C03: Setting up NVidia GPU using class DevContainers

In this video, I discuss using NVidia GPU compute from inside of a Docker class DevContainer. Docker and the remote development extensions come ready to perform gpu passthrough out of the box now. Using containers and GPU passthrough if you have an NVidia GPU, is a good way to organize and work on different projects on the same system needing access to GPU compute. In this video I show how to enable GPU passthrough for your docker container when starting it. Also the correct way to install the tensorflow/keras packages for the container. We also look at running cpu and gpu compute in an iPython notebook, and a few other tips and tricks. Resources: East Texas A&M University: https://www.tamuc.edu/ ETAMU Department of Computer Science: https://www.tamuc.edu/computer-science-and-information-systems/ 00:00 Introduction 00:46 Some advice on getting NVidia resources for your own hardware 05:05 Looking at .devcontainer/Dockerfile and .devcontainer/devcontainer.json GPU setup 12:39 Modify and rebuild devcontainer to enable GPU passthrough and install tensorflow with cuda support 18:31 Howto test tensorflow gpu access from an iPython notebook 25:10 Summary