def get_arguments()

in conv_split_awa_hybrid.py [0:0]
52 lines of code
1 McCabe index (conditional complexity)

def get_arguments():
    """Parse all the arguments provided from the CLI.

    Returns:
      A list of parsed arguments.
    """
    parser = argparse.ArgumentParser(description="Script for split AWA Hybrid experiment.")
    parser.add_argument("--cross-validate-mode", action="store_true",
            help="If option is chosen then snapshoting after each batch is disabled")
    parser.add_argument("--online-cross-val", action="store_true",
            help="If option is chosen then enable the online cross validation of the learning rate")
    parser.add_argument("--train-single-epoch", action="store_true", 
            help="If option is chosen then train for single epoch")
    parser.add_argument("--set-hybrid", action="store_true", 
            help="If option is chosen then train using hybrid model")
    parser.add_argument("--eval-single-head", action="store_true",
            help="If option is chosen then evaluate on a single head setting.")
    parser.add_argument("--arch", type=str, default=ARCH,
                        help="Network Architecture for the experiment.\
                                \n \nSupported values: %s"%(VALID_ARCHS))
    parser.add_argument("--num-runs", type=int, default=NUM_RUNS,
                       help="Total runs/ experiments over which accuracy is averaged.")
    parser.add_argument("--train-iters", type=int, default=TRAIN_ITERS,
                       help="Number of training iterations for each task.")
    parser.add_argument("--batch-size", type=int, default=BATCH_SIZE,
                       help="Mini-batch size for each task.")
    parser.add_argument("--random-seed", type=int, default=RANDOM_SEED,
                       help="Random Seed.")
    parser.add_argument("--learning-rate", type=float, default=LEARNING_RATE,
                       help="Starting Learning rate for each task.")
    parser.add_argument("--optim", type=str, default=OPTIM,
                        help="Optimizer for the experiment. \
                                \n \nSupported values: %s"%(VALID_OPTIMS))
    parser.add_argument("--imp-method", type=str, default=IMP_METHOD,
                       help="Model to be used for LLL. \
                        \n \nSupported values: %s"%(MODELS))
    parser.add_argument("--synap-stgth", type=float, default=SYNAP_STGTH,
                       help="Synaptic strength for the regularization.")
    parser.add_argument("--fisher-ema-decay", type=float, default=FISHER_EMA_DECAY,
                       help="Exponential moving average decay for Fisher calculation at each step.")
    parser.add_argument("--fisher-update-after", type=int, default=FISHER_UPDATE_AFTER,
                       help="Number of training iterations after which the Fisher will be updated.")
    parser.add_argument("--do-sampling", action="store_true",
                       help="Whether to do sampling")
    parser.add_argument("--mem-size", type=int, default=SAMPLES_PER_CLASS,
                       help="Number of samples per class from previous tasks.")
    parser.add_argument("--is-herding", action="store_true",
                        help="Herding based sampling")
    parser.add_argument("--data-dir", type=str, default=DATA_DIR,
                       help="Directory from where the AWA data will be read.\
                               NOTE: Provide path till <AWA_DIR>/Animals_with_Attributes2")
    parser.add_argument("--init-checkpoint", type=str, default=RESNET18_IMAGENET_CHECKPOINT,
                       help="TF checkpoint file containing initialization for ImageNet.\
                               NOTE: NPZ file for VGG and TF Checkpoint for ResNet")
    parser.add_argument("--log-dir", type=str, default=LOG_DIR,
                       help="Directory where the plots and model accuracies will be stored.")
    return parser.parse_args()