BAAI

/

bge-large-en

like 224

Follow

Beijing Academy of Artificial Intelligence 3.76k

Hi there,

I've tried deploying it to sage maker. It was success but the problem is i am getting completely different embeddings from hugginface and sagemaker same model. For example, for same sentence, here is the first 5 indexes of the sagemaker embedding:
"[0.1533402055501938, 0.06615851819515228, -0.13619430363178253, 0.053574495017528534, -0.057980358600616455]"
and here are the ones from hugginface:
"[-0.0020180027931928635, 0.013546918518841267, 0.004793777596205473, -0.017005406320095062, -0.005101366899907589]"

Any way to make sure the embeddings stay consistent, so that i don't have to re-do all the data I already have in my VectorStore.

If someone can provide me a script for sagemaker deployment which would result in same values as huggingface, it would be so great.

Thanks in advance.