Rewrite the generation of csv files (splits) into a parser format

[martinwholtmon]

Instead of having a script for each dataset, instead make a parser for each dataset, register the parser and process the dataset.

Create a generic class that represent a parser

from abc import ABC, abstractmethod

class Parser(ABC):
    @abstractmethod
    def __init__(self, data_path: str, fold: int, val_split: float):
        self.data_path = data_path
        self.val_split = val_split
        self.classes: dict = self._get_classes()

    @abstractmethod
    def _get_classes(self):
        """Get the class idx and class names.
        
        Returns:
            dict[str, int]: class name, class id
        """

    @abstractmethod
    def process(self):
        """Process the dataset, generating the train,val,test splits"""

    def split_train_val(
        self, train_data: pd.DataFrame
    ) -> Tuple[pd.DataFrame, pd.DataFrame]:
        total_len = len(train_data)
        val_len = int(total_len * self.val_split)
        train_len = total_len - val_len

        shuffled = train_data.sample(frac=1).reset_index(drop=True)
        return shuffled.iloc[:train_len], shuffled.iloc[train_len:]

    def save_csv(self, data: pd.DataFrame, file_name: str):
        data.to_csv(
            os.path.join(self.data_path, file_name),
            sep=" ",
            index=False,
            header=False,
        )

Based on this class, abstract from it and create the parser for each dataset. Create a main that instanciates the correct parser based on the arguments etc..