TensorRT3.0 install error on Ubuntu 16.04 (Depends: cuda-cublas-9-0 but it is not installable)

I’m having trouble in installing TensorRT 3.0 on Ubuntu 16.04

I proceeded as described in official TensorRT 3.0 installation guide(.pdf)
However, after 3 command below:

  1. sudo dpkg -i nv-tensorrt-repo-ubuntu1604-ga-cuda9.0-trt3.0.4-20180208_1-1_amd64.deb
  2. sudo apt-get update
  3. sudo apt-get install tensorrt

There always error such as

========================== error ===========================
The following packages have unmet dependencies:
libnvinfer4 : Depends: cuda-cublas-9-0 but it is not installable
E: Unmet dependencies. Try ‘apt-get -f install’ with no packages (or specify a solution).

My current environment is
. Ubuntu 16.04 (PC)
. Titan Xp
. cuda 9.0
. cudnn 7.0.5

How can I install cuda-cublas-9-0 again? I already installed cuda 9.0.

(In my terminal)
$ nvcc --version

nvcc: NVIDIA ® Cuda compiler driver
Copyright © 2005-2017 NVIDIA Corporation
Built on Fri_Sep__1_21:08:03_CDT_2017
Cuda compilation tools, release 9.0, V9.0.176

It seems you need cuda-cublas-9-0.
What if you install cuda toolkit 9.1 from .run file? Also make sure you installed last driver from www.nvidia.com/drivers. Otherwise 9.0 toolkit installation from .run file will do.

What is the output of

$ dpkg-query -W | grep cuda-cublas

Reference: https://devtalk.nvidia.com/default/topic/1027490/gpu-accelerated-libraries/tensorrt-3-0-installation-with-cuda-toolkit-9-1-cublas-error/post/5232334/#5232334

Hi, Andrey1984
Thank you for your reply and Sorry for late.

In my system, there is no output when I type “dpkg-query -W | grep cuda-cublas” on terminal.

But I’m curious about your recommendation, ‘intalling cuda toolkit 9.1 from .run file’, because I’m using CUDA 9.0 on the ground of compatibility with Tensorflow r1.5

Is any other way that I can install cuda-cublas-9-0 ???

For reference:

Hello, I think Andrey is right.
I have the same issue on Ubuntu 16.04 with cuda 9.0 .
My previous CUDA environment was installed by .run file. So I remove it and reinstall CUDA by .deb file which fixed my problem.

consider the legacy releases from the versions different from the latest https://developer.nvidia.com/cuda-toolkit-archive