以下摘自github hoszbh的回答 https://github.com/BVLC/caffe/issues/5729 modifying cudnn_conv_layer.cpp and aligning the address to be multiples of 32. You can insert tow lines of code before size_t total_max_workspace = ... as follow: size_t m=32; max_workspace = (max_workspace + m-1) / m * m;