unsatisfied condition: cuda >= 10.1 about gpu-jupyter

[mathematicalmichael]

I'd guess it's because of the cuda mismatches. I think the version is specified in the docker build, not auto-selected based on the version on your host.

[ChristophSchranz]

Thanks for sharing so much details. Yes, IMHO @mathematicalmichael is right, if you have installed cuda9 there seem to be problems to install a driver for cuda>=10.1 on top of it.
Have you already tried to decrease the CUDA driver version in src/Dockerfile.header? Please let me know if this works out for you!

You might have to adapt some lines in src/Dockerfile.gpulibs too, as e.g. Pytorch requires a specific cuda version:

gpu-jupyter/src/Dockerfile.gpulibs

Lines 19 to 22 in 1c8b32a

	RUN conda install --quiet --yes \
	pytorch \
	torchvision \
	cudatoolkit=10.1 -c pytorch

[ChristophSchranz]

Closing the issue as the cause is clear and there were no comments anymore.