agoralab
OmegaViT (ΩViT) is a cutting-edge vision transformer architecture that combines multi-query attention, rotary embeddings, state space modeling, and mixture of experts to achieve superior performance across various computer vision tasks.
An open source implementation of the paper: "AN EVOLVED UNIVERSAL TRANSFORMER MEMORY"
