Abstracting Sparse DNN Acceleration via Structured Sparse Tensor Decomposition