generate_dataset_binary.py

import argparse
from data_loader.dataset import MAGDataset

def main(args):
    binary_dataset = MAGDataset(name=args.taxon_name, path=args.data_dir, embed_suffix=args.embed_suffix, raw=True,
                                existing_partition=args.existing_partition, partition_pattern=args.partition_pattern,
                                shortest_path=args.shortest_path)

if __name__ == '__main__':
    args = argparse.ArgumentParser(description='Generate binary data from one input taxonomy')
    args.add_argument('-t', '--taxon_name', required=True, type=str, help='taxonomy name')
    args.add_argument('-d', '--data_dir', required=True, type=str, help='path to data directory')
    args.add_argument('-es', '--embed_suffix', default="", type=str, help='embed suffix indicating a specific initial embedding vectors')
    args.add_argument('-p', '--existing_partition', default=0, type=int, help='whether to use the existing train/validation/test partition files')
    args.add_argument('-pp', '--partition_pattern', default='leaf', type=str, help='how to partition existing train/validation/test files', choices=['internal', 'leaf'])
    args.add_argument('-sp', '--shortest_path', default=0, type=int, help='whether to use shortes path margin loss')
    args = args.parse_args()
    args.existing_partition = (args.existing_partition == 1)
    main(args)