Minor rules/values update, and working example

Author: RXGottlieb

src/interval/interval.jl

@@ -17,7 +17,7 @@ function var_names(::IntervalTransform, s::Term{Real, Base.ImmutableDict{DataTyp
         sL = genvar(Symbol(string(get_name(s))*"_lo"), arg_list)
         sU = genvar(Symbol(string(get_name(s))*"_hi"), arg_list)
     end
-    return sL, sU
+    return Symbolics.value(sL), Symbolics.value(sU)
 end
 function var_names(::IntervalTransform, s::Real)
     return s, s
@@ -43,12 +43,12 @@ function var_names(::IntervalTransform, s::Term{Real, Nothing}) #Any terms like
 
     sL = s.f(var_lo)
     sU = s.f(var_hi)
-    return sL, sU
+    return Symbolics.value(sL), Symbolics.value(sU)
 end
 function var_names(::IntervalTransform, s::Sym) #The parameters
     sL = genparam(Symbol(string(get_name(s))*"_lo"))
     sU = genparam(Symbol(string(get_name(s))*"_hi"))
-    return sL, sU
+    return Symbolics.value(sL), Symbolics.value(sU)
 end
 
 function translate_initial_conditions(::IntervalTransform, prob::ODESystem, new_eqs::Vector{Equation})

src/relaxation/relaxation.jl

@@ -13,7 +13,7 @@ function var_names(::McCormickTransform, s::Term{Real, Base.ImmutableDict{DataTy
         scv = genvar(Symbol(string(get_name(s))*"_cv"), arg_list)
         scc = genvar(Symbol(string(get_name(s))*"_cc"), arg_list)
     end
-    return scv, scc
+    return Symbolics.value(scv), Symbolics.value(scc)
 end
 function var_names(::McCormickTransform, s::Real)
     return s, s
@@ -39,12 +39,12 @@ function var_names(::McCormickTransform, s::Term{Real, Nothing}) #Any terms like
 
     scv = s.f(var_cv)
     scc = s.f(var_cc)
-    return scv, scc
+    return Symbolics.value(scv), Symbolics.value(scc)
 end
 function var_names(::McCormickTransform, s::Sym) #The parameters
     scv = genparam(Symbol(string(get_name(s))*"_cv"))
     scc = genparam(Symbol(string(get_name(s))*"_cc"))
-    return scv, scc
+    return Symbolics.value(scv), Symbolics.value(scc)
 end
 
 function translate_initial_conditions(::McCormickTransform, prob::ODESystem, new_eqs::Vector{Equation})

src/relaxation/rules.jl

@@ -142,6 +142,17 @@ function transform_rule(::McCormickTransform, ::typeof(*), zL, zU, zcv, zcc, xL,
 
 end
 
+function transform_rule(::McCormickTransform, ::typeof(min), zL, zU, zcv, zcc, xL, xU, xcv, xcc, yL, yU, ycv, ycc)
+    rcv = Equation(zcv, min(xcv, ycv))
+    rcc = Equation(zcc, min(xcc, ycc))
+    return rcv, rcc
+end
+
+function transform_rule(::McCormickTransform, ::typeof(max), zL, zU, zcv, zcc, xL, xU, xcv, xcc, yL, yU, ycv, ycc)
+    rcv = Equation(zcv, min(xcv, ycv))
+    rcc = Equation(zcc, min(xcc, ycc))
+    return rcv, rcc
+end
 
 #=
 TODO: Add other operators. It's probably helpful to break the McCormick overload and McCormick + Interval Outputs

works.jl

@@ -0,0 +1,34 @@
+
+# This case WORKED at one point, so it should work whenever you're testing something.
+using ModelingToolkit, OrdinaryDiffEq
+using DifferentialEquations: solve
+using DiffEqGPU
+using BenchmarkTools
+@parameters p[1:2] t
+@variables x[1:2](t)
+D = Differential(t)
+x0 = [1.0; 0.0]
+tspan = (0.0, 35.0)
+p_start = [0.020; 0.025]
+eqns = [D(x[1]) ~ -p[1]*x[1] + p[2]*x[2],
+        D(x[2]) ~ p[1]*x[1] - p[2]*x[2]]
+
+@named test_works = ODESystem(eqns, t, x, p, default_u0=Dict( x[i] .=> x0[i] for i in 1:2), default_p = Dict( p[i] .=> p_start[i] for i in 1:2))
+tested_MC_works = apply_transform(McCormickIntervalTransform(), test_works) #This one breaks
+set_bounds(tested_MC_works, p[1], (0.015, 0.025))
+ode_problem_works = eval(ODEProblemExpr(structural_simplify(tested_MC_works), tested_MC_works.defaults, tspan))
+
+# Can create u's and p's to interpolate into the prob_func definition like this:
+p_order = tested_MC_works.ps.value
+u_order = structural_simplify(tested_MC_works).states.value
+u_list = []
+p_list = []
+for i = 1:11
+    push!(u_list, [1.0, 1.0, 1.0, 1.0, 0.0, 0.0, 0.0, 0.0+(i-1)*0.1])
+    push!(p_list, [0.01, 0.01, 0.01, 0.01, 0.01+(i-1)*0.001, 0.01-(i-1)*0.001, 0.01, 0.01])
+end
+prob_func_works = (prob, i, repeat) -> remake(prob, u0=:($u_list)[i], p = :($p_list)[i])
+
+# And this ensemble problem works with EnsembleGPUArray, currently.
+ensemble_works = EnsembleProblem(ode_problem_works, prob_func=prob_func_works)
+@benchmark solve(ensemble_works, Tsit5(), EnsembleGPUArray(), trajectories=11)