利用 cross-attention(交叉注意力层)来实现多模态训练:class-condition, text-to-image, layout-to-image 资料 Diffusion Probabilistic Models: Theory and Applications Hugging Face blog : Annotated Diffusion What are Diffusion Models Generative Modeling by Estimating Gradients of the Data Distribution HuggingFace Diff Models # Definition