import argparse

def init():
    parser = argparse.ArgumentParser(description="PyTorch")
    parser.add_argument('--path_to_images', type=str, default='/fast_dataset/shuai/avadata/images',
                        help='directory to images')
    parser.add_argument('--path_to_save_csv', type=str, default="./dataset/",
                        help='directory to csv_folder')
    parser.add_argument('--experiment_dir_name', type=str, default='.',
                        help='directory to project')
    parser.add_argument('--path_to_model_weight', type=str, default='SRCC_758_LCC_765.pth',
                        help='directory to pretrain model')
    parser.add_argument('--init_lr_res365_last', type=int, default=0.0003, help='learning_rate')
    parser.add_argument('--init_lr_mobileNet', type=int, default=0.0003, help='learning_rate')
    parser.add_argument('--init_lr_head', type=int, default=0.0003, help='learning_rate')
    parser.add_argument('--init_lr_head_rgb', type=int, default=0.0003, help='learning_rate')
    parser.add_argument('--init_lr_hypernet', type=int, default=0.0003, help='learning_rate')
    parser.add_argument('--init_lr_tygertnet', type=int, default=0.0003, help='learning_rate')
    parser.add_argument('--init_lr', type=int, default=0.0003, help='learning_rate')
    parser.add_argument('--num_epoch', type=int, default=20, help='epoch num for train'
                        )
    parser.add_argument('--batch_size', type=int, default=48, help='16how many pictures to process one time'
                        )
    parser.add_argument('--num_workers', type=int, default=6, help='num_workers',
                        )
    parser.add_argument('--gpu_id', type=str, default='0', help='which gpu to use')
    args = parser.parse_args()
    return args