maheshdev209/fhehp

1import os2import shutil3from pathlib import Path4from typing import List, Tuple, Union5 6import numpy7import pandas8 9from concrete.ml.sklearn import XGBClassifier as ConcreteXGBoostClassifier10 11# Max Input to be displayed on the HuggingFace space brower using Gradio12# Too large inputs, slow down the server: https://github.com/gradio-app/gradio/issues/187713INPUT_BROWSER_LIMIT = 40014 15# Store the server's URL16SERVER_URL = "http://localhost:8000/"17 18CURRENT_DIR = Path(__file__).parent19DEPLOYMENT_DIR = CURRENT_DIR / "deployment_files"20KEYS_DIR = DEPLOYMENT_DIR / ".fhe_keys"21CLIENT_DIR = DEPLOYMENT_DIR / "client_dir"22SERVER_DIR = DEPLOYMENT_DIR / "server_dir"23 24ALL_DIRS = [KEYS_DIR, CLIENT_DIR, SERVER_DIR]25 26# Columns that define the target27TARGET_COLUMNS = ["prognosis_encoded", "prognosis"]28 29TRAINING_FILENAME = "./data/Training_preprocessed.csv"30TESTING_FILENAME = "./data/Testing_preprocessed.csv"31 32# pylint: disable=invalid-name33 34 35def pretty_print(36    inputs, case_conversion=str.title, which_replace: str = "_", to_what: str = " ", delimiter=None37):38    """39    Prettify and sort the input as a list of string.40 41    Args:42        inputs (Any): The inputs to be prettified.43 44    Returns:45        List: The prettified and sorted list of inputs.46 47    """48    # Flatten the list if required49    pretty_list = []50    for item in inputs:51        if isinstance(item, list):52            pretty_list.extend(item)53        else:54            pretty_list.append(item)55 56    # Sort57    pretty_list = sorted(list(set(pretty_list)))58    # Replace59    pretty_list = [item.replace(which_replace, to_what) for item in pretty_list]60    pretty_list = [case_conversion(item) for item in pretty_list]61    if delimiter:62        pretty_list = f"{delimiter.join(pretty_list)}."63 64    return pretty_list65 66 67def clean_directory() -> None:68    """69    Clear direcgtories70    """71    print("Cleaning...\n")72    for target_dir in ALL_DIRS:73        if os.path.exists(target_dir) and os.path.isdir(target_dir):74            shutil.rmtree(target_dir)75        target_dir.mkdir(exist_ok=True, parents=True)76 77 78def get_disease_name(encoded_prediction: int, file_name: str = TRAINING_FILENAME) -> str:79    """Return the disease name given its encoded label.80 81    Args:82        encoded_prediction (int): The encoded prediction83        file_name (str): The data file path84 85    Returns:86        str: The according disease name87    """88    df = pandas.read_csv(file_name, usecols=TARGET_COLUMNS).drop_duplicates()89    disease_name, _ = df[df[TARGET_COLUMNS[0]] == encoded_prediction].values.flatten()90    return disease_name91 92 93def load_data() -> Union[Tuple[pandas.DataFrame, numpy.ndarray], List]:94    """95    Return the data96 97    Args:98        None99 100    Return:101        The train, testing set and valid symptoms.102    """103    # Load data104    df_train = pandas.read_csv(TRAINING_FILENAME)105    df_test = pandas.read_csv(TESTING_FILENAME)106 107    # Separate the traget from the training / testing set:108    # TARGET_COLUMNS[0] -> "prognosis_encoded" -> contains the numeric label of the disease109    # TARGET_COLUMNS[1] -> "prognosis"         -> contains the name of the disease110 111    y_train = df_train[TARGET_COLUMNS[0]]112    X_train = df_train.drop(columns=TARGET_COLUMNS, axis=1, errors="ignore")113 114    y_test = df_test[TARGET_COLUMNS[0]]115    X_test = df_test.drop(columns=TARGET_COLUMNS, axis=1, errors="ignore")116 117    return (118        (X_train, X_test),119        (y_train, y_test),120        X_train.columns.to_list(),121        df_train[TARGET_COLUMNS[1]].unique().tolist(),122    )123 124 125def load_model(X_train: pandas.DataFrame, y_train: numpy.ndarray):126    """127    Load a pre-trained serialized model128 129    Args:130        X_train (pandas.DataFrame): Training set131        y_train (numpy.ndarray): Targets of the training set132 133    Return:134        The Concrete ML model and its circuit135    """136    # Parameters137    concrete_args = {"max_depth": 1, "n_bits": 3, "n_estimators": 3, "n_jobs": -1}138    classifier = ConcreteXGBoostClassifier(**concrete_args)139    # Train the model140    classifier.fit(X_train, y_train)141    # Compile the model142    circuit = classifier.compile(X_train)143 144    return classifier, circuit145