Due to upload restrictions, more complete projects including saved models can be referenced: https://drive.google.com/drive/folders/1nnZvxmeSdWrgihmc9epwNJ8Vch8SxS1S?usp=drive_link
Single GPU can directly run python RTSP_train.py or python RTSP_test.py
Multiple GPUs need
RTSP_train.py:
os.environ['TF_CPP_MIN_LOG_LEVEL']='3' # numbers of GPUs
os.environ['CUDA_VISIBLE_DEVICES'] = '0,1,2' # id of GPUs
cd ./Robust-VRP/Robust-TSP/TSP/RTSP
python -m torch.distributed.launch --nproc_per_node=3 --master_port=11267 RTSP_train.py
Single GPU and multi-GPU are adjusted according to the scale and memory size, and the corresponding RTSPTrainer.py, RTSP_train.py, utils.py need to be modified.