Reproducing the Sparse Huffman Address Map Compression for Deep Neural Networks