【论文阅读】Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffma