dust

pinballrt.dust.load(filename, device='cpu')

Load a Dust object from a file or state_dict.

Parameters:

filenamestr, dict, or Dust

The filename to load the Dust object from, or a state_dict, or a Dust object.

devicestr

The device to load the Dust object onto.

class pinballrt.dust.Dust(lam=None, kabs=None, ksca=None, amax=None, p=None, device='cpu')

Bases: LightningDataModule

Initialize the Dust module with wavelength, absorption, and scattering coefficients.

Parameters:

lamastropy.units.Quantity

Wavelengths at which the dust opacities are defined.

kabsastropy.units.Quantity

Absorption coefficients of the dust.

kscaastropy.units.Quantity

Scattering coefficients of the dust.

amaxastropy.units.Quantity

Maximum grain size for the size distribution.

pfloat

Power-law index for the grain size distribution.

devicestr

Device to run the computations on (e.g., “cpu” or “cuda”).

Methods Summary

learn([model, nsamples, test_split, ...])	Learn a model for either the random_nu function or the ml_step function.
fit([epochs, batch_size, num_workers, ckpt_path])	Run the model fit.
test_model([plot])	Test the model fit.
run_dust_simulation([nphotons, tau_range, ...])	Run a dust simulation that can be used to learn an ML-step model with the given parameters.
plot_opacity_model([model])	Plot the learned opacity model against the interpolated opacity.
plot_pmo_model()	Plot the learned Planck mean opacity model against the interpolated Planck mean opacity.
plot_random_nu_model([nsamples])	Plot samples drawn from the learned random_nu model against samples drawn from the random_nu_manual function.
plot_ml_step([tau, temperature, amax, p, ...])	Plot the samples drawn from a sphere with the provided optical depth, temperature, and frequency.
save(filename)	Save the Dust object to a file.

Methods Documentation

learn(model='random_nu', nsamples=200000, test_split=0.1, valid_split=0.2, hidden_units=(48, 48, 48), tau_range=(3.0, 10000.0), temperature_range=(<Quantity 0.1 K>, <Quantity 10000. K>), amax_range=(<Quantity 1. micron>, <Quantity 10. cm>), p_range=(2.5, 4.5), nu_range=None, overwrite=False)

Learn a model for either the random_nu function or the ml_step function.

Parameters:

modelstr

The model to learn. Either “random_nu” or “ml_step”.

nsamplesint

The total number of samples to generate for the learning process (including validation and testing)

test_splitfloat

The fraction of the samples to use for testing.

valid_splitfloat

The fraction of the remaining samples (after testing) to use for validation.

hidden_unitstuple

The number of hidden units in each layer of the neural network.

tau_rangetuple

The range of optical depths to sample from for the ml_step model.

temperature_rangetuple

The range of temperatures to sample from for the ml_step model.

amax_rangetuple

The range of maximum grain sizes to sample from for the ml_step model.

p_rangetuple

The range of power-law indices to sample from for the ml_step model.

nu_rangetuple

The range of frequencies to sample from (in GHz) for the ml_step model. If None, use the full range of the dust opacities.

fit(epochs=10, batch_size=100, num_workers=1, ckpt_path=None)

Run the model fit.

test_model(plot=False)

Test the model fit.

run_dust_simulation(nphotons=1000, tau_range=(3.0, 10000.0), temperature_range=(<Quantity 0.1 K>, <Quantity 10000. K>), amax_range=(<Quantity 1. micron>, <Quantity 10. cm>), p_range=(2.5, 4.5), nu_range=None, use_ml_step=False, position=0)

Run a dust simulation that can be used to learn an ML-step model with the given parameters.

Parameters:

nphotonsint

The number of photons to simulate.

tau_rangetuple

The range of optical depths to sample from (in log10).

temperature_rangetuple

The range of temperatures to sample from (in log10).

amax_rangetuple

The range of maximum grain sizes to sample from (in log10).

p_rangetuple

The range of power-law indices to sample from.

nu_rangetuple

The range of frequencies to sample from (in GHz). If None, use the full range of the dust opacities.

plot_opacity_model(model='kabs')

Plot the learned opacity model against the interpolated opacity.

Parameters:

modelstr

The model to plot. Either “kabs” or “ksca”.

plot_pmo_model()

Plot the learned Planck mean opacity model against the interpolated Planck mean opacity.

plot_random_nu_model(nsamples=100000)

Plot samples drawn from the learned random_nu model against samples drawn from the random_nu_manual function.

Parameters:

nsamplesint

The number of samples to draw from each model for the plot.

plot_ml_step(tau=30.0, temperature=<Quantity 100. K>, amax=<Quantity 1. micron>, p=3.5, nu=<Quantity 1000. GHz>, nsamples=1000, plot_columns=array(['log10_nu', 'log10_Eabs', 'log10_tau', 'yaw', 'pitch',        'direction_yaw', 'direction_pitch'], dtype='<U15'))

Plot the samples drawn from a sphere with the provided optical depth, temperature, and frequency.

Parameters:

taufloat

The log-optical depth to use for the samples.

temperaturefloat

The temperature to use for the samples.

amaxfloat

The maximum grain size to use for the samples.

pfloat

The power-law index to use for the samples.

nufloat

The frequency to use for the samples.

nsamplesint

The number of samples to generate.

plot_columnsnumpy.ndarray

The columns to plot in the triangle plots.

save(filename)

Save the Dust object to a file.

Parameters:

filenamestr

The filename to save the Dust object to.