Working on NN quantization/pruning