| from langchain_community.embeddings import HuggingFaceBgeEmbeddings | |
| from langchain_community.embeddings import HuggingFaceEmbeddings | |
| def get_embeddings_function(version = "v1.2",query_instruction = "Represent this sentence for searching relevant passages: "): | |
| if version == "v1.2": | |
| # https://huggingface.co/BAAI/bge-base-en-v1.5 | |
| # Best embedding model at a reasonable size at the moment (2023-11-22) | |
| model_name = "BAAI/bge-base-en-v1.5" | |
| encode_kwargs = {'normalize_embeddings': True,"show_progress_bar":False} # set True to compute cosine similarity | |
| print("Loading embeddings model: ", model_name) | |
| embeddings_function = HuggingFaceBgeEmbeddings( | |
| model_name=model_name, | |
| encode_kwargs=encode_kwargs, | |
| query_instruction=query_instruction, | |
| ) | |
| else: | |
| embeddings_function = HuggingFaceEmbeddings(model_name = "sentence-transformers/multi-qa-mpnet-base-dot-v1") | |
| return embeddings_function | |