hf_backend.py

# hf_backend.py
import time, logging
from typing import Any, Dict, AsyncIterable

import torch
from transformers import AutoTokenizer, AutoModelForCausalLM
from backends_base import ChatBackend, ImagesBackend
from config import settings

logger = logging.getLogger(__name__)

try:
    import spaces
    from spaces.zero import client as zero_client
except ImportError:
    spaces, zero_client = None, None

# --- Model setup (CPU-safe load, real inference on GPU only) ---
MODEL_ID = settings.LlmHFModelID or "Qwen/Qwen2.5-1.5B-Instruct"
logger.info(f"Preloading tokenizer for {MODEL_ID} on CPU (ZeroGPU safe)...")

tokenizer, model, load_error = None, None, None
try:
    tokenizer = AutoTokenizer.from_pretrained(MODEL_ID, trust_remote_code=True, use_fast=False)
    model = AutoModelForCausalLM.from_pretrained(
        MODEL_ID,
        torch_dtype=torch.float32,   # dummy dtype for CPU preload
        trust_remote_code=True,
    )
    model.eval()
except Exception as e:
    load_error = f"Failed to load model/tokenizer: {e}"
    logger.exception(load_error)


# ---------------- Chat Backend ----------------
class HFChatBackend(ChatBackend):
    async def stream(self, request: Dict[str, Any]) -> AsyncIterable[Dict[str, Any]]:
        if load_error:
            raise RuntimeError(load_error)

        messages = request.get("messages", [])
        prompt = messages[-1]["content"] if messages else "(empty)"
        temperature = float(request.get("temperature", settings.LlmTemp or 0.7))
        max_tokens = int(request.get("max_tokens", settings.LlmOpenAICtxSize or 512))

        rid = f"chatcmpl-hf-{int(time.time())}"
        now = int(time.time())

        if not spaces:
            raise RuntimeError("ZeroGPU (spaces) is required but not available!")

        # --- Inject X-IP-Token into global headers ---
        x_ip_token = request.get("x_ip_token")
        if x_ip_token and zero_client:
            zero_client.HEADERS["X-IP-Token"] = x_ip_token
            logger.debug("Injected X-IP-Token into ZeroGPU headers")

        # --- Define the GPU-only inference function ---
        @spaces.GPU(duration=120)
        def run_once(prompt: str) -> str:
            device = "cuda"   # force CUDA
            dtype = torch.float16

            model.to(device=device, dtype=dtype).eval()
            inputs = tokenizer(prompt, return_tensors="pt").to(device)

            with torch.inference_mode(), torch.autocast(device_type=device, dtype=dtype):
                outputs = model.generate(
                    **inputs,
                    max_new_tokens=max_tokens,
                    temperature=temperature,
                    do_sample=True,
                )
            return tokenizer.decode(outputs[0], skip_special_tokens=True)

        try:
            text = run_once(prompt)
            yield {
                "id": rid,
                "object": "chat.completion.chunk",
                "created": now,
                "model": MODEL_ID,
                "choices": [
                    {"index": 0, "delta": {"content": text}, "finish_reason": "stop"}
                ],
            }
        except Exception:
            logger.exception("HF inference failed")
            raise


# ---------------- Stub Images Backend ----------------
class StubImagesBackend(ImagesBackend):
    """
    Stub backend for images since HFChatBackend is text-only.
    Returns a transparent 1x1 PNG placeholder.
    """
    async def generate_b64(self, request: Dict[str, Any]) -> str:
        logger.warning("Image generation not supported in HF backend.")
        return (
            "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR4nGP4BwQACfsD/etCJH0AAAAASUVORK5CYII="
        )