Tensorflow 在 Ubuntu 16.04 下環境問題解決方案

Tensorflow 需求

在 Ubuntu 跑 Tensorflow ，尤其是支援 gpu 版本，要符合這幾個需求:

• cuDNN 6.0
• CUDA 8.0
• Nvidia Driver

安裝詳情請參考 Tensorflow: https://www.tensorflow.org/install/install_linux

安裝後，需要優先設置 export 路徑，放在 .bashrc 裡面也可以:

export PATH="/home/vk/anaconda3/bin:$PATH"
export PATH=/usr/local/cuda-9.0/bin${PATH:+:${PATH}}
export LD_LIBRARY_PATH=/usr/local/cuda-9.0/lib64${LD_LIBRARY_PATH:+:${LD_LIBRARY_PATH}}
export CUDA_HOME=/usr/local/cuda-9.0
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64/

Nvidia-Docker-Plugin 找不到

如果 Nvidia-docker 出現 nvidia-docker-plugin 錯誤，解決方案: https://github.com/NVIDIA/nvidia-docker/issues/255

1. 移除 nvidia-docker: rm -rf /var/lib/nvidia-docker
2. 重新安裝 nvidia-docker (https://github.com/NVIDIA/nvidia-docker/wiki)
3. 重新開機

Tensorflow 無法引用 audio_ops

若出現 from tensorflow.contrib.framework.python.ops import audio_ops as contrib_audio 錯誤，表示你的 tensorflow 是新的版本，需要另外多安裝一個 tf-nightly:

pip install tf-nightly

Anaconda 環境下的 Lib 庫需求

• pip 安裝 tensorflow / tensorflow-gpu
• conda 也要安裝 tensorflow / tensorflow-gpu
• 找不到 conda 指令，請先 export:
  

  export PATH="/home/vk/anaconda3/bin:$PATH"
  

Reference:
https://github.com/tensorflow/tensorflow/issues/13031
https://github.com/NVIDIA/nvidia-docker/wiki
https://github.com/NVIDIA/nvidia-docker/issues/255
https://blog.gtwang.org/programming/tensorboard-tensorflow-visualization-tutorial/
http://playground.tensorflow.org/#activation=tanh&batchSize=11&dataset=gauss&regDataset=reg-plane&learningRate=0.03&regularizationRate=0&noise=5&networkShape=2,2&seed=0.44761&showTestData=false&discretize=false&percTrainData=90&x=true&y=false&xTimesY=false&xSquared=false&ySquared=false&cosX=false&sinX=false&cosY=false&sinY=false&collectStats=false&problem=classification&initZero=false&hideText=false
https://www.tensorflow.org/install/install_linux