One Model to Train them All: Hierarchical Self-Distillation for Enhanced Early Layer Embeddings

Open in new window