megengine.functional.clip¶

clip(x, lower=None, upper=None)[源代码]¶

Clamps all elements in input tensor into the range [ lower, upper ] and returns a resulting tensor:

\[\begin{split}y_i = \begin{cases} \text{lower} & \text{if } x_i < \text{lower} \\ x_i & \text{if } \text{lower} \leq x_i \leq \text{upper} \\ \text{upper} & \text{if } x_i > \text{upper} \end{cases}\end{split}\]

参数

x (Tensor) – (Tensor): The input tensor.
lower – (Numberic，optional): lower-bound of the range to be clamped to.
upper – (Numberic，optional): upper-bound of the range to be clamped to.

注解

If both lower and upper are None, raises an AssertionError.
If lower is bigger than upper, the result is same as clip(Tensor(), upper, upper).

返回类型: Tensor
返回: output clamped tensor. The result must have a data type determined by 类型提升规则.

实际案例

>>> import numpy as np
>>> x = Tensor([0,1,2,3,4])
>>> F.clip(x, 2, 4)
Tensor([2 2 2 3 4], dtype=int32, device=xpux:0)

>>> x = Tensor([0,1,2,3,4])
>>> F.clip(x, 4, 3)
Tensor([3 3 3 3 3], dtype=int32, device=xpux:0)

>>> x = F.arange(5)
>>> F.clip(x, lower=3)
Tensor([3. 3. 3. 3. 4.], device=xpux:0)

>>> x = F.arange(5, dtype=np.int32)
>>> F.clip(x, upper=2.1)
Tensor([0.  1.  2.  2.1 2.1], device=xpux:0)

megengine.functional.floor

megengine.functional.exp