Torch model to Pytorch Model

vision
1

Hi guys,

I would to create a neural network model in Pytorch using an example of implementation of Torch model (.lua model).

However, I do not understand the synthax in this code and I do not find it in any Torch tutorials

input = - nn.SpatialConvolution(2, 64, 7, 7, 1, 1, 3, 3)
s1 = input - nn.SpatialBatchNormalization(64, 1e-4)
        - nn.ReLU()
        - nn.SpatialMaxPooling(2, 2, 2, 2)
        - nn.SpatialConvolution(64, 128, 5, 5, 1, 1, 2, 2)
        - nn.SpatialBatchNormalization(128)
        - nn.ReLU()
s2 = s1
        - nn.SpatialMaxPooling(2, 2, 2, 2)
        - nn.SpatialConvolution(128, 256, 5, 5, 1, 1, 2, 2)
        - nn.SpatialBatchNormalization(256)
        - nn.ReLU()
s3 = s2
        - nn.SpatialMaxPooling(2, 2, 2, 2)
        - nn.SpatialConvolution(256, 256, 3, 3, 1, 1, 1, 1)
        - nn.SpatialBatchNormalization(256)
        - nn.ReLU()
        - nn.SpatialConvolution(256, 256, 3, 3, 1, 1, 1, 1)
        - nn.SpatialBatchNormalization(256)
        - nn.ReLU()

What does the dash - mean, for instance, in the first lane ?

Thanks in advance for your help