# Qwen Image Edit Lite ZeroGPU Demo

This repository is prepared for a Hugging Face ZeroGPU Space.

## What changed

- The source app was migrated into this repo.
- `benchmark/` was removed.
- Base models are configured as Hugging Face download URLs in `config.ini`.
- LoRAs are also configured as Hugging Face URLs and cached locally under `models/loras`.
- Gated or private LoRAs are authenticated through the `HF_TOKEN` Space secret.
- `attn_mode` is fixed to `xformers`.

## Space entrypoint

- `app.py`
- `requirements.txt`
- `README.md` front matter for Gradio Spaces

## Token options

- Register `HF_TOKEN` as a Space secret