Official PyTorch Implementation of Gated DeltaNet-2: Decoupling Erase and Write in Linear Attention - View it on GitHub