Great question — answering as the UphillTech team:
If you download the repo for the embedding model, you can then pass the path to the root of the repo instead of the name of the model to SentenceTransformerEmbeddingFunction(model_name) from chromadb.utils.embedding_functions.sentence_transformer_embedding_function import ( SentenceTransformerEmbeddingFunction ) chromadb_embedding = SentenceTransformerEmbeddingFunction(model_path) A similar solution works for llamaindex: from llama_index.embeddings.huggingface import HuggingFaceEmbedding llamaindex_embedding = HuggingFaceEmbedding(model_path) and for directive usage of the model for ranking: from sentence_transformers import SentenceTransformer embedding = SentenceTransformer(model_path)...
Full answer by Danny Varod: https://stackoverflow.stackexchange.com/a/79282075 (CC BY-SA 4.0).
Need it done for your business? UphillTech.
