This is our Pytorch implementation for the paper: "M2GNN: Metapath and Multi-interest Aggregated Graph Neural Network for Tag-based Cross-domain Recommendation"
The proposed model is called M2GNN.
Submitted at 1/31/2023.
DPBJ dataset will be released later because it is being deencrypted.
Amazon dataset has been released.
Training M2GNN on DPBJ dataset needs at least 4 A100 devices, the start-up command is
python -m torch.distributed.launch --nproc_per_node 4 main_M2GNN.py --max_K 6 --gamma 7
Training M2GNN on Amazon dataset needs at least 2 V100 devices, the start-up command is
python -m torch.distributed.launch --nproc_per_node 2 main_M2GNN_amazon.py --max_K 6 --gamma 7
You can also see the training log to check the convergence process of the model.