I'm curious. How do people protocol their experiments? When I started, I used to just keep the cells but that lead to very long and impossible to parse Jupyter

I'm curious. How do people protocol their experiments? When I started, I used to just keep the cells but that lead to very long and impossible to parse Jupyter notebooks. I have since opted for keeping a journal.txt file in Atom where I write down hpyerparameter configurations, epochs run and results (for ML). But that feels a bit awkward as well.