Use embedding models to pre-compute the query embeddings to make clustering experiments quicker using cached embeddings