BrunoScience / BrunoScience.github.io

The best that most of us can hope to achieve in science is simply to misunderstand at a deeper level.
https://BrunoScience.github.io
Apache License 2.0
0 stars 1 forks source link

The NVIDIA documentation on WSL 2 is ... full of JUNK ... clues, but a LOT is wrong. #24

Open MarkBruns opened 1 year ago

MarkBruns commented 1 year ago

The [Wikipedia page on CUDA] is perhaps more useful than the NVIDIA documentation ... FOR BEGINNERS using things like Ubuntu on WSL 2

Generally, the NVIDIA documentation which is FULL of information which no longer applies because of changes that have happened ... the NVIDIA documentation [particularly for things that were introduced two years ago and are now several revision behind] is FULL of problems ... it has clues ... but not real information.

Believe it or not ... the Microsoft WSL documentation on GPU compute was sort of useful ... except that installing Docker Desktop [for Windows 11] is a really bad idea.

But even the Microsoft WSL documentation on GPU compute actually did the following command right.

docker run --gpus all -it --shm-size=1g --ulimit memlock=-1 --ulimit stack=67108864 nvcr.io/nvidia/tensorflow:22.10.1-tf2-py3

Might have to see what ExplainShell.com has to say about it ... not much ... it just says "docker" is a "System tray for KDE3/GNOME2 docklet applications" ... so interpretting docker commands is largely a matter of looking at the NVIDIA documentation and developer forums, the Docker documentation, the WSL GPU documentations, RAPIDS documentation, StackOverflow, Ubuntu forums and easily 100 other sources of clues.