handler.py

from transformers import AutoTokenizer, AutoModelForSequenceClassification
from torch.nn.functional import softmax

import torch
from typing import Any, Dict, List


class EndpointHandler:
    def __init__(self, path=""):
        self.tokenizer = AutoTokenizer.from_pretrained(path)
        self.model = AutoModelForSequenceClassification.from_pretrained(path)
        self.tokenizer.pad_token = self.tokenizer.eos_token
        self.model.config.pad_token_id = self.tokenizer.pad_token_id

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

        tokens = self.tokenizer(
            batch_of_strings, padding=True, truncation=True, return_tensors="pt"
        )

        # Calculate the loss
        with torch.no_grad():
            outputs = self.model(**tokens)

        probabilities = softmax(outputs.logits, dim=1)

        return {
            "predictions": [pred[0] for pred in probabilities.tolist()],
        }