Christopher Rackauckas

Oh yes, the signatures look like they are missing u,p,t, things.

Yes it looks like the ArrayPartition case is not handled here so this needs a fixing.

Yup it looks like this is only implemented for AbstractTimeseriesSolutions right now, so non-time series objects (steady states) are missing this plot. This needs to get fixed.

Yeah, I think this is an oversight. The ensemble stuff all assumes a timeseries and we need some extra dispatches.

Yes, this is a good idea and definitely something we should add. I think the first place to support it would be NeuralPDE.jl

@kanav99 has looked into this a bit. https://github.com/SciML/OrdinaryDiffEq.jl/pull/783 . I think it's a good time to finish this though. We should just clean up the major upstream libraries that we...

@YingboMa @devmotion @frankschae @isaacsas @kanav99 @Vaibhavdixit02 let's just do this. We should remove the `kwargs...` of every integrator. In many repos it will warn already if an argument is allowed...

If it's a kwarg, no one will ever do it. I can't tell you how many kwarg typos I've seen recently. I think people need to start using named tuples...

``` function foo(args...; diffeq_kwargs = (saveat=0.1, alg=Tsit5()), other_kwargs...) something_else(;other_kwargs...) prob = ODEProblem(...; diffeq_kwargs...) solve(prob) ... ```

Is that a standard usage? It is not the way backward error analysis is normally considered since it's the pullback of a perturbation of the solution, and that pullback is...