Recommended threshold
#1
by
johnny-godoy
- opened
When using this model for semantic caching, is there a recommended value for the similarity threshold for cache hits? Or is the recommendation to always tune it to the specific dataset?
on our Semantic Caching service (Redis LangCache), we have seen 0.9 as a good starting point and then folks can tune it based on desired accuracy and CHR.
johnny-godoy
changed discussion status to
closed
@johnny-godoy We are actively working on bringing newer and better models so we will appreciate any feedback that helps us improve our quality.