ValueError: Query vector size 1024 does not match index column size 1536

[6643789wsx]

{
    "name": "ValueError",
    "message": "Query vector size 1024 does not match index column size 1536",
    "stack": "---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
Cell In[65], line 1
----> 1 result = await search_engine.asearch(\"Tell me about Agent Mercer\")
      2 print(result.response)

File D:\\Documents\\graphrag\\graphrag\\query\\structured_search\\local_search\\search.py:66, in LocalSearch.asearch(self, query, conversation_history, **kwargs)
     63 start_time = time.time()
     64 search_prompt = \"\"
---> 66 context_text, context_records = self.context_builder.build_context(
     67     query=query,
     68     conversation_history=conversation_history,
     69     **kwargs,
     70     **self.context_builder_params,
     71 )
     72 log.info(\"GENERATE ANSWER: %s. QUERY: %s\", start_time, query)
     73 try:

File D:\\Documents\\graphrag\\graphrag\\query\\structured_search\\local_search\\mixed_context.py:139, in LocalSearchMixedContext.build_context(self, query, conversation_history, include_entity_names, exclude_entity_names, conversation_history_max_turns, conversation_history_user_turns_only, max_tokens, text_unit_prop, community_prop, top_k_mapped_entities, top_k_relationships, include_community_rank, include_entity_rank, rank_description, include_relationship_weight, relationship_ranking_attribute, return_candidate_context, use_community_summary, min_community_rank, community_context_name, column_delimiter, **kwargs)
    134     pre_user_questions = \"\
\".join(
    135         conversation_history.get_user_turns(conversation_history_max_turns)
    136     )
    137     query = f\"{query}\
{pre_user_questions}\"
--> 139 selected_entities = map_query_to_entities(
    140     query=query,
    141     text_embedding_vectorstore=self.entity_text_embeddings,
    142     text_embedder=self.text_embedder,
    143     all_entities=list(self.entities.values()),
    144     embedding_vectorstore_key=self.embedding_vectorstore_key,
    145     include_entity_names=include_entity_names,
    146     exclude_entity_names=exclude_entity_names,
    147     k=top_k_mapped_entities,
    148     oversample_scaler=2,
    149 )
    151 # build context
    152 final_context = list[str]()

File D:\\Documents\\graphrag\\graphrag\\query\\context_builder\\entity_extraction.py:55, in map_query_to_entities(query, text_embedding_vectorstore, text_embedder, all_entities, embedding_vectorstore_key, include_entity_names, exclude_entity_names, k, oversample_scaler)
     51 matched_entities = []
     52 if query != \"\":
     53     # get entities with highest semantic similarity to query
     54     # oversample to account for excluded entities
---> 55     search_results = text_embedding_vectorstore.similarity_search_by_text(
     56         text=query,
     57         text_embedder=lambda t: text_embedder.embed(t),
     58         k=k * oversample_scaler,
     59     )
     60     for result in search_results:
     61         matched = get_entity_by_key(
     62             entities=all_entities,
     63             key=embedding_vectorstore_key,
     64             value=result.document.id,
     65         )

File D:\\Documents\\graphrag\\graphrag\\vector_stores\\lancedb.py:120, in LanceDBVectorStore.similarity_search_by_text(self, text, text_embedder, k, **kwargs)
    118 query_embedding = text_embedder(text)
    119 if query_embedding:
--> 120     return self.similarity_search_by_vector(query_embedding, k)
    121 return []

File D:\\Documents\\graphrag\\graphrag\\vector_stores\\lancedb.py:99, in LanceDBVectorStore.similarity_search_by_vector(self, query_embedding, k, **kwargs)
     89     docs = (
     90         self.document_collection.search(query=query_embedding)
     91         .where(self.query_filter, prefilter=True)
     92         .limit(k)
     93         .to_list()
     94     )
     95 else:
     96     docs = (
     97         self.document_collection.search(query=query_embedding)
     98         .limit(k)
---> 99         .to_list()
    100     )
    101 return [
    102     VectorStoreSearchResult(
    103         document=VectorStoreDocument(
   (...)
    111     for doc in docs
    112 ]

File e:\\usual_app\\anaconda\\envs\\GraphRAG\\Lib\\site-packages\\lancedb\\query.py:303, in LanceQueryBuilder.to_list(self)
    295 def to_list(self) -> List[dict]:
    296     \"\"\"
    297     Execute the query and return the results as a list of dictionaries.
    298 
   (...)
    301     fields are returned whether or not they're explicitly selected.
    302     \"\"\"
--> 303     return self.to_arrow().to_pylist()

File e:\\usual_app\\anaconda\\envs\\GraphRAG\\Lib\\site-packages\\lancedb\\query.py:528, in LanceVectorQueryBuilder.to_arrow(self)
    519 def to_arrow(self) -> pa.Table:
    520     \"\"\"
    521     Execute the query and return the results as an
    522     [Apache Arrow Table](https://arrow.apache.org/docs/python/generated/pyarrow.Table.html#pyarrow.Table).
   (...)
    526     vector and the returned vectors.
    527     \"\"\"
--> 528     return self.to_batches().read_all()

File e:\\usual_app\\anaconda\\envs\\GraphRAG\\Lib\\site-packages\\lancedb\\query.py:558, in LanceVectorQueryBuilder.to_batches(self, batch_size)
    545     vector = [v.tolist() for v in vector]
    546 query = Query(
    547     vector=vector,
    548     filter=self._where,
   (...)
    556     with_row_id=self._with_row_id,
    557 )
--> 558 result_set = self._table._execute_query(query, batch_size)
    559 if self._reranker is not None:
    560     rs_table = result_set.read_all()

File e:\\usual_app\\anaconda\\envs\\GraphRAG\\Lib\\site-packages\\lancedb\\table.py:1623, in LanceTable._execute_query(self, query, batch_size)
   1619 def _execute_query(
   1620     self, query: Query, batch_size: Optional[int] = None
   1621 ) -> pa.RecordBatchReader:
   1622     ds = self.to_lance()
-> 1623     return ds.scanner(
   1624         columns=query.columns,
   1625         filter=query.filter,
   1626         prefilter=query.prefilter,
   1627         nearest={
   1628             \"column\": query.vector_column,
   1629             \"q\": query.vector,
   1630             \"k\": query.k,
   1631             \"metric\": query.metric,
   1632             \"nprobes\": query.nprobes,
   1633             \"refine_factor\": query.refine_factor,
   1634         },
   1635         with_row_id=query.with_row_id,
   1636         batch_size=batch_size,
   1637     ).to_reader()

File e:\\usual_app\\anaconda\\envs\\GraphRAG\\Lib\\site-packages\\lance\\dataset.py:336, in LanceDataset.scanner(self, columns, filter, limit, offset, nearest, batch_size, batch_readahead, fragment_readahead, scan_in_order, fragments, prefilter, with_row_id, use_stats)
    320 builder = (
    321     ScannerBuilder(self)
    322     .columns(columns)
   (...)
    333     .use_stats(use_stats)
    334 )
    335 if nearest is not None:
--> 336     builder = builder.nearest(**nearest)
    337 return builder.to_scanner()

File e:\\usual_app\\anaconda\\envs\\GraphRAG\\Lib\\site-packages\\lance\\dataset.py:2150, in ScannerBuilder.nearest(self, column, q, k, metric, nprobes, refine_factor, use_index)
   2146     raise TypeError(
   2147         f\"Query column {column} must be a vector. Got {column_field.type}.\"
   2148     )
   2149 if len(q) != column_type.list_size:
-> 2150     raise ValueError(
   2151         f\"Query vector size {len(q)} does not match index column size\"
   2152         f\" {column_type.list_size}\"
   2153     )
   2155 if k is not None and int(k) <= 0:
   2156     raise ValueError(f\"Nearest-K must be > 0 but got {k}\")

ValueError: Query vector size 1024 does not match index column size 1536"
}

[AlonsoGuevara]

Hi @6643789wsx

Are you still having this issue? It makes sense since 1536 is default size for OpenAI text-embedding-3-small andtext-embedding-ada-002. Are you using a different embedding model?

[goodmaney]

Hi @6643789wsx

Are you still having this issue? It makes sense since 1536 is default size for OpenAI text-embedding-3-small andtext-embedding-ada-002. Are you using a different embedding model?

https://github.com/microsoft/graphrag/issues/451 look like it happens on local embedding. I dont find where to change the size

[AlonsoGuevara]

Hi!

We are consolidating alternate model and local embedding issues here: https://github.com/microsoft/graphrag/issues/657