app.py

import os
import torch
from fastapi import FastAPI, Request
from fastapi.responses import FileResponse, JSONResponse
from transformers import AutoTokenizer, AutoModelForCausalLM
from peft import PeftModel
import uvicorn

# ------------------------------
# Config
# ------------------------------
BASE_MODEL = "deepseek-ai/deepseek-coder-6.7b-base"
ADAPTER_PATH = "Agasthya0/colabmind-coder-6.7b-ml-qlora"

# ------------------------------
# Load Model + Tokenizer
# ------------------------------
print("🚀 Loading tokenizer...")
tokenizer = AutoTokenizer.from_pretrained(BASE_MODEL)

print("🧠 Loading base model in 4-bit...")
model = AutoModelForCausalLM.from_pretrained(
    BASE_MODEL,
    load_in_4bit=True,
    device_map="auto",
    torch_dtype=torch.float16
)

print("🔗 Attaching LoRA adapter...")
model = PeftModel.from_pretrained(model, ADAPTER_PATH)

# ------------------------------
# Inference Function
# ------------------------------
def generate_code(prompt: str):
    if not prompt.strip():
        return "⚠️ Please enter a prompt."

    inputs = tokenizer(prompt, return_tensors="pt").to(model.device)
    with torch.no_grad():
        outputs = model.generate(
            **inputs,
            max_new_tokens=512,
            temperature=0.2,
            top_p=0.9,
            do_sample=True,
            pad_token_id=tokenizer.eos_token_id
        )
    return tokenizer.decode(outputs[0], skip_special_tokens=True)

# ------------------------------
# FastAPI App
# ------------------------------
app = FastAPI()

@app.get("/")
def serve_frontend():
    return FileResponse("index.html")

@app.post("/run/predict")
async def predict(request: Request):
    data = await request.json()
    prompt = data.get("data", [""])[0]
    output = generate_code(prompt)
    return JSONResponse({"data": [output]})

# ------------------------------
# Run (for local debugging, Spaces ignores this)
# ------------------------------
if __name__ == "__main__":
    uvicorn.run(app, host="0.0.0.0", port=7860)