Set up DL environment and versions/dependencies

When the server is available, the next step is to set up a compatible and robust environment that supports the DL platform.

The most commonly used system is Ubuntu 20.04 LTS, Ubuntu 18.04 LTS, and 16.04 LTS. What's need to be managed include: PyTorch, TensorFlow, CUDA, cuDNN, and NVIDIA drivers.

Lambda stack

The easiest way is to use lambda stack directly: https://lambdalabs.com/lambda-stack-deep-learning-software#server-installation

It includes the setup with python virtual environments, on ubuntu 20.04/18.04 servers, ubuntu 16.04, or some docker container.

Others:

if you already have a system set up, and here is some information that might be helpful:

• conda takes care of the version compatibility, even on cudnn. preferred it over pip.

• be cautious about CUDA 10.1 and PyTorch.