minGRU (Feng et al.) in PyTorch, plus a measured ladder of state-tracking recoveries — signed, rotation, Givens, and delta (DeltaNet-style Householder products) mixers with time decay, all under one parallel scan — with fused Triton GPU kernels as an optional backend
cuda pytorch recurrent-neural-networks gru triton rnn householder gpu-kernels givens-rotations parallel-scan expressivity sequence-modeling paper-implementation deltanet time-decay state-tracking linear-rnn mingru orthogonal-rnn
-
Updated
Jul 31, 2026 - Python