DMSG

This is the code associated with the ICLR 2025 paper “Towards Continuous Reuse of Graph Models via Holistic Memory“ by Ziyue Qiao, et al. The code is based on the DGL library and is implemented in PyTorch.

Running the Example:

The data is downloaded using the dgl library, as found in the "data_prepare(args)" function within "pipeline.py".

python train.py --dataset CoraFull-CL

Citation

If this code assists in your research, please consider citing our paper:

@inproceedings{dmsg,
  title={Towards Continuous Reuse of Graph Models via Holistic Memory},
  author={Ziyue Qiao, Junren Xiao, Qingqiang Sun, Meng Xiao, Xiao Luo, Hui Xiong},
  booktitle={International Conference on Learning Representations},
  year={2025}
}