4 ms·
What is being used to actually create the embeddings?
by osigurdson 3y ago
What is being used to actually create the embeddings?
- dmezzetti 3y agoThe library is open source. Here's where you can see how they're creating the embeddings. https://github.com/kagisearch/vectordb/blob/main/vectordb/embedding.py#L36 https://github.com/kagisearch/vectordb/blob/main/vectordb/em...
- simonw 3y agohttps://github.com/kagisearch/vectordb/blob/453bb658bb7108389cbbdbf35051d22696408a1c/vectordb/embedding.py#L37 https://github.com/kagisearch/vectordb/blob/453bb658bb710838... Looks like it uses one of these, depending on your settings: Fast model: google/universal-sentence-encoder/4 Multilingual model: universal-sentence-encoder-multilingual-large/3 Normal model (Alternative): BAAI/bge-small-en-v1.5 Best model: BAAI/bge-base-en-v1.5