# CogView3-Plus-3B

<p style="text-align: center;">
  <div align="center">
  <img src=https://github.com/THUDM/CogView3/raw/main/resources/logo.svg width="50%"/>
  </div>
  <p align="center">
  <a href="README.md">📄 Read in English</a> | 
  <a href="https://huggingface.co/spaces/THUDM-HF-SPACE/CogView-3-Plus">🤗 Hugging Face Space | </a> 
  <a href="https://github.com/THUDM/CogView3">🌐 Github </a> | 
  <a href="https://arxiv.org/pdf/2403.05121">📜 arxiv </a>
</p>
<p align="center">
📍 前往<a href="https://chatglm.cn/main/gdetail/65a232c082ff90a2ad2f15e2?fr=osm_cogvideox&lang=zh"> 清言 </a> 和 <a href="https://open.bigmodel.cn/?utm_campaign=open&_channel_track_key=OWTVNma9"> API平台</a> 体验更大规模的商业版视频生成模型。
</p>

## 推理要求和模型介绍

该模型是 CogView3 的 DiT 版本图像生成模型，支持从 512 到 2048 范围内的图像生成。

+ 分辨率: 长宽均需满足 512px - 2048px 之间，均需被32整除。
+ 推理速度: 1s / step (在 A100 进行测试)
+ 精度: BF16 / FP32 (不支持FP16，会出现溢出导致纯黑图片)

## 显存消耗

我们在A100设备上对几个常见分辨率的显存消耗进行了测试，`batchsize=1, BF16`, 如下表所示：

| 分辨率         | enable_model_cpu_offload OFF | enable_model_cpu_offload ON |
|-------------|------------------------------|-----------------------------|
| 512 * 512   | 19GB                         | 11GB                        |
| 720 * 480   | 20GB                         | 11GB                        |
| 1024 * 1024 | 23GB                         | 11GB                        |
| 1280 * 720  | 24GB                         | 11GB                        |
| 2048 * 2048 | 25GB                         | 11GB                        |

## 快速开始

首先，确保从源代码安装`diffusers`库。

```shell
pip install git+https://github.com/huggingface/diffusers.git
```

接着，运行以下代码：

```python
from diffusers import CogView3PlusPipeline
import torch

pipe = CogView3PlusPipeline.from_pretrained("THUDM/CogView3-Plus-3B", torch_dtype=torch.float16).to("cuda")

# Open it for reduce GPU memory usage
pipe.enable_model_cpu_offload()
pipe.vae.enable_slicing()
pipe.vae.enable_tiling()

prompt = "A vibrant cherry red sports car sits proudly under the gleaming sun, its polished exterior smooth and flawless, casting a mirror-like reflection. The car features a low, aerodynamic body, angular headlights that gaze forward like predatory eyes, and a set of black, high-gloss racing rims that contrast starkly with the red. A subtle hint of chrome embellishes the grille and exhaust, while the tinted windows suggest a luxurious and private interior. The scene conveys a sense of speed and elegance, the car appearing as if it's about to burst into a sprint along a coastal road, with the ocean's azure waves crashing in the background."
image = pipe(
    prompt=prompt,
    guidance_scale=7.0,
    num_images_per_prompt=1,
    num_inference_steps=50,
    width=1024,
    height=1024,
).images[0]

image.save("cogview3.png")
```

更多内容以及下载 SAT 原始权重，请前往我们的 [github](https://github.com/THUDM/CogView3)。

## 引用

🌟 如果您发现我们的工作有所帮助，欢迎引用我们的文章，留下宝贵的stars

```
@article{zheng2024cogview3,
  title={Cogview3: Finer and faster text-to-image generation via relay diffusion},
  author={Zheng, Wendi and Teng, Jiayan and Yang, Zhuoyi and Wang, Weihan and Chen, Jidong and Gu, Xiaotao and Dong, Yuxiao and Ding, Ming and Tang, Jie},
  journal={arXiv preprint arXiv:2403.05121},
  year={2024}
}
```

## 模型协议

该模型基于 [Apache 2.0 License](LICENSE) 协议发布。