537de129aff815d5690e46858c7e5c1de0d4abd6,torch_geometric/nn/modules/lin.py,Lin,reset_parameters,#Lin#,34

Before Change

        self.reset_parameters()

    def reset_parameters(self):
        stdv = 1. / math.sqrt(self.in_features)

        self.weight.data.uniform_(-stdv, stdv)
        if self.bias is not None:
            self.bias.data.uniform_(-stdv, stdv)

    def forward(self, input):
        return lin(input, self.weight, self.bias)

    def __repr__(self):

After Change

    def reset_parameters(self):
        size = self.in_features
        uniform(size, self.weight, self.bias)

    def forward(self, input):
        return lin(input, self.weight, self.bias)

Instances