You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Q1: shall we design KV memory management before kernel design to include some features like paged mem?
Q2: shall we somehow decouple the mem management design and perf test for cuda kernel?