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
I wanted to ask whether achieving the results reported in the paper requires a learning rate schedule (e.g. cosine) or if a constant learning rate, as in the released config files, is sufficient.
Hi. Thank you for the great work!
I wanted to ask whether achieving the results reported in the paper requires a learning rate schedule (e.g. cosine) or if a constant learning rate, as in the released config files, is sufficient.