17x speedup without needing to modify the underlying CUDA; Nous's design is simply a shortcut for long-text reasoning, and the detail of switching back to full attention at the end of training is very stable.

View Original
Original content no longer visible
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments
  • Pinned