lambda_e
This keyword sets the weight \(\lambda_e\) of the loss term associated with the energy in the loss function.
Syntax
lambda_e <weight>
Parameters
Here, <weight> represents \(\lambda_e\), which must satisfy \(\lambda_e \geq 0\) and defaults to \(\lambda_e = 1.0\).
Output
None.
Examples
lambda_e 1.0
Notes
It might be beneficial to use a two-step training scheme, where \(\lambda_e\) takes a small value (such as 0.1) in the first training and a large value (such as 10) in the second (restarting from the first). During the second training, the energy error might decrease appreciably, while the force and virial errors are not much affected.