Cogview4 6B

🚀 CogView4-6B

CogView4-6B 是一个文本到图像的模型，可根据输入的文本生成相应的图像。该模型在多个基准测试中取得了优异的成绩，具有较高的图像生成质量和准确性。

🤗 模型空间 | 🌐 Github 仓库 | 📜 CogView3 论文

🚀 快速开始

首先，确保你从源代码安装 diffusers 库。

pip install git+https://github.com/huggingface/diffusers.git
cd diffusers
pip install -e .

然后，运行以下代码：

from diffusers import CogView4Pipeline

pipe = CogView4Pipeline.from_pretrained("THUDM/CogView4-6B", torch_dtype=torch.bfloat16)

# 开启以减少 GPU 内存使用
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=3.5,
    num_images_per_prompt=1,
    num_inference_steps=50,
    width=1024,
    height=1024,
).images[0]

image.save("cogview4.png")

✨ 主要特性

推理要求和模型介绍

• 分辨率：宽度和高度必须在 512px 到 2048px 之间，且能被 32 整除，并确保最大像素数不超过 2^21 像素。
• 精度：支持 BF16 / FP32（不支持 FP16，因为会导致溢出，生成全黑图像）

使用 BF16 精度和 batchsize=4 进行测试，内存使用情况如下表所示：

模型指标

我们在多个基准测试中进行了测试，并取得了以下分数：

DPG-Bench

GenEval

T2I-CompBench

中文文本准确性评估

📄 许可证

该模型根据 Apache 2.0 许可证 发布。

📚 引用

🌟 如果你觉得我们的工作有帮助，请考虑引用我们的论文并留下宝贵的 Star。

@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}
}