comfyui安装速通

1.先去下个cuda装上，这一步全傻瓜式安装就不说了(后续的安装都是假定你的cuda版本为为cu128和win10环境)

https://developer.nvidia.com/cuda-downloads?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exe_local

PS.记得把谷歌的网页翻译关了，不然你选择版本点不动

2.nvcc --version确认一下你cuda是不是安装成功了(别做了半天发现白忙活)

3.装个UV，建议先装个scoop直接一键解决，就是scoop一键安装脚本要安装时改一下powershell的安全策略比较麻烦就不贴了(反正后续只要命令行能用uv就行，不关scoop什么事情)

4.在磁盘里找块喜欢的地方打开命令行到这uv init && uv venv(我怕给我覆盖掉就不演示了)

5.改一下pyproject.toml再uv sync(其实本来应该先改这个文件再创虚拟环境的，但是问题不大凑合能用就行)

查看代码

[project]
name = "cuda"
version = "0.1.0"
description = "Add your description here"
readme = "README.md"
requires-python = "==3.12.9"
dependencies = [
    "aiohttp>=3.11.8",
    "av>=14.1.0",
    "comfyui-frontend-package==1.17.11",
    "comfyui-workflow-templates==0.1.3",
    "einops>=0.8.1",
    "kornia>=0.7.1",
    "numpy>=1.25.0",
    "pillow>=11.2.1",
    "psutil>=7.0.0",
    "pydantic~=2.0",
    "pyyaml>=6.0.2",
    "safetensors>=0.4.2",
    "scipy>=1.15.2",
    "sentencepiece>=0.2.0",
    "soundfile>=0.13.1",
    "spandrel>=0.4.1",
    "tokenizers>=0.13.3",
    "torchsde>=0.2.6",
    "tqdm>=4.67.1",
    "transformers>=4.28.1",
    "yarl>=1.18.0",
    "torch>=2.7.0",
    "torchaudio>=2.6.0",
    "torchvision>=0.21.0",
]

[tool.uv.sources]
torch = [
  { index = "pytorch-cu128", marker = "sys_platform == 'linux' or sys_platform == 'win32'" },
]
torchvision = [
  { index = "pytorch-cu128", marker = "sys_platform == 'linux' or sys_platform == 'win32'" },
]

[[tool.uv.index]]
name="pytorch-cu128"
url = "https://download.pytorch.org/whl/cu128"
explicit = true

如果出现的全是是绿色的加号就说明pytorch环境就算完事了

6.找个风水宝地克隆一下comfyui项目(git默认你是配好的，因为说起来很麻烦)

git https://github.com/comfyanonymous/ComfyUI.git

 PS.这个克隆过来默认就有comfyui这层文件夹所以咱玩了个套娃

7.写个bat脚本一键启动，前面的call就是启动环境，路径填你pytorch环境的位置(你用uv venv的会生成.venv文件夹，那个activate.bat就是启动你虚拟环境的程序)

后面那一段就是uv的运行main.py，就和python -m main.py一样的意思，后面启动参数不喜欢不填也行

call "B:\Cuda\.venv\Scripts\activate" && uv run ./ComfyUI/main.py --auto-launch
pause