Dual-pronged deep learning preprocessing on heterogeneous platforms with CPU, GPU and CSD

Re-assign community