Speculative Decoding • Accelerating LLMs, Part 2

In this post we continue our series on how to accelerate LLMs. Previously we covered the FlashAttention algorithm in Part 1. Follow along… Continue reading on Medium »