neuraltrain.models.common.LayerScale

class neuraltrain.models.common.LayerScale(channels: int, init: float = 0.1, boost: float = 5.0)

Layer scale from [Touvron et al 2021] (https://arxiv.org/pdf/2103.17239.pdf). This rescales diagonaly residual outputs close to 0 initially, then learnt.

forward(x)

Define the computation performed at every call.

Should be overridden by all subclasses.

Note

Although the recipe for forward pass needs to be defined within this function, one should call the Module instance afterwards instead of this since the former takes care of running the registered hooks while the latter silently ignores them.