Directories trained_weights/ = state dict of full model (planner_net) extracted_weights/ = state dict of encoder part (percept_net) models/ = network architectures Extracting / running See extract_network.py