Medusa: Simple Framework for Accelerating LLM Generation with Multiple Decoding Heads
Primary LanguageJupyter NotebookApache License 2.0Apache-2.0