Medusa: Simple Framework for Accelerating LLM Generation with Multiple Decoding Heads
Latest commits.
Builders behind this project.