I have Cuda 10.0 installed in a Tensorflow GPU container in Docker (image tag is tensorflow/tensorflow1.15.2-gpu-py3)
root@d7284da3dc5c://tmp/external# nvcc --versionnvcc: NVIDIA (R) Cuda compiler driverCopyright (c) 2005-2018 NVIDIA CorporationBuilt on Sat_Aug_25_21:08:01_CDT_2018Cuda compilation tools, release 10.0, V10.0.130However, when I run dpkg -i libcudnn7_7.6.5.32-1+cuda10.0_amd64.deb on the debian package I found for the most recent 10.0 on the Nvidia developer site for Ubuntu 18.04 I get the following:
root@d7284da3dc5c://tmp/external# dpkg -i libcudnn7_7.6.5.32-1+cuda10.0_amd64.deb(Reading database ... 39767 files and directories currently installed.)Preparing to unpack libcudnn7_7.6.5.32-1+cuda10.0_amd64.deb ...Unpacking libcudnn7 (7.6.5.32-1+cuda10.0) over (7.6.5.32-1+cuda10.0) ...Setting up libcudnn7 (7.6.5.32-1+cuda10.0) ...Processing triggers for libc-bin (2.27-3ubuntu1) .../sbin/ldconfig.real: /usr/lib/x86_64-linux-gnu/libcuda.so.1 is not a symbolic linkNow, I have seen people using tar gzipped versions of the cudNN files but I'm not exactly sure where to get those from the nvidia developer site. I'm not sure if that would help me get the files and copy them over as I have seen in other articles.
However, with my method it doesn't seem like anything installs after I run the dpkg -i command but maybe I am just overlooking some things. Any idea what could be a solution to this problem?