4_tensorflow.md

安裝 TensorFlow

1. 安裝依賴套件

   $ sudo apt-get install default-jdk libcupti-dev
   $ export JAVA_HOME='/usr/lib/jvm/java-8-openjdk-arm64/'

2. 取得 TensorFlow 編譯腳本

   $ git clone git://github.com/jetsonhacks/installTensorFlowTX2
   $ cd installTensorFlowTX2

3. 執行編譯腳本

   $ ./installPrerequisitesPy3.sh
   $ ./cloneTensorFlow.sh
   $ ./setTensorFlowEVPy3.sh
   $ ./buildTensorFlow.sh
   $ ./packageTensorFlow.sh

4. 安裝 TensorFlow

   1. 建立虛擬開發環境

      $ virtualenv TensorFlow

   2. 進入虛擬開發環境

      $ cd TensorFlow
      $ source bin/active

   3. 安裝 TensorFlow 到虛擬環境

      pip3 install $HOME/<TensorFlow 的 .whl 安裝封包>

   4. 測試 TensorFlow

      1. Hello World

         import tensorflow as tf
         hello = tf.constant('Hello, TensorFlow on NVIDIA Jetson TX2!')
         sess = tf.Session()
         print(sess.run(hello))

         輸出

         Hello, TensorFlow on NVIDIA Jetson TX2!
         

      2. 運算單元

         import tensorflow as tf
         # Creates a graph.
         a = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[2, 3], name='a')
         b = tf.constant([1.0, 2.0, 3.0, 4.0, 5.0, 6.0], shape=[3, 2], name='b')
         c = tf.matmul(a, b)
         # Creates a session with log_device_placement set to True.
         sess = tf.Session(config=tf.ConfigProto(log_device_placement=True))
         # Runs the op.
         print(sess.run(c))

         輸出

         name: NVIDIA Tegra X2
         major: 6 minor: 2 memoryClockRate (GHz) 1.3005
         MatMul: (MatMul): /job:localhost/replica:0/task:0/gpu:0
         b: (Const): /job:localhost/replica:0/task:0/gpu:0
         a: (Const): /job:localhost/replica:0/task:0/gpu:0
         [[ 22.  28.]
         [ 49.  64.]]