`LayerNorm`

Layer normalisation with an optional affine transform.

class LayerNorm : Module— python/vkml/nn.py:814

Calls layer_norm for the normalisation, then applies weight and bias itself — the operator is the normalisation alone, and this layer owns the affine part.

Construction¶

`init`¶

def __init__self, normalized_shape, eps: float=1e-05, elementwise_affine: bool=True— python/vkml/nn.py:821

Forward¶

`forward`¶

def forwardself, x: V.Tensor -> V.Tensor— python/vkml/nn.py:835

Internals¶

`repr`¶

def __repr__self -> str— python/vkml/nn.py:841

See also layer_norm, rms_norm, BatchNorm2d, Module

LayerNorm