Hossein Pourbozorg

I think the problem comes from ```julia sensealg = InterpolatingAdjoint(autojacvec=false) ``` because ```julia sensealg = InterpolatingAdjoint() ``` works fine.

Issues related to `sciml_train` should be closed or moved after this PR merged. https://github.com/search?q=sciml_train+repo%3ASciML%2FDiffEqFlux.jl+state%3Aopen+in%3Atitle&type=Issues

I tried other AD: `Zygote`, `ForwardDiff`, `Tracker`, `FiniteDifferences`, `FiniteDiff` all of them work without error.

There are differences between Julia v1.9 and v1.10 ambiguity report. IMO, there should be an Aqua test in CI, so we can have an updated list of ambiguities for both...

Why not add a deprecation warning first, then convert its usage?

I updated the code, and removed `ComponentArrays`. Maybe the problem is this line: https://github.com/FluxML/Zygote.jl/blob/e2d7839a476fc1cda407500d9a2ca125bd6e2aa5/src/lib/base.jl#L70 I don't know what `Base.skip_deleted` do, but it can produce zero like: ```julia julia> d =...

Okay, I will try to reduce it to `Zygote` and `stdlib`.

Not using the default `solvealg` of `SciMLSensitivity` fix the problem: ```julia icnf = construct(RNODE, nn, 1; compute_mode=ZygoteMatrixMode, sol_kwargs=Dict(:sensealg => InterpolatingAdjoint())) ``` I tested `InterpolatingAdjoint()`, `QuadratureAdjoint()`, `BacksolveAdjoint()`, and `ForwardDiffSensitivity()`; all of...

Oh, I think I found the problem, it's having an empty `Dict`. code: ```julia using Zygote, DifferentialEquations, SciMLSensitivity function f_aug(u, p, t) p .* u end function f_sol(u, p, sol_kwargs=Dict())...

MWE: ```julia using Zygote function f_sol(u, sol_kwargs=Dict()) sum(u; sol_kwargs...) end u2 = rand(Float32, 2) diff_loss(x) = f_sol(x) diff_loss(u2) Zygote.jacobian(diff_loss, u2) ``` error: ```julia ERROR: LoadError: BoundsError: attempt to access 16-element...