TensorFlow 使用

• 搭建你的TensorFlow开发环境

 tensorflow：

tensorflow 一Current release for CPU-only (recommendedfor beginners) 
tensorflow-gpu 一Current release with GPU support (Ubuntu and Windows) 
tf-nightly —Nightly build for CPU-only (unstable) 
tf-nightly-gpu —Nightly build with GPU support (unstable, Ubuntu and Windows)

创建Python虚拟环境

C:Users67001>pip install virtualenv

# 创建一个名字为envname的虚拟环境
D:Program FilesAnaconda3virtual_Env>virtualenv TestEnv
Using base prefix 'd:\program files\anaconda3'
  No LICENSE.txt / LICENSE found in source
New python executable in D:Program FilesAnaconda3virtual_EnvTestEnvScriptspython.exe
Installing setuptools, pip, wheel...
done.

virtualenv -p python2 envname # 如果安装了多个python版本，如py2和py3，需要指定使用哪个创建虚拟环境

# 进入虚拟环境文件
D:Program FilesAnaconda3virtual_Env>cd TestEnv
# 进入相关的启动文件夹
D:Program FilesAnaconda3virtual_EnvTestEnv>cd Scripts
# 启动虚拟环境
D:Program FilesAnaconda3virtual_EnvTestEnvScripts>activate

(TestEnv) D:Program FilesAnaconda3virtual_EnvTestEnvScripts>

deactivate # 退出虚拟环境

在虚拟环境下安装tensorflow：

(TestEnv) D:Program FilesAnaconda3virtual_EnvTestEnvScripts>pip install tensorflow

 或者：国内安装比较快

(TestEnv) D:Program FilesAnaconda3virtual_EnvTestEnvScripts>pip install -i https://pypi.tuna.tsinghua.edu.cn/simple/ --upgrade tensorflow

• “Hello TensorFlow”

>>> import tensorflow as tf
# 定义常量操作 hello
>>> hello = tf.constant("Hello TensorFlow")
# 创建一个会话
>>> sess = tf.Session()
2019-07-24 15:31:55.832973: I tensorflow/core/platform/cpu_feature_guard.cc:142] 
Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
#执行常量操作hello并打印到标准输出
>>> print(sess.run(hello))
b'Hello TensorFlow'

 

支持AVX2指令集的CPUs

• Intel
• Haswell processor, Q2 2013
• Haswell E processor, Q3 2014
• Broadwell processor, Q4 2014
• Broadwell E processor, Q3 2016
• Skylake processor, Q3 2015
• Kaby Lake processor, Q3 2016(ULV mobile)/Q1 2017(desktop/mobile)
• Skylake-X processor, Q2 2017
• Coffee Lake processor, Q4 2017
• Cannon Lake processor, expected in 2018
• Cascade Lake processor, expected in 2018
• Ice Lake processor, expected in 2018


• AMD
• Excavator processor and newer, Q2 2015
• Zen processor, Q1 2017
• Zen+ processor, Q2 2018

• 在交互式环境中使用 TensorFlow

pip install jupyter

(TestEnv) D:Program FilesAnaconda3virtual_EnvTestEnvScripts>pip install ipykernel

(TestEnv) D:Program FilesAnaconda3virtual_EnvTestEnvScripts>python -m ipykernel install --user --name=TestEnv
Installed kernelspec TestEnv in C:Users67001AppDataRoamingjupyterkernels	estenv

(TestEnv) D:Program FilesAnaconda3virtual_EnvTestEnvScripts>jupyter kernelspec list
Available kernels:
  testenv    C:Users67001AppDataRoamingjupyterkernels	estenv
  python3    d:program filesanaconda3virtual_env	estenvsharejupyterkernelspython3

(TestEnv) D:Program FilesAnaconda3virtual_EnvTestEnvScripts>jupyter notebook

多层感知机模型示例

Neural Network Overview

MNIST Dataset Overview

MNIST图像数据集使用形如［28，28］的二阶数组来表示每张图像，数组中的每个元素对应一个像素点。

该数据集中的图像都是256阶灰度图，像素值0表示白色（背景），255表示黑色（前景）。

由于每张图像的尺寸都是28x28像素，为了方便连续存储，我们可以将形如［28，28］

的二阶数组“摊平”成形如［784］的一阶数组。数组中的784个元素共同组成了一个784维的向量。

More info: http://yann.lecun.com/exdb/mnist/

from __future__ import print_function

# 导入 MNIST 数据集
from tensorflow.examples.tutorials.mnist import input_data
mnist = input_data.read_data_sets("/tmp/data/", one_hot=True)

import tensorflow as tf

# 超参数
learning_rate = 0.1
num_steps = 500
batch_size = 128
display_step = 100

# 神经网络参数
n_hidden_1 = 256 # 第一层神经元个数
n_hidden_2 = 256 # 第二层神经元个数
num_input = 784 # MNIST 输入数据(图像大小: 28*28)
num_classes = 10 # MNIST 手写体数字类别 (0-9)

# 输入到数据流图中的训练数据
X = tf.placeholder("float", [None, num_input])
Y = tf.placeholder("float", [None, num_classes])


# 权重和偏置
weights = {
    'h1': tf.Variable(tf.random_normal([num_input, n_hidden_1])),
    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])),
    'out': tf.Variable(tf.random_normal([n_hidden_2, num_classes]))
}
biases = {
    'b1': tf.Variable(tf.random_normal([n_hidden_1])),
    'b2': tf.Variable(tf.random_normal([n_hidden_2])),
    'out': tf.Variable(tf.random_normal([num_classes]))
}



# 权重和偏置
weights = {
    'h1': tf.Variable(tf.random_normal([num_input, n_hidden_1])),
    'h2': tf.Variable(tf.random_normal([n_hidden_1, n_hidden_2])),
    'out': tf.Variable(tf.random_normal([n_hidden_2, num_classes]))
}
biases = {
    'b1': tf.Variable(tf.random_normal([n_hidden_1])),
    'b2': tf.Variable(tf.random_normal([n_hidden_2])),
    'out': tf.Variable(tf.random_normal([num_classes]))
}



# 定义神经网络
def neural_net(x):
    # 第一层隐藏层（256个神经元）
    layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'])
    # 第二层隐藏层（256个神经元）
    layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'])
    # 输出层
    out_layer = tf.matmul(layer_2, weights['out']) + biases['out']
    return out_layer



# 构建模型
logits = neural_net(X)

# 定义损失函数和优化器
loss_op = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(
    logits=logits, labels=Y))
optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate)
train_op = optimizer.minimize(loss_op)

# 定义预测准确率
correct_pred = tf.equal(tf.argmax(logits, 1), tf.argmax(Y, 1))
accuracy = tf.reduce_mean(tf.cast(correct_pred, tf.float32))

# 初始化所有变量（赋默认值）
init = tf.global_variables_initializer()



# 开始训练
with tf.Session() as sess:

    # 执行初始化操作
    sess.run(init)

    for step in range(1, num_steps+1):
        batch_x, batch_y = mnist.train.next_batch(batch_size)
        # 执行训练操作，包括前向和后向传播
        sess.run(train_op, feed_dict={X: batch_x, Y: batch_y})
        if step % display_step == 0 or step == 1:
            # 计算损失值和准确率
            loss, acc = sess.run([loss_op, accuracy], feed_dict={X: batch_x,
                                                                 Y: batch_y})
            print("Step " + str(step) + ", Minibatch Loss= " + 
                  "{:.4f}".format(loss) + ", Training Accuracy= " + 
                  "{:.3f}".format(acc))

    print("Optimization Finished!")

    # 计算测试数据的准确率
    print("Testing Accuracy:", 
        sess.run(accuracy, feed_dict={X: mnist.test.images,
                                      Y: mnist.test.labels}))

• 在容器中使用 TensorFlow

VM vs Docker Container