def parse_arguments()

in src/run.py [0:0]

• 22 lines of code
• 2 McCabe index (conditional complexity)

def parse_arguments():
    parser = argparse.ArgumentParser()

    # data and I/O
    parser.add_argument("--data_path", type=str, default="/root/downloads/imagenet")
    parser.add_argument("--ckpt_path", type=str, default="/root/downloads/model.ckpt-1000000")
    parser.add_argument("--color_cluster_path", type=str, default="/root/downloads/kmeans_centers.npy")
    parser.add_argument("--save_dir", type=str, default="/root/save/")

    # model
    parser.add_argument("--n_embd", type=int, default=512)
    parser.add_argument("--n_head", type=int, default=8)
    parser.add_argument("--n_layer", type=int, default=24)
    parser.add_argument("--n_px", type=int, default=32, help="image height or width in pixels")
    parser.add_argument("--n_vocab", type=int, default=512, help="possible values for each pixel")

    parser.add_argument("--bert", action="store_true", help="use the bert objective (defaut: autoregressive)")
    parser.add_argument("--bert_mask_prob", type=float, default=0.15)
    parser.add_argument("--clf", action="store_true", help="add a learnable classification head")

    # parallelism
    parser.add_argument("--n_sub_batch", type=int, default=8, help="per-gpu batch size")
    parser.add_argument("--n_gpu", type=int, default=8, help="number of gpus to distribute training across")

    # mode
    parser.add_argument("--eval", action="store_true", help="evaluates the model, requires a checkpoint and dataset")
    parser.add_argument("--sample", action="store_true", help="samples from the model, requires a checkpoint and clusters")

    # reproducibility
    parser.add_argument("--seed", type=int, default=42, help="seed for random, np, tf")

    args = parser.parse_args()
    print("input args:\n", json.dumps(vars(args), indent=4, separators=(",", ":")))
    return args