Need help in storing and loading document store

[sohanasarah]

Hi I am using InMemory Document Store and an Embedding retriever for Q/A pipeline.

from haystack.document_stores import InMemoryDocumentStore

document_store = InMemoryDocumentStore(embedding_dim =768, use_bm25=True)
document_store.write_documents(docs_processed)

from haystack.nodes import EmbeddingRetriever

retriever_model_path = 'downloaded_models\local\my_local_multi-qa-mpnet-base-dot-v1'
retriever = EmbeddingRetriever(document_store=document_store,
                               embedding_model=retriever_model_path,
                               use_gpu=True)  

document_store.update_embeddings(retriever=retriever)

As the embedding takes a while, I want to load the embeddings and later use them again in the retriever. (in rest API side). I dont want to use ElasticSearch or Faiss. How can I achieve this using In Memory Store? I tried to use pickle, but there is no way to store the embeddings. Again, in the embedding retriever there is no load function.

[anakin87]

Hello @sohanasarah!
I answered on StackOverflow.