torch 默认的初始化方法

查了下torch里的源码，如下：

conv

主要代码为：
可见，使用的是 kaiming_uniform 和 uniform初始化

def reset_parameters(self) -> None:
    init.kaiming_uniform_(self.weight, a=math.sqrt(5))
    if self.bias is not None:
        fan_in, _ = init._calculate_fan_in_and_fan_out(self.weight)
        bound = 1 / math.sqrt(fan_in)
        init.uniform_(self.bias, -bound, bound)

batchnorm

主要代码为：
可见是[0, 1]的正态分布。

def reset_running_stats(self) -> None:
    if self.track_running_stats:
        self.running_mean.zero_()
        self.running_var.fill_(1)
        self.num_batches_tracked.zero_()

nn.linear

主要代码为：
可知使用的是 kaiming_uniform_ 和 uniform_

def reset_parameters(self) -> None:
    init.kaiming_uniform_(self.weight, a=math.sqrt(5))
    if self.bias is not None:
        fan_in, _ = init._calculate_fan_in_and_fan_out(self.weight)
        bound = 1 / math.sqrt(fan_in)
        init.uniform_(self.bias, -bound, bound)

kaiming_uniform_可以参考

https://zhuanlan.zhihu.com/p/305055975

conv 源码

class _ConvNd(Module):

    __constants__ = ['stride', 'padding', 'dilation', 'groups',
                     'padding_mode', 'output_padding', 'in_channels',
                     'out_channels', 'kernel_size']
    __annotations__ = {'bias': Optional[torch.Tensor]}

    _in_channels: int
    out_channels: int
    kernel_size: Tuple[int, ...]
    stride: Tuple[int, ...]
    padding: Tuple[int, ...]
    dilation: Tuple[int, ...]
    transposed: bool
    output_padding: Tuple[int, ...]
    groups: int
    padding_mode: str
    weight: Tensor
    bias: Optional[Tensor]

    def __init__(self,
                 in_channels: int,
                 out_channels: int,
                 kernel_size: _size_1_t,
                 stride: _size_1_t,
                 padding: _size_1_t,
                 dilation: _size_1_t,
                 transposed: bool,
                 output_padding: _size_1_t,
                 groups: int,
                 bias: Optional[Tensor],
                 padding_mode: str) -> None:
        super(_ConvNd, self).__init__()
        if in_channels % groups != 0:
            raise ValueError('in_channels must be divisible by groups')
        if out_channels % groups != 0:
            raise ValueError('out_channels must be divisible by groups')
        valid_padding_modes = {'zeros', 'reflect', 'replicate', 'circular'}
        if padding_mode not in valid_padding_modes:
            raise ValueError("padding_mode must be one of {}, but got padding_mode='{}'".format(
                valid_padding_modes, padding_mode))
        self.in_channels = in_channels
        self.out_channels = out_channels
        self.kernel_size = kernel_size
        self.stride = stride
        self.padding = padding
        self.dilation = dilation
        self.transposed = transposed
        self.output_padding = output_padding
        self.groups = groups
        self.padding_mode = padding_mode
        # `_reversed_padding_repeated_twice` is the padding to be passed to
        # `F.pad` if needed (e.g., for non-zero padding types that are
        # implemented as two ops: padding + conv). `F.pad` accepts paddings in
        # reverse order than the dimension.
        self._reversed_padding_repeated_twice = _reverse_repeat_tuple(self.padding, 2)
        if transposed:
            self.weight = Parameter(torch.Tensor(
                in_channels, out_channels // groups, *kernel_size))
        else:
            self.weight = Parameter(torch.Tensor(
                out_channels, in_channels // groups, *kernel_size))
        if bias:
            self.bias = Parameter(torch.Tensor(out_channels))
        else:
            self.register_parameter('bias', None)
        self.reset_parameters()

    def reset_parameters(self) -> None:  # 这里
        init.kaiming_uniform_(self.weight, a=math.sqrt(5))
        if self.bias is not None:
            fan_in, _ = init._calculate_fan_in_and_fan_out(self.weight)
            bound = 1 / math.sqrt(fan_in)
            init.uniform_(self.bias, -bound, bound)

batchnorm 源码

class _NormBase(Module):
    """Common base of _InstanceNorm and _BatchNorm"""
    _version = 2
    __constants__ = ['track_running_stats', 'momentum', 'eps',
                     'num_features', 'affine']
    num_features: int
    eps: float
    momentum: float
    affine: bool
    track_running_stats: bool
    # WARNING: weight and bias purposely not defined here.
    # See https://github.com/pytorch/pytorch/issues/39670

    def __init__(
        self,
        num_features: int,
        eps: float = 1e-5,
        momentum: float = 0.1,
        affine: bool = True,
        track_running_stats: bool = True
    ) -> None:
        super(_NormBase, self).__init__()
        self.num_features = num_features
        self.eps = eps
        self.momentum = momentum
        self.affine = affine
        self.track_running_stats = track_running_stats
        if self.affine:
            self.weight = Parameter(torch.Tensor(num_features))
            self.bias = Parameter(torch.Tensor(num_features))
        else:
            self.register_parameter('weight', None)
            self.register_parameter('bias', None)
        if self.track_running_stats:
            self.register_buffer('running_mean', torch.zeros(num_features))
            self.register_buffer('running_var', torch.ones(num_features))
            self.register_buffer('num_batches_tracked', torch.tensor(0, dtype=torch.long))
        else:
            self.register_parameter('running_mean', None)
            self.register_parameter('running_var', None)
            self.register_parameter('num_batches_tracked', None)
        self.reset_parameters()

    def reset_running_stats(self) -> None:   # 这里
        if self.track_running_stats:
            self.running_mean.zero_()
            self.running_var.fill_(1)
            self.num_batches_tracked.zero_()

    def reset_parameters(self) -> None:
        self.reset_running_stats()
        if self.affine:
            init.ones_(self.weight)
            init.zeros_(self.bias)

nn.linear 源码

class Linear(Module):
    r"""Applies a linear transformation to the incoming data: :math:`y = xA^T + b`

    This module supports :ref:`TensorFloat32<tf32_on_ampere>`.

    Args:
        in_features: size of each input sample
        out_features: size of each output sample
        bias: If set to ``False``, the layer will not learn an additive bias.
            Default: ``True``

    Shape:
        - Input: :math:`(N, *, H_{in})` where :math:`*` means any number of
          additional dimensions and :math:`H_{in} = \text{in\_features}`
        - Output: :math:`(N, *, H_{out})` where all but the last dimension
          are the same shape as the input and :math:`H_{out} = \text{out\_features}`.

    Attributes:
        weight: the learnable weights of the module of shape
            :math:`(\text{out\_features}, \text{in\_features})`. The values are
            initialized from :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})`, where
            :math:`k = \frac{1}{\text{in\_features}}`
        bias:   the learnable bias of the module of shape :math:`(\text{out\_features})`.
                If :attr:`bias` is ``True``, the values are initialized from
                :math:`\mathcal{U}(-\sqrt{k}, \sqrt{k})` where
                :math:`k = \frac{1}{\text{in\_features}}`

    Examples::

        >>> m = nn.Linear(20, 30)
        >>> input = torch.randn(128, 20)
        >>> output = m(input)
        >>> print(output.size())
        torch.Size([128, 30])
    """
    __constants__ = ['in_features', 'out_features']
    in_features: int
    out_features: int
    weight: Tensor

    def __init__(self, in_features: int, out_features: int, bias: bool = True) -> None:
        super(Linear, self).__init__()
        self.in_features = in_features
        self.out_features = out_features
        self.weight = Parameter(torch.Tensor(out_features, in_features))
        if bias:
            self.bias = Parameter(torch.Tensor(out_features))
        else:
            self.register_parameter('bias', None)
        self.reset_parameters()

    def reset_parameters(self) -> None:
        init.kaiming_uniform_(self.weight, a=math.sqrt(5))
        if self.bias is not None:
            fan_in, _ = init._calculate_fan_in_and_fan_out(self.weight)
            bound = 1 / math.sqrt(fan_in)
            init.uniform_(self.bias, -bound, bound)