CUDA version mismatch in docker enviroment

[xsir317]

After i build the docker image( run docker build . ) , i found that the installed CUDA is 9.0 ,but the driver is 10.0.

and the df_console.py failed with "CUDA driver version is insufficient for CUDA runtime version"

in df_console.py , print('CUDA version', torch.version.cuda) # this is 10.0

i checked conda list , there is pytorch-nightly 1.0.0.dev20190226 py3.7_cuda10.0.130_cudnn7.4.2_0 pytorch

and the CUDA version in Dockerfile is ENV CUDA_VERSION 9.0.176

so this could be a mismatch . is this a bug?

[bytesoftly]

Is a mismatch but I found installing nvidia driver 410 (i.e. for CUDA 10) on my host machine, running with nvidia-docker works.

[JeremyCollinsMPI]

I'm having the same problem, it would be good if the docker file could be fixed. I'm trying to fix it at the moment. @bytesoftly Could you possibly provide a docker file for your solution?