Set up
When running a pydeeplearn program you might have to set up some environment variables, depending on your configuration. If want to use the GPU for training/testing a model, you have to ensure that theano knows where your CUDA installation is (for detailed instructions see below).
Setting up theano to work on the GPU
PATH needs to contain the path to nvcc (usually /usr/local/cuda/bin)
CUDA_PATH needs to contain the path to cuda (usually /usr/local/cuda/)
LD_LIBR
详细参照包内文档