Bug: when do the normalization of embeddings.

[Bannerli]

I have found an error in the source code, which targets to do the normalization of embeddings but encountering the ZeronDivision.

INFO: Reading settings from settings.yaml
Error embedding chunk {'OpenAIEmbedding': 'Internal Server Error'}
Traceback (most recent call last):
File "", line 198, in _run_module_as_main
File "", line 88, in _run_code
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/graphrag/query/main.py", line 75, in
run_local_search(
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/graphrag/query/cli.py", line 154, in run_local_search
result = search_engine.search(query=query)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/graphrag/query/structured_search/local_search/search.py", line 118, in search
context_text, context_records = self.context_builder.build_context(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/graphrag/query/structured_search/local_search/mixed_context.py", line 139, in build_context
selected_entities = map_query_to_entities(
^^^^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/graphrag/query/context_builder/entity_extraction.py", line 55, in map_query_to_entities
search_results = text_embedding_vectorstore.similarity_search_by_text(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/graphrag/vector_stores/lancedb.py", line 118, in similarity_search_by_text
query_embedding = text_embedder(text)
^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/graphrag/query/context_builder/entity_extraction.py", line 57, in
text_embedder=lambda t: text_embedder.embed(t),
^^^^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/graphrag/query/llm/oai/embedding.py", line 96, in embed
chunk_embeddings = np.average(chunk_embeddings, axis=0, weights=chunk_lens)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/opt/miniconda3/envs/rag/lib/python3.11/site-packages/numpy/lib/function_base.py", line 550, in average
raise ZeroDivisionError(
ZeroDivisionError: Weights sum to zero, can't be normalized

[jamesfyda]

I encounter the same situation while using local opensource embedding server with all-MiniLM-L6-v2 model

[Kingatlas115]

same issue. i can get global search working by modifying the openai_embeddings file. but no local search. hoping this is fixed soon

[AlonsoGuevara]

Hi! We are centralizing other LLM discussions in these threads:
Other LLM/Api bases: #339,
Ollama: #345
Local embeddings: #370

I'll resolve this issue so we can keep the focus on those threads

[mbt1909432]

My way can work, try this

file:.venv/Lib/site-packages/graphrag/query/llm/oai/embedding.py

def embed(self, text: str, **kwargs: Any) -> list[float]:
"""
Embed text using OpenAI Embedding's sync function.

    For text longer than max_tokens, chunk texts into max_tokens, embed each chunk, then combine using weighted average.
    Please refer to: https://github.com/openai/openai-cookbook/blob/main/examples/Embedding_long_inputs.ipynb
    """
    token_chunks = chunk_text(
        text=text, token_encoder=self.token_encoder, max_tokens=self.max_tokens
    )
    chunk_embeddings = []
    chunk_lens = []
    for chunk in token_chunks:

        chunk = self.token_encoder.decode(chunk)
        try:
            # embedding, chunk_len = self._embed_with_retry(chunk, **kwargs)
            # Make sure to `pip install openai` first
            from openai import OpenAI
            client = OpenAI(base_url="http://localhost:1234/v1", api_key="lm-studio")

            def get_embedding(text, model="nomic-ai/nomic-embed-text-v1.5-GGUF"):
                text = text.replace("\n", " ")
                return client.embeddings.create(input=[text], model=model).data[0].embedding

            print("__________")
            print(chunk)
            chunk_embeddings.append(get_embedding(chunk))
            chunk_lens.append(len(chunk))
        # TODO: catch a more specific exception
        except Exception as e:  # noqa BLE001
            self._reporter.error(
                message="Error embedding chunk",
                details={self.__class__.__name__: str(e)},
            )

            continue
    chunk_embeddings = np.average(chunk_embeddings, axis=0, weights=chunk_lens)
    chunk_embeddings = chunk_embeddings / np.linalg.norm(chunk_embeddings)
    return chunk_embeddings.tolist()

you should integrate ml studio api callback into this function