docker: Error response from daemon: Unknown runtime specified nvidia.

Describe the bug

Docker does not run with the runtime parameter nvidia from the command: make start-gpu

What is the current behavior? Using the make start-gpu I get the following error:

docker: Error response from daemon: Unknown runtime specified nvidia.
See 'docker run --help'.
make: *** [Makefile:50: start-gpu] Error 125

The whole dump of the error message is available from here.

If the current behavior is a bug, please provide the steps to reproduce.

I am not 100% sure how it could be reproduced but this thread might provide some insights.

Expected behavior

The docker must be run and all requirements must be installed without a problem via the installation/build script prior to line 50 in the Makefile. It appears that installing nvidia-docker2 is what is required as mentioned here. And the installation could be followed from here.

Screenshots

NA

Other relevant information: poetry version: 1.1.11 python version: 3.8.5 Operating System: 5.8.0-38-generic #43~20.04.1-Ubuntu Additional tools: NA

Additional context

NA

[Hartorn]

Yeah exactly @Optimox On the host computer, you need NVIDIA-container-toolkit, or NVIDIA docker2 (quite the same)

Inside the container, Cuda must be installed (and it is)

But as far as i know, you need the packages around container toolkit

See https://docs.nvidia.com/datacenter/cloud-native/container-toolkit/install-guide.html

[Optimox]

When installing with pip install pytorch-tabnet you get the latest stable release.

If you wish to install the develop branch you can do this: !pip install "git+https://github.com/dreamquark-ai/tabnet.git@develop#egg=pytorch_tabnet" --upgrade