Medusa: Framework for Accelerating LLM Generation with Multiple Decoding Heads (sites.google.com) 1 points by azeirah 2y ago ↗ HN
1 comment
[ 4.0 ms ] story [ 74.2 ms ] thread