search

USGS-R / river-dl

Deep learning model for predicting environmental variables on river systems
Creative Commons Zero v1.0 Universal
21 stars 15 forks source link

alternative way of no-pretrain experiment #144

Closed jsadler2 closed 2 years ago

jsadler2 commented 2 years ago

@jdiaz4302 this is another way of setting up the experiment which, in my opinion, would be a bit clearer and cleaner. It basically makes two changes

  1. the merging code is in the actual Snakefile. This is good because, if possible, it's nice not to have to touch the .py files in river-dl for a single experiment
  2. it introduces two "branches" (PBIn and pretrain) by having another wildcard.

Here's what the dag looks like: image

Happy to talk through it or answer questions if you have them.

jsadler2 commented 2 years ago

PS. even though I generated the DAG, I did not run or test the actual code. Just wanted to show you how you could set up the experiment in Snakemake. I think you'd have to do some modification on the merging function.