Hi, I am using custom model from

At a glance

The community member is using a custom Hugging Face model to generate embeddings and storing the documents along with their embeddings in a Qdrant vector store. They want to create a retriever that uses the same embedding model as the one used for indexing, and then use a GPT model to synthesize the final response. The comments suggest that the community member is unsure about the relationship between the index, retriever, and embedding model, and is seeking clarification on how to create a retriever that utilizes the same embedding model as the one used for indexing.

UUzair

Hi, I am using custom model from hugging face to generate embedding and then I am using Qdrant to store my documents along with their embeddings as shown below

Plain Text

MODEL = "mixedbread-ai/mxbai-embed-large-v1"

reader = SimpleDirectoryReader(input_dir="/content/data")
documents = reader.load_data()
embed_model = HuggingFaceEmbedding(model_name=MODEL)
vector_store = QdrantVectorStore(client=QDRANT_CLIENT, collection_name=collection_name)
storage_context = StorageContext.from_defaults(vector_store=vector_store)
index = VectorStoreIndex.from_documents(
    documents,
    storage_context=storage_context,
    embed_model=embed_model,
)

Now I want to make retriever which should be of the same model as of embedding model. And then to synthesize the final response I want to use gpt, How can I make retriever with the same embedding model that I used for document embeddings?

3 comments

kkapa.ai

Let me check the knowledge base for you. One moment! :books:

kkapa.ai

Sorry @Uzair, I seem to be having a bit of trouble. Don't worry though, I've pinged my support and they will try to fix me up! :technologist:

Perhaps try again in 15mins?

UUzair

Can someone correct my concepts here (if I am wrong),

After creating an index, the retriever is utilized to fetch the most relevant texts (nodes), after which this information is passed to the synthesizer along with the prompt (if provided) to generate the final response. Right?

And the query engine is built on top of the index. Does this mean that if we utilize a model from Hugging Face to generate embeddings for indexing, and then we use this index to create a retriever, will this retriever be utilizing our embedding model or not?

Plain Text

retriever = VectorIndexRetriever(
   index=index,
    similarity_top_k=5,
)

Add a reply

Find answers from the community

Hi, I am using custom model from