Merge pull request #1096 from prbzrg/clear-12

Remove Deps:  `StochasticDiffEq` and `OrdinaryDiffEq`

Project.toml

@@ -24,7 +24,6 @@ GPUArraysCore = "46192b85-c4d5-4398-a991-12ede77f4527"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
 LinearSolve = "7ed4a6bd-45f5-4d41-b270-4a48e9bafcae"
 Markdown = "d6f4376e-aef5-505a-96c1-9c027394607a"
-OrdinaryDiffEq = "1dea7af3-3e70-54e6-95c3-0bf5283fa5ed"
 Parameters = "d96e819e-fc66-5662-9728-84c9c7592b0a"
 PreallocationTools = "d236fae5-4411-538c-8e31-a6e3d9e00b46"
 QuadGK = "1fd47b50-473d-5c70-9696-f719f8f3bcdc"
@@ -40,7 +39,6 @@ SparseDiffTools = "47a9eef4-7e08-11e9-0b38-333d64bd3804"
 StaticArrays = "90137ffa-7385-5640-81b9-e52037218182"
 StaticArraysCore = "1e83bf80-4336-4d27-bf5d-d5a4f845583c"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
-StochasticDiffEq = "789caeaf-c7a9-5a7d-9973-96adeb23e2a0"
 SymbolicIndexingInterface = "2efcf032-c050-4f8e-a9bb-153293bab1f5"
 Tracker = "9f7883ad-71c0-57eb-9f7f-b5c9e6d3789c"
 TruncatedStacktraces = "781d530d-4396-4725-bb49-402e4bee1e77"
@@ -99,7 +97,7 @@ StaticArrays = "1.8.0"
 StaticArraysCore = "1.4"
 Statistics = "1.10"
 SteadyStateDiffEq = "2.0.1"
-StochasticDiffEq = "6.63.2"
+StochasticDiffEq = "6.69"
 SymbolicIndexingInterface = "0.3.26"
 Test = "1.10"
 Tracker = "0.2.30"
@@ -120,6 +118,7 @@ NLsolve = "2774e3e8-f4cf-5e23-947b-6d7e65073b56"
 NonlinearSolve = "8913a72c-1f9b-4ce2-8d82-65094dcecaec"
 Optimization = "7f7a1694-90dd-40f0-9382-eb1efda571ba"
 OptimizationOptimisers = "42dfb2eb-d2b4-4451-abcd-913932933ac1"
+OrdinaryDiffEq = "1dea7af3-3e70-54e6-95c3-0bf5283fa5ed"
 Pkg = "44cfe95a-1eb2-52ea-b672-e2afdf69b78f"
 SafeTestsets = "1bc83da4-3b8d-516f-aca4-4fe02f6d838f"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
@@ -128,4 +127,4 @@ StochasticDiffEq = "789caeaf-c7a9-5a7d-9973-96adeb23e2a0"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
 
 [targets]
-test = ["AlgebraicMultigrid", "Aqua", "Calculus", "ComponentArrays", "DelayDiffEq", "Distributed", "Lux", "ModelingToolkit", "NLsolve", "NonlinearSolve", "Optimization", "OptimizationOptimisers", "Pkg", "SafeTestsets", "SparseArrays", "SteadyStateDiffEq", "StochasticDiffEq", "Test"]
+test = ["AlgebraicMultigrid", "Aqua", "Calculus", "ComponentArrays", "DelayDiffEq", "Distributed", "Lux", "ModelingToolkit", "NLsolve", "NonlinearSolve", "Optimization", "OptimizationOptimisers", "OrdinaryDiffEq", "Pkg", "SafeTestsets", "SparseArrays", "SteadyStateDiffEq", "StochasticDiffEq", "Test"]

src/SciMLSensitivity.jl

@@ -7,7 +7,6 @@ using Adapt
 using LinearSolve
 using Parameters: @unpack
 import Accessors: @reset
-using StochasticDiffEq
 import DiffEqNoiseProcess
 import RandomNumbers: Xorshifts
 using Random

src/adjoint_common.jl

@@ -199,7 +199,7 @@ function adjointdiffcache(g::G, sensealg, discrete, sol, dgdu::DG1, dgdp::DG2, f
                 paramjac_config = tape
             end
         elseif noiseterm &&
-               (!StochasticDiffEq.is_diagonal_noise(prob) || isnoisemixing(sensealg))
+               (!SciMLBase.is_diagonal_noise(prob) || isnoisemixing(sensealg))
             tape = nothing
             paramjac_config = tape
         else
@@ -256,7 +256,7 @@ function adjointdiffcache(g::G, sensealg, discrete, sol, dgdu::DG1, dgdp::DG2, f
             if isinplace
                 for i in 1:m
                     function noisetape(indx)
-                        if StochasticDiffEq.is_diagonal_noise(prob)
+                        if SciMLBase.is_diagonal_noise(prob)
                             ReverseDiff.GradientTape((y, _p, [_t])) do u, p, t
                                 du1 = p !== nothing && p !== DiffEqBase.NullParameters() ?
                                       similar(p, size(u)) : similar(u)
@@ -283,7 +283,7 @@ function adjointdiffcache(g::G, sensealg, discrete, sol, dgdu::DG1, dgdp::DG2, f
             else
                 for i in 1:m
                     function noisetapeoop(indx)
-                        if StochasticDiffEq.is_diagonal_noise(prob)
+                        if SciMLBase.is_diagonal_noise(prob)
                             ReverseDiff.GradientTape((y, _p, [_t])) do u, p, t
                                 unwrappedf(u, p, first(t))[indx]
                             end
@@ -303,7 +303,7 @@ function adjointdiffcache(g::G, sensealg, discrete, sol, dgdu::DG1, dgdp::DG2, f
             end
         elseif autojacvec isa Bool
             if isinplace
-                if StochasticDiffEq.is_diagonal_noise(prob)
+                if SciMLBase.is_diagonal_noise(prob)
                     pf = DiffEqBase.ParamJacobianWrapper(unwrappedf, _t, y)
                     if isnoisemixing(sensealg)
                         uf = DiffEqBase.UJacobianWrapper(unwrappedf, _t, p)
@@ -321,7 +321,7 @@ function adjointdiffcache(g::G, sensealg, discrete, sol, dgdu::DG1, dgdp::DG2, f
                 paramjac_noise_config = build_param_jac_config(
                     sensealg, pf, y, SciMLStructures.replace(Tunable(), p, tunables))
             else
-                if StochasticDiffEq.is_diagonal_noise(prob)
+                if SciMLBase.is_diagonal_noise(prob)
                     pf = ParamGradientWrapper(unwrappedf, _t, y)
                     if isnoisemixing(sensealg)
                         uf = DiffEqBase.UDerivativeWrapper(unwrappedf, _t, p)
@@ -333,7 +333,7 @@ function adjointdiffcache(g::G, sensealg, discrete, sol, dgdu::DG1, dgdp::DG2, f
                 paramjac_noise_config = nothing
                 jac_noise_config = nothing
             end
-            if StochasticDiffEq.is_diagonal_noise(prob)
+            if SciMLBase.is_diagonal_noise(prob)
                 pJ = similar(u0, numindvar, numparams)
                 if isnoisemixing(sensealg)
                     J = similar(u0, numindvar, numindvar)

src/backsolve_adjoint.jl

@@ -36,7 +36,7 @@ function (S::ODEBacksolveSensitivityFunction)(du, u, p, t)
     if S.noiseterm
         if length(u) == length(du)
             vecjacobian!(dλ, y, λ, p, t, S, dgrad = dgrad, dy = dy)
-        elseif length(u) != length(du) && StochasticDiffEq.is_diagonal_noise(prob) &&
+        elseif length(u) != length(du) && SciMLBase.is_diagonal_noise(prob) &&
                !isnoisemixing(S.sensealg)
             vecjacobian!(dλ, y, λ, p, t, S, dy = dy)
             jacNoise!(λ, y, p, t, S, dgrad = dgrad)
@@ -82,7 +82,7 @@ function split_states(du, u, t, S::ODEBacksolveSensitivityFunction; update = tru
         dgrad = @view du[(idx + 1):(end - idx)]
         dy = @view du[(end - idx + 1):end]
 
-    elseif length(u) != length(du) && StochasticDiffEq.is_diagonal_noise(prob) &&
+    elseif length(u) != length(du) && SciMLBase.is_diagonal_noise(prob) &&
            !isnoisemixing(S.sensealg)
         # Diffusion term, diagonal noise, length(du) =  u*m
         idx1 = [length(u) * (i - 1) + i for i in 1:idx] # for diagonal indices of [1:idx,1:idx]
@@ -92,7 +92,7 @@ function split_states(du, u, t, S::ODEBacksolveSensitivityFunction; update = tru
         dgrad = @view du[(idx + 1):(end - idx), 1:idx]
         dy = @view du[idx2]
 
-    elseif length(u) != length(du) && StochasticDiffEq.is_diagonal_noise(prob) &&
+    elseif length(u) != length(du) && SciMLBase.is_diagonal_noise(prob) &&
            isnoisemixing(S.sensealg)
         # Diffusion term, diagonal noise, (as above but can handle mixing noise terms)
         idx2 = [(length(u) + 1) * i - idx for i in 1:idx] # for diagonal indices of [end-idx+1:end,1:idx]
@@ -288,7 +288,8 @@ end
     len = length(u0) + numparams
     λ = one(eltype(u0)) .* similar(tunables, len)
 
-    if StochasticDiffEq.alg_interpretation(sol.alg) == :Stratonovich
+    if SciMLBase.alg_interpretation(sol.alg) ==
+       SciMLBase.AlgorithmInterpretation.Stratonovich
         sense_drift = ODEBacksolveSensitivityFunction(g, sensealg, discrete, sol,
             dgdu_continuous, dgdp_continuous,
             sol.prob.f, alg)
@@ -344,7 +345,7 @@ end
     _sol = deepcopy(sol)
     backwardnoise = reverse(_sol.W)
 
-    if StochasticDiffEq.is_diagonal_noise(sol.prob) && sol.W.u[end] isa Number
+    if SciMLBase.is_diagonal_noise(sol.prob) && sol.W.u[end] isa Number
         # scalar noise case
         noise_matrix = nothing
     else

src/derivative_wrappers.jl

@@ -779,7 +779,7 @@ function _jacNoise!(λ, y, p, t, S::TS, isnoise::Bool, dgrad, dλ,
             end
         end
 
-        if StochasticDiffEq.is_diagonal_noise(prob)
+        if SciMLBase.is_diagonal_noise(prob)
             pJt = transpose(λ) .* transpose(pJ)
             recursive_copyto!(dgrad, pJt)
         else
@@ -796,7 +796,7 @@ function _jacNoise!(λ, y, p, t, S::TS, isnoise::Bool, dgrad, dλ,
     end
 
     if dλ !== nothing &&
-       (isnoisemixing(sensealg) || !StochasticDiffEq.is_diagonal_noise(prob))
+       (isnoisemixing(sensealg) || !SciMLBase.is_diagonal_noise(prob))
         @unpack J, uf, f_cache, jac_noise_config = S.diffcache
         if dy !== nothing
             if inplace_sensitivity(S)
@@ -813,7 +813,7 @@ function _jacNoise!(λ, y, p, t, S::TS, isnoise::Bool, dgrad, dλ,
                 if dy !== nothing
                     ForwardDiff.jacobian!(J, uf, dy, y)
                 else
-                    if StochasticDiffEq.is_diagonal_noise(prob)
+                    if SciMLBase.is_diagonal_noise(prob)
                         dy = similar(y)
                     else
                         dy = similar(prob.noise_rate_prototype)
@@ -832,7 +832,7 @@ function _jacNoise!(λ, y, p, t, S::TS, isnoise::Bool, dgrad, dλ,
             #  jacobian!(J, uf, y, nothing, sensealg, nothing)
         end
 
-        if StochasticDiffEq.is_diagonal_noise(prob)
+        if SciMLBase.is_diagonal_noise(prob)
             Jt = transpose(λ) .* transpose(J)
             recursive_copyto!(dλ, Jt)
         else
@@ -866,7 +866,7 @@ function _jacNoise!(λ, y, p, t, S::TS, isnoise::ReverseDiffVJP, dgrad, dλ,
         ReverseDiff.value!(tp, p)
         ReverseDiff.value!(tt, [t])
         ReverseDiff.forward_pass!(tapei)
-        if StochasticDiffEq.is_diagonal_noise(prob)
+        if SciMLBase.is_diagonal_noise(prob)
             ReverseDiff.increment_deriv!(output, λ[i])
         else
             ReverseDiff.increment_deriv!(output, λ)
@@ -881,7 +881,7 @@ function _jacNoise!(λ, y, p, t, S::TS, isnoise::ReverseDiffVJP, dgrad, dλ,
         end
         ReverseDiff.pull_value!(output)
 
-        if StochasticDiffEq.is_diagonal_noise(prob)
+        if SciMLBase.is_diagonal_noise(prob)
             dλ !== nothing && (dλ[:, i] .= vec(ReverseDiff.deriv(tu)))
             dy !== nothing && (dy[i] = ReverseDiff.value(output))
         else
@@ -909,7 +909,7 @@ function _jacNoise!(λ, y, p, t, S::TS, isnoise::ZygoteVJP, dgrad, dλ,
     noise_rate_prototype = prob.noise_rate_prototype
     m = noise_rate_prototype === nothing ? length(y) : size(noise_rate_prototype)[2]
 
-    if StochasticDiffEq.is_diagonal_noise(prob)
+    if SciMLBase.is_diagonal_noise(prob)
         if VERSION < v"1.9" # pre "stack" function
             for i in 1:m
                 if inplace_sensitivity(S)

src/interpolating_adjoint.jl

@@ -127,7 +127,7 @@ function (S::ODEInterpolatingAdjointSensitivityFunction)(du, u, p, t)
     if S.noiseterm
         if length(u) == length(du)
             vecjacobian!(dλ, y, λ, p, t, S, dgrad = dgrad)
-        elseif length(u) != length(du) && StochasticDiffEq.is_diagonal_noise(prob) &&
+        elseif length(u) != length(du) && SciMLBase.is_diagonal_noise(prob) &&
                !isnoisemixing(S.sensealg)
             vecjacobian!(dλ, y, λ, p, t, S)
             jacNoise!(λ, y, p, t, S, dgrad = dgrad)
@@ -245,7 +245,7 @@ function split_states(du, u, t, S::TS;
         dλ = @view du[1:idx]
         dgrad = @view du[(idx + 1):end]
 
-    elseif length(u) != length(du) && StochasticDiffEq.is_diagonal_noise(prob) &&
+    elseif length(u) != length(du) && SciMLBase.is_diagonal_noise(prob) &&
            !isnoisemixing(S.sensealg)
         idx1 = [length(u) * (i - 1) + i for i in 1:idx] # for diagonal indices of [1:idx,1:idx]
 
@@ -519,7 +519,7 @@ end
     _sol = deepcopy(sol)
     backwardnoise = reverse(_sol.W)
 
-    if StochasticDiffEq.is_diagonal_noise(sol.prob) && sol.W.u[end] isa Number
+    if SciMLBase.is_diagonal_noise(sol.prob) && sol.W.u[end] isa Number
         # scalar noise case
         noise_matrix = nothing
     else

src/sde_tools.jl

@@ -12,7 +12,7 @@ end
 function StochasticTransformedFunction(sol, f, g, corfunc_analytical = nothing)
     @unpack prob = sol
 
-    if StochasticDiffEq.is_diagonal_noise(prob)
+    if SciMLBase.is_diagonal_noise(prob)
         gtmp = copy(sol.u[end])
     else
         gtmp = similar(prob.p, size(prob.noise_rate_prototype))

test/sde_scalar_ito.jl

@@ -67,7 +67,8 @@ probStrat = SDEProblem(SDEFunction(fStrat, σ),
 )
 
 # solve Strat sense
-solStrat = solve(probStrat, RKMil(interpretation = :Stratonovich), dt = dt,
+solStrat = solve(probStrat,
+    RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich), dt = dt,
     adaptive = false, save_noise = true, saveat = dt)
 
 # check that forward solution agrees

test/sde_stratonovich.jl

@@ -83,7 +83,9 @@ p2 = [1.01, 0.87]
         Random.seed!(seed)
         tmp_prob = remake(prob_oop_sde, u0 = eltype(p).(prob_oop_sde.u0), p = p,
             tspan = eltype(p).(prob_oop_sde.tspan))
-        sol = solve(tmp_prob, RKMil(interpretation = :Stratonovich), dt = tend / 10000,
+        sol = solve(tmp_prob,
+            RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich),
+            dt = tend / 10000,
             adaptive = false, sensealg = DiffEqBase.SensitivityADPassThrough(),
             saveat = tarray)
         A = convert(Array, sol)
@@ -109,7 +111,8 @@ end
 
     Random.seed!(seed)
     prob_oop_sde2 = SDEProblem(f_oop_linear, σ_oop_linear, u₀, trange, p2)
-    sol_oop_sde2 = solve(prob_oop_sde2, RKMil(interpretation = :Stratonovich), dt = dt1,
+    sol_oop_sde2 = solve(prob_oop_sde2,
+        RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich), dt = dt1,
         adaptive = false, save_noise = true)
 
     res_sde_u02, res_sde_p2 = adjoint_sensitivities(sol_oop_sde2, EulerHeun(), t = tarray,
@@ -158,7 +161,9 @@ end
         tmp_prob = remake(prob_oop_sde2, u0 = eltype(p).(prob_oop_sde2.u0), p = p,
             tspan = eltype(p).(prob_oop_sde2.tspan)            #,abstol=abstol, reltol=reltol
         )
-        sol = solve(tmp_prob, RKMil(interpretation = :Stratonovich), dt = dt1,
+        sol = solve(tmp_prob,
+            RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich),
+            dt = dt1,
             adaptive = false, sensealg = DiffEqBase.SensitivityADPassThrough(),
             saveat = tarray)
         A = convert(Array, sol)
@@ -326,7 +331,9 @@ end
     function GSDE3(u)
         Random.seed!(seed)
         tmp_prob = remake(prob_oop_sde2, u0 = u)
-        sol = solve(tmp_prob, RKMil(interpretation = :Stratonovich), dt = dt1,
+        sol = solve(tmp_prob,
+            RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich),
+            dt = dt1,
             adaptive = false, saveat = tarray)
         A = convert(Array, sol)
         res = g(A, nothing, nothing)
@@ -545,7 +552,9 @@ end
         Random.seed!(seed)
         tmp_prob = remake(prob_oop_sde, u0 = eltype(p).(prob_oop_sde.u0), p = p,
             tspan = eltype(p).(prob_oop_sde.tspan))
-        sol = solve(tmp_prob, RKMil(interpretation = :Stratonovich), dt = 5e-4,
+        sol = solve(tmp_prob,
+            RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich),
+            dt = 5e-4,
             adaptive = false, sensealg = DiffEqBase.SensitivityADPassThrough(),
             saveat = tarray)
         A = convert(Array, sol)
@@ -556,7 +565,7 @@ end
     Random.seed!(seed)
     res_sde_trackeru0, res_sde_trackerp = Zygote.gradient(
         (u0, p) -> sum(Array(solve(prob_oop_sde,
-            RKMil(interpretation = :Stratonovich),
+            RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich),
             dt = 5e-4,
             adaptive = false,
             u0 = u0,

test/sde_transformation_test.jl

@@ -45,7 +45,7 @@ prob_strat = SDEProblem{false}(
     tspan,
     p)
 Random.seed!(seed)
-sol_strat = solve(prob_strat, RKMil(interpretation = :Stratonovich), adaptive = false,
+sol_strat = solve(prob_strat, RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich), adaptive = false,
     dt = 0.0001, save_noise = true)
 prob_strat1 = SDEProblem{false}(
     SDEFunction((u, p, t) -> transformed_function(u, p, t) .+
@@ -56,7 +56,7 @@ prob_strat1 = SDEProblem{false}(
     tspan,
     p)
 Random.seed!(seed)
-sol_strat1 = solve(prob_strat1, RKMil(interpretation = :Stratonovich), adaptive = false,
+sol_strat1 = solve(prob_strat1, RKMil(interpretation = SciMLBase.AlgorithmInterpretation.Stratonovich), adaptive = false,
     dt = 0.0001, save_noise = true)
 
 # Test if we recover Ito solution in Stratonovich sense