def train_dt_procedure(
    session: Session,
    training_table: str,
    feature_cols: list,
    target_col: str,
    model_name: str,
) -> T.Variant:

    """
    This will be our training procedure. Later we will register this as snowflake procedure.

    training_table: snowflake table name to be used for training task
    feature_cols: list of columns to be used in training
    target_col: target column to be used
    model_name: model name to used for model saving purpose

    """

    #convert as pandas DF, rest of the steps similar to the local model training and saving.
    local_training_data = session.table(training_table).to_pandas()

    from sklearn.tree import DecisionTreeClassifier

    X = local_training_data[feature_cols]
    y = local_training_data[target_col]

    model = DecisionTreeClassifier()
    model.fit(X, y)

    #do what ever you want to do with model, even the hyperparameter tuning..
    # here I'll get feature importance
    feat_importance = pd.DataFrame(
        model.feature_importances_, feature_cols, columns=["FeatImportance"]
    ).to_dict()

    from joblib import dump

    dump(model, "/tmp/" + model_name)
    session.file.put(
        "/tmp/" + model_name, "@ML_MODELS", auto_compress=False, overwrite=True
    )
    return feat_importance