tf.clip_norm_clipnorm

作者：羊村懒王 | 2024-03-25 17:49:18

踩

clipnorm

t= [ $x_{1}$ , $x_{2}$ ,… $x_{2}$ ]

l2= $\sqrt{\displaystyle\sum\limits_{i=0}^n x_i^2}$
norm = 2.0

clip_norm $=\frac{t*norm}{l2}$

import numpy as np

t=np.array([[1, 2, 3, 4, 5]])
l2norm4t = np.linalg.norm(t)
clip_norm = 2.0
transformed_t = t *clip_norm/l2norm4t
print(transformed_t)

1
2
3
4
5
6
7
8

[[0.26967996 0.5393599  0.80903983 1.0787199  1.3483998 ]]
1

some_nums = tf.constant([[1, 2, 3, 4, 5]], dtype=tf.float32)
print(tf.clip_by_norm(some_nums, 2.0).numpy())

1
2
3

[[0.26967996 0.5393599  0.80903983 1.0787199  1.3483998 ]]
1

声明：本文内容由网友自发贡献，不代表【wpsshop博客】立场，版权归原作者所有，本站不承担相应法律责任。如您发现有侵权的内容，请联系我们。转载请注明出处：https://www.wpsshop.cn/w/羊村懒王/article/detail/311554

tf.clip_norm_clipnorm

clip_norm = t ∗ n o r m l 2 =\frac{t*norm}{l2} =l2t∗norm​

clip_norm $=\frac{t*norm}{l2}$