Hi there, from what I learned is that OpenAI embedding models use average pooling to get the sentence embeddings, is there any possible way that we can get the token-level embeddings and test on different pooling methods ourselves?? thank you
1 Like