Pycuda installation 2018.1.1 - Cannot install as per instructions - Error

Autonomous Machines Jetson & Embedded Systems Jetson TX2
1

Hi,

Ubuntu 16.04, Jetson TX2.
Host: Ubuntu 16.04
cuda: 9.2
Python: python3.5

I have installed Tensor RT, Cuda 9.2, and tensor flow. I installed pycuda as per instructions in Installing PyCUDA on Linux - Andreas Klöckner's Former Wiki . The following is the issue:

  1. I have already installed numpy using pip3 command
  2. When I typed the command su -c “python distribute_setup.py” I get an error stating there is no file in that name.

Alternatively, I tried running the ‘setup.py.’ It started throwing error saying it cannot find ‘numpy’ module.

Note: I have created a CUDA_ROOT ajavascript:void(0);nd PATH variable to cuda 9.2’s root directory. In my case, it is /usr/local/cuda-9.2/bin.

Any help to solve this problem will be highly appreciated! Thanks

2

Moving this to the Jetson TX2 forum for better visibility.

3

Hi,

We don’t have a CUDA 9.2 release for Jetson platform.
For TX2, please use JetPack3.3 to setup the CUDA, cuDNN and TensorRT environment:
[url]https://developer.nvidia.com/embedded/jetpack-3_3[/url]

More, please noticed that TensorRT python API doesn’t support Jetson TX2.
Thanks.