/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

Primary LanguagePython

  • 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