kyegomez / Mixture-of-Depths

Implementation of the paper: "Mixture-of-Depths: Dynamically allocating compute in transformer-based language models"
MIT License
55 stars 2 forks source link