handler.py

import json
import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
from typing import Dict, List, Any

# Replace with actual GraniteMoeForCausalLM import if available
# from granitemoe import GraniteMoeForCausalLM

class EndpointHandler:
    def __init__(self, path: str = ""):
        self.tokenizer = AutoTokenizer.from_pretrained(path)
        self.model = AutoModelForCausalLM.from_pretrained(
            path,
            torch_dtype=torch.bfloat16,
            device_map="auto"
        )
        self.model.eval()

    def __call__(self, data: Dict[str, Any]) -> Dict[str, Any]:
        inputs = data.get("inputs", "")
        parameters = data.get("parameters", {})
        input_ids = self.tokenizer(inputs, return_tensors="pt").input_ids.to(self.model.device)
        max_length = parameters.get("max_length", 100)
        temperature = parameters.get("temperature", 1.0)
        top_p = parameters.get("top_p", 1.0)
        do_sample = parameters.get("do_sample", True)
        with torch.no_grad():
            outputs = self.model.generate(
                input_ids,
                max_length=max_length,
                temperature=temperature,
                top_p=top_p,
                do_sample=do_sample,
                pad_token_id=self.tokenizer.pad_token_id,
                eos_token_id=self.tokenizer.eos_token_id
            )
        generated_text = self.tokenizer.decode(outputs[0], skip_special_tokens=True)
        return {"generated_text": generated_text}