def get_config()

in src/controlnet_aux/zoe/zoedepth/utils/config.py [0:0]

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

def get_config(model_name, mode='train', dataset=None, **overwrite_kwargs):
    """Main entry point to get the config for the model.

    Args:
        model_name (str): name of the desired model.
        mode (str, optional): "train" or "infer". Defaults to 'train'.
        dataset (str, optional): If specified, the corresponding dataset configuration is loaded as well. Defaults to None.
    
    Keyword Args: key-value pairs of arguments to overwrite the default config.

    The order of precedence for overwriting the config is (Higher precedence first):
        # 1. overwrite_kwargs
        # 2. "config_version": Config file version if specified in overwrite_kwargs. The corresponding config loaded is config_{model_name}_{config_version}.json
        # 3. "version_name": Default Model version specific config specified in overwrite_kwargs. The corresponding config loaded is config_{model_name}_{version_name}.json
        # 4. common_config: Default config for all models specified in COMMON_CONFIG

    Returns:
        easydict: The config dictionary for the model.
    """


    check_choices("Model", model_name, ["zoedepth", "zoedepth_nk"])
    check_choices("Mode", mode, ["train", "infer", "eval"])
    if mode == "train":
        check_choices("Dataset", dataset, ["nyu", "kitti", "mix", None])

    config = flatten({**COMMON_CONFIG, **COMMON_TRAINING_CONFIG})
    config = update_model_config(config, mode, model_name)

    # update with model version specific config
    version_name = overwrite_kwargs.get("version_name", config["version_name"])
    config = update_model_config(config, mode, model_name, version_name)

    # update with config version if specified
    config_version = overwrite_kwargs.get("config_version", None)
    if config_version is not None:
        print("Overwriting config with config_version", config_version)
        config = update_model_config(config, mode, model_name, config_version)

    # update with overwrite_kwargs
    # Combined args are useful for hyperparameter search
    overwrite_kwargs = split_combined_args(overwrite_kwargs)
    config = {**config, **overwrite_kwargs}

    # Casting to bool   # TODO: Not necessary. Remove and test
    for key in KEYS_TYPE_BOOL:
        if key in config:
            config[key] = bool(config[key])

    # Model specific post processing of config
    parse_list(config, "n_attractors")

    # adjust n_bins for each bin configuration if bin_conf is given and n_bins is passed in overwrite_kwargs
    if 'bin_conf' in config and 'n_bins' in overwrite_kwargs:
        bin_conf = config['bin_conf']  # list of dicts
        n_bins = overwrite_kwargs['n_bins']
        new_bin_conf = []
        for conf in bin_conf:
            conf['n_bins'] = n_bins
            new_bin_conf.append(conf)
        config['bin_conf'] = new_bin_conf

    if mode == "train":
        orig_dataset = dataset
        if dataset == "mix":
            dataset = 'nyu'  # Use nyu as default for mix. Dataset config is changed accordingly while loading the dataloader
        if dataset is not None:
            config['project'] = f"MonoDepth3-{orig_dataset}"  # Set project for wandb

    if dataset is not None:
        config['dataset'] = dataset
        config = {**DATASETS_CONFIG[dataset], **config}
        

    config['model'] = model_name
    typed_config = {k: infer_type(v) for k, v in config.items()}
    # add hostname to config
    config['hostname'] = platform.node()
    return edict(typed_config)