Update OpenAI embedding model for semantic similarity #83
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
This PR updates the default embedding model for
semantic_similarity
fromtext-embedding-ada-002
totext-embedding-3-small
.OpenAI announced new embedding models last week, and the new model
text-embedding-3-small
worked better than the old onetext-embedding-ada-002
for my use cases.(The reduced price is another strong point for the new model.)
Especially, although
text-embedding-ada-002
had an issue: the cosine similarities tend to be skewed quite heavily towards higher numbers (as commented in docs), new modeltext-embedding-3-small
seems to resolve this issue.The another new model
text-embedding-3-large
is not good for the default model, because of the increased price.Supplemental data
Here is the result of my test cases(Japanese):
Test script here, or see my writeup (Japanese).
Thank you.