Simple embedding -> text model trained on a tiny subset of Wikipedia sentences (~7m) embedded via text-embedding-ada-002.
Used to demonstrate the arithmetic properties of sentence embeddings.
Built on nanoGPT.
-