01 tensorflow calculation model (I) - calculation diagram

TensorFlow Calculation model —— Calculation chart

About TensorFlow

TensorFlow It's a data flow graph （data flow graphs）, Open source software library for numerical calculation . node （Nodes） Show the mathematical operation in the figure , The lines in the picture （edges） It represents the multidimensional data array associated with each other among nodes , That is, tensor （tensor）. Its flexible architecture allows you to perform calculations on multiple platforms , For example, one or more of the desktop computers CPU（ or GPU）, The server , Mobile devices, etc .TensorFlow By the first Google Brain groups （ Affiliated to the Google Machine Intelligence Research Institute ） The researchers and engineers of , For machine learning and deep neural network research , But the versatility of this system makes it widely used in other computing fields

The calculation chart is TensorFlow The most basic concept in ,TensorFlow All calculations in will be converted into nodes on the calculation graph .

The concept of computational graph

TensorFlow The information you can get from its name is , It contains two of the most important concepts ——Tensor and Flow.

• Tensor The concept of

Tensor Tensor . The concept of tensor has different interpretations in mathematics or physics , But in TensorFlow It does not emphasize its own meaning , It is simply understood as a multidimensional array .

• Flow The concept of

if TensorFlow Medium Tensor Shows its data structure , that Flow It reflects its calculation model . Chinese will Flow Translated into “ flow ”, It intuitively expresses the process of mutual transformation between tensors through calculation .

• TensorFlow Calculation diagram in

TensorFlow It is a programming system that expresses calculation in the form of calculation diagram .TensorFlow Each calculation in is a node on the calculation graph , And the edge between nodes describes the dependency between calculations .

• Use of calculation diagram

TensorFlow The general flow of the program

First of all , Define all calculations in the calculation diagram

second , Perform calculations

• Calculation definition stage

Log in and copy

      
      

       
       import tensorflow as tf
       
       
g1 = tf.Graph()
       
       
with g1.as_default():
       
       
    #  In the calculation diagram “g1” Define variables in “v”, And set the initial value to 0.
       
       
    v = tf.get_variable("v",initializer=tf.zeros_initializer()(shape=[1]))
       
       
    
       
       
g2 = tf.Graph()
       
       
with g2.as_default():
       
       
    #  In the calculation diagram “g1” Define variables in “v”, And set the initial value to 1.
       
       
    v = tf.get_variable("v",initializer=tf.ones_initializer()(shape=[1]))
       
       
    
       
       
#  In the calculation diagram g1 Read variables in “v” The value of .
       
       
with tf.Session(graph=g1) as sess:
       
       
    tf.global_variables_initializer().run()
       
       
    with tf.variable_scope("",reuse=True):
       
       
        #  In the calculation diagram g1 in , Variable “v” The value of is 0, So the following line will output [0.].
       
       
        print(sess.run(tf.get_variable("v")))
       
       
        
       
       
#  In the calculation diagram g2 Read variables in “v” The value of .
       
       
with tf.Session(graph=g2) as sess:
       
       
    tf.global_variables_initializer().run()
       
       
    with tf.variable_scope("",reuse=True):
       
       
        #  In the calculation diagram g1 in , Variable “v” The value of is 0, So the following line will output [1.].
       
       
        print(sess.run(tf.get_variable("v")))
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
       
       
5.
       
       
6.
       
       
7.
       
       
8.
       
       
9.
       
       
10.
       
       
11.
       
       
12.
       
       
13.
       
       
14.
       
       
15.
       
       
16.
       
       
17.
       
       
18.
       
       
19.
       
       
20.
       
       
21.
       
       
22.
       
       
23.
       
       
24.
      
      

The above code produces two calculation diagrams , In each calculation diagram, a name is defined as “v” The variable of . In the calculation diagram g1 in , take v Initialize to 0; In the calculation diagram g2 in , take v Initialize to 1. It can be seen that when running different calculation diagrams , Variable v The value of is also different .TensorFlow The computational graph in can not only be used to isolate tensors and calculations , It also provides mechanisms for managing tensor and computation . The calculation diagram can be obtained by tf.Graph.device Function to specify the device that runs the calculation . This is TensorFlow Use GPU Provides a mechanism . The following program can soak the addition calculation in GPU On .

Log in and copy

      
      

       
       g = tf.Graph()
       
       
#  Run the specified computing device .
       
       
with g.device('/gpu:0'):
       
       
    result = a+b
       
       
    print(result)
       
       
    
      
      
      
      

       
       
1.
       
       
2.
       
       
3.
       
       
4.
       
       
5.
       
       
6.