mindspore.ops.clip_by_value

mindspore.ops.clip_by_value(x, clip_value_min=None, clip_value_max=None)[source]

Clips tensor values to a specified min and max.

Limits the value of \(x\) to a range, whose lower limit is clip_value_min and upper limit is clip_value_max .

\[\begin{split}out_i= \left\{ \begin{array}{align} clip\_value\_max & \text{ if } x_i\ge clip\_value\_max \\ x_i & \text{ if } clip\_value\_min \lt x_i \lt clip\_value\_max \\ clip\_value\_min & \text{ if } x_i \le clip\_value\_min \\ \end{array}\right.\end{split}\]

Note

clip_value_min needs to be less than or equal to clip_value_max . The data type of x, clip_value_min and clip_value_max should support implicit type conversion and cannot all be bool type.

Parameters

x (Tensor) – Input data. The shape is \((N,*)\) where \(*\) means, any number of additional dimensions.
clip_value_min (Tensor) – The minimum value. clip_value_min and clip_value_max cannot be all None. Default: None.
clip_value_max (Tensor) – The maximum value. clip_value_min and clip_value_max cannot be all None. Default: None.

Returns

Tensor, a clipped Tensor. The data type is the one with higher precision or higher digits among the x, clip_value_min and clip_value_max .

Supported Platforms:: Ascend GPU CPU

Examples

>>> from mindspore import Tensor, ops
>>> import numpy as np
>>> min_value = Tensor(5, mindspore.float32)
>>> max_value = Tensor(20, mindspore.float32)
>>> x = Tensor(np.array([[1., 25., 5., 7.], [4., 11., 6., 21.]]), mindspore.float32)
>>> output = ops.clip_by_value(x, min_value, max_value)
>>> print(output)
[[ 5. 20.  5.  7.]
 [ 5. 11.  6. 20.]]