deepSIP

class deepSIP.model.deepSIP(spec_len=1024, seed=100, drop_rate=0.02)

Bases: object

class for deploying trained deepSIP models

Parameters:

spec_len : int, optional

number of wavelength bins for pre-processed spectra (must match what was used in training models)

seed : int, optional

seed for random number generator

Attributes:

models : pd.DataFrame

models, along with metadata and utilities, indexed by purpose

device : torch.device

device type being used (GPU if available, else CPU)

Methods

predict(self, spectra[, threshold, mcnum, …])

make predictions with trained models

predict(self, spectra, threshold=0.9, mcnum=30, status=False)

make predictions with trained models

Parameters:

spectra : np.ndarray or pd.DataFrame

pre-preocessed spectra if np.ndarray else pd.DataFrame with columns of [SN, filename, z] and optionally obsframe as bool

threshold : float, optional

minimum threshold for ‘in’ classification by Domain model

mcnum : int, optional

number of stochastic forward passes to perform

status : bool, optional

show status bars

Returns:

pd.DataFrame

predictions generated by each model