Detailed Usage Notes

Installation

The dependencies can be installed into your currently active python3 environment using the command python3 setup.py install

Execution

VlaPy can be executed using the start_run method within the manager module. For example, run_vlapy.py has the following code

k0 = np.random.uniform(0.3, 0.4, 1)[0]
log_nu_over_nu_ld = None

all_params_dict = initializers.make_default_params_dictionary()
all_params_dict = initializers.specify_epw_params_to_dict(
    k0=k0, all_params_dict=all_params_dict
)
all_params_dict = initializers.specify_collisions_to_dict(
    log_nu_over_nu_ld=log_nu_over_nu_ld, all_params_dict=all_params_dict
)

all_params_dict["vlasov-poisson"]["time"] = "leapfrog"
all_params_dict["vlasov-poisson"]["edfdv"] = "exponential"
all_params_dict["vlasov-poisson"]["vdfdx"] = "exponential"

all_params_dict["fokker-planck"]["type"] = "lb"

pulse_dictionary = {
    "first pulse": {
        "start_time": 0,
        "t_L": 6,
        "t_wL": 2.5,
        "t_R": 20,
        "t_wR": 2.5,
        "w0": all_params_dict["w_epw"],
        "a0": 1e-7,
        "k0": k0,
    }
}

mlflow_exp_name = "landau-damping"

uris = {
    "tracking": "local",
}

print_to_screen.print_startup_message(
    mlflow_exp_name, all_params_dict, pulse_dictionary
)

that_run = manager.start_run(
    all_params=all_params_dict,
    pulse_dictionary=pulse_dictionary,
    diagnostics=landau_damping.LandauDamping(
        vph=all_params_dict["v_ph"],
        wepw=all_params_dict["w_epw"],
    ),
    uris=uris,
    name=mlflow_exp_name,
)

Through this interface, the user can specify the initial conditions of the domain in space and time. The user can also specify the settings of the wave-driver module, the collision frequency nu, and custom diagnostic routines.

Wave Driver

The wave driver module uses a 5th order polynomial that is given in ref. [@Joglekar2018]. The implementation can be found in vlapy/manager.py.

Simulation Management

VlaPy leverages MLflow to manage data corresponding to different categories and types of simulations. Before starting a simulation, VlaPy enables you to provide an experiment name (in MLFlow parlance). Each simulation run with this experiment name will be stored here.

Each simulation creates a temporary directory for the simulation files. Once completed, MLFlow will move the simulation folder into a centralized datastore corresponding to the specified experiment name. This datastore can be accessed through a web-browser based UI.

To start the MLFlow UI server, type mlflow ui into the terminal and then navigate to localhost:5000 in your web browser. If you have run the default run_vlapy.py script a few times, you will see a page like this one below

Diagnostics

The diagnostics module is designed to provide flexibility to the user. The user is free to design their own diagnostics that are called at the end of the simulation. The current implementation relies on a diagnostics.<CUSTOMCLASS>(storage_manager) call that performs all the necessary diagnostics.

For example, please refer to the Landau damping diagnostics in diagnostics/landau_damping.py where the electric field damping rate and oscillation frequency are calculated, and plots are made of the time-evolution of the electric field to be eventually stored by the run manager object in a location of its choosing.