handler.py

from transformers import AutoTokenizer, AutoModelForCausalLM
import torch

class ModelHandler:
    def __init__(self):
        self.initialized = False

    def initialize(self, model_dir: str):
        self.tokenizer = AutoTokenizer.from_pretrained(model_dir)
        self.model = AutoModelForCausalLM.from_pretrained(model_dir, torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32)
        self.model.eval()
        if torch.cuda.is_available():
            self.model.to("cuda")
        self.initialized = True

    def predict(self, inputs: dict):
        if not self.initialized:
            raise RuntimeError("Model not initialized")

        messages = inputs.get("messages", [])
        max_tokens = inputs.get("max_tokens", 512)
        temperature = inputs.get("temperature", 0.7)

        # Convert OpenAI-style messages into a single prompt
        prompt = self._build_prompt(messages)

        # Tokenize
        input_ids = self.tokenizer(prompt, return_tensors="pt").input_ids
        if torch.cuda.is_available():
            input_ids = input_ids.to("cuda")

        # Generate
        output_ids = self.model.generate(
            input_ids,
            max_new_tokens=max_tokens,
            temperature=temperature,
            do_sample=True,
            pad_token_id=self.tokenizer.eos_token_id,
        )

        response = self.tokenizer.decode(output_ids[0], skip_special_tokens=True)

        # Return just the newly generated portion
        generated_text = response[len(prompt):].strip()

        return {
            "id": "chatcmpl-fakeid",
            "object": "chat.completion",
            "choices": [
                {
                    "index": 0,
                    "message": {
                        "role": "assistant",
                        "content": generated_text
                    },
                    "finish_reason": "stop"
                }
            ],
            "model": "your-model-id",
        }

    def _build_prompt(self, messages):
        prompt = ""
        for msg in messages:
            role = msg["role"]
            content = msg["content"]
            if role == "user":
                prompt += f"User: {content}\n"
            elif role == "assistant":
                prompt += f"Assistant: {content}\n"
        prompt += "Assistant:"
        return prompt