Eagle-3 Speculative Decoding for LLM Inference (5.6x speedup)github.com2 points·summarity··0 commentsOpen articleSaveView on HN