handler.py

from typing import Dict, List, Any
from transformers import AutoModelForSeq2SeqLM, AutoTokenizer, pipeline
import torch

class EndpointHandler:
    def __init__(self, path=""):
        # load model and processor from path
        self.model =  AutoModelForSeq2SeqLM.from_pretrained(path, device_map="auto")
        self.tokenizer = AutoTokenizer.from_pretrained(path)

        self.pipeline = pipeline(task="text-generation", tokenizer=self.tokenizer, device=0, device_map="auto", framework="pt", model=self.model, max_length=512)

    def __call__(self, data: Dict[str, Any]) -> Dict[str, str]:
        """
        Args:
            data (:obj:):
                includes the deserialized image file as PIL.Image
        """
        # process input
        inputs = data.pop("inputs", data)
        parameters = data.pop("parameters", None)

        # preprocess
        input_ids = self.tokenizer(inputs, return_tensors="pt").input_ids

        # postprocess the prediction
        prediction = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
        # pass inputs with all kwargs in data
        if parameters is not None:
            outputs = self.model.generate(inputs, device=0, **parameters)
        else:
            outputs = self.model.generate(inputs, device=0)

        # postprocess the prediction
        prediction = self.tokenizer.decode(outputs[0], skip_special_tokens=True)

        return [{"generated_text": prediction}]