TokenGate: Adapting LLMs for pre-fill with learned skimming for ~11x speedup

 (github.com)

1 points | by anandvc 7 hours ago ago

1 comments