nvcc issue

[GeorvityLabs]

@iot-salzburg ,

i tried to run nvcc --version from within the notebook , but it shows it isn't there.
any tips on how to fix this issue?

[Steven Ding]

same - we have to download and install. did it get deleted during the build?

[Christoph]

Hi,
unfortunately, I can't reproduce this issue.
nvcc is installed within this lines:

gpu-jupyter/src/Dockerfile.gpulibs

Lines 43 to 48 in 5d8443c

 # reinstall nvcc with cuda-nvcc to install ptax 
 USER $NB_UID 
 RUN conda install -c nvidia cuda-nvcc -y && \ 
     conda clean --all -f -y && \ 
     fix-permissions $CONDA_DIR && \ 
     fix-permissions /home/$NB_USER 

Note that this Dockerfile part is stacked together into .build/Dockerfile.
The reason why nvcc is installed this way is discussed in #106.

• Does the error still occur after rebooting?
• What's the output of nvcc --version on the host machine?
• Can you manually install the lines and does it work then?
• What image version do you use? Does it also occur with v1.5_cuda-11.6_ubuntu-20.04

[Christoph]

Closing as there is no response and the issue seems to not occur anymore on recent versions.