def reset_parameters(self):
if self.in_channels <= 0:
pass
elif self.weight_initializer == 'glorot':
inits.glorot(self.weight)
elif self.weight_initializer == 'uniform':
bound = 1.0 / math.sqrt(self.weight.size(-1))
torch.nn.init.uniform_(self.weight.data, -bound, bound)
elif self.weight_initializer == 'kaiming_uniform':
inits.kaiming_uniform(self.weight,
fan=self.in_channels,
a=math.sqrt(5))
elif self.weight_initializer is None:
inits.kaiming_uniform(self.weight,
fan=self.in_channels,
a=math.sqrt(5))
else:
raise RuntimeError(f"Linear layer weight initializer "
f"'{self.weight_initializer}' is not supported")
if self.bias is None or self.in_channels <= 0:
pass
elif self.bias_initializer == 'zeros':
inits.zeros(self.bias)
elif self.bias_initializer is None:
inits.uniform(self.in_channels, self.bias)
else:
raise RuntimeError(f"Linear layer bias initializer "
f"'{self.bias_initializer}' is not supported")
def __init__(self, in_channels, k, bias=True, **kwargs):
super(SACNConv, self).__init__(aggr='add', **kwargs)
assert k > 0
self.in_channels = in_channels
self.relationships_M = Parameter(uniform((in_channels, in_channels)))
self.chebyshev_coefficients = Parameter(uniform((k, )))
def test_inits():
x = torch.empty(1, 4)
uniform(size=4, tensor=x)
assert x.min() >= -0.5
assert x.max() <= 0.5
glorot(x)
assert x.min() >= -1.25
assert x.max() <= 1.25
zeros(x)
assert x.tolist() == [[0, 0, 0, 0]]
ones(x)
assert x.tolist() == [[1, 1, 1, 1]]
def reset_parameters(self):
if self.att is None:
size = self.num_relations * self.in_channels
else:
size = self.num_bases * self.in_channels
inits.uniform(size, self.att)
inits.uniform(size, self.basis)
inits.uniform(size, self.root)
inits.uniform(size, self.bias)
def __init__(self,
in_channels,
out_channels,
k,
linear_transform=False,
bias=True,
**kwargs):
super(LAGNConv, self).__init__(aggr='add', **kwargs)
num_basis_functions = kwargs.get("num_basis_functions", 10)
assert k > 0
self.in_channels = in_channels
self.out_channels = out_channels
self.inner_coefficients = Parameter(uniform((k, num_basis_functions)))
self.outer_coefficients = Parameter(uniform((k, num_basis_functions)))
self.lin = lambda x: x
if linear_transform:
self.lin = torch.nn.Linear(in_channels, out_channels, bias=bias)
def reset_parameters(self):
if self.in_channels > 0:
if self.weight_initializer == 'glorot':
inits.glorot(self.weight)
elif (self.weight_initializer == 'kaiming_uniform'
or self.weight_initializer is None):
inits.kaiming_uniform(self.weight,
fan=self.in_channels,
a=math.sqrt(5))
else:
raise RuntimeError(
f"Linear layer weight initializer "
f"'{self.weight_initializer}' is not supported")
if self.in_channels > 0 and self.bias is not None:
if self.bias_initializer == 'zeros':
inits.zeros(self.bias)
elif self.bias_initializer is None:
inits.uniform(self.in_channels, self.bias)
else:
raise RuntimeError(
f"Linear layer bias initializer "
f"'{self.bias_initializer}' is not supported")
def reset_parameters(self):
size = self.out_channels
uniform(size, self.weight)
self.rnn.reset_parameters()
def reset_parameters(self):
uniform(self.in_channels, self.weight)
uniform(self.in_channels, self.bias)
def reset_parameters(self):
glorot(self.weight)
zeros(self.bias)
uniform(1, self.alpha)
def reset_parameters(self):
inits.uniform(self.node_channels * self.out_channels, self.node_weight)
inits.uniform(self.edge_channels * self.out_channels, self.edge_weight)
inits.uniform(self.node_channels * self.out_channels,
self.neighbor_weight)
inits.zeros(self.bias)
def reset_parameters(self):
uniform(self.in_channels, self.weight)
self.lin1.reset_parameters()
self.lin2.reset_parameters()
def reset_parameters(self):
size = self.in_channels * 144 * self.K
uniform(size, self.weight)
uniform(size, self.bias)
def reset_parameters(self):
reset(self.nn)
uniform(self.in_channels, self.root)
uniform(self.in_channels, self.bias)
def reset_parameters(self):
size = self.num_bases * self.in_channels
uniform(size, self.basis)
uniform(size, self.att)
uniform(size, self.root)
uniform(size, self.bias)
def reset_parameters(self):
size = self.in_channels
uniform(size, self.Wc)
uniform(size, self.Wn)
uniform(size, self.bias)
def reset_parameters(self):
size = self.weight.size(0)
uniform(size, self.weight)
def reset_parameters(self):
uniform(self.weights.size(0), self.weights)
def reset_parameters(self):
if self.weights is not None:
uniform(self.weights.size(0), self.weights)
def reset_parameters(self):
size = self.num_relations * self.in_dim
glorot(self.basis)
uniform(size, self.bias)
def reset_parameters(self):
# uniform(self.in_channels, self.weight)
self.weight.data = nn.init.xavier_uniform_(
self.weight.data, gain=nn.init.calculate_gain('relu'))
uniform(self.in_channels, self.bias)
def reset_parameters(self):
size = self.heads * self.in_channels
uniform(size, self.weight)
uniform(size, self.att_weight)
uniform(size, self.bias)
def reset_parameters(self):
size = self.edge_attr_dim * self.in_channels
uniform(size, self.weight)
uniform(size, self.att_weight)
uniform(size, self.bias)
def reset_parameters(self):
size = self.in_channels
uniform(size, self.weight)
def reset_parameters(self):
uniform(self.out_channels, self.weight)
self.rnn.reset_parameters()
def reset_parameters(self):
inits.kaiming_uniform(self.weight, fan=self.in_channels,
a=math.sqrt(5))
inits.uniform(self.in_channels, self.bias)
def reset_parameters(self):
size = self.in_channels
uniform(size, self.fc.weight)
uniform(size, self.fc_attention.weight)
uniform(size, self.fc_edge_attr.weight)
