Speculative Speculative Decoding: How to Parallelize Drafting and ... for 2x Faster LLM Inference



DOWNLOAD





Bagikan