Enhance error messages to be more informative and user-friendly

[mariegrho]

A collection of common pymob error messages that would benefit from clearer explanations and suggested fixes.

[mariegrho]

ValueError: vmap in_axes must be an int, None, or a tuple of entries corresponding to the positional arguments passed to the function, but got len(in_axes)=4, len(args)=3

• Thrown by evaluator()
• Caused by

1. an empty Y_0 argument in the JaxSolver class: result = loop_eval(*Y_0, *ode_args, *pp_args, *x_in_flat)
2. mismatch of n_ode_states and len(Y_0)

• Solution:

1. define y0 value in sim.config.simulation: sim.config.simulation.y0 = ["expression=expression"]
2. define n_ode_states: sim.config.simulation.n_ode_states = 1

[mariegrho]

AssertionError: Pymob is trying to look up the values of ['expression'] in the reference data, but reference_data=None. Provide reference data if necessary and check if the right input key is used (currently: input='y0').

• Thrown by SimulationBase.parse_input: y0 = sim.parse_input("y0", drop_dims=["time"])
• Caused by missing assingment of the reference_data
• Solution: Assign the observations as reference data; y0 = sim.parse_input("y0", reference_data=sim.observations, drop_dims=[]).isel(time=[0])

[mariegrho]

KeyError: "Dimensions of the input array (1) and the specified dimensions on the data variable 'expression(dimensions=['time'])' did not match. Is a dimension missing from the data variable? You can update it by using sim.config.data_structure.expression.dimensions = [...]"`

• Thrown by: y0 = sim.parse_input("y0", reference_data=sim.observations, drop_dims=["time"])
• Caused by missing dimension of the y0-Dataset -> empty dataset
• Solution: y0 = sim.parse_input("y0", reference_data=sim.observations, drop_dims=[]).isel(time=[0])

[mariegrho]

TypeError: unsupported operand type(s) for *: 'DynamicJaxprTracer' and 'tuple'

• Thrown by evaluator()
• Caused by wrong model definition -> the model needs to return a tuple
• Solution: change the structure of the model function to return a tuple:

def gene_expression_baseline(t, y, t_mzt, alpha_zyg, k_deg):
    E, = y
    s = 5.0
    alpha = alpha_zyg / (1 + jnp.exp(-s * (t - t_mzt)))
    dydt = alpha - k_deg * E

    return dydt,

[flo-schu]

Amazing, thanks @mariegrho 🎉 keep 'em coming ;)

[flo-schu]

Another Error was discovered in mempyguts.
https://gitlab.uni-osnabrueck.de/memuos/mempyguts/-/issues/18