Code for Fast Ternary Large Language Model Inference with Addition-Based Sparse GEMM on Edge Devices - View it on GitHub