编辑 pyproject.toml 持久化
将以下内容追加到 pyproject.toml ，直接运行 uv add torch torchvision 即可安装对应的 CUDA 版本。
适合多人协同开发，他人只需 uv sync 即可完美同步环境。
toml
[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
快速安装
快速安装 cuda 版本的 torch，无法添加到 pyproject.toml 。uv sync 就没了。cu128 可换为 auto 以自动指定。
bash
uv pip install torch torchvision --torch-backend=cu128
Refs: https://docs.astral.sh/uv/guides/integration/pytorch/#the-uv-pip-interface
