---
tags:
- image-to-image
- vae
---
MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization
[](https://arxiv.org/abs/2507.07997)
[Mingkai Jia](https://scholar.google.com/citations?user=fcpTdvcAAAAJ&hl=zh-CN)1,2, [Wei Yin](https://yvanyin.net/)2*§, [Xiaotao Hu](https://huxiaotaostasy.github.io/)1,2, [Jiaxin Guo](https://wrld.github.io/)3, [Xiaoyang Guo](https://xy-guo.github.io/)2
[Qian Zhang](https://scholar.google.com.hk/citations?hl=zh-CN&user=pCY-bikAAAAJ)2, [Xiao-Xiao Long](https://www.xxlong.site/)4, [Ping Tan](https://scholar.google.com/citations?user=XhyKVFMAAAAJ&hl=en)1
[HKUST](https://hkust.edu.hk/)1, [Horizon Robotics](https://en.horizon.auto/)2, [CUHK](https://cuhk.edu.hk/)3, [NJU](https://www.nju.edu.cn/)4
* Corresponding Author, § Project Leader
## 🚀News
- ```[August 2025]``` Achieve SOTA at paperwithcode leaderboards: Image Reconstruction on ImageNet and UHDBench.
- ```[August 2025]``` Released Inference Code
- ```[August 2025]``` Released [model zoo](https://huggingface.co/mkjia/MGVQ/tree/main).
- ```[August 2025]``` Released dataset for ultra-high-definition image reconstruction evaluation. Our proposed super-resolution image reconstruction [UHDBench dataset](https://huggingface.co/datasets/mkjia/UHDBench/tree/main) is released.
- ```[July 2025]``` Released [paper](https://arxiv.org/abs/2507.07997).
## 🔨TO DO LIST
- [ ] Training code.
- [ ] More demos.
- [x] Models & Evaluation code.
- [x] Huggingface models.
- [x] Release zero-shot reconstruction benchmarks.
## 🙈 Model Zoo
| Model | Downsample | Groups | Codebook Size | Training Data | Link |
|---|---|---|---|---|---|
|mgvq-f8c32-g4|8|4|32768|imagenet| [link](https://huggingface.co/mkjia/MGVQ/blob/main/mgvq_f8c32_g4.pt) |
|mgvq-f8c32-g8|8|8|16384|imagenet| [link](https://huggingface.co/mkjia/MGVQ/blob/main/mgvq_f8c32_g8.pt) |
|mgvq-f16c32-g4|16|4|32768|imagenet| [link](https://huggingface.co/mkjia/MGVQ/blob/main/mgvq_f16c32_g4.pt) |
|mgvq-f16c32-g8|16|8|16384|imagenet| [link](https://huggingface.co/mkjia/MGVQ/blob/main/mgvq_f16c32_g8.pt) |
|mgvq-f16c32-g4-mix|16|4|32768|mix| [link](https://huggingface.co/mkjia/MGVQ/blob/main/mgvq_f16c32_g4_mix.pt) |
|mgvq-f32c32-g8-mix|32|8|16384|mix| [link](https://huggingface.co/mkjia/MGVQ/blob/main/mgvq_f32c32_g8_mix.pt) |
## 🔑 Quick Start
### Installation
```bash
git clone https://github.com/MKJia/MGVQ.git
cd MGVQ
pip3 install requirements.txt
```
### Download models
Download the pretrained models from our [model zoo](https://huggingface.co/mkjia/MGVQ/tree/main) to your `/path/to/your/ckpt`.
### Data Preparation
Try our UHDBench dataset on [huggingface](https://huggingface.co/datasets/mkjia/UHDBench/tree/main) and download to your `/path/to/your/dataset`.
### Evaluation on Reconstruction
Remember to change the paths of `ckpt` and `dataset_root`, and make sure you are evaluating the expected `model` on `dataset`.
```bash
cd evaluation
python3 eval_recon.sh
```
### Generation Demo&Evaluation
You can download the pretrained GPT model for generation on [huggingface](https://huggingface.co/datasets/mkjia/MGVQ/blob/main/MGVQ_GPT_XXL.pt), and test it with our `mgvq-f16c32-g4` [tokenizer model](https://huggingface.co/mkjia/MGVQ/blob/main/mgvq_f16c32_g4.pt) for demo image sampling. Remember to change the paths of `gpt_ckpt` and `vq_ckpt`.
```
cd evaluation
python3 demo_gen.sh
```
We also provide our .npz file on [huggingface](https://huggingface.co/datasets/mkjia/MGVQ/blob/main/GPT_XXL_300ep_topk_12.npz) sampled by `sample_c2i_ddp.py` for evaluation.
```
cd evaluation
python3 evaluator.py /path/to/your/VIRTUAL_imagenet256_labeled.npz /path/to/your/GPT_XXL_300ep_topk_12.npz
```
## 🗄️Demos
- 🔥 Qualitative reconstruction images with $16$ x downsampling on $2560$ x $1440$ UHDBench dataset.
- 🔥 Qualitative class-to-image generation of Imagenet. The classes are dog(Golden Retriever and Husky), cliff, and bald eagle.
- 🔥 Reconstruction evaluation on 256×256 ImageNet benchmark.
- 🔥 Zero-shot reconstruction evaluation with a downsample ratio of 16 on 512×512 datasets.
- 🔥 Zero-shot reconstruction evaluation with a downsample ratio of 16 on 2560×1440 datasets.
## 🗄️Demos
## 📌 Citation
If the paper and code from `MGVQ` help your research, we kindly ask you to give a citation to our paper ❤️. Additionally, if you appreciate our work and find this repository useful, giving it a star ⭐️ would be a wonderful way to support our work. Thank you very much.
```bibtex
@article{jia2025mgvq,
title={MGVQ: Could VQ-VAE Beat VAE? A Generalizable Tokenizer with Multi-group Quantization},
author={Jia, Mingkai and Yin, Wei and Hu, Xiaotao and Guo, Jiaxin and Guo, Xiaoyang and Zhang, Qian and Long, Xiao-Xiao and Tan, Ping},
journal={arXiv preprint arXiv:2507.07997},
year={2025}
}
```
## License
This repository is under the MIT License. For more license questions, please contact Mingkai Jia (mjiaab@connect.ust.hk) and Wei Yin (yvanwy@outlook.com).