| from pathlib import Path |
| import zipfile |
|
|
| base = Path("/mnt/data/nova_ai_chatbot_pro") |
| base.mkdir(exist_ok=True) |
|
|
| app = r'''import os |
| import threading |
| import gradio as gr |
| import torch |
| from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer |
|
|
| MODEL_ID = os.getenv("MODEL_ID", "Qwen/Qwen2.5-0.5B-Instruct") |
| SYSTEM_PROMPT = ( |
| "You are Nova, a helpful, thoughtful AI assistant. " |
| "Be clear, friendly, accurate, and concise. If you are unsure, say so." |
| ) |
|
|
| _tokenizer = None |
| _model = None |
| _load_error = None |
|
|
|
|
| def get_model(): |
| global _tokenizer, _model, _load_error |
| if _model is not None: |
| return _tokenizer, _model |
| if _load_error: |
| raise RuntimeError(_load_error) |
| try: |
| _tokenizer = AutoTokenizer.from_pretrained(MODEL_ID) |
| kwargs = {"low_cpu_mem_usage": True} |
| if torch.cuda.is_available(): |
| kwargs["torch_dtype"] = torch.float16 |
| kwargs["device_map"] = "auto" |
| else |
