def __init__()

in models/cifar/resnext_cnsn.py [0:0]

• 55 lines of code
• 5 McCabe index (conditional complexity)

  def __init__(self,
               inplanes,
               planes,
               cardinality,
               base_width,
               norm_func,
               pos, beta, crop, cnsn_type,
               stride=1,
               downsample=None):
    super(ResNeXtBottleneckCustom, self).__init__()

    dim = int(math.floor(planes * (base_width / 64.0)))

    self.conv_reduce = nn.Conv2d(
        inplanes,
        dim * cardinality,
        kernel_size=1,
        stride=1,
        padding=0,
        bias=False)
    self.bn_reduce = norm_func(dim * cardinality)

    self.conv_conv = nn.Conv2d(
        dim * cardinality,
        dim * cardinality,
        kernel_size=3,
        stride=stride,
        padding=1,
        groups=cardinality,
        bias=False)
    self.bn = norm_func(dim * cardinality)

    self.conv_expand = nn.Conv2d(
        dim * cardinality,
        planes * 4,
        kernel_size=1,
        stride=1,
        padding=0,
        bias=False)
    self.bn_expand = norm_func(planes * 4)

    self.downsample = downsample

    assert cnsn_type in ['sn', 'cn', 'cnsn']

    if 'cn' in cnsn_type:
        print('using CrossNorm with crop: {}'.format(crop))
        crossnorm = CrossNorm(crop=crop, beta=beta)
    else:
        crossnorm = None

    if 'sn' in cnsn_type:
        print('using SelfNorm')
        if pos in ['pre', 'identity']:
            selfnorm = SelfNorm(inplanes)
        else:
            selfnorm = SelfNorm(planes * 4)
    else:
        selfnorm = None

    self.cnsn = CNSN(crossnorm=crossnorm, selfnorm=selfnorm)

    self.pos = pos
    # if pos is not None:
    print('{} in residual module: {}'.format(cnsn_type, pos))
    assert pos in ['residual', 'identity', 'pre', 'post']