Search code examples
pythonazuredockernvidiacvxpy

How to install CVXPY / SCS to use with GPU?

All the dependencies are installed in a docker container, but when I run the example with GPU=True it returns "Status: Unbounded" and with GPU=False it works ok. The scs-python GPU tests return "Status: Unbounded". Also, nvidia-smi displays GPU usage. What might be the issue?

Dockerfile: https://gist.github.com/Cerebrock/eef5d33ece1782d3efdf0b35d29fb48e

Example: https://gist.github.com/Cerebrock/885023ec2faf1d1c6bf9b1a4df12df36

OS: Ubuntu 16.04 on Azure, image NVIDIA GPU Cloud image, NVIDIA Tesla P100

Solution

  • I had the same problem and I was able to fix it building SCS-PYTHON like:

    python3 setup.py install --scs --gpu --int

    If you look at the setup.py help it says:

    GPU code always uses 32 bit ints

    python3 setup.py install --help