Module PINNacle.src.optimizer.lr_adaptor
Classes
LR_Adaptor(optimizer, loss_weight, num_pde, alpha=0.1, mode='max')
: Callback for learning rate annealing algorithm of physics-informed neural networks.
loss_weight - initial loss weights
num_pde - the number of the PDEs (boundary conditions excluded)
alpha - parameter of moving average
mode - "max" (PINN-LA), "mean" (PINN-LA-2)
### Ancestors (in MRO)
* torch.optim.optimizer.Optimizer
### Methods
`step(self, closure)`
: