How to use the search by "cosine similarity"?

[guillaumeOfr]

Hello everyone,

Here are the steps I followed :

1. I created a Chroma base and a collection
2. After, following the advice of the issue [Question]What is the algorithm used by Chroma to calculate vector similarity? #213 , I modified the source code by changing "l2" to "cosine" at these lines :

• https://github.com/chroma-core/chroma/blob/0.3.11/chromadb/db/index/hnswlib.py#L26
• https://github.com/chroma-core/chroma/blob/0.3.11/chromadb/db/index/hnswlib.py#L155

4. When I do a search with collection.query(...) I get distances < -1 (around -15, for example)

How can I use the search by "cosine similarity" in my collection ?

Thanks in advance,

[jeffchuber]

Hi @guillaumeOfr ! You are totally right. Chroma needs this and we are going to implement it. It is tracked in this previous issue, #157, so will close this one for cleanliness. I will tag you when the other issue is closed!