基于可变精度脉动阵列结构的卷积神经网络加速器

秦婷,李钊,闫玉婷,杜亚飞

Convolutional neural network accelerator based on variable-precision systolic-array structure

Ting QIN,Zhao LI,Yuting YAN,Yafei DU

表 1 CIFAR-100数据集上不同模型的性能量化实验结果

Tab.1 Experimental results on quantization performance for different models on CIFAR-100 dataset

模型 AccFP/% AccINT/% L RC ResNet20 74.46 73.74 0.72 7.8× ResNet56 74.56 74.09 0.47 16.2×