FlopPITy 0.1

normalising Flow exoPlanet Parameter Inference Toolkyt

FlopPITy

normalizing Flow exoplanet Parameter Inference Toolkyt

FlopPITy allows the user to easily perform atmospheric retrievals using SNPE-C (citation) and neural spline flows (citation).

Installation guide

Currently FlopPITy doesn't work with python 3.13

$ conda create -n floppity_env python==3.12.9
$ conda activate floppity_env
$ pip install floppity

Basic usage:

• First, import FlopPITy:

from floppity import Retrieval

• Now you can initialize the retrieval class with a simulator. A simulator is a function that takes in parameters and returns spectra, look below to see specifically how it needs to be written. Functionality for ARCiS and PICASO comes built-in (you need to install them separately). Look further down for examples.

R = Retrieval(your_simulator_function, 'emis')

• Read in observations and define parameters to retrieve:

R.get_obs({obs_0:'path/to/obs_0', obs_1:'path/to/obs_1',..., obs_n:np.array(shape=[n_wvl,>3])})
    
R.add_parameter(par_0, min, max)
R.add_parameter(par_1, min, max)
...
R.add_parameter(par_m, min, max)

• You can now run the retrieval, indicating the number of rounds and samples per round:

R.run(n_rounds=10, n_samples=1000, simulator_kwargs=simulator_kwargs)

• Great! You can now inspect your posterior:

fig = R.plot_corner()

ARCiS example:

• Firstly, initialize your retrieval object:

from floppity import Retrieval
from floppity.simulators import read_ARCiS_input, ARCiS

R = Retrieval(ARCiS, 'emis')

• For ARCiS, the observations and parameters can be read from the ARCiS input file:

pars, obs_list = read_ARCiS_input('path/to/ARCiS/input')
R.get_obs(obs_list)
R.parameters=pars

• The input file and output directory need to be passed in a dictionary:

ARCiS_kwargs= dict(
                    ARCiS_dir = "/path/to/ARCiS/executable", #only needs to be set if ARCiS is not on the default path
                    input_file = 'path/to/ARCiS/input',
                    output_dir = 'path/to/output',
                  )

• You can now run the retrieval as usual:

R.run(n_rounds=10, n_samples=1000, simulator_kwargs=ARCiS_kwargs)

PICASO example:

• Running a retrieval with PICASO is very similar (this only works with the gridtree branch):

from floppity import Retrieval
from floppity.simulators import read_PICASO_config, PICASO

R = Retrieval(PICASO, 'emis')

pars, obs_list = read_PICASO_config('path/to/config.toml')
R.get_obs(obs_list)
R.parameters=pars

• The configuration file needs to be passed as a kwarg:

PICASO_kwargs= dict(
                    config_file = 'path/to/config.toml'
                  )

• You can now run the retrieval as usual:

R.run(n_rounds=10, n_samples=1000, simulator_kwargs=PICASO_kwargs)

Writing a simulator

Writing a simulator to work for FlopPITy is relatively straightforward. All that's needed is a function that takes in observations and parameters and returns spectra. The spectra need to be returned in a dictionary where each key represents each of the observations simulated (e.g. simulated['prism'] contains PRISM spectra and simulated['lrs'] contains MIRI/LRS spectra):

def simulator(obs, parameters, **kwargs):
    wvl_prism = obs['prism'][:,0]
    wvl_lrs = obs['lrs'][:,0]
    ...
    wvl_n = obs[n][:,0]

    spectra={}
    spectra['prism'] = # array of shape (ndims, len(wvl_prism))
    spectra['lrs'] = # array of shape (ndims, len(wvl_lrs))
    ...
    spectra[n] = # array of shape (ndims, len(wvl_n))

    return spectra

Advanced options:

• Additional post processing parameters (currently RV, vrot, offset and scaling) can be added, for example:

R.add_parameter('RV', -100, 100, post_process=True) # km/s