Stable Diffusion pytorch

### 使用PyTorch实现Stable Diffusion

为了使用PyTorch实现Stable Diffusion模型，需先安装必要的依赖库并配置环境。具体来说，可以按照以下方式操作：

对于GPU支持的选择适当版本的PyTorch进行安装[^2]：

pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/cu113

接着获取预训练好的Stable Diffusion权重文件[^1]:

import requests
from pathlib import Path

url = "https://huggingface.co/runwayml/stable-diffusion-v1-5/blob/main/v1-5-pruned.safetensors"
path = Path("v1-5-pruned.safetensors")

response = requests.get(url, stream=True)
if response.status_code == 200:
    with open(path, 'wb') as f:
        for chunk in response.iter_content(1024):
            f.write(chunk)
else:
    raise Exception(f"Failed to download model weights from {url}")

加载模型时需要注意，在基于PyTorch后端的情况下应将`jit_compile`参数设为`False`以确保兼容性[^3]:

from diffusers import StableDiffusionPipeline

pipeline = StableDiffusionPipeline.from_pretrained(
    pretrained_model_name_or_path="./",
    local_files_only=True,
    revision="fp16", 
    torch_dtype=torch.float16,
).to("cuda")

最后通过调用管道对象来生成图像:

prompt = "A beautiful landscape painting of mountains and rivers under sunset."
image = pipeline(prompt=prompt).images[0]
image.save("output.png")