Reference / API

Models

Loading/saving a model

ParsnipModel(path, bands[, device, threads, ...])	Generative model of transient light curves
load_model([path, device, threads])	Load a ParSNIP model.
ParsnipModel.save()	Save the model
ParsnipModel.to(device[, force])	Send the model to the specified device

Interacting with a dataset

ParsnipModel.preprocess(dataset[, ...])	Preprocess an lcdata dataset
ParsnipModel.augment_light_curves(light_curves)	Augment a set of light curves
ParsnipModel.get_data_loader(dataset[, augment])	Get a PyTorch DataLoader for an lcdata Dataset
ParsnipModel.fit(dataset[, max_epochs, ...])	Fit the model to a dataset
ParsnipModel.score(dataset[, rounds, ...])	Evaluate the loss function on a given dataset.

Generating model predictions

ParsnipModel.predict(light_curves[, augment])	Generate predictions for a light curve or set of light curves.
ParsnipModel.predict_dataset(dataset[, augment])	Generate predictions for a dataset
ParsnipModel.predict_dataset_augmented(dataset)	Generate predictions for a dataset with augmentation
ParsnipModel.predict_light_curve(light_curve)	Predict the flux of a light curve on a grid
ParsnipModel.predict_spectrum(light_curve, time)	Predict the spectrum of a light curve at a given time
ParsnipModel.predict_sncosmo(light_curve[, ...])	Package the predictions for a light curve as an sncosmo model

Individual parts of the model

ParsnipModel.forward(light_curves[, sample, ...])	Run a set of light curves through the full ParSNIP model
ParsnipModel.encode(input_data)	Predict the latent variables for a set of light curves
ParsnipModel.decode(encoding, ref_times, ...)	Predict the light curves for a given set of latent variables
ParsnipModel.decode_spectra(encoding, ...[, ...])	Predict the spectra at a given set of latent variables
ParsnipModel.loss_function(result[, ...])	Compute the loss function for a set of light curves

Datasets

Loading datasets

load_dataset(path[, kind, in_memory, ...])	Load a dataset using the lcdata package.
load_datasets(dataset_paths[, kind, ...])	Load a list of datasets and merge them
parse_dataset(dataset[, path_or_name, kind, ...])	Parse a dataset from the lcdata package.

Parsers for specific instruments

parse_plasticc(dataset[, reject_invalid, ...])	Parse a PLAsTiCC dataset
parse_ps1(dataset[, reject_invalid, ...])	Parse a PanSTARRS-1 dataset
parse_ztf(dataset[, reject_invalid, ...])	Parse a ZTF dataset

Tools for manipulating datasets

split_train_test(dataset)	Split a dataset into training and testing parts.
get_bands(dataset)	Retrieve a list of bands in a dataset

Plotting

plot_light_curve(light_curve[, model, ...])	Plot a light curve
plot_representation(predictions, plot_labels)	Plot the representation of a ParSNIP model
plot_spectrum(light_curve, model, time[, ...])	Plot the spectrum of a light curve predicted by a ParSNIP model
plot_spectra(light_curve, model[, times, ...])	Plot the spectral time series of a light curve predicted by a ParSNIP model
plot_sne_space(light_curve, model, name[, ...])	Compare a ParSNIP spectrum prediction to a real spectrum from sne.space
plot_confusion_matrix(predictions, ...[, ...])	Plot a confusion matrix
get_band_plot_color(band)	Return the plot color for a given band.
get_band_plot_marker(band)	Return the plot marker for a given band.

Classification

Classifier()	LightGBM classifier that operates on ParSNIP predictions
extract_top_classifications(classifications)	Extract the top classification for each row a classifications Table.
weighted_multi_logloss(true_types, ...)	Calculate a weighted log loss metric.

SNCosmo Interface

ParsnipSncosmoSource([model])	SNCosmo interface for a ParSNIP model
ParsnipModel.predict_sncosmo(light_curve[, ...])	Package the predictions for a light curve as an sncosmo model

Custom Neural Network Layers

ResidualBlock(in_channels, out_channels, ...)	1D residual convolutional neural network block
Conv1dBlock(in_channels, out_channels, dilation)	1D convolutional neural network block
GlobalMaxPoolingTime(*args, **kwargs)	Time max pooling layer for 1D sequences

Settings

parse_settings(bands[, settings, ...])	Parse the settings for a ParSNIP model
parse_int_list(text)	Parse a string into a list of integers
build_default_argparse(description)	Build an argparse object that can handle all of the ParSNIP model settings.
update_derived_settings(settings)	Update the derived settings for a model
update_settings_version(settings)	Update settings to a new version

Light curve utilities

preprocess_light_curve(light_curve, settings)	Preprocess a light curve for the ParSNIP model
time_to_grid(time, reference_time)	Convert a time in the original units to one on the internal ParSNIP grid
grid_to_time(grid_time, reference_time)	Convert a time on the internal grid to a time in the original units
get_band_effective_wavelength(band)	Calculate the effective wavelength of a band
calculate_band_mw_extinctions(bands)	Calculate the Milky Way extinction corrections for a set of bands
should_correct_background(bands)	Determine if we should correct the background levels for a set of bands

General utilities

nmad(x)	Calculate the normalize median absolute deviation (NMAD)
frac_to_mag(fractional_difference)	Convert a fractional difference to a difference in magnitude.
parse_device(device)	Figure out which PyTorch device to use