TensorFlow Graphs



In [1]:

    
import tensorflow as tf



In [2]:

    
n1 = tf.constant(1)
n2 = tf.constant(2)



In [3]:

    
n3 = n1 + n2



In [4]:

    
# Using with auto-closes the session
with tf.Session() as sess:
    result = sess.run(n3)
print(result)

When you start TF, a default Graph is created, you can create additional graphs easily:



In [5]:

    
print(tf.get_default_graph())









    



<tensorflow.python.framework.ops.Graph object at 0x0000011BEFB3ADA0>



In [6]:

    
g = tf.Graph()



In [7]:

    
print(g)









    



<tensorflow.python.framework.ops.Graph object at 0x0000011C6F652438>

Setting a graph as the default:



In [8]:

    
graph_one = tf.get_default_graph()
graph_two = tf.Graph()



In [9]:

    
graph_one is tf.get_default_graph()









    Out[9]:





True



In [10]:

    
graph_two is tf.get_default_graph()









    Out[10]:





False



In [11]:

    
with graph_two.as_default():
    print(graph_two is tf.get_default_graph())









    



True



In [12]:

    
graph_two is tf.get_default_graph()









    Out[12]:





False