Martin Robinson

agree that we still need the documentation in the ask/tell methods to make this clear

theoretically you can apply any current input protocol to any of the models. You might run into numerical issues however. If you have run into a specific issue with one...

Thanks @Siel . what is your usecase? What equations are you solving, what does your noise term look like? What do you currently use for this (packages and methods)? Any...

that reminds me, we should get rid of python-idaklu, I don't think it serves a useful purpose anymore. I'll add an issue

could have the python and jax sensitivity functions also return the same as return_jacp_stacked=True so that everything is aligned

In the above code, if you replace `t_eval` with `t_data` in the call to `solver.jaxify` the results look better: ``` jax_all_grad [[0.0085677 0.08058828] [0.01686409 0.08268554] [0.02149744 0.0836817 ] [0.02467562 0.08446308]...

> I’d like to work on the Jaxified idaklu solver issue and help resolve it. Hi @RohitP2005, this issue would require quite detailed knowledge of both JAX and the pybamm...

Main drawback of this approach is that it would be very inefficient if you wanted to use dense matrices for the jacobian, not sure if that would be an issue...

another downside would be if a user wanted to rapidly change the number of inputs, in which case you would have to setup new equations every time which would be...

any comments welcome on this approach, the above is my current thinking but happy to consider alternatives before I dive into making changes: @BradyPlanden @rtimms @jsbrittain @valentinsulzer.