Fix SDEProblem constructor issue #2281
Conversation
Signed-off-by: ErikQQY <[email protected]>
| @@ -583,7 +583,7 @@ function DiffEqBase.SDEProblem{iip}(sys::SDESystem, u0map = [], tspan = get_tspa | |||
| noise_rate_prototype = zeros(eltype(u0), size(noiseeqs)) | |||
| end | |||
|
|
|||
| SDEProblem{iip}(f, u0, tspan, p; callback = cbs, | |||
| SDEProblem{iip}(f, f.g, u0, tspan, p; callback = cbs, | |||
There was a problem hiding this comment.
Yeah, but from testing locally
f, u0, p = process_DEProblem(SDEFunction{iip}, sys, u0map, parammap; check_length, kwargs...)
SDEProbelm(f, u0, tspan, p; ...)would cause the same error in CI
but if we changed to
f, u0, p = process_DEProblem(SDEFunction{iip}, sys, u0map, parammap; check_length, kwargs...)
SDEProbelm(f, f.g u0, tspan, p; ...) # or SDEProbelm(f.f, f.g u0, tspan, p; ...)working fine
There was a problem hiding this comment.
But that's the constructor you deleted? Why not just use the SDEProblem(::SDEFunction, u0, tspan, p) dispatch?
There was a problem hiding this comment.
What is the f generated from process_DEProblem here? I think it should be working fine with the v2.0 SciMLBase, which directly build SDEProblem on SDEFunction by using SDEProblem(f, u0, tspan, p).
There was a problem hiding this comment.
It's generating an SDEFunction.
There was a problem hiding this comment.
It's strange here, I tested f here and indeed I got an SDEFunction, but the constructor SDEProblem(f::SDEFunction, u0, tspan, p; ...) still error here
There was a problem hiding this comment.
yes, that's what I was asking before merge that you were testing. Did you @which see that everything is going to the right dispatches? I thought you needed tspan::Tuple or something so that it could dispatch correctly.
There was a problem hiding this comment.
Don't sure if this guess is right, I am wondering since this is a DiffEqBase.SDEProblem dispatch for SDESystem type, what is the "version" of SDEProblem used here(MTK is still use 1.103.0 version DiffEqBase)?
ModelingToolkit.jl/src/systems/diffeqs/sdesystem.jl
Lines 586 to 587 in 1aa4806
There was a problem hiding this comment.
It's forcing SciMLBase v2. The StochasticDiffEq issues are the same problem without MTK involved. I don't think they are hitting the correct dispatches.
Should fix the problems in StochasticDiffEq