Notable Vehicles Pickup and Delivery Problem-NIS

Dependencies

• Python>=3.8
• PyTorch>=1.7
• tensorboard_logger
• tqdm

Usage

Training

NVRP examples

21 nodes:

python run.py --problem nvrp --graph_size 20 --shared_critic

51 nodes:

python run.py --problem nvrp --graph_size 50 --shared_critic

101 nodes:

python run.py --problem nvrp --graph_size 100 --shared_critic

NVTA examples

21 nodes:

python run.py --problem nvta --graph_size 20 --shared_critic

51 nodes:

python run.py --problem nvta --graph_size 50 --shared_critic

101 nodes:

python run.py --problem nvta --graph_size 100 --shared_critic

Examples

For inference 2,000 NVTA instances with 100 nodes and no data augment (NIS):

python run.py --eval_only --no_saving --no_tb --problem nvta --graph_size 100 --val_m 1 --val_dataset './datasets/pdp_100.pkl' --load_path './pre-trained/nis/pdtspl_100/epoch-198.pt' --val_size 2000 --val_batch_size 2000 --T_max 3000 --shared_critic

For inference 2,000 NVTA instances with 100 nodes using the augments (NIS-A):

python run.py --eval_only --no_saving --no_tb --problem nvta --graph_size 100 --val_m 50 --val_dataset './NIS-datasets/pdp_100.pkl' --load_path './NIS-pretrained-model/nis/nvta_100/epoch-198.pt' --val_size 2000 --val_batch_size 200 --T_max 3000 --shared_critic

Run python run.py -h for detailed help on the meaning of each argument.

Acknowledgements

We appreciate the code and framework that have provided assistance to this repository.