USGS-R / river-dl

Deep learning model for predicting environmental variables on river systems
Creative Commons Zero v1.0 Universal
21 stars 14 forks source link

Need to remove hold-out reach observations from calculating metrics for train and val partitions #185

Closed SimonTopp closed 2 years ago

SimonTopp commented 2 years ago

After #178, we now can mask out segments for validation and testing, but since those segment id's and dates still exist in prepped.npz, when we join them to the observation file during combine_metrics they'll wind up included in the training and validation performance metrics. TLDR, we need to pass those hold-out segments to combine_metrics and mask them out so they're only included in the metrics for the appropriate partition.

SimonTopp commented 2 years ago

Addressed in #190