Function Differences with tf.clip_by_value

tf.clip_by_value

tf.clip_by_value(t, clip_value_min, clip_value_max, name=None) -> Tensor

For more information, see tf.clip_by_value.

mindspore.ops.clip_by_value

mindspore.ops.clip_by_value(x, clip_value_min=None, clip_value_max=None) -> Tensor

For more information, see mindspore.ops.clip_by_value.

Differences

TensorFlow: Given a tensor t, this operation returns a tensor of the same type and shape as t. Any value less than clip_value_min in t is set to clip_value_min, and any value greater than clip_value_max is set to clip_value_max. When clip_value_min is greater than clip_value_max, the value of the tensor will be set to clip_value_min.

MindSpore: When clip_value_min is less than or equal to clip_value_max, MindSpore API implements the same function as TensorFlow. When clip_value_min is greater than clip_value_max, the value of the tensor element will be set to clip_value_max.

Categories	Subcategories	TensorFlow	MindSpore	Differences
Parameters	Parameter 1	t	x	Same function, different parameter names
	Parameter 2	clip_value_min	clip_value_min	-
	Parameter 3	clip_value_max	clip_value_max	-
	Parameter 4	name	-	Not involved

Code Example 1

When clip_value_min is less than or equal to clip_value_max, the two APIs achieve the same function and have the same usage.

# TensorFlow
import tensorflow as tf
t = tf.constant([[1., 25., 5., 7.], [4., 11., 6., 21.]])
t2 = tf.clip_by_value(t, clip_value_min=5, clip_value_max=22)
print(t2.numpy())
#[[ 5. 22.  5.  7.]
# [ 5. 11.  6. 21.]]

# MindSpore
import mindspore
from mindspore import Tensor, ops
import numpy as np
input = Tensor(np.array([[1., 25., 5., 7.], [4., 11., 6., 21.]]), mindspore.float32)
output = ops.clip_by_value(input, clip_value_min=5, clip_value_max=22)
print(output)
#[[ 5. 22.  5.  7.]
# [ 5. 11.  6. 21.]]

Code Example 2

When clip_value_min is greater than clip_value_max, TensorFlow will set the value of the tensor to clip_value_min and MindSpore will set it to clip_value_max.

# TensorFlow
import tensorflow as tf
t = tf.constant([[1., 25., 5., 7.], [4., 11., 6., 21.]])
t2 = tf.clip_by_value(t, clip_value_min=22, clip_value_max=5)
print(t2.numpy())
#[[ 22. 22. 22. 22.]
# [ 22. 22. 22. 22.]]

# MindSpore
import mindspore
from mindspore import Tensor, ops
import numpy as np
input = Tensor(np.array([[1., 25., 5., 7.], [4., 11., 6., 21.]]), mindspore.float32)
output = ops.clip_by_value(input, clip_value_min=22, clip_value_max=5)
print(output)
#[[ 5. 5. 5. 5.]
# [ 5. 5. 5. 5.]]