• import tensorflow as tf
  
  
• 
  
  with tf.name_scope("data-set"):
  
  
• 
  
      rlog = "./singlecell"
  
  
• 
  
      x = tf.constant(2.0, name = "input")
  
  
• 
  
      w = tf.Variable(0.8, name = "weight")
  
  
• 
  
      y_predict = tf.multiply(w, x, name = "output")
  
  
• 
  
      y = tf.constant(0.0, name = "real_result")
  
  
• 
  
      loss = tf.pow(y_predict - y, 2, name = "loss")
  
  
• 
  
  
  
  
• 
  
  with tf.name_scope("train"):
  
  
• 
  
      train_step = tf.train.GradientDescentOptimizer(0.001).minimize(loss)
  
  
• 
  
  with tf.name_scope("summar"):
  
  
• 
  
      for value in [x, w, y_predict, y, loss]:
  
  
• 
  
          tf.summary.scalar(value.op.name, value)
  
  
• 
  
      summaries = tf.summary.merge_all()
  
  
• 
  
  
  
  
• 
  
  ss = tf.Session()
  
  
• 
  
  xsum = tf.summary.FileWriter(rlog, ss.graph)
  
  
• 
  
  init = tf.global_variables_initializer()
  
  
• 
  
  ss.run(init)
  
  
• 
  
  
  
  
• 
  
  with tf.name_scope("session"):
  
  
• 
  
      for i in range(1000):
  
  
• 
  
          xdat = ss.run(summaries)
  
  
• 
  
          xsum.add_summary(xdat, i)
  
  
• 
  
          xdat = ss.run(train_step)
  
  
• 
  
          x2 = ss.run(x)
  
  
• 
  
          y2 = ss.run(y_predict)
  
  
• 
  
          w2 = ss.run(w)
  
  
• 
  
          s2 = ss.run(loss)
  
  
• 
  
          if i % 100 == 0:
  
  
• 
  
              print(i, "#, y2:",y2, ", s2:", s2, ", x2:", x2, ", w2:", w2 )
  
  
• 
  
  ss.close()
  
  
  

结果是：

• 
  
  0 #, y2: 1.5872 , s2: 2.5192 , x2: 2.0 , w2: 0.7936
  
  
• 
  
  100 #, y2: 0.710885 , s2: 0.505357 , x2: 2.0 , w2: 0.355442
  
  
• 
  
  200 #, y2: 0.318395 , s2: 0.101375 , x2: 2.0 , w2: 0.159197
  
  
• 
  
  300 #, y2: 0.142605 , s2: 0.020336 , x2: 2.0 , w2: 0.0713023
  
  
• 
  
  400 #, y2: 0.0638705 , s2: 0.00407944 , x2: 2.0 , w2: 0.0319353
  
  
• 
  
  500 #, y2: 0.0286067 , s2: 0.000818343 , x2: 2.0 , w2: 0.0143034
  
  
• 
  
  600 #, y2: 0.0128125 , s2: 0.000164161 , x2: 2.0 , w2: 0.00640627
  
  
• 
  
  700 #, y2: 0.00573855 , s2: 3.2931e-05 , x2: 2.0 , w2: 0.00286928
  
  
• 
  
  800 #, y2: 0.00257022 , s2: 6.60601e-06 , x2: 2.0 , w2: 0.00128511
  
  
• 
  
  900 #, y2: 0.00115116 , s2: 1.32518e-06 , x2: 2.0 , w2: 0.000575581
  
  
  

在cmd窗口输入：

tensorboard -logdir=C:\Users\wx\tensorflowcode\singlecell

然后在浏览器的地址栏中输入：

http://computer:6006

可以看到如下结果：

整个训练的数据流图如下：

奈斯