search

SciML / OrdinaryDiffEq.jl

High performance ordinary differential equation (ODE) and differential-algebraic equation (DAE) solvers, including neural ordinary differential equations (neural ODEs) and scientific machine learning (SciML)
https://diffeq.sciml.ai/latest/
Other
521 stars 198 forks source link

don't reset accept-step so SciMLBase.check_error can look at it #2210

Closed oscardssmith closed 1 month ago

oscardssmith commented 1 month ago

This line was unnecessarily setting accept_step to false which made check_error think that your step was always rejected. This was what caused the DelayDiffEq failures on https://github.com/SciML/SciMLBase.jl/pull/693. The problem, however isn't delay problem specific, it just turns out delay diffeqs are much more likely to have tiny timesteps.