This is almost certainly a transient issue, but I wanted present my workaround for anyone having this issue. As of RAPIDS 22.02, the wsl 2 installation of cudatoolkit doesn’t seem to be working.
This workaround may amount to a hack or may have unforeseen side-effects (I’m marginally informed and a bit out-of-depth)
from Ubuntu-20.04 wsl shell:
nano ~/.bashrc
at the bottom add in (you may want to double check this is the location of your libcuda.so files by searching in wsl)
LD_LIBRARY_PATH=/usr/lib/wsl/lib
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH
ctrl-x to save
then from wsl prompt:
sudo ldconfig
At this point, it’s probably a good idea to exit → exit and restart wsl. Then
jupyter notebook
Hope that helps!