LSTM_NET

LSTM_NET.py

### build model ###
class RNN(nn.Module):
    def __init__(self, vocab_size, embed_size, num_hiddens, num_layers,
                 bidirectional, weight, labels, **kwargs):
        super(RNN, self).__init__(**kwargs)
        self.num_hiddens = num_hiddens
        self.num_layers = num_layers
        self.bidirectional = bidirectional
        self.embedding = nn.Embedding.from_pretrained(weight)
        self.embedding.weight.requires_grad = False
        self.encoder = nn.LSTM(input_size=embed_size, hidden_size=self.num_hiddens,
                               num_layers=num_layers, bidirectional=self.bidirectional,
                               dropout=0.3)

        if self.bidirectional:
            self.linear1 = nn.Linear(num_hiddens * 4, labels)
        else:
            self.linear1 = nn.Linear(num_hiddens * 2, labels)



    def forward(self, inputs):
        embeddings = self.embedding(inputs)
        states, hidden = self.encoder(embeddings.permute([1, 0, 2]))
        encoding = torch.cat([states[0], states[-1]], dim=1) #if it's bidirectional, choose first and last output
        outputs = self.linear1(encoding)

        return outputs

num_epochs = 10
num_hiddens = 100
num_layers = 2
bidirectional = True
labels = 2
device = torch.device("cuda:1" if torch.cuda.is_available() else "cpu")
net = RNN(vocab_size=(vocab_size+1), embed_size=embed_size,
                   num_hiddens=num_hiddens, num_layers=num_layers,
                   bidirectional=bidirectional, weight=weight,
                   labels=labels)

print(net)