INNER CODE UNIT · Python

CNN20d

RichardS0268/CNN-for-Trading · model.py:50

class CNN20d(nn.Module):
    # Input: [N, (1), 64, 60]; Output: [N, 2]
    # Three Convolution Blocks
    
    def init_weights(self, m):
        if isinstance(m, nn.Linear) or isinstance(m, nn.Conv2d):
            torch.nn.init.xavier_uniform(m.weight)
            m.bias.data.fill_(0.01)
    
    def __init__(self):
        super(CNN20d, self).__init__()
        self.conv1 = nn.Sequential(OrderedDict([
            ('Conv', nn.Conv2d(1, 64, (5, 3), padding=(3, 1), stride=(3, 1), dilation=(2, 1))), # output size: [N, 64, 21, 60]
            ('BN', nn.BatchNorm2d(64, affine=True)),
            ('ReLU', nn.ReLU()),
            ('Max-Pool', nn.MaxPool2d((2,1))) # output size: [N, 64, 10, 60]
        ]))
        self.conv1 = self.conv1.apply(self.init_weights)

View source record →