Closed raybellwaves closed 7 months ago
I left some thoughts on the PR: https://github.com/rapidsai/docs/pull/510#issuecomment-2065170581
Overall, RAPIDS aims to support the most stable and well-recognized installation methods like pip
and conda
. If we see increased interest in using other tools, we can definitely re-evaluate the RAPIDS documentation in the future. For now, I'll close this issue since I don't think there are short-term actions. However, we can continue conversation and I'll reference this issue in the future if we see interest grow as tools become more mainstream. We welcome issue reports from users of uv
, pixi
, etc. if there are ways that RAPIDS packages need to be improved.
Is your feature request related to a problem? Please describe.
Wonder if it's worth adding other install methods into the selector tool (https://docs.rapids.ai/install#selector) such as uv (https://github.com/astral-sh/uv) and pixi (https://pixi.sh/latest/)?
Describe the solution you'd like Perhaps additional boxes in the selector tool or comments around replacements e.g.
uv pip
forpip
.Describe alternatives you've considered I understand adding a few new package management tools who release often runs the risk that it may break things. You also need additional documentation on using them and troubleshooting. They also may not be mature enough that's they are worth including.
Additional context Recently saw ibis has pixi in their install docs https://ibis-project.org/install
Recently got a nice speed up (275%) using
uv pip
vspip
on google colab (https://colab.research.google.com/drive/1d3_gb_4y2Xgvl6xBBdknfwUV8Heq-nzJ)pixi may offer speed up downloading conda packages but I haven't worked a clean rapids install yet (https://github.com/prefix-dev/pixi/discussions/1029)