Evaluating the performance of embeddings


I’m a newbie here, so sorry if my question is dumb.
I have been using embeddings and nearest neighbor search to buil a recommendation system (inspired by the following code openai-cookbook/Recommendation_using_embeddings.ipynb at main · openai/openai-cookbook · GitHub )
I would like to measure the performance of this model using a metric like accuracy or loss function or others. I have been searching for documentation around this but nothing, isn’t evaluation the performance a must do step in creating AI models?
Is there a documentation I can read on the subject ?
Any recommendation / remark is welcome !

Thank you in advance !

1 Like