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

• 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