Medusa: Simple LLM Inference Acceleration Framework with Multiple Decoding Heads

Re-assign community