Pytorch netwrok with variable number of hidden layers

Question

I want to create a class that creates a simple network with X fully connected layers, where X is an input given by the user. I tried this using the setattr/getattr but for some reason is not working.

class MLP(nn.Module):
      def __init__(self,in_size, out_size,n_layers, hidden_size):
        super(MLP,self).__init__()
        self.n_layers=n_layers

        for i in range(n_layers):
            if i==0:
                layer_in_size = in_size
            else:
                layer_in_size = hidden_size

            if i==(n_layers-1):
                layer_out_size = out_size
            else:
                layer_out_size = hidden_size

            setattr(self,'dense_{}'.format(i), nn.Linear(layer_in_size,layer_out_size))

        def forward(self,x):
            out = x
            for i in range(self.n_layers):
                if i==(self.n_layers-1):
                    out = getattr(self,'dense_{}'.format(i),out)
                else:
                    out = F.relu(getattr(self,'dense_{}'.format(i),out))

            return out

This is the error im getting when trying a forward pass with the net:

enter image description here

Some insights of what's the issue will be helpful.

Warkaz · Accepted Answer · 2022-03-06 10:28:20Z

1

This seems like a problem with forward implementation with the mod2 function. Try the pytorch functions (torch.fmod and torch.remainder) or if you don't need the backprop capabilities try to do .detach() before the mod2 function.

answered Mar 6, 2022 at 10:28

Warkaz

9237 silver badges19 bronze badges

Sign up to request clarification or add additional context in comments.

Collectives™ on Stack Overflow

Pytorch netwrok with variable number of hidden layers

1 Answer 1

Comments

Your Answer

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Your Answer

Sign up or log in

Post as a guest

Related