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
Hi, thanks for you great work! The issue I am concerned about is the deployed parallelism when compared to Lookahead. As far as I know, Lookahead currently does not supports tensor parallelism which is important in LLM serving. Do you achieve this on Lookahead, or are you just comparing it with pipeline parallelism version?
The text was updated successfully, but these errors were encountered:
Hi, thanks for you great work! The issue I am concerned about is the deployed parallelism when compared to Lookahead. As far as I know, Lookahead currently does not supports tensor parallelism which is important in LLM serving. Do you achieve this on Lookahead, or are you just comparing it with pipeline parallelism version?
The text was updated successfully, but these errors were encountered: