Parallelizing Neural Networks Development Environment C++11 CUDA 9.0 Usage Compile and run the train.cu file: nvcc --std=c++11 train.cu -o train ./train To execute only on the CPU, change the last argument of the Train() function in train.cu to true.