I’m no ML expert so take what I say with a grain of salt.

Two resources that might be useful are AWS’ SageMaker documentation and the Machine Learning Engineering book by Andriy Burkov. This book doesn’t really go into detail on logging though. One way to evaluate a model is to run a SageMaker processing job that saves the performance metrics in a json file in S3 somewhere. More info on processing jobs: https://docs.aws.amazon.com/sagemaker/latest/dg/processing-j... . AWS has various services for logging which you can look into. This will mostly apply to orgs using AWS, but it might give a sense of how things can be done more generally.

For now, the plan is to move from Jupyter back to a text editor. Jupyter is very forgiving of mistakes. The model didn't work? Change some parameters and rerun the training cell. This is amazing for new folks, who are being bombarded by new information, and (it sounds like) for experienced folks who have already developed great habits around ML projects. But I think intermediate folks need a little friction to help hammer home why best practice is best practice.

I'm hoping the text editor + project directory approach helps force ML projects away from a single file and towards some sort of codified project structure. Sometimes it just feels like there's too much information in a file and it becomes hard to assign it to a location mentally (a bit like reading a physical copy of a tough book vs a kindle copy). Any advice or thoughts on this would be appreciated!