[官方]Tensorflow-GPU+CUDA+Cudnn

借鉴tensorflow官网的tensorflow-gpu的安装教程：https://www.tensorflow.org/install/gpu

使用 apt 安装 CUDA

此部分介绍如何针对 Ubuntu 16.04 和 18.04 安装 CUDA 10（TensorFlow 13.0 及更高版本）和 CUDA 9。这些说明可能适用于其他 Debian 系发行版。

注意：安全启动使 NVIDIA 驱动程序的安装过程变复杂，并且不在这些说明的讨论范围内。

Ubuntu 18.04 (CUDA 10)

    # Add NVIDIA package repositories
    wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.0.130-1_amd64.deb
    sudo dpkg -i cuda-repo-ubuntu1804_10.0.130-1_amd64.deb
    sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub
    sudo apt-get update
    wget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1804/x86_64/nvidia-machine-learning-repo-ubuntu1804_1.0.0-1_amd64.deb
    sudo apt install ./nvidia-machine-learning-repo-ubuntu1804_1.0.0-1_amd64.deb sudo apt-get update

    # Install NVIDIA driver
    sudo apt-get install --no-install-recommends nvidia-driver-410
    # Reboot. Check that GPUs are visible using the command: nvidia-smi

    # Install development and runtime libraries (~4GB)
    sudo apt-get install --no-install-recommends \
        cuda-10-0 \
        libcudnn7=7.4.1.5-1+cuda10.0  \
        libcudnn7-dev=7.4.1.5-1+cuda10.0
# Install TensorRT. Requires that libcudnn7 is installed above.
    sudo apt-get update && \
            sudo apt-get install nvinfer-runtime-trt-repo-ubuntu1804-5.0.2-ga-cuda10.0 \
            && sudo apt-get update \
            && sudo apt-get install -y --no-install-recommends libnvinfer-dev=5.0.2-1+cuda10.0
    

Ubuntu 16.04 (CUDA 10)

    # Add NVIDIA package repositories
    # Add HTTPS support for apt-key
    sudo apt-get install gnupg-curlwget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_10.0.130-1_amd64.debsudo dpkg -i cuda-repo-ubuntu1604_10.0.130-1_amd64.debsudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub
sudo apt-get updatewget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.deb
sudo apt install ./nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.deb sudo apt-get update
 # Install NVIDIA Driver
 # Issue with driver install requires creating /usr/lib/nvidia
    sudo mkdir /usr/lib/nvidiasudo apt-get install --no-install-recommends nvidia-410
    # Reboot. Check that GPUs are visible using the command: nvidia-smi

    # Install development and runtime libraries (~4GB)
    sudo apt-get install --no-install-recommends \
        cuda-10-0 \
        libcudnn7=7.4.1.5-1+cuda10.0  \
        libcudnn7-dev=7.4.1.5-1+cuda10.0

 # Install TensorRT. Requires that libcudnn7 is installed above.
sudo apt-get update && \
            sudo apt-get install nvinfer-runtime-trt-repo-ubuntu1604-5.0.2-ga-cuda10.0 \
            && sudo apt-get update \
            && sudo apt-get install -y --no-install-recommends libnvinfer-dev=5.0.2-1+cuda10.0

Ubuntu 16.04（CUDA 9.0，TensorFlow 1.13.0 以下版本）

 # Add NVIDIA package repository
sudo apt-key adv --fetch-keys http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/7fa2af80.pub
wget http://developer.download.nvidia.com/compute/cuda/repos/ubuntu1604/x86_64/cuda-repo-ubuntu1604_9.1.85-1_amd64.deb
sudo apt install ./cuda-repo-ubuntu1604_9.1.85-1_amd64.deb  wget http://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu1604/x86_64/nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.debsudo apt install ./nvidia-machine-learning-repo-ubuntu1604_1.0.0-1_amd64.debsudo apt update
# Install CUDA and tools. Include optional NCCL 2.x
    sudo apt install cuda9.0 cuda-cublas-9-0 cuda-cufft-9-0 cuda-curand-9-0 \
        cuda-cusolver-9-0 cuda-cusparse-9-0 libcudnn7=7.2.1.38-1+cuda9.0 \
        libnccl2=2.2.13-1+cuda9.0 cuda-command-line-tools-9-0

 # Optional: Install the TensorRT runtime (must be after CUDA install) sudo apt updatesudo apt install libnvinfer4=4.1.2-1+cuda9.0