2 StableRep: Synthetic Images from Text-to-Image Models Make Strong Visual Representation Learners · 5 authors 1