Short description explaining the high-level reason for the new issue.
An LSTM can be trained with any combination of forcings, so don't hard code them. Let the BMI determine the forcings to include from the training configuration. But we can still initialize everything, and just not include them in the input tensor.
Current behavior
def initialize_forcings(self):
#------------------------------------------------------------
if 'total_precipitation' in self.cfg_train['dynamic_inputs']:
self.total_precipitation = 0
#------------------------------------------------------------
if 'temperature' in self.cfg_train['dynamic_inputs']:
self.temperature = 0
Expected behavior
for forcing_name in self.cfg_train['dynamic_inputs']:
setattr(self, forcing_name, 0)
Short description explaining the high-level reason for the new issue. An LSTM can be trained with any combination of forcings, so don't hard code them. Let the BMI determine the forcings to include from the training configuration. But we can still initialize everything, and just not include them in the input tensor.
Current behavior
Expected behavior