deazy

↗ HN profile [ 975 ms ] full profile
Karma
0
Created
()
Submissions
0
  1. Sharing my fictional work, with the world. This game has players from every country on earth and their top players, question is who is really wining the game and who is sure to lose every thing ?

  2. I was exploring this conceptual architecture for neural networks (transformers/attention) based long-context models, its conceptual but grounded in sound existing research and architecture implementations on specialized…

  3. Gemini's Long Context MoE Architecture (Hypothesized): Sharing how I think (hypothesis) Gemini models achieve their 1-10 Million long context window. With details to clues to support the same. Ensemble of Expert (EoE)…