SMG: The Case for Disaggregating CPU from GPU in LLM Serving

(pytorch.org)

2 points | by gmays 9 hours ago ago