fix: remake initialization problem during DAE initialization #2209
Conversation
AayushSabharwal
force-pushed
the
as/initprob-params
branch
from
4d8276a to
18a22ef
Compare
AayushSabharwal
force-pushed
the
as/initprob-params
branch
from
18a22ef to
e3b7d7a
Compare
AayushSabharwal
force-pushed
the
as/initprob-params
branch
from
f0720b7 to
e3b7d7a
Compare
| if initializeprob.f.sys !== nothing && prob.f.sys !== nothing | ||
| initu0vars = variable_symbols(initializeprob) | ||
| initu0order = variable_index.((initializeprob,), initu0vars) | ||
| # Variable symbols are not guaranteed to be in order | ||
| invpermute!(initu0vars, initu0order) | ||
| initu0 = getu(prob.f.initializeprob, initu0vars)(prob) | ||
| initp = remake_buffer(initializeprob, parameter_values(initializeprob), | ||
| Dict(sym => getu(prob, sym)(prob) for sym in parameter_symbols(initializeprob))) | ||
| initializeprob = remake(initializeprob; u0 = initu0, p = initp) |
There was a problem hiding this comment.
I'm not sure this all makes sense. First of all, it shouldn't come from prob since you can have callback reinitialization, so it should come from integrator.u. But also, this is just very heavy and dynamic: can't we just add a initializeprobsetu in the ODEFunction which sets up the initalizeprob based on integrator.u, and then also have one for p that updates based on any p updates (and also this should update t, which is a psuedo-parameter of the initialization)
There was a problem hiding this comment.
That makes sense. Since this involves SciMLBase changes (changing ODEFunction and DAEFunction fields) I'll merge this PR with #2228 since that depends on this PR and modifies those structs anyway.
|
Superseded by #2228 |
Close SciML/ModelingToolkit.jl#2717
Requires
remake_bufferModelingToolkit.jl#2729Checklist
contributor guidelines, in particular the SciML Style Guide and
COLPRAC.
Additional context
Add any other context about the problem here.