Apollo: Problem of Install Nvidia driver in docker libcudnn7

[c-xyli]

I was following the script of run_perception_module on_your_local_computer

In step 5 After download, install cuDNN by
sudo dpkg -i libcudnn7-dev_7.1.1.5-1+cuda8.0_amd64.deb
[Error:]

(Reading database ... 191546 files and directories currently installed.)
Preparing to unpack libcudnn7-dev_7.1.1.5-1+cuda8.0_amd64.deb ...
Unpacking libcudnn7-dev (7.1.1.5-1+cuda8.0) over (7.1.1.5-1+cuda8.0) ...
dpkg: dependency problems prevent configuration of libcudnn7-dev:
libcudnn7-dev depends on libcudnn7 (= 7.1.1.5-1+cuda8.0); however:
Package libcudnn7 is not installed.

dpkg: error processing package libcudnn7-dev (--install):
dependency problems - leaving unconfigured
Errors were encountered while processing:
libcudnn7-dev

I have tried the solution of changing other type of installation package / exporting environment path / change it to local machine rather than docker image, all not working.

Please help me with the correct installation setting of the libcudnn7

[c-xyli]

I found the solution with the help of my colleague. The problem comes from the lack of runtime library. However the version of 7.1.15 provided by the how_to manuscript has disappeared (or abandoned) in nvidia.

The solution is to downward version of libcudnn7, take an example of Download cuDNN v7.0.5 (Dec 5, 2017), for CUDA 8.0 list in Nvidia Driver Download Page, download the package of

• cuDNN v7.0.5 Runtime Library for Ubuntu16.04 (Deb)
• cuDNN v7.0.5 Developer Library for Ubuntu16.04 (Deb)

to the /apollo root folder and run the upcoming scripts

sudo dpkg -i libcudnn7*.deb

so as to install the runtime and developer library together.

[snuffysasa]

thanks this helped

[somubenz]

Change you current path to your downloaded folder then try to install.
It works for me.