stylegan2_ada_pytorch/parser.py

# Copyright (c) Facebook, Inc. and its affiliates.
# All rights reserved.

# This source code is licensed under the license found in the
# LICENSE file in the root directory of this source tree.

from argparse import ArgumentParser

def get_parser():
    parser = ArgumentParser(description="StyleGANv2")
    # General options.
    parser.add_argument(
        "--json_config",
        type=str,
        default="",
        help="Json config from where to load the configuration parameters.",
    )

    parser.add_argument(
        "--exp_name", help="Experiment name", required=False, default="default_name"
    )
    parser.add_argument(
        "--base_root",
        help="Where to save the results",
        required=False,
        default="training-runs",
        metavar="DIR",
    )
    parser.add_argument(
        "--slurm_logdir",
        help="Where to save the logs from SLURM",
        required=False,
        default="training-runs",
        metavar="DIR",
    )
    parser.add_argument(
        "--partition",
        help="Partition name for SLURM",
        required=False,
        default="learnlab",
    )
    parser.add_argument(
        "--slurm_time",
        help="Time in minutes that an experiment runs in SLURM",
        default=3200,
        type=int,
        metavar="INT",
    )
    parser.add_argument(
        "--gpus", help="Number of GPUs to use [default: 1]", type=int, metavar="INT"
    )
    parser.add_argument(
        "--nodes",
        help="Number of nodes to use [default: 1]",
        type=int,
        metavar="INT",
        default=1,
    )
    parser.add_argument(
        "--snap", help="Snapshot interval [default: 50 ticks]", type=int, metavar="INT"
    )
    parser.add_argument(
        "--seed", help="Random seed [default: 0]", type=int, metavar="INT"
    )
    parser.add_argument(
        "--port",
        help="Port number for DDP connection [default: 40000]",
        type=int,
        default=40000,
        metavar="INT",
    )
    parser.add_argument(
        "--dry-run", help="Print training options and exit", type=bool, metavar="BOOL"
    )

    # Dataset.
    parser.add_argument(
        "--data_root",
        help="Path where to find the data",
        metavar="PATH",
        required=False,
        default=None,
    )
    parser.add_argument(
        "--data",
        help="Training data (directory or zip)",
        metavar="PATH",
        required=False,
        default="datasets/cocostuff_128.zip",
    )
    # parser.add_argument('--cond',
    #                     help='Train conditional model based on dataset labels [default: false]',
    #                     type=bool, metavar='BOOL')
    parser.add_argument(
        "--class_cond",
        help="Use class labels to condition model [default: false]",
        type=bool,
        metavar="BOOL",
    )
    parser.add_argument(
        "--subset",
        help="Train with only N images [default: all]",
        type=int,
        metavar="INT",
    )
    parser.add_argument(
        "--mirror",
        help="Enable dataset x-flips [default: false]",
        type=bool,
        metavar="BOOL",
    )
    parser.add_argument(
        "--label_dim",
        help="nb of classes when using class conditioning",
        default=1000,
        type=int,
        metavar="INT",
    )

    # IC-GAN options for Dataset.
    parser.add_argument(
        "--root_feats",
        help="Training data features as instance conditioning (hdf5 file)",
        metavar="PATH",
        required=False,
        default="",
    )
    parser.add_argument(
        "--root_nns",
        help="NN Training data for each instance conditioning (hdf5 file)",
        metavar="PATH",
        required=False,
        default="",
    )
    parser.add_argument(
        "--instance_cond",
        help="Use instance features to condition model [default: false]",
        type=bool,
        metavar="BOOL",
    )
    parser.add_argument(
        "--feature_augmentation",
        help="Use horizontal flips in instances to obtain instance features [default: false]",
        type=bool,
        metavar="BOOL",
    )

    # Base config.
    parser.add_argument(
        "--cfg",
        help="Base config [default: auto]",
        choices=["auto", "stylegan2", "paper256", "paper512", "paper1024", "cifar"],
    )
    parser.add_argument("--gamma", help="Override R1 gamma", type=float)
    parser.add_argument(
        "--kimg", help="Override training duration", type=int, metavar="INT"
    )
    parser.add_argument("--batch", help="Override batch size", type=int, metavar="INT")
    parser.add_argument("--lrate", help="Override lrate", type=float)
    parser.add_argument("--num_channel_g", help="Override width of generator", type=int)
    parser.add_argument(
        "--num_channel_d", help="Override width of discriminator", type=int
    )
    parser.add_argument(
        "--channel_max_g", help="Override max width of generator", type=int
    )
    parser.add_argument(
        "--channel_max_d", help="Override max width of discriminator", type=int
    )
    parser.add_argument(
        "--hidden_dim_c",
        help="Override embedding size in maping network for class conditioning",
        type=int,
    )
    parser.add_argument(
        "--hidden_dim_h",
        help="Override embedding size in maping network for class conditioning",
        type=int,
    )
    parser.add_argument(
        "--es_patience",
        help="Early stopping patience",
        type=int,
        default=100000000,
        metavar="INT",
    )

    # Discriminator augmentation.
    parser.add_argument(
        "--aug",
        help="Augmentation mode [default: ada]",
        choices=["noaug", "ada", "fixed"],
    )
    parser.add_argument(
        "--p", help="Augmentation probability for --aug=fixed", type=float
    )
    parser.add_argument("--target", help="ADA target value for --aug=ada", type=float)
    parser.add_argument(
        "--augpipe",
        help="Augmentation pipeline [default: bgc]",
        choices=[
            "blit",
            "geom",
            "color",
            "filter",
            "noise",
            "cutout",
            "bg",
            "bgc",
            "bgcf",
            "bgcfn",
            "bgcfnc",
        ],
    )

    # Transfer learning.
    parser.add_argument(
        "--resume", help="Resume training [default: noresume]", metavar="PKL"
    )
    parser.add_argument(
        "--freezed", help="Freeze-D [default: 0 layers]", type=int, metavar="INT"
    )

    # Performance options.
    parser.add_argument(
        "--fp32", help="Disable mixed-precision training", type=bool, metavar="BOOL"
    )
    parser.add_argument(
        "--nhwc", help="Use NHWC memory format with FP16", type=bool, metavar="BOOL"
    )
    parser.add_argument(
        "--nobench", help="Disable cuDNN benchmarking", type=bool, metavar="BOOL"
    )
    parser.add_argument(
        "--allow-tf32",
        help="Allow PyTorch to use TF32 internally",
        type=bool,
        metavar="BOOL",
    )
    parser.add_argument(
        "--workers",
        help="Override number of DataLoader workers",
        type=int,
        metavar="INT",
    )

    ## Experiment setup
    parser.add_argument(
        "--slurm",
        help="Using SLURM to launch the experiment in a cluster",
        type=bool,
        metavar="BOOL",
    )
    return parser


# ----------------------------------------------------------------------------