Python NN.bias_initialization Examples

Example #1

File: wide_resnet_with_1x1.py Project: nutszebra/is_receptive_field_necessary_to_big

 def weight_initialization(self):
     self.conv1.weight.data = torch.FloatTensor(
         NN.weight_relu_initialization(self.conv1))
     self.conv1.bias.data = torch.FloatTensor(
         NN.bias_initialization(self.conv1, constant=0))
     self.conv2.weight.data = torch.FloatTensor(
         NN.weight_relu_initialization(self.conv2))
     self.conv2.bias.data = torch.FloatTensor(
         NN.bias_initialization(self.conv2, constant=0))

Example #2

File: wide_resnet_with_1x1.py Project: nutszebra/is_receptive_field_necessary_to_big

 def weight_initialization(self):
     self.conv1.weight.data = torch.FloatTensor(
         NN.weight_relu_initialization(self.conv1))
     self.conv1.bias.data = torch.FloatTensor(
         NN.bias_initialization(self.conv1, constant=0))
     self.linear.weight.data = torch.FloatTensor(
         np.random.uniform(-0.1, 0.1, self.linear.weight.data.shape))
     self.linear.bias.data = torch.FloatTensor(
         NN.bias_initialization(self.linear, constant=0.0))