google-deepmind/recurrentgemma
Open weights language model from Google DeepMind, based on Griffin.
218 best points · 38 comments · 2 Hacker News posts · 685 GitHub stars · View on GitHub
Hacker News discussions
- Implementation of Google's Griffin Architecture – RNN LLM (218 points, 38 comments, by milliondreams, 2024-04-10)
- RecurrentGemma: Mixture local attention and linear recurrences (3 points, 0 comments, by tosh, 2024-04-09)