handler.py

from typing import Dict, List, Any
from transformers import pipeline

class EndpointHandler():
    def __init__(self, path=""):
        self.pipeline = pipeline("text-classification", model=path)

    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:
        """
       data args:
            inputs (:obj: `str`)
            date (:obj: `str`)
      Return:
            A :obj:`list` | `dict`: will be serialized and returned
        """
        # get inputs
        inputs = data.pop("inputs", data)

        # run normal prediction
        prediction = self.pipeline(inputs)

        # Dictionary to map labels
        label_mapping = {
            'LABEL_0': 'credit_card',
            'LABEL_1': 'credit_reporting',
            'LABEL_2': 'debt_collection',
            'LABEL_3': 'mortgages_and_loans',
            'LABEL_4': 'retail_banking'
        }

        # Apply the mapping to the output
        mapped_output = [{'label': label_mapping.get(item['label'], item['label']), 'score': item['score']} for item in
                         prediction]

        return mapped_output