Eagle-3 Speculative Decoding for LLM Inference (5.6x speedup)github.com/SafeAILab2 pointssummaritya year ago