cuda-python
cuda-python copied to clipboard
Update `pyproject.toml` for `uv`
Capturing feedbacks provided by @xiakun-lu offline.
The NCCL team noticed that uv sync complains nccl4py[cu12] and nccl4py[cu13] are incompatible (uv venv && uv pip install -e . works out of box):
[project.optional-dependencies]
cu12 = [
"nvidia-nccl-cu12",
"cuda-bindings~=12.0",
]
cu13 = [
"nvidia-nccl-cu13",
"cuda-bindings~=13.0",
]
With CPython 3.14.0:
Creating virtual environment at: .venv
× No solution found when resolving dependencies for split (markers: python_full_version >= '3.13'):
╰─:arrow_forward: Because nccl4py[cu13] depends on cuda-bindings>=13.0,<14.dev0 and nccl4py[cu12] depends on cuda-bindings>=12.0,<13.dev0, we can conclude that nccl4py[cu12] and nccl4py[cu13] are incompatible.
And because your project requires nccl4py[cu12] and nccl4py[cu13], we can conclude that your project's requirements are unsatisfiable.
It can be fixed by adding this configuration:
[tool.uv]
conflicts = [
[
{ extra = "cu12" },
{ extra = "cu13" },
]
]