Tensorflow development template

Tensorflow development template

1. 加载 训练数据，测试数据

2. 定义 超参数

如 学习率，迭代次数，batch_size

3. 定义 Graph

定义 训练数据 占位符

x = tf.placeholder("float", [None, n_input])
y = tf.placeholder("float", [None, n_classes])

定义 模型参数

# Store layers weight & bias
weights = {
  'h1': tf.Variable(tf.random_normal([n_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, n_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([n_classes]))
}

定义 模型

# Create model
def multilayer_perceptron(x, weights, biases):
  # Hidden layer with RELU activation
  layer_1 = tf.add(tf.matmul(x, weights['h1']), biases['b1'])
  layer_1 = tf.nn.relu(layer_1)

  # Hidden layer with RELU activation
  layer_2 = tf.add(tf.matmul(layer_1, weights['h2']), biases['b2'])
  layer_2 = tf.nn.relu(layer_2)

  # Output layer with linear activation
  out_layer = tf.matmul(layer_2, weights['out']) + biases['out']

  return out_layer

# Construct model
pred = multilayer_perceptron(x, weights, biases)

定义 损失函数

cost = tf.reduce_mean(tf.nn.softmax_cross_entropy_with_logits(pred, y))

定义 优化函数

optimizer = tf.train.AdamOptimizer(learning_rate=learning_rate).minimize(cost)

定义 初始化变量init

# Initializing the variables
init = tf.initialize_all_variables()

定义 模型存储器

# 'Saver' op to save and restore all the variables
saver = tf.train.Saver()

4. 训练模型

启动 session

with tf.Session() as sess:

执行 初始化变量init

# Initialize variables
sess.run(init)

迭代执行 优化函数，并输出 当前的 损失值， 模型参数

# Training cycle
for epoch in range(3):

5. 测试模型

定义 correct_prediction

# Test model
correct_prediction = tf.equal(tf.argmax(pred, 1), tf.argmax(y, 1))

定义 accuracy

# Calculate accuracy
accuracy = tf.reduce_mean(tf.cast(correct_prediction, "float"))

执行 accuracy， 并输出

print "Accuracy:", accuracy.eval({x: mnist.test.images, y: mnist.test.labels})

6. 保存或加载模型

# Save model weights to disk
save_path = saver.save(sess, model_path)
# Restore model weights from previously saved model
load_path = saver.restore(sess, model_path)