Lucid3.0
fn

clamp

→Tensor
clamp(input: Tensor, min: Scalar | None, max: Scalar | None)
source

Element-wise clamping to a closed interval.

Constrains each entry of input to lie within [min, max]. Either bound may be None to leave that side unbounded.

Parameters

inputTensor
Input tensor.
minscalar or None
Lower bound. None disables the lower clip.
maxscalar or None
Upper bound. None disables the upper clip.

Returns

Tensor

Tensor with the same shape and dtype as input, with values constrained to [min, max].

Notes

Mathematical definition:

outi=min⁡(max⁡(inputi,min),max)\text{out}_i = \min(\max(\text{input}_i, \text{min}), \text{max})

Gradient flows through where the input lies strictly inside the interval and is zero at the clipped boundaries. Equivalent to clip.

Examples

>>> import lucid
>>> x = lucid.tensor([-3.0, 0.0, 3.0])
>>> lucid.clamp(x, -1.0, 1.0)
Tensor([-1., 0., 1.])