handler.py

from transformers import AutoModelForCausalLM, AutoTokenizer
import torch
import os

class EndpointHandler:
    def __init__(self, path=""):
        self.tokenizer = AutoTokenizer.from_pretrained(path)
        self.model = AutoModelForCausalLM.from_pretrained(path, torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32)
        self.model.eval()

    def __call__(self, inputs: dict):
        prompt = inputs.get("inputs", "")
        if not prompt:
            return {"error": "No input provided."}
        input_ids = self.tokenizer(prompt, return_tensors="pt").input_ids
        outputs = self.model.generate(input_ids=input_ids, max_new_tokens=100)
        generated = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
        return {"generated_text": generated}