Stepped MoE: reducing memory movement in large mixture-of-experts models

 (arxiv.org)

2 points | by arnav1993k 5 hours ago ago

2 comments

  • Iolaum 5 hours ago ago

    This may be a very good paper but when I saw Apple regarding the authors org I stopped.

  • arnav1993k 5 hours ago ago

    [flagged]