Multi-Edge Interaction Network with residual connections.

Encode-Process-Decode GraphNet model.

Exponential learning rate scheduler Based on procedure described in LTS If min_lr==0 and decay_steps==1, same as torch.optim.lr_scheduler

Args: optimizer (Optimizer): Wrapped optimizer. multiplier: target learning rate = base lr * multiplier if multiplier > 1.0. if m

Model for fluid simulation.

Model for static cloth simulation.

Feature normalizer that accumulates statistics online.