MNIST training starter kit with Ray, Lightning and PyTorch

Some time ago, I shared a starter kit for reinforcement learning research with Ray and RLlib. The following is a productive tech stack for doing deep learning research.

• Ray
• PyTorch Lightning
• PyTorch
• TensorBoard
• JupyterLab

What makes this stack challenging to use is that the constituent packages are updated frequently. In my experience, if I just install the latest versions of each package, sometimes the stack runs out-of-the-box and sometimes not.

Here, I share a working snapshot of the tech stack. There is an example MNIST training program and a requirements.txt for reproducting the run-time environment.

Python 3.8.16 is used.