Include functionality for modifying parameters

[FinbarArgus]

For replacing the Python OpenCOR API with LibOpenCOR, we need to be able to run the following commands:

self.simulation = oc.open_simulation(cellml_path)
self.simulation.valid():
self.simulation.data()
self.data.odeSolverProperties():
self.data.set_ode_solver_property(key, value)
self.data.set_point_interval(self.dt)  # time interval for data storage
self.data.set_starting_point(0)
self.data.set_ending_point(self.stop_time)

self.simulation.run()
self.simulation.reset()
self.simulation.release_all_values()
self.simulation.clear_results()
self.simulation.reset(False) # True resets everything, False resets only the states

self.simulation.results().states():
self.simulation.results().states()[variable_name].values()[-self.n_steps - 1:].copy())
self.simulation.results().algebraic():
self.simulation.results().algebraic()[variable_name].values()[-self.n_steps-1:].copy())
self.data.constants():
self.data.constants()[variable_name])

self.data.states()[param_name] = param_vals[JJ]
self.data.constants()[param_name] = param_vals[JJ]

Or equivalent commands.

For context see https://github.com/FinbarArgus/circulatory_autogen/blob/devel/src/utilities/opencor_helper.py

[agarny]

FTR, for ISAN 2024, we quickly added a way to change model parameters by adding removeAllInitialConditions() and addInitialCondition(const std::string &pParameter, double pValue) to SedInstance. Corresponding methods were also added in the private implementation of SedInstanceTask, as well as applyInitialConditions() which was called during the initialisation of a model during a uniform time course (only supported case at this stage).

It was done this way in the interest of time since the values for the states, rates, constants, computed constants, and algebraic variables are all stored in the private implementation of SedInstanceTask. So, we probably ought to have some API on SedInstanceTask to modify the value of those different types of variables.

[FinbarArgus]

Hi @agarny has there been any progress on this?

I'm keen to switch Circulatory Autogen over to using libOpenCOR when possible.

[agarny]

Actually, you can do this now. Not using the kind of API you mentioned in your original message, but through SED-ML. I will provide you with some example on Monday. In the meantime, you might want to look at https://github.com/opencor/webapp/releases/tag/v0.20250709.0 for the desktop version and app / appdev for the Web app version of OpenCOR (appdev gets updated every time I push something while app gets updated everytime I generate a new version of OpenCOR). Use it with the Lorenz model and play with the model parameters. Whenever you change a value, the model gets recomputed, i.e. do what you originally asked in your first message.

[FinbarArgus]

Awesome, looking forward to the info on Monday, thanks!

[agarny]

Hi @FinbarArgus, as I was implying, libOpenCOR (and, therefore, the new OpenCOR) handles simulations through SED-ML (while in legacy OpenCOR, we were merely serialising a simulation to SED-ML). This means that we use SED-ML to do what you originally requested. SED-ML easily allows to make such changes. Ok, some code now!

import libopencor as loc


def output_results(instance_task, filename):
    with open(filename, "w") as file:
        # Header.

        file.write(instance_task.voi_name)

        for i in range(instance_task.state_count):
            file.write("," + instance_task.state_name(i))

        # Values.
        # Note: only 1 in 100 values are written to the file to reduce its size.

        file.write("\n")

        for i in range(0, len(instance_task.voi), 100):
            file.write(str(instance_task.voi[i]))

            for j in range(instance_task.state_count):
                file.write("," + str(instance_task.state(j)[i]))

            file.write("\n")



# Load and instantiate the Lorenz model.

file = loc.File("https://github.com/opencor/webapp/raw/refs/heads/main/tests/models/lorenz.omex")
document = loc.SedDocument(file)
model = document.models[0]
instance = document.instantiate()

# Run the model using its default settings.

instance.run()

# Save the simulation results.

instance_task = instance.tasks[0]

output_results(instance_task, "lorenz_default.csv")

# Rerun the model with different settings. To run the instance means running the model using its default settings + any
# changes made to it. Here, we change the value of the rho parameter from its default value of 28 to 9.

model.add_change(loc.SedChangeAttribute("main", "rho", "9"))

instance.run()

output_results(instance_task, "lorenz_rho_9.csv")

# Rerun the model with yet again different settings. First, we remove all previous changes made to the model, then we
# change the value of the rho parameter from its default value of 28 to 39.

model.remove_all_changes() # Reset the model to its default settings.
model.add_change(loc.SedChangeAttribute("main", "rho", "39"))

instance.run()

output_results(instance_task, "lorenz_rho_39.csv")

HTH, Alan.

[FinbarArgus]

Hi Alan,

Thanks, this looks great. Are the equivalents of the following available? So that we can run for a certain amount of time, change params, then keep running.

self.data.odeSolverProperties():
self.data.set_ode_solver_property(key, value)
self.data.set_point_interval(self.dt) # time interval for data storage
self.data.set_starting_point(0)
self.data.set_ending_point(self.stop_time)

and

self.simulation.reset(False) # True resets everything, False resets only the states

Just checking everything is there before we swap over to libOpenCOR.

[agarny]

Hi Finbar, sorry, those are not yet available. They are obviously on my TODO list, but not currently at the top of it. Right now, I am working on finishing the interpreter for libOpenCOR. Peter wants it and needs it to support DAEs (and I also need to add support for external variables). Once this is done, we (you, Peter, @nickerso, and I) should probably sit down and set some priorities. I clearly have mine, but they may not necessarily align with yours. :)

[FinbarArgus]

Okay, sounds good!

[FinbarArgus]

Hi Alan,

Just leaving a comment here to let you know where the issue we discussed yesterday is.

It looks like changing parameter values is possible from your above comments, but doing the following is needed so that we can run for a certain time, change parameter values, then keep running.:

self.data.odeSolverProperties():
self.data.set_ode_solver_property(key, value)
self.data.set_point_interval(self.dt) # time interval for data storage
self.data.set_starting_point(0)
self.data.set_ending_point(self.stop_time)

and

self.simulation.reset(False) # True resets everything, False resets only the states

Cheers,
Finbar

[agarny]

Thanks, @FinbarArgus, I have added this issue (and issue #224 on external variables) to my list of things to do in the "near" future.