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
One thing that vectorscale handled quite well was adding ad hoc extra filters to a global index. What's the best approach to making an index that is fast for searching on a mix of classical filters and search, like when you have a cartesian product of xx_type columns that may or may not be filtered on followed by a vector search within that class.
In ML curation, examples would be training vs eval vs verification set, which sensor an embedding came from, crop type, true annotated object class, etc etc. Those are a bit separate from the multitenant usecase but could probably be enabled in postgres by making the planner estimate costs properly with amoptionalkey/amsearcharray/amcostestimate and by allowing it to push down the iteration over the posting lists into the index before reranking
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
Uh oh!
There was an error while loading. Please reload this page.
One thing that vectorscale handled quite well was adding ad hoc extra filters to a global index. What's the best approach to making an index that is fast for searching on a mix of classical filters and search, like when you have a cartesian product of xx_type columns that may or may not be filtered on followed by a vector search within that class.
In ML curation, examples would be training vs eval vs verification set, which sensor an embedding came from, crop type, true annotated object class, etc etc. Those are a bit separate from the multitenant usecase but could probably be enabled in postgres by making the planner estimate costs properly with amoptionalkey/amsearcharray/amcostestimate and by allowing it to push down the iteration over the posting lists into the index before reranking
All reactions