create multiple convolutions to work on a single input in parallel.

multi_resolution.py

def makeSomeLayers(in_channels,out_channels,resolutions):
    layers = []
    for r in resolutions:
        layers.append(nn.Conv2d(in_channels=in_channels,out_channels=out_channels,kernel_size=r))
    return layers


class Net(nn.Module):

    def __init__(self):
        super(Net, self).__init__()
       
        in_channels = 3
        out_channels = 64
        resolutions = [3,5,7,9,11]
        self.convs = nn.ModuleList(makeSomeLayers(in_channels,out_channels,resolutions))
    
    
    def forward(self,x):
        # call the various defined operations on  the input...
        outs = [op(x) for op in self.convs]
        
        
        # do something with output...
        x = .....
        
        return x