Code configuration

YAML configuration for training process

Visit Training and Dataset Generation for explicit documentation of configuration options and examples to follow.

Specific configurations to keep in mind:

For file_train,

  • If you're following the training steps, split the training and testing data 50/50.

  • If you're using pre-trained model weights, make sure at least one file is in the training list.

For files_train and files_test , you can exclude the file extension (especially if there's a file for each band), as the code will append one based on the reader_type selected.

  • For example, if you have 5 NetCDF files corresponding to one day:

SNPP_VIIRS.20250623.L3m.DAY.RRS.Rrs_410.4km.NRT.nc
SNPP_VIIRS.20250623.L3m.DAY.RRS.Rrs_443.4km.NRT.nc
SNPP_VIIRS.20250623.L3m.DAY.RRS.Rrs_486.4km.NRT.nc
SNPP_VIIRS.20250623.L3m.DAY.RRS.Rrs_551.4km.NRT.nc
SNPP_VIIRS.20250623.L3m.DAY.RRS.Rrs_671.4km.NRT.nc

  • Instead of including them all in the YAML, simply include:

SNPP_VIIRS.20250623.L3m.DAY.

More specific readers and reader keyword arguments can be found in the above link.

E.g. configurations

PreviousTraining Dataset PreprocessingNextTraining Command

Last updated