Tags
1 page
Speculative Decoding
Speculative Decoding: 2x to 4x speedup of LLMs without quality loss