tensorflow clip_by_norm in pytorch and numpy

clip_by_norm.py

import numpy as np
import torch
import tensorflow as tf
import numpy as np
import torch
import tensorflow as tf

# Numpy
np.random.seed(42)
x = np.random.randn(10, 10)
l2_norm = np.linalg.norm(x) 
print(l2_norm) # 9.095637931826113

clip_norm = 5.0
transformed_x = x * clip_norm / l2_norm
l2_norm = np.linalg.norm(transformed_x) 
print(l2_norm) # 5.0

# pytorch
torch_x = torch.tensor(x)
l2_norm = torch.norm(torch_x, p=2)
transformed_x = torch_x * clip_norm / l2_norm
l2_norm = torch.norm(transformed_x, p=2)
print(l2_norm) # 5.0

# tensorflow
transformed_x = tf.clip_by_norm(x, 5.0)
print(tf.norm(transformed_x, ord=2)) # 5.0