Tf.constant usage

The function prototype

tf.constant(
    value,
    dtype=None,
    shape=None,
    name='Const',
    verify_shape=False
)

• The first value is required , It can be a number , It can also be a list ; In order to view the results, you must create a session , And use the value function eval() To view the created tensor Value ; The last four parameters can be written or not
• The second parameter represents the data type , It can generally be tf.float32, tf.float64 etc.
• The third parameter represents the tensor “ shape ”, That is, the dimension and the size of each dimension . If the third parameter is specified , When the first parameter value When it's numbers , All elements of the tensor are filled with this number ;( And when the first parameter value When it's a list , Note that the length of the list must be less than or equal to the third parameter shape Size ( That is, the product of the dimensions ), Otherwise, an error will be reported ;); This is because the function generates a shape The tensor of size , And then use value The values in this list are filled one by one shape The elements in . Here the list size is 7, and shape The size is 2*3=6, Cannot fill correctly , So something went wrong . And if the list size is smaller than shape size , Then the remaining tensor elements will be filled with the last element of the list ：
• Fourth parameter name It can be anything , Mainly string .
• Fifth parameter verify_shape The default is False, If it's changed to True Words mean inspection value The shape and shape of shape Is it consistent with , If it does not match, an error will be reported .

Code

import tensorflow._api.v2.compat.v1 as tf
tf.disable_v2_behavior()

''' tf.constant( value, dtype=None, shape=None, name='Const', verify_shape=False ) '''

#  The first parameter is a value 
tmp = tf.constant(1)
sess = tf.Session()
with sess.as_default():
    print(tmp.eval())
#  The first parameter is a list 
tmp = tf.constant([1, 2])
sess = tf.Session()
with sess.as_default():
    print(tmp.eval())
#  Specify the second parameter , Specify the data type 
tmp = tf.constant([1, 2], dtype = tf.float32)
sess = tf.Session()
with sess.as_default():
    print(tmp.eval())
#  Specify the third parameter :  For specifying tensors " shape "
tmp = tf.constant(3, shape=[2, 2])
sess = tf.Session()
with sess.as_default():
    print(tmp.eval())
#  Specify the fourth parameter :  character string 
tmp = tf.constant([2, 3], dtype = tf.float32, shape=[2,2], name = "LiQuan")
print(tmp)

Output :

''' [1. 2.] [[3 3] [3 3]] Tensor("LiQuan:0", shape=(2, 2), dtype=float32) '''