Browse Teach Curate

Cross-Attention Mechanism

An attention operation in which the queries come from one sequence or feature map while the keys and values come from another, letting one representation attend to and be modulated by a different source, such as conditioning image features on external metadata.

jeremy Video

Cross Attention using Stable Diffusion Text Conditioning in Deep Learning

Cross attention is a mechanism within deep learning that enables global routing of information between primary input features and conditional embeddings by projecting query vectors from one domain an…