How do I use tensorflow.is to get the embeddings of text and use knn to classify them