Implementation for our paper Sinkhorn Transformer for Multi-Behaviour Recommendation.

The code is built on Pytorch and the RecBole library. Use following command to install the requeiremnts:

pip install -r requirements.txt

python runSTMBR.py --model=[STMBR] --dataset=[tmall_beh] --gpu_id=[0] --batch_size=[2048], where [value] is the default value.

• Note that we modified the evaluation sampling setting in recbole/sampler/sampler.py to make it static.
• The model code is at recbole/model/sequential_recommender/stmbr.py.
• Feel free to explore other RecBole library baseline models to compare the performances.