FPGA加速调研

普遍量化使用DoReFa Net量化的方法

在模型上大多是AlexNet, VGG16, ShuffleNetV2

  • TensorFlow to cloud FPGAs: Tradeoffs for accelerating deep neural networks

image-20210518183206640

  • FPGA-based training accelerator utilizing sparseness of convolutional neural network

    SPARSE CNN TRAINING ACCELERATOR

image-20210518183324590

  • A high-performance CNN processor based on FPGA for mobilenets

image-20210519101210759

  • RNA: An accurate residual network accelerator for quantized and reconstructed deep neural networks

image-20210519104459231

image-20210519101642780

  • Synetgy: Algorithm-hardware Co-design for ConvNet accelerators on embedded FPGAS

使用改进的ShuffleNets V2,用移位代替乘法

image-20210519104532571

Last modification:May 19, 2021
恰饭环节