QQ, I turned on logging when using a

PPwnosaurusRex

QQ, I turned on logging when using a Huggingface embedding and I see it always connects to Huggingface, even if I already have the model cached. Looks like it grabs the tokenizer_config and config.json each time, even though I see those files in the cache folder. Any way to tell it to stop?

Plain Text

from llama_index.embeddings.huggingface import HuggingFaceEmbedding
# loads BAAI/bge-small-en
# embed_model = HuggingFaceEmbedding()
# loads BAAI/bge-small-en-v1.5
embed_model = HuggingFaceEmbedding(model_name="BAAI/bge-small-en-v1.5")

8 comments

LLogan M

its basically just checking if anything changed. AFAIK the only way to get it to stop doing that is specifying the actual model path as the name, and not just the model name

PPwnosaurusRex

LLogan M

maybe theres some huggingface env that stops that, not totally sure

PPwnosaurusRex

Thanks!

PPwnosaurusRex

One more if you have a sec...

PPwnosaurusRex

https://docs.llamaindex.ai/en/stable/examples/vector_stores/postgres/#improving-hybrid-search-with-queryfusionretriever, the section above talks about setting hybrid_search=True...but for queryfusion it doesn't matter right?

LLogan M

Oh its specifically talking about the hybrid mode built into postgres

LLogan M

(which isn't very good imo lol)

Add a reply

Find answers from the community

QQ, I turned on logging when using a