def __init__(self, model, input_dim, output_dim, hidden_dim, activation='sigmoid'):
super().__init__(model, input_dim, output_dim, hidden_dim)
if self.method is NeuralLogicInferenceMethod.MLP:
self.layer.add_module(str(len(self.layer)), get_activation(activation))
else:
raise NotImplementedError('Unknown logic inference method: {}.'.format(self.method))
def __init__(self, in_channels, out_channels, kernel_size, stride=1,
padding_mode='default', padding=0, border_mode='zero',
dilation=1, groups=1,
batch_norm=None, dropout=None, bias=None, activation=None):
if bias is None:
bias = (batch_norm is None)
nr_dims = type(self).__nr_dims__
clz_name = 'Conv{}d'.format(nr_dims)
modules = [getattr(conv, clz_name)(
in_channels, out_channels, kernel_size,
stride=stride, padding_mode=padding_mode, padding=padding, border_mode=border_mode,
dilation=dilation, groups=groups, bias=bias
)]
if batch_norm is not None and batch_norm is not False:
modules.append(get_batcnnorm(batch_norm, out_channels, nr_dims))
if dropout is not None and dropout is not False:
modules.append(get_dropout(dropout, nr_dims))
if activation is not None and activation is not False:
modules.append(get_activation(activation))
super().__init__(*modules)
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.padding_mode = padding_mode
self.padding = padding
self.border_mode = border_mode
self.dilation = dilation
self.groups = groups
def __init__(self, in_channels, out_channels, kernel_size, stride=None,
padding_mode='same', padding=0, border_mode=None,
dilation=1, groups=1,
output_size=None, scale_factor=None, resize_mode='nearest',
batch_norm=None, dropout=None, bias=None, activation=None,
algo='resizeconv'):
super().__init__()
self.in_channels = in_channels
self.out_channels = out_channels
self.kernel_size = kernel_size
self.stride = stride
self.padding_mode = padding_mode
self.padding = padding
self.border_mode = border_mode
self.dilation = dilation
self.groups = groups
if bias is None:
bias = (batch_norm is None)
self.algo = algo = DeconvAlgo.from_string(algo)
nr_dims = type(self).__nr_dims__
if algo is DeconvAlgo.CONVTRANSPOSE:
clz_name = 'ConvTranspose{}d'.format(nr_dims)
assert scale_factor is not None
assert stride is None, 'Can not set strides for Conv-Transpose based Deconv.'
assert border_mode is None, 'Can not set strides for Conv-Transpose based Deconv.'
self.deconv = getattr(conv, clz_name)(
in_channels, out_channels, kernel_size, stride=scale_factor,
padding_mode=padding_mode, padding=padding, border_mode='zero',
dilation=dilation, groups=groups, bias=bias
)
elif algo is DeconvAlgo.RESIZECONV:
clz_name = 'ResizeConv{}d'.format(nr_dims)
stride = stride or 1
border_mode = border_mode or 'replicate'
self.deconv = getattr(conv, clz_name)(
in_channels, out_channels, kernel_size, stride=stride,
padding_mode=padding_mode, padding=padding, border_mode=border_mode,
dilation=dilation, groups=groups, bias=bias,
output_size=output_size, scale_factor=scale_factor, resize_mode=resize_mode
)
self.output_size = output_size
self.scale_factor = scale_factor
post_modules = []
if batch_norm is not None and batch_norm is not False:
post_modules.append(get_batcnnorm(batch_norm, out_channels, nr_dims))
if dropout is not None and dropout is not False:
post_modules.append(get_dropout(dropout, nr_dims))
if activation is not None and activation is not False:
post_modules.append(get_activation(activation))
self.post_process = nn.Sequential(*post_modules)
def __init__(self, in_features, out_features, batch_norm=None, dropout=None, bias=None, activation=None):
if bias is None:
bias = (batch_norm is None)
modules = [nn.Linear(in_features, out_features, bias=bias)]
if batch_norm is not None and batch_norm is not False:
modules.append(get_batcnnorm(batch_norm, out_features, 1))
if dropout is not None and dropout is not False:
modules.append(get_dropout(dropout, 1))
if activation is not None and activation is not False:
modules.append(get_activation(activation))
super().__init__(*modules)
self.in_features = in_features
self.out_features = out_features
