[Hybrid] Simplify accepted token counting in spec decode for hybrid models#38372

njhill merged 2 commits into

vllm-project:mainvllm-project/vllm:mainfrom

fuscof-ibm:optimize_update_statesfuscof-ibm/vllm:optimize_update_statesCopy head branch name to clipboard