Here’s another example of a basic PyTorch workflow that goes a bit further than the useful baseline example provided by @fnands
It includes data exploration, feature selection (i.e., choosing which months of data to use), custom transforms, and vegetation indices commonly used in remote sensing. I chose to avoid frameworks like PyTorch Lightning so that the workflow is more explicit.
There are plenty of areas for optimization, but I don’t plan to make further improvements so my hope is that others can find something useful from the work. Feel free to use any of the code, and provide attribution if your karma needs a boost 