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
[This is a low priority issue, we can solve it alongside something else]
The hivemind-server (cli) currently defines expert names in the form of expert_pattern (e.g. ffn.[0:256].[0:256]), while RemoteMixtureOfExperts needs explicit grid sizes.
It would be great to switch to a common interface (e.g. use expert patterm in RemoteMixtureOfExperts)
The text was updated successfully, but these errors were encountered:
[This is a low priority issue, we can solve it alongside something else]
The
hivemind-server
(cli) currently defines expert names in the form ofexpert_pattern
(e.g.ffn.[0:256].[0:256]
), while RemoteMixtureOfExperts needs explicit grid sizes.It would be great to switch to a common interface (e.g. use expert patterm in RemoteMixtureOfExperts)
The text was updated successfully, but these errors were encountered: