fix autodiff of remake

src/remake.jl

@@ -51,14 +51,19 @@ function remake(prob::ODEProblem; f = missing,
     if tspan === missing
         tspan = prob.tspan
     end
-    defs = Dict()
-    if hasproperty(prob.f, :sys)
-        if hasfield(typeof(prob.f.sys), :ps)
-            defs = mergedefaults(defs, prob.p, getfield(prob.f.sys, :ps))
-        end
-        if hasfield(typeof(prob.f.sys), :u0)
-            defs = mergedefaults(defs, prob.u0, getfield(prob.f.sys, :u0))
+
+    if (p !== missing && eltype(p) <: Pair) || (u0 !== missing && eltype(u0) <: Pair)
+        defs = Dict{Any, Any}()
+        if hasproperty(prob.f, :sys)
+            if hasfield(typeof(prob.f.sys), :ps)
+                defs = mergedefaults(defs, prob.p, getfield(prob.f.sys, :ps))
+            end
+            if hasfield(typeof(prob.f.sys), :u0)
+                defs = mergedefaults(defs, prob.u0, getfield(prob.f.sys, :u0))
+            end
         end
+    else
+        defs = nothing
     end
 
     if p === missing

src/solutions/optimization_solutions.jl

@@ -45,7 +45,8 @@ $(TYPEDEF)
 
 Representation the default cache for an optimization problem defined by an `OptimizationProblem`.
 """
-mutable struct DefaultOptimizationCache{F <: OptimizationFunction, P} <: AbstractOptimizationCache
+mutable struct DefaultOptimizationCache{F <: OptimizationFunction, P} <:
+               AbstractOptimizationCache
     f::F
     p::P
 end

test/downstream/Project.toml

@@ -2,5 +2,7 @@
 BoundaryValueDiffEq = "764a87c0-6b3e-53db-9096-fe964310641d"
 ModelingToolkit = "961ee093-0014-501f-94e3-6117800e7a78"
 OrdinaryDiffEq = "1dea7af3-3e70-54e6-95c3-0bf5283fa5ed"
+SciMLSensitivity = "1ed8b502-d754-442c-8d5d-10ac956f44a1"
 Sundials = "c3572dad-4567-51f8-b174-8c6c989267f4"
 Unitful = "1986cc42-f94f-5a68-af5c-568840ba703d"
+Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"

test/downstream/remake_autodiff.jl

@@ -0,0 +1,27 @@
+using OrdinaryDiffEq, ModelingToolkit, Zygote, SciMLSensitivity
+
+@variables t
+D = Differential(t)
+function lotka_volterra(; name = name)
+    states = @variables x(t)=1.0 y(t)=1.0
+    params = @parameters p1=1.5 p2=1.0 p3=3.0 p4=1.0
+    eqs = [
+        D(x) ~ p1 * x - p2 * x * y,
+        D(y) ~ -p3 * y + p4 * x * y,
+    ]
+    return ODESystem(eqs, t, states, params; name = name)
+end
+
+@named lotka_volterra_sys = lotka_volterra()
+prob = ODEProblem(lotka_volterra_sys, [], (0.0, 10.0), [])
+sol = solve(prob, Tsit5(), reltol = 1e-6, abstol = 1e-6)
+
+function sum_of_solution(u0, p)
+    _prob = remake(prob, u0 = u0, p = p)
+    sum(solve(_prob, Tsit5(), reltol = 1e-6, abstol = 1e-6, saveat = 0.1,
+              sensealg = BacksolveAdjoint(autojacvec = ZygoteVJP())))
+end
+
+u0 = [1.0 1.0]
+p = [1.5 1.0 1.0 1.0]
+du01, dp1 = Zygote.gradient(sum_of_solution, u0, p)