I have been trying to use something similar to metadata filtering with Colbert, and from the function description of
RAGPretrainedModel.from_pretrained("colbert-ir/colbertv2.0").search()
I've seen that we can focus the retrieval search on a list of documents stated in docs_id.
The code I am using is the following (for testing purposes):
RAG = RAGPretrainedModel.from_pretrained("colbert-ir/colbertv2.0")
index_path = RAG.index(index_name=index_name, collection=collection, document_ids = document_id)
p = RAG.encode(document_id[:20])
out_list = RAG.search("How do I change a password?", doc_ids = p)
However, I still get documents in the results (out_list) that are not in the list provided in variable p.
Is there another way of doing this correctly, or is the feature not fully implemented yet? I haven't been able to find some examples or better documentation on this.
Thanks in advance!
I have been trying to use something similar to metadata filtering with Colbert, and from the function description of
RAGPretrainedModel.from_pretrained("colbert-ir/colbertv2.0").search()I've seen that we can focus the retrieval search on a list of documents stated in docs_id.The code I am using is the following (for testing purposes):
However, I still get documents in the results (out_list) that are not in the list provided in variable p. Is there another way of doing this correctly, or is the feature not fully implemented yet? I haven't been able to find some examples or better documentation on this. Thanks in advance!