雲服務：Google Cloud Platform
作業系統：ubuntu 18.04 LTS
顯示卡：TESLA T4

安裝 GPU

以下皆在 root 下執行

shell
1
$ sudo su -
shell
1
2
3
$ sudo apt-get update
$ sudo apt-get install -y ubuntu-drivers-common
$ ubuntu-drivers devices

輸出看到可以安裝 440 的驅動版本

shell
1
2
3
4
5
6
== /sys/devices/pci0000:00/0000:00:04.0 ==
modalias : pci:v000010DEd00001EB8sv000010DEsd000012A2bc03sc02i00
vendor   : NVIDIA Corporation
driver   : nvidia-driver-435 - distro non-free
driver   : nvidia-driver-440 - distro non-free recommended
driver   : xserver-xorg-video-nouveau - distro free builtin

安裝驅動

shell
1
$ sudo ubuntu-drivers autoinstall

安裝 cuda 10.2 & cuDNN

安裝 cuda 10.2 軟件包

shell
1
2
3
4
$ cd /tmp
$ wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/cuda-repo-ubuntu1804_10.2.89-1_amd64.deb
$ sudo dpkg -i cuda-repo-ubuntu1804_10.2.89-1_amd64.deb
$ sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu1804/x86_64/7fa2af80.pub

安裝 cuDNN v7.6.5 軟件包

cuDNN 軟件包到 這裡下載，載 Runtime 與 Developer

  • cuDNN Runtime Library for Ubuntu18.04 (Deb)
  • cuDNN Developer Library for Ubuntu18.04 (Deb)

安裝軟件包

shell
1
2
$ sudo dpkg -i libcudnn7_7.6.5.32-1+cuda10.2_amd64.deb
$ sudo dpkg -i libcudnn7-dev_7.6.5.32-1+cuda10.2_amd64.deb

執行安裝後重開機

shell
1
2
3
4
5
6
$ sudo apt-get update
$ sudo apt-get install --no-install-recommends \
   cuda-10-2 \
   libcudnn7=7.6.5.32-1+cuda10.2  \
   libcudnn7-dev=7.6.5.32-1+cuda10.2
$ sudo reboot

重開機後測試

plaintext
1
$ nvidia-smi

安裝 Miniconda

shell
1
2
$ cd /tmp && curl -O https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh
$ sudo bash Miniconda3-latest-Linux-x86_64.sh

安裝位置 /opt/miniconda3，並將路徑新增至 .bashrc

shell
1
export PATH="/opt/miniconda3/bin:$PATH"

安裝 tensorflow

shell
1
2
3
pip install --upgrade --user pip
pip install tensorflow-gpu
python -c 'import tensorflow as tf; tf.test.gpu_device_name()'

輸出結果為