handler.py

from typing import  Dict, List, Any
from optimum.onnxruntime import ORTModelForSequenceClassification
from transformers import AutoTokenizer
from optimum.pipelines import pipeline


import torch

if torch.backends.cudnn.is_available():
    print("cudnn:", torch.backends.cudnn.version())

class EndpointHandler():
    def __init__(self, path=""):

        on_cuda = torch.cuda.is_available()
        # load the optimized model

        provider = "CPUExecutionProvider"
        if on_cuda:
            provider = "CUDAExecutionProvider"
        
        model = ORTModelForSequenceClassification.from_pretrained(
            path,
            export=False,
            provider=provider,
        )
        tokenizer = AutoTokenizer.from_pretrained(path)
        
        device = -1
        if on_cuda:
            device = 0
        # create inference pipeline
        self.pipeline = pipeline("text-classification", model=model, tokenizer=tokenizer, device=device)


    def __call__(self, data: Any) -> List[List[Dict[str, float]]]:
        """
        Args:
            data (:obj:):
                includes the input data and the parameters for the inference.
        Return:
            A :obj:`list`:. The object returned should be a list of one list like [[{"label": 0.9939950108528137}]] containing :
                - "label": A string representing what the label/class is. There can be multiple labels.
                - "score": A score between 0 and 1 describing how confident the model is for this label/class.
        """
        inputs = data.pop("inputs", data)
        parameters = data.pop("parameters", dict())

        prediction = self.pipeline(inputs, **parameters)

        return prediction