AttributeError: 'RNN' object has no attribute 'n_layers'

1

Greetings PyTorch Community,

I am trying to implement an RNN model using GRU cell. Look for some guidance on why the init_hidden() function returning the error.

class RNN(nn.Module):
    def __init__(
        self, 
        batch_size,
        e_dim = 200, 
        n_hidden = 6, 
        n_class = 6,
        n_layers = 1
    ):
        super(RNN, self).__init__()
        self.n_layers = n_layers
        self.e_dim = e_dim
        self.n_hidden = n_hidden
        self.batch_size = batch_size
        self.n_class = n_class
        
        self.gru = nn.GRU(
            self.e_dim, 
            self.n_hidden,
            num_layers = n_layers,
            batch_first = True
        )
        self.fc = nn.Linear(self.n_hidden, self.n_class)

    def init_hidden(self):
        return torch.randn(self.n_layers, self.batch_size, self.hidden_size)

    def forward(self, inputs):
        # Avoid breaking if the last batch has a different size
        batch_size = inputs.size(0)
        if batch_size != self.batch_size:
            self.batch_size = batch_size
            
        output, hidden = self.gru(inputs, self.init_hidden())
        output = self.fc(output[:, :, -1]).squeeze()
        return output
2

I don’t get your mentioned error message, but instead self.hidden_size is undefined.
I assume you meant to use self.n_hidden instead.