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 am trying to write a custom metric that maintains some state that is a List[str]. I want to be able to sync across ranks and concatenate the lists belonging to each rank. Reading through sync_dist it's unclear to me where such a synchronization would occur since by default function being applied is gather_all_tensors and there wouldn't be any tensors in the lists.
Is my understanding correct? Is there a different dist_sync_fn I could use to ensure correct syncing of non tensor lists?
reacted with thumbs up emoji reacted with thumbs down emoji reacted with laugh emoji reacted with hooray emoji reacted with confused emoji reacted with heart emoji reacted with rocket emoji reacted with eyes emoji
-
I am trying to write a custom metric that maintains some state that is a List[str]. I want to be able to sync across ranks and concatenate the lists belonging to each rank. Reading through sync_dist it's unclear to me where such a synchronization would occur since by default function being applied is
gather_all_tensors
and there wouldn't be any tensors in the lists.Is my understanding correct? Is there a different dist_sync_fn I could use to ensure correct syncing of non tensor lists?
Beta Was this translation helpful? Give feedback.
All reactions