relu6

→Tensor

relu6(x: Tensor, inplace: bool = False)

source

Rectified linear unit clipped at six.

Introduced for quantisation-friendly mobile architectures (MobileNetV1 / V2): bounding ReLU above keeps activations inside a small fixed range so that low-precision integer arithmetic does not overflow.

Parameters

xTensor

Input tensor of any shape; activation is element-wise.

inplacebool= False

Accepted for API compatibility; currently ignored.

Returns

Tensor

Activated tensor with the same shape as x, values in $[0, 6]$ .

Notes

\text{ReLU6}(x) = \min(\max(0, x), 6)

Derivative is $1$ on $(0, 6)$ and $0$ outside — so beyond suffering from "dead neurons" on the negative side, ReLU6 also kills gradients above $6$ . The trade-off is worthwhile for int8 quantisation where the symmetric $[0, 6]$ range maps cleanly to a small dynamic-range integer codebook.

Examples

>>> import lucid
>>> from lucid.nn.functional import relu6
>>> x = lucid.tensor([-1.0, 0.0, 3.0, 7.0])
>>> relu6(x)
Tensor([0.0000, 0.0000, 3.0000, 6.0000])