handler.py

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


class EndpointHandler:
    def __init__(self, path=""):
        self.tokenizer = AutoTokenizer.from_pretrained(path)
        self.model = AutoModelForSequenceClassification.from_pretrained(path)
        self.model.eval()

    def __call__(self, data: Dict[str, Any]) -> List[Dict]:
        inputs_text = data.pop("inputs", data)
        if isinstance(inputs_text, str):
            inputs_text = [inputs_text]

        encoded = self.tokenizer(
            inputs_text,
            return_tensors="pt",
            padding=True,
            truncation=True,
            return_token_type_ids=False,
        )

        with torch.no_grad():
            logits = self.model(**encoded).logits

        scores = torch.softmax(logits, dim=-1)
        id2label = self.model.config.id2label

        results = []
        for row in scores:
            label_scores = sorted(
                [{"label": id2label[i], "score": float(row[i])} for i in range(len(row))],
                key=lambda x: -x["score"],
            )
            results.append(label_scores)

        return results if len(results) > 1 else results[0]