Feature

feature_subsetting

This file contains implementation for model training on feature subsets.

class scalr.feature.feature_subsetting.FeatureSubsetting(feature_subsetsize: int, chunk_model_config: dict, chunk_model_train_config: dict, train_data: AnnData | AnnCollection, val_data: AnnData | AnnCollection, target: str, mappings: dict, dirpath: str = None, device: str = 'cpu', num_workers: int = 1, sample_chunksize: int = None)[source]

Bases: object

Class for FeatureSubsetting.

It trains a model for each subsetted datasets, each containing feature_subsetsize genes as features.

get_updated_configs()[source]: Returns updated configs.

train_chunked_models() → list[Module][source]

Trains a model for each subset data.

Returns:: List of models for each subset.
Return type:: list[nn.Module]

write_feature_subsetted_data()[source]: Write chunks of feature-subsetted data, to enable parallel training of models using different chunks of data.