# LLaVA-NeXT: A Strong Zero-shot Video Understanding Model

## 论文

`LLaVA-NeXT: A Strong Zero-shot Video Understanding Model`

* https://llava-vl.github.io/blog/2024-04-30-llava-next-video/

## 模型结构

参考[README.md](../README.md)

## 算法原理

参考[README.md](../README.md)

## 数据集

无

## 训练

无

## 推理

### 原生

```bash
cd ..

bash scripts/video/demo/video_demo.sh /path/to/LLaVA-NeXT-Video-7B-DPO vicuna_v1 32 2 average no_token True playground/demo/xU25MMA2N4aVtYay.mp4
```

### hf

```bash
python inference_hf.py
```

## result

![alt text](readme_imgs/result.png)

### 精度

无

## 应用场景

参考[README.md](../README.md)

## 预训练权重

|model|url|
|:---:|:---:|
|LLaVA-NeXT-Video-7B-DPO|[hf](https://huggingface.co/lmms-lab/LLaVA-NeXT-Video-7B-DPO) \| [SCNet]() |
|LLaVA-NeXT-Video-7B-hf|[hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-hf) \| [SCNet]() |
|LLaVA-NeXT-Video-7B-32K-hf|[hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-32K-hf) \| [SCNet]() |
|LLaVA-NeXT-Video-7B-DPO-hf|[hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-7B-DPO-hf) \| [SCNet]() |
|LLaVA-NeXT-Video-34B-hf|[hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-34B-hf) \| [SCNet]() |
|LLaVA-NeXT-Video-34B-DPO-hf|[hf](https://huggingface.co/llava-hf/LLaVA-NeXT-Video-34B-DPO-hf) \| [SCNet]() |

模型下载后保存至`ckpts`（需自行创建）.

## 源码仓库及问题反馈

参考[README.md](../README.md)

## 参考资料

* https://github.com/LLaVA-VL/LLaVA-NeXT/blob/main/docs/LLaVA-NeXT-Video.md
* https://llava-vl.github.io/blog/2024-04-30-llava-next-video/