Romain Veltz

I am not sure ranges follow the requirements of the package (axpy,...): https://jutho.github.io/KrylovKit.jl/latest/#Package-features-and-alternatives

It's been a while I had forgotten about this. I just tested with CUDA and KK0.5.3, similar problem. It is a serious problem because you basically have the only iterative...

sure! that's why I did not bother you earlier! My research rests a lot on `eigsolve`

Any clue about this issue? It seems in [366](https://github.com/JuliaDiffEq/DiffEqBase.jl/issues/366), the options are not set properly either.

> Ideally we'd represent the full (possibly infinite) system of equations for the chemical master equation symbolically, but then specialize to a concrete system based on the initial condition or...

> Also note, `rx.rate` is not the full state-dependent transition rate for a multiparticle reaction like `A --> B`, which should be `rx.rate*A(t)`. This is given by the associated `jumpratelaw`....

Thank you!

My issue is to build a jump problem (here a PDMP) from several reaction_networks choosing which one is used as an ODE or as a Jump part.

Any chance to have this improved? The issue is that `generate_iterator` does not accept the `verbose` keyword. It has to be either shaved from `f.kwargs` , or passed as above,...

I have another issue pending to be submitted to `DiffEqBase` which could explain this. Somehow, I got the feeling that the tolerance `tol` is not well handled in `function (f::LinSolveIterativeSolvers)(x,A,b,update_matrix=false;...