Christopher Rackauckas

icon indicating a website link www.chrisrackauckas.com icon indicating an email [email protected]

icon company Massachusetts Institute of Technology icon location Cambridge, MA icon location Applied Mathematics Instructor at MIT, researching numerical differential equations and their applications to scientific machine learning (SciML)

Results 4655 comments of Christopher Rackauckas

Bayesian inference updates in manifest.toml and project.toml

What do you get locally?

Parameter estimation

I'm pretty sure this one needs updates to the callback syntax? Are you running these and making sure they are working? You keep telling me to review PRs that are...

Parameter estimation

I don't know, that would take some time to dive in, but I assume the majority of fixing this benchmark is likely figuring that one out.

Parameter estimation

This is mixing dynamical and parameter estimation. Closing because this doesn't have any real changes in it. @ParamThakkar123 I'm going to recommend this project as a SciML Small Grant for...

(potential regression?) Case of parameter fitting suddenly throw error (`MethodError: no method matching +(::@NamedTuple{…}, ::Base.RefValue{…})`)

This is known. @DhairyaLGandhi is on it.

(potential regression?) Case of parameter fitting suddenly throw error (`MethodError: no method matching +(::@NamedTuple{…}, ::Base.RefValue{…})`)

https://github.com/FluxML/Zygote.jl/pull/1574 it's a Zygote issue. Will get a patch in.

Implicit Function Adjoint Rules for Enzyme

Note that NonlinearSolve is currently missing Enzyme overloads. Only LinearSolve and ODEProblems have gotten that treatment so far.

Implicit Function Adjoint Rules for Enzyme

It's just like the ODE infrastructure in that it just lowers to call SciMLSensitivity for all of the real work. The current dispatch is in DiffEqBase.jl right next to the...

Implicit Function Adjoint Rules for Enzyme

Yup and then the adjoint just passes it on: https://github.com/SciML/DiffEqBase.jl/blob/08fbaf0561d797359027849bac7797ce10948846/ext/DiffEqBaseChainRulesCoreExt.jl#L22C28-L29 . Looking at the ODE one, the NonlinearProblem one might just work if NonlinearProblem is made a mutable struct.

Implicit Function Adjoint Rules for Enzyme

No, which makes this a major PITA. With DiffEqGPU we had to create a separate immutable problem that gets used internally to keep CUDA kernel compatibility. It converts under the...