Add dynamics checkpoints and visualizations (steps 0–29000, 16×A100 DDP training)

Add config: training_large_lam.yaml

Add config: dynamics_large_lam.yaml

initial commit