Python – tensorflow.clip_by_norm()

TensorFlow 是由 Google 设计的开源Python库，用于开发机器学习模型和深度学习神经网络.

clip_by_norm()用于将张量值裁剪为最大 L2 范数。

Syntax: tensorflow.clip_by_norm(t, clip_norm, axes, name)

Parameters:

• t: It is the input tensor that need to be clipped.
• clip_norm: It is 0-D scalar tensor which defines the maximum clipping value.
• axes(optional): It’s 1-D vector tensor which defines the dimension to be used for calculating L2norm. If none is provided all dimensions will be used.
• name(optional): It defines the name for the operation.

Returns:

It returns a Tensor.

示例 1：

Python3

# Importing the library
import tensorflow as tf
  
# Initializing the input tensor
t = tf.constant([1, 2, 3, 4], dtype = tf.float64)
clip_norm = .8
  
# Printing the input tensor
print('t: ', t)
print('clip_norm: ', clip_norm)
  
# Calculating tangent
res = tf.clip_by_norm(t, clip_norm)
  
# Printing the result
print('Result: ', res)

输出：

t:  tf.Tensor([1. 2. 3. 4.], shape=(4, ), dtype=float64)
clip_norm:  0.8
Result:  tf.Tensor([0.14605935 0.2921187  0.43817805 0.58423739], shape=(4, ), dtype=float64)

示例 2：

Python3

# Importing the library
import tensorflow as tf
  
# Initializing the input tensor
t = tf.constant([1, 2, 3, 4], dtype = tf.float64)
clip_norm = 5.2
  
  
# Printing the input tensor
print('t: ', t)
print('clip_norm: ', clip_norm)
  
# Calculating tangent
res = tf.clip_by_norm(t, clip_norm)
  
# Printing the result
print('Result: ', res)

输出：

t:  tf.Tensor([1. 2. 3. 4.], shape=(4, ), dtype=float64)
clip_norm:  5.2
Result:  tf.Tensor([0.94938577 1.89877153 2.8481573  3.79754307], shape=(4, ), dtype=float64)