Towards Fast GPU-based Sparse DNN Inference: A Hybrid Compute Model | IEEE Conference Publication | IEEE Xplore