CBiT

Implementation of the paper "Contrastive Learning with Bidirectional Transformers for Sequential Recommendation".

Run beauty

python main.py --template train_bert --dataset_code beauty

Run toys

python main.py --template train_bert --dataset_code toys

Run ml-1m

python main.py --template train_bert --dataset_code ml-1m

Test a pretrained checkpoint on beauty

python test.py --template test_bert

• The checkpoint file is stored at /experiments/test_2022-04-26_0/models/best_acc_model.pth . You may need to download the checkpoint file manually from git LFS.

Acknowledgements

Training pipeline is implemented based on this repo https://github.com/jaywonchung/BERT4Rec-VAE-Pytorch . We would like to thank the contributors for their work.

Citation

Please cite our paper if you find our codes useful:

@inproceedings{CBiT,
  author    = {Hanwen Du and
               Hui Shi and
               Pengpeng Zhao and
               Deqing Wang and
               Victor S. Sheng and
               Yanchi Liu and
               Guanfeng Liu and
               Lei Zhao},
  title     = {Contrastive Learning with Bidirectional Transformers for Sequential Recommendation},
  booktitle = {CIKM},
  year      = {2022}
}