[NeurIPS 2024] Learning-to-Cache: Accelerating Diffusion Transformer via Layer Caching - View it on GitHub