Source code for torch_geometric.nn.norm.layer_norm

from torch_geometric.typing import OptTensor

import torch
from torch.nn import Parameter
from torch import Tensor
from torch_scatter import scatter
from torch_geometric.utils import degree

from ..inits import ones, zeros

[docs]class LayerNorm(torch.nn.Module): r"""Applies layer normalization over each individual example in a batch of node features as described in the `"Layer Normalization" <>`_ paper .. math:: \mathbf{x}^{\prime}_i = \frac{\mathbf{x} - \textrm{E}[\mathbf{x}]}{\sqrt{\textrm{Var}[\mathbf{x}] + \epsilon}} \odot \gamma + \beta The mean and standard-deviation are calculated across all nodes and all node channels separately for each object in a mini-batch. Args: in_channels (int): Size of each input sample. eps (float, optional): A value added to the denominator for numerical stability. (default: :obj:`1e-5`) affine (bool, optional): If set to :obj:`True`, this module has learnable affine parameters :math:`\gamma` and :math:`\beta`. (default: :obj:`True`) """ def __init__(self, in_channels, eps=1e-5, affine=True): super(LayerNorm, self).__init__() self.in_channels = in_channels self.eps = eps if affine: self.weight = Parameter(torch.Tensor([in_channels])) self.bias = Parameter(torch.Tensor([in_channels])) else: self.register_parameter('weight', None) self.register_parameter('bias', None) self.reset_parameters()
[docs] def reset_parameters(self): ones(self.weight) zeros(self.bias)
[docs] def forward(self, x: Tensor, batch: OptTensor = None) -> Tensor: """""" if batch is None: x = x - x.mean() out = x / (x.std(unbiased=False) + self.eps) else: batch_size = int(batch.max()) + 1 norm = degree(batch, batch_size, dtype=x.dtype).clamp_(min=1) norm = norm.mul_(x.size(-1)).view(-1, 1) mean = scatter(x, batch, dim=0, dim_size=batch_size, reduce='add').sum(dim=-1, keepdim=True) / norm x = x - mean[batch] var = scatter(x * x, batch, dim=0, dim_size=batch_size, reduce='add').sum(dim=-1, keepdim=True) var = var / norm out = x / (var.sqrt()[batch] + self.eps) if self.weight is not None and self.bias is not None: out = out * self.weight + self.bias return out
def __repr__(self): return f'{self.__class__.__name__}({self.in_channels})'