- create rough architecture of HSTU
- finish position-based relative attetion bias (rab^{p})
- finish position-based relative attetion bias (rab^{t})
- create reco and ranking trainers
- create datasets
roman-dusek/GR-HSTU
Unofficial implementation of generative recommender (GR) with Hierarchical Sequential Transduction Unit (HSTU) from Actions Speak Louder than Words: Trillion-Parameter Sequential Transducers for Generative Recommendations
Python