NRp5C/src/lib/lib.jl:206
 [28] (::Zygote.var"#2169#back#296"{…})(Δ::ODESolution{…})
    @ Zygote ~/.julia/packages/ZygoteRules/M4xmc/src/adjoint.jl:72
 [29] #solve#51
    @ ~/.julia/packages/DiffEqBase/frOsk/src/solve.jl:1015 [inlined]
 [30] (::Zygote.Pullback{…})(Δ::ODESolution{…})
    @ Zygote ~/.julia/packages/Zygote/NRp5C/src/compiler/interface2.jl:0
 [31] #294
    @ ~/.julia/packages/Zygote/NRp5C/src/lib/lib.jl:206 [inlined]
 [32] #2169#back
    @ ~/.julia/packages/ZygoteRules/M4xmc/src/adjoint.jl:72 [inlined]
 [33] solve
    @ ~/.julia/packages/DiffEqBase/frOsk/src/solve.jl:1005 [inlined]
 [34] (::Zygote.Pullback{…})(Δ::ODESolution{…})
    @ Zygote ~/.julia/packages/Zygote/NRp5C/src/compiler/interface2.jl:0
 [35] my_f
    @ ~/GitHub/Research/Undef/Autodiff QuantumToolbox/autodiff.jl:135 [inlined]
 [36] (::Zygote.Pullback{Tuple{…}, Tuple{…}})(Δ::Float64)
    @ Zygote ~/.julia/packages/Zygote/NRp5C/src/compiler/interface2.jl:0
 [37] (::Zygote.var"#78#79"{Zygote.Pullback{Tuple{…}, Tuple{…}}})(Δ::Float64)
    @ Zygote ~/.julia/packages/Zygote/NRp5C/src/compiler/interface.jl:91
 [38] gradient(f::Function, args::Float64)
    @ Zygote ~/.julia/packages/Zygote/NRp5C/src/compiler/interface.jl:148
 [39] top-level scope
    @ ~/GitHub/Research/Undef/Autodiff QuantumToolbox/autodiff.jl:143
Some type information was truncated. Use `show(err)` to see complete types.

Enzyme Error

autodiff(Reverse, my_f, Active(1.9))
ERROR: Constant memory is stored (or returned) to a differentiable variable.
As a result, Enzyme cannot provably ensure correctness and throws this error.
This might be due to the use of a constant variable as temporary storage for active memory (https://enzyme.mit.edu/julia/stable/faq/#Runtime-Activity).
If Enzyme should be able to prove this use non-differentable, open an issue!
To work around this issue, either:
 a) rewrite this variable to not be conditionally active (fastest, but requires a code change), or
 b) set the Enzyme mode to turn on runtime activity (e.g. autodiff(set_runtime_activity(Reverse), ...) ). This will maintain correctness, but may slightly reduce performance.
Mismatched activity for:   store atomic {} addrspace(10)* addrspacecast ({}* inttoptr (i64 138286044220400 to {}*) to {} addrspace(10)*), {} addrspace(10)* addrspace(11)* %48 release, align 8, !dbg !410, !tbaa !336, !alias.scope !340, !noalias !343 const val: {} addrspace(10)* addrspacecast ({}* inttoptr (i64 138286044220400 to {}*) to {} addrspace(10)*)
 value=[1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0, 1.0] of type Vector{Float64}
 llvalue={} addrspace(10)* addrspacecast ({}* inttoptr (i64 138286044220400 to {}*) to {} addrspace(10)*)

Stacktrace:
 [1] _
   @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:118
 [2] ODEProblem (repeats 2 times)
   @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:111
 [3] #ODEProblem#318
   @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:194
 [4] ODEProblem
   @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:193
 [5] _
   @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:140
 [6] ODEProblem
   @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:136
 [7] my_f
   @ ~/GitHub/Research/Undef/Autodiff QuantumToolbox/autodiff.jl:134

Stacktrace:
  [1] _
    @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:118 [inlined]
  [2] ODEProblem (repeats 2 times)
    @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:111 [inlined]
  [3] #ODEProblem#318
    @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:194 [inlined]
  [4] ODEProblem
    @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:193 [inlined]
  [5] _
    @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:140 [inlined]
  [6] ODEProblem
    @ ~/.julia/packages/SciMLBase/hJh6T/src/problems/ode_problems.jl:136 [inlined]
  [7] my_f
    @ ~/GitHub/Research/Undef/Autodiff QuantumToolbox/autodiff.jl:134 [inlined]
  [8] diffejulia_my_f_49982wrap
    @ ~/GitHub/Research/Undef/Autodiff QuantumToolbox/autodiff.jl:0
  [9] macro expansion
    @ ~/.julia/packages/Enzyme/BRtTP/src/compiler.jl:8137 [inlined]
 [10] enzyme_call
    @ ~/.julia/packages/Enzyme/BRtTP/src/compiler.jl:7703 [inlined]
 [11] CombinedAdjointThunk
    @ ~/.julia/packages/Enzyme/BRtTP/src/compiler.jl:7476 [inlined]
 [12] autodiff
    @ ~/.julia/packages/Enzyme/BRtTP/src/Enzyme.jl:491 [inlined]
 [13] autodiff
    @ ~/.julia/packages/Enzyme/BRtTP/src/Enzyme.jl:537 [inlined]
 [14] autodiff(mode::ReverseMode{false, false, FFIABI, false, false}, f::typeof(my_f), args::Active{Float64})
    @ Enzyme ~/.julia/packages/Enzyme/BRtTP/src/Enzyme.jl:504
 [15] top-level scope
    @ ~/GitHub/Research/Undef/Autodiff QuantumToolbox/autodiff.jl:147

Environment (please complete the following information):

Status `~/GitHub/Research/Undef/Autodiff QuantumToolbox/Project.toml`
  [6e4b80f9] BenchmarkTools v1.5.0
  [7da242da] Enzyme v0.13.12
  [1dea7af3] OrdinaryDiffEq v6.89.0
  [6c2fb7c5] QuantumToolbox v0.20.0 `~/.julia/dev/QuantumToolbox`
  [1ed8b502] SciMLSensitivity v7.69.0
  [53ae85a6] SciMLStructures v1.5.0
  [e88e6eb3] Zygote v0.6.72
Status `~/GitHub/Research/Undef/Autodiff QuantumToolbox/Manifest.toml`
  [47edcb42] ADTypes v1.9.0
  [621f4979] AbstractFFTs v1.5.0
  [1520ce14] AbstractTrees v0.4.5
  [7d9f7c33] Accessors v0.1.38
  [79e6a3ab] Adapt v4.1.1
  [66dad0bd] AliasTables v1.1.3
  [ec485272] ArnoldiMethod v0.4.0
  [4fba245c] ArrayInterface v7.16.0
  [4c555306] ArrayLayouts v1.10.4
  [a9b6321e] Atomix v0.1.0
  [6e4b80f9] BenchmarkTools v1.5.0
  [e2ed5e7c] Bijections v0.1.9
  [62783981] BitTwiddlingConvenienceFunctions v0.1.6
  [fa961155] CEnum v0.5.0
  [2a0fbf3d] CPUSummary v0.2.6
  [7057c7e9] Cassette v0.3.14
  [082447d4] ChainRules v1.71.0
  [d360d2e6] ChainRulesCore v1.25.0
  [fb6a15b2] CloseOpenIntervals v0.1.13
  [861a8166] Combinatorics v1.0.2
  [38540f10] CommonSolve v0.2.4
  [bbf7d656] CommonSubexpressions v0.3.1
  [f70d9fcc] CommonWorldInvalidations v1.0.0
  [34da2185] Compat v4.16.0
  [b0b7db55] ComponentArrays v0.15.17
  [b152e2b5] CompositeTypes v0.1.4
  [a33af91c] CompositionsBase v0.1.2
  [2569d6c7] ConcreteStructs v0.2.3
  [187b0558] ConstructionBase v1.5.8
  [adafc99b] CpuId v0.3.1
  [9a962f9c] DataAPI v1.16.0
  [864edb3b] DataStructures v0.18.20
  [e2d170a0] DataValueInterfaces v1.0.0
  [2b5f629d] DiffEqBase v6.158.3
  [459566f4] DiffEqCallbacks v4.0.0
  [77a26b50] DiffEqNoiseProcess v5.23.0
  [163ba53b] DiffResults v1.1.0
  [b552c78f] DiffRules v1.15.1
  [a0c0ee7d] DifferentiationInterface v0.6.18
  [b4f34e82] Distances v0.10.12
  [31c24e10] Distributions v0.25.112
  [ffbed154] DocStringExtensions v0.9.3
  [5b8099bc] DomainSets v0.7.14
  [7c1d4256] DynamicPolynomials v0.6.0
  [4e289a0a] EnumX v1.0.4
  [7da242da] Enzyme v0.13.12
  [f151be2c] EnzymeCore v0.8.5
  [d4d017d3] ExponentialUtilities v1.26.1
  [e2ba6199] ExprTools v0.1.10
⌅ [6b7a57c9] Expronicon v0.8.5
  [7a1cc6ca] FFTW v1.8.0
  [7034ab61] FastBroadcast v0.3.5
  [9aa1b823] FastClosures v0.3.2
  [29a986be] FastLapackInterface v2.0.4
  [a4df4552] FastPower v1.1.1
  [1a297f60] FillArrays v1.13.0
  [6a86dc24] FiniteDiff v2.26.0
  [1fa38f19] Format v1.3.7
  [f6369f11] ForwardDiff v0.10.36
  [f62d2435] FunctionProperties v0.1.2
  [069b7b12] FunctionWrappers v1.1.3
  [77dc65aa] FunctionWrappersWrappers v0.1.3
  [d9f16b24] Functors v0.4.12
⌅ [0c68f7d7] GPUArrays v10.3.1
⌅ [46192b85] GPUArraysCore v0.1.6
  [61eb1bfa] GPUCompiler v1.0.1
  [14197337] GenericLinearAlgebra v0.3.14
  [c145ed77] GenericSchur v0.5.4
  [86223c79] Graphs v1.12.0
  [3e5b6fbb] HostCPUFeatures v0.1.17
  [34004b35] HypergeometricFunctions v0.3.24
  [7869d1d1] IRTools v0.4.14
  [615f187c] IfElse v0.1.1
  [40713840] IncompleteLU v0.2.1
  [d25df0c9] Inflate v0.1.5
  [18e54dd8] IntegerMathUtils v0.1.2
  [8197267c] IntervalSets v0.7.10
  [3587e190] InverseFunctions v0.1.17
  [92d709cd] IrrationalConstants v0.2.2
  [82899510] IteratorInterfaceExtensions v1.0.0
  [692b3bcd] JLLWrappers v1.6.1
  [682c06a0] JSON v0.21.4
  [ccbc3e58] JumpProcesses v9.14.0
  [ef3ab10e] KLU v0.6.0
  [63c18a36] KernelAbstractions v0.9.29
  [ba0b0d4f] Krylov v0.9.8
  [929cbde3] LLVM v9.1.3
  [b964fa9f] LaTeXStrings v1.4.0
  [23fbe1c1] Latexify v0.16.5
  [10f19ff3] LayoutPointers v0.1.17
  [5078a376] LazyArrays v2.2.1
  [2d8b4e74] LevyArea v1.0.0
  [87fe0de2] LineSearch v0.1.4
  [d3d80556] LineSearches v7.3.0
  [7ed4a6bd] LinearSolve v2.36.2
  [2ab3a3ac] LogExpFunctions v0.3.28
  [bdcacae8] LoopVectorization v0.12.171
  [d8e11817] MLStyle v0.4.17
  [1914dd2f] MacroTools v0.5.13
  [d125e4d3] ManualMemory v0.1.8
  [bb5d69b7] MaybeInplace v0.1.4
  [e1d29d7a] Missings v1.2.0
  [46d2c3a1] MuladdMacro v0.2.4
  [102ac46a] MultivariatePolynomials v0.5.7
  [d8a4904e] MutableArithmetics v1.5.2
  [d41bc354] NLSolversBase v7.8.3
  [2774e3e8] NLsolve v4.5.1
  [872c559c] NNlib v0.9.24
  [77ba4419] NaNMath v1.0.2
  [8913a72c] NonlinearSolve v3.15.1
  [d8793406] ObjectFile v0.4.2
  [6fe1bfb0] OffsetArrays v1.14.1
  [429524aa] Optim v1.9.4
  [3bd65402] Optimisers v0.3.3
  [bac558e1] OrderedCollections v1.6.3
  [1dea7af3] OrdinaryDiffEq v6.89.0
  [89bda076] OrdinaryDiffEqAdamsBashforthMoulton v1.1.0
  [6ad6398a] OrdinaryDiffEqBDF v1.1.2
  [bbf590c4] OrdinaryDiffEqCore v1.9.0
  [50262376] OrdinaryDiffEqDefault v1.1.0
  [4302a76b] OrdinaryDiffEqDifferentiation v1.1.0
  [9286f039] OrdinaryDiffEqExplicitRK v1.1.0
  [e0540318] OrdinaryDiffEqExponentialRK v1.1.0
  [becaefa8] OrdinaryDiffEqExtrapolation v1.2.1
  [5960d6e9] OrdinaryDiffEqFIRK v1.2.0
  [101fe9f7] OrdinaryDiffEqFeagin v1.1.0
  [d3585ca7] OrdinaryDiffEqFunctionMap v1.1.1
  [d28bc4f8] OrdinaryDiffEqHighOrderRK v1.1.0
  [9f002381] OrdinaryDiffEqIMEXMultistep v1.1.0
  [521117fe] OrdinaryDiffEqLinear v1.1.0
  [1344f307] OrdinaryDiffEqLowOrderRK v1.2.0
  [b0944070] OrdinaryDiffEqLowStorageRK v1.2.1
  [127b3ac7] OrdinaryDiffEqNonlinearSolve v1.2.1
  [c9986a66] OrdinaryDiffEqNordsieck v1.1.0
  [5dd0a6cf] OrdinaryDiffEqPDIRK v1.1.0
  [5b33eab2] OrdinaryDiffEqPRK v1.1.0
  [04162be5] OrdinaryDiffEqQPRK v1.1.0
  [af6ede74] OrdinaryDiffEqRKN v1.1.0
  [43230ef6] OrdinaryDiffEqRosenbrock v1.2.0
  [2d112036] OrdinaryDiffEqSDIRK v1.1.0
  [669c94d9] OrdinaryDiffEqSSPRK v1.2.0
  [e3e12d00] OrdinaryDiffEqStabilizedIRK v1.1.0
  [358294b1] OrdinaryDiffEqStabilizedRK v1.1.0
  [fa646aed] OrdinaryDiffEqSymplecticRK v1.1.0
  [b1df2697] OrdinaryDiffEqTsit5 v1.1.0
  [79d7bb75] OrdinaryDiffEqVerner v1.1.1
  [90014a1f] PDMats v0.11.31
  [65ce6f38] PackageExtensionCompat v1.0.2
  [d96e819e] Parameters v0.12.3
  [69de0a69] Parsers v2.8.1
  [e409e4f3] PoissonRandom v0.4.4
  [f517fe37] Polyester v0.7.16
  [1d0040c9] PolyesterWeave v0.2.2
  [f27b6e38] Polynomials v4.0.11
  [85a6dd25] PositiveFactorizations v0.2.4
  [d236fae5] PreallocationTools v0.4.24
  [aea7be01] PrecompileTools v1.2.1
  [21216c6a] Preferences v1.4.3
  [27ebfcd6] Primes v0.5.6
  [43287f4e] PtrArrays v1.2.1
  [1fd47b50] QuadGK v2.11.1
  [6c2fb7c5] QuantumToolbox v0.20.0 `~/.julia/dev/QuantumToolbox`
  [74087812] Random123 v1.7.0
  [e6cf234a] RandomNumbers v1.6.0
  [c1ae055f] RealDot v0.1.0
  [3cdcf5f2] RecipesBase v1.3.4
  [731186ca] RecursiveArrayTools v3.27.2
  [f2c3362d] RecursiveFactorization v0.2.23
  [189a3867] Reexport v1.2.2
  [ae029012] Requires v1.3.0
  [ae5879a3] ResettableStacks v1.1.1
  [37e2e3b7] ReverseDiff v1.15.3
  [79098fc4] Rmath v0.8.0
  [47965b36] RootedTrees v2.23.1
  [7e49a35a] RuntimeGeneratedFunctions v0.5.13
  [94e857df] SIMDTypes v0.1.0
  [476501e8] SLEEFPirates v0.6.43
  [0bca4576] SciMLBase v2.58.0
  [19f34311] SciMLJacobianOperators v0.1.0
  [c0aeaf25] SciMLOperators v0.3.11
  [1ed8b502] SciMLSensitivity v7.69.0
  [53ae85a6] SciMLStructures v1.5.0
  [6c6a2e73] Scratch v1.2.1
  [efcf1570] Setfield v1.1.1
  [727e6d20] SimpleNonlinearSolve v1.12.3
  [699a6c99] SimpleTraits v0.9.4
  [ce78b400] SimpleUnPack v1.1.0
  [a2af1166] SortingAlgorithms v1.2.1
  [9f842d2f] SparseConnectivityTracer v0.6.8
  [47a9eef4] SparseDiffTools v2.23.0
  [dc90abb0] SparseInverseSubset v0.1.2
  [0a514795] SparseMatrixColorings v0.4.8
  [e56a9233] Sparspak v0.3.9
  [276daf66] SpecialFunctions v2.4.0
  [aedffcd0] Static v1.1.1
  [0d7ed370] StaticArrayInterface v1.8.0
  [90137ffa] StaticArrays v1.9.8
  [1e83bf80] StaticArraysCore v1.4.3
  [10745b16] Statistics v1.11.1
  [82ae8749] StatsAPI v1.7.0
  [2913bbd2] StatsBase v0.34.3
  [4c63d2b9] StatsFuns v1.3.2
  [789caeaf] StochasticDiffEq v6.70.0
  [7792a7ef] StrideArraysCore v0.5.7
  [09ab397b] StructArrays v0.6.18
  [53d494c1] StructIO v0.3.1
  [2efcf032] SymbolicIndexingInterface v0.3.34
  [19f23fe9] SymbolicLimits v0.2.2
  [d1185830] SymbolicUtils v3.7.2
  [0c5d862f] Symbolics v6.17.0
  [3783bdb8] TableTraits v1.0.1
  [bd369af6] Tables v1.12.0
  [8ea1fca8] TermInterface v2.0.0
  [8290d209] ThreadingUtilities v0.5.2
  [a759f4b9] TimerOutputs v0.5.25
  [9f7883ad] Tracker v0.2.35
  [d5829a12] TriangularSolve v0.2.1
  [781d530d] TruncatedStacktraces v1.4.0
  [3a884ed6] UnPack v1.0.2
  [a7c27f48] Unityper v0.1.6
  [013be700] UnsafeAtomics v0.2.1
  [d80eeb9a] UnsafeAtomicsLLVM v0.2.1
  [3d5dd08c] VectorizationBase v0.21.70
  [19fa3120] VertexSafeGraphs v0.2.0
  [e88e6eb3] Zygote v0.6.72
  [700de1a5] ZygoteRules v0.2.5
⌅ [7cc45869] Enzyme_jll v0.0.157+0
  [f5851436] FFTW_jll v3.3.10+1
  [1d5cc7b8] IntelOpenMP_jll v2024.2.1+0
  [dad2f222] LLVMExtra_jll v0.0.34+0
  [856f044c] MKL_jll v2024.2.0+0
  [efe28fd5] OpenSpecFun_jll v0.5.5+0
  [f50d1b31] Rmath_jll v0.5.1+0
  [1317d2d5] oneTBB_jll v2021.12.0+0
  [0dad84c5] ArgTools v1.1.2
  [56f22d72] Artifacts v1.11.0
  [2a0f44e3] Base64 v1.11.0
  [ade2ca70] Dates v1.11.0
  [8ba89e20] Distributed v1.11.0
  [f43a241f] Downloads v1.6.0
  [7b1f6079] FileWatching v1.11.0
  [9fa8497b] Future v1.11.0
  [b77e0a4c] InteractiveUtils v1.11.0
  [4af54fe1] LazyArtifacts v1.11.0
  [b27032c2] LibCURL v0.6.4
  [76f85450] LibGit2 v1.11.0
  [8f399da3] Libdl v1.11.0
  [37e2e46d] LinearAlgebra v1.11.0
  [56ddb016] Logging v1.11.0
  [d6f4376e] Markdown v1.11.0
  [a63ad114] Mmap v1.11.0
  [ca575930] NetworkOptions v1.2.0
  [44cfe95a] Pkg v1.11.0
  [de0858da] Printf v1.11.0
  [9abbd945] Profile v1.11.0
  [9a3f8284] Random v1.11.0
  [ea8e919c] SHA v0.7.0
  [9e88b42a] Serialization v1.11.0
  [1a1011a3] SharedArrays v1.11.0
  [6462fe0b] Sockets v1.11.0
  [2f01184e] SparseArrays v1.11.0
  [4607b0f0] SuiteSparse
  [fa267f1f] TOML v1.0.3
  [a4e569a6] Tar v1.10.0
  [8dfed614] Test v1.11.0
  [cf7118a7] UUIDs v1.11.0
  [4ec0a83e] Unicode v1.11.0
  [e66e0078] CompilerSupportLibraries_jll v1.1.1+0
  [deac9b47] LibCURL_jll v8.6.0+0
  [e37daf67] LibGit2_jll v1.7.2+0
  [29816b5a] LibSSH2_jll v1.11.0+1
  [c8ffd9c3] MbedTLS_jll v2.28.6+0
  [14a3606d] MozillaCACerts_jll v2023.12.12
  [4536629a] OpenBLAS_jll v0.3.27+1
  [05823500] OpenLibm_jll v0.8.1+2
  [bea87d4a] SuiteSparse_jll v7.7.0+0
  [83775a58] Zlib_jll v1.2.13+1
  [8e850b90] libblastrampoline_jll v5.11.0+0
  [8e850ede] nghttp2_jll v1.59.0+0
  [3f19e933] p7zip_jll v17.4.0+2
Julia Version 1.11.1
Commit 8f5b7ca12ad (2024-10-16 10:53 UTC)
Build Info:
  Official https://julialang.org/ release
Platform Info:
  OS: Linux (x86_64-linux-gnu)
  CPU: 32 × 13th Gen Intel(R) Core(TM) i9-13900KF
  WORD_SIZE: 64
  LLVM: libLLVM-16.0.6 (ORCJIT, alderlake)
Threads: 16 default, 0 interactive, 8 GC (on 32 virtual cores)
Environment:
  JULIA_EDITOR = code
  JULIA_NUM_THREADS = 16
albertomercurio commented 3 weeks ago

I think that the Zygote problem is related to the update of the ScalarOperator here. Where the struct contains a Float64, but during the differentiation it is trying to convert its field val to a Dual number, thus changing the type of the structure.

I found a possible partial fix for the out-of-place case. If I consider the out-of-place ODEProblem prob = ODEProblem{false}(U, u0, tspan, [γ]), and I update of the ScalarOperator is updated in the out-of-place case

function SciMLOperators.update_coefficients(L::ScalarOperator, u, p, t; kwargs...)
  return ScalarOperator(L.update_func(L.val, u, p, t; kwargs...), L.update_func)
end

instead of the current implementation

function update_coefficients!(L::ScalarOperator, u, p, t; kwargs...)
    L.val = L.update_func(L.val, u, p, t; kwargs...)
    nothing
end

function update_coefficients(L::ScalarOperator, u, p, t; kwargs...)
    update_coefficients!(L, u, p, t; kwargs...)
    L
end

Then it works

Zygote.gradient(my_f, 1.9) # (-0.17161488226273966,)

However, it still fails for the Enzyme case and for the in-place version of Zygote (which would be much more efficient I guess).

I will make a PR to SciMLOperators.jl for fix at least this case.

albertomercurio commented 1 week ago

I found that the in-place version works when using ComplexF64 types

using LinearAlgebra
using SparseArrays
using OrdinaryDiffEq
using SciMLOperators
using Zygote
using Enzyme
using SciMLSensitivity

##

T = ComplexF64
const N = 10
const u0 = ones(T, N)
# H_tmp = rand(T, N, N)
H_tmp = sprand(T, N, N, 0.5)
const H = H_tmp + H_tmp'
const U = ScalarOperator(one(params[1]), coef) * MatrixOperator(Diagonal(H)) + MatrixOperator(Diagonal(H))

coef(a, u, p, t) = - p[1]

function my_f(params)
    tspan = (0.0, 1.0)
    # prob = ODEProblem{true}(U, u0, tspan, [γ], sensealg = InterpolatingAdjoint(autojacvec=false))
    prob = ODEProblem{true}(U, u0, tspan, params)
    sol = solve(prob, Tsit5())
    return real(sol.u[end][end])
end

params = T[1]
my_f(params) # 0.25621142049273665

##

Zygote.gradient(my_f, params)

But I get the following warnings during the differentiation

┌ Warning: Potential performance improvement omitted. ReverseDiffVJP tried and failed in the automated AD choice algorithm. To show the stack trace, set SciMLSensitivity.STACKTRACE_WITH_VJPWARN[] = true. To turn off this printing, add `verbose = false` to the `solve` call.
└ @ SciMLSensitivity ~/.julia/packages/SciMLSensitivity/ME3jV/src/concrete_solve.jl:67

┌ Warning: Reverse-Mode AD VJP choices all failed. Falling back to numerical VJPs
└ @ SciMLSensitivity ~/.julia/packages/SciMLSensitivity/ME3jV/src/concrete_solve.jl:207
(ComplexF64[-1.9743163253371472 + 0.0im],)

First, how can I know the sensealg (and its options) that is automatically chosen?

Then, although I think the problem is related to ReverseDiff.jl, I followed this page and run

tspan = (0.0, 1.0)
# prob = ODEProblem{true}(U, u0, tspan, [γ], sensealg = InterpolatingAdjoint(autojacvec=false))
prob = ODEProblem{true}(U, u0, tspan, params)

u0 = prob.u0
p = prob.p
tmp2 = Enzyme.make_zero(p)
t = prob.tspan[1]
du = zero(u0)

if DiffEqBase.isinplace(prob)
    _f = prob.f
else
    _f = (du, u, p, t) -> (du .= prob.f(u, p, t); nothing)
end

_tmp6 = Enzyme.make_zero(_f)
tmp3 = zero(u0)
tmp4 = zero(u0)
ytmp = zero(u0)
tmp1 = zero(u0)

Enzyme.autodiff(Enzyme.Reverse, Enzyme.Duplicated(_f, _tmp6),
    Enzyme.Const, Enzyme.Duplicated(tmp3, tmp4),
    Enzyme.Duplicated(ytmp, tmp1),
    Enzyme.Duplicated(p, tmp2),
    Enzyme.Const(t))

which returns nothing for every variable

((nothing, nothing, nothing, nothing),)
albertomercurio commented 1 week ago

Now that I better understand how Enzyme.jl works, this is not a bad thing. It means that Enzyme can differentiate the function, right? So I don't understand why I see a warning on ReverseDiffVJP.

I guess that I can try using autojacvec=EnzymeVJP(), but I don't know what is the used sensealg.

Moreover, can I use Enzyme instead of Zygote to directly differentiate the my_f function? What are the advantages and the disadvantages?

albertomercurio commented 1 week ago

If I try sensealg = BacksolveAdjoint(autojacvec=EnzymeVJP()), I get the error

ERROR: MethodError: no method matching augmented_primal(::EnzymeCore.EnzymeRules.RevConfigWidth{…}, ::Const{…}, ::Type{…}, ::Duplicated{…}, ::Duplicated{…}, ::Duplicated{…}, ::Const{…}, ::Const{…})
The function `augmented_primal` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  augmented_primal(::EnzymeCore.EnzymeRules.RevConfig, ::Const{typeof(mul!)}, ::Type{RT}, ::Annotation{<:StridedVecOrMat}, ::Const{<:Union{SparseArrays.AbstractSparseMatrixCSC{Tv, Ti}, SubArray{Tv, 2, <:SparseArrays.AbstractSparseMatrixCSC{Tv, Ti}, Tuple{Base.Slice{Base.OneTo{Int64}}, I}} where I<:(AbstractUnitRange{<:Integer})} where {Tv, Ti}}, ::Annotation{<:StridedVecOrMat}, ::Annotation{<:Number}, ::Annotation{<:Number}) where RT
   @ Enzyme ~/.julia/packages/Enzyme/azJki/src/internal_rules.jl:732
  augmented_primal(::Any, ::Const{typeof(QuadGK.quadgk)}, ::Type{RT}, ::Any, ::Annotation{T}...; kws...) where {RT, T}
   @ QuadGKEnzymeExt ~/.julia/packages/QuadGK/BjmU0/ext/QuadGKEnzymeExt.jl:6
  augmented_primal(::Any, ::Const{typeof(NNlib._dropout!)}, ::Type{RT}, ::Any, ::OutType, ::Any, ::Any, ::Any) where {OutType, RT}
   @ NNlibEnzymeCoreExt ~/.julia/packages/NNlib/CkJqS/ext/NNlibEnzymeCoreExt/NNlibEnzymeCoreExt.jl:318
ChrisRackauckas commented 1 week ago

It looks like you're hitting a missing rule in enzyme for sparse mul! https://github.com/EnzymeAD/Enzyme.jl/issues/2013

wsmoses commented 6 days ago

what's the full types of the method match failure? And what version of enzyme are you using (it is the latest)?

albertomercurio commented 6 days ago 
LoadError: MethodError: no method matching augmented_primal(::EnzymeCore.EnzymeRules.RevConfigWidth{1, false, false, (false, false, false, false, false, false), false}, ::Const{typeof(mul!)}, ::Type{Const{Vector{ComplexF64}}}, ::Duplicated{Vector{ComplexF64}}, ::Duplicated{SparseMatrixCSC{ComplexF64, Int64}}, ::Duplicated{Vector{ComplexF64}}, ::Const{Bool}, ::Const{Bool})
The function `augmented_primal` exists, but no method is defined for this combination of argument types.

Closest candidates are:
  augmented_primal(::EnzymeCore.EnzymeRules.RevConfig, ::Const{typeof(mul!)}, ::Type{RT}, ::Annotation{<:StridedVecOrMat}, ::Const{<:Union{SparseArrays.AbstractSparseMatrixCSC{Tv, Ti}, SubArray{Tv, 2, <:SparseArrays.AbstractSparseMatrixCSC{Tv, Ti}, Tuple{Base.Slice{Base.OneTo{Int64}}, I}} where I<:(AbstractUnitRange{<:Integer})} where {Tv, Ti}}, ::Annotation{<:StridedVecOrMat}, ::Annotation{<:Number}, ::Annotation{<:Number}) where RT
   @ Enzyme ~/.julia/packages/Enzyme/azJki/src/internal_rules.jl:732
  augmented_primal(::Any, ::Const{typeof(QuadGK.quadgk)}, ::Type{RT}, ::Any, ::Annotation{T}...; kws...) where {RT, T}
   @ QuadGKEnzymeExt ~/.julia/packages/QuadGK/BjmU0/ext/QuadGKEnzymeExt.jl:6
  augmented_primal(::Any, ::Const{typeof(NNlib._dropout!)}, ::Type{RT}, ::Any, ::OutType, ::Any, ::Any, ::Any) where {OutType, RT}
   @ NNlibEnzymeCoreExt ~/.julia/packages/NNlib/CkJqS/ext/NNlibEnzymeCoreExt/NNlibEnzymeCoreExt.jl:318`

And this is my versioninfo

Status `~/GitHub/Research/Undef/Autodiff QuantumToolbox/Project.toml`
  [6e4b80f9] BenchmarkTools v1.5.0
  [13f3f980] CairoMakie v0.12.16
  [b0b7db55] ComponentArrays v0.15.19
  [7da242da] Enzyme v0.13.15
  [f6369f11] ForwardDiff v0.10.38
  [1dea7af3] OrdinaryDiffEq v6.90.1
  [33c8b6b6] ProgressLogging v0.1.4
  [6c2fb7c5] QuantumToolbox v0.21.5 `~/.julia/dev/QuantumToolbox`
  [731186ca] RecursiveArrayTools v3.27.3
  [37e2e3b7] ReverseDiff v1.15.3
⌃ [0bca4576] SciMLBase v2.61.0
⌃ [1ed8b502] SciMLSensitivity v7.71.1
  [5d786b92] TerminalLoggers v0.1.7
  [e88e6eb3] Zygote v0.6.73
Info Packages marked with ⌃ have new versions available and may be upgradable.