How can I resolve the unexpected 'token' keyword error in HuggingFaceInstructEmbeddings?

I'm deploying a RAG system and encountering the following error when trying to initialize HuggingFaceInstructEmbeddings:

PYTHONCopied
Traceback (most recent call last):
  File "...", line ..., in <module>
    ...
TypeError: INSTRUCTOR._load_sbert_model() got an unexpected keyword argument 'token'

This error occurs in the following context within my function:

PYTHONCopied
def create_or_get_vector_store(chunks: list) -> FAISS:
    """
    Funzione per creare o caricare il database vettoriale dalla memoria locale.

    Returns:
        FAISS: Vector store
    """

    # Automatically choose device: CUDA if available, otherwise CPU
    device = 'cuda' if torch.cuda.is_available() else 'cpu'

    # Initialize HuggingFaceInstructEmbeddings with the chosen device
    embeddings = HuggingFaceInstructEmbeddings(
        query_instruction="Represent the query for retrieval: ",
        model_name="hkunlp/instructor-large",  # Example model name
        model_kwargs={'device': device}, 
        encode_kwargs={'normalize_embeddings': True}  # Optional: to normalize the embeddings
    )

    # Check if the vector store exists; if not, create it
    if not os.path.exists("./db"):
        print("CREATING DB")
        # Create the FAISS vector store from the documents (chunks)
        vectorstore = FAISS.from_documents(chunks, embeddings)
        # Save the vector store locally
        vectorstore.save_local("./db")
    else:
        print("LOADING DB")
        # Load the pre-existing vector store
        vectorstore = FAISS.load_local("./db", embeddings)

    return vectorstore

I suspect this error is related to a version mismatch between LangChain’s community integrations and the underlying sentence-transformers library, with an unexpected keyword being passed (i.e., 'token').

I have verified that I am not explicitly supplying a 'token' parameter, and I plan to check the following:

• Package versions for langchain, sentence-transformers, and related dependencies
• Whether an update or downgrade of sentence-transformers (for example, downgrading to 2.2.2) resolves the issue
• The latest langchain-community packages for any related fixes

Has anyone encountered this error before or have suggestions on how to resolve this keyword conflict? Any help would be appreciated!