Anton Troynikov
Anton Troynikov
Just ran accuracy eval with the llama implementation and recall results are the same to within 0.002.
Have given some thought to that and I think the best way is to expose a 're-embed' method on collections which allows users to swap embedding functions, which will re-embed...
I've split the inference environment implementation from this issue, see https://github.com/chroma-core/chroma/issues/2682.
I think this is fine, we should communicate the bump in the migration log and explain to people how to get the improvement.
Just to confirm, what are the inconsistencies that you're seeing? Is it that a different set of results is returned between different initializations, or is it that the same query...