handler.py

from typing import Dict, List, Any
from transformers import AutoTokenizer, AutoModelForSequenceClassification


class EndpointHandler:
    def __init__(self,
                 path=""
                 ):
        # Preload all the elements you are going to need at inference.
        # pseudo:
        self.model = AutoModelForSequenceClassification.from_pretrained(path)
        self.tokenizer = AutoTokenizer.from_pretrained("roberta-large", padding_side='right')

    def __call__(self, data: Dict[str, Any]) -> List[Dict[str, Any]]:

        input_dict = data.pop("inputs", data)

        self.model.eval()
        input = self.tokenizer(input_dict['answer'],
                               input_dict['source'],
                               truncation=True,
                               max_length=None,
                               return_tensors="pt"
                               )
        # input.to(device)
        # with torch.no_grad():
        #     output = model(**input)
        output = model(**input)

        prediction = output.logits.argmax(dim=-1)

        #smax = nn.Softmax(dim=1)
        #score = smax(output.logits)

        return [{
                 "label": prediction.item(),
                 #"score": score[0][0].item()
                }
                ]