使用 Pytorch 前向传播 RNN

Question

我正在尝试创建一个 RNN 前向传递方法，该方法可以采用可变的输入、隐藏和输出大小并创建所需的 rnn 单元。对我来说，似乎我正在将正确的变量传递给 self.rnn_cell——x 和前一个隐藏层的输入值。但是，我收到的错误包含在下面。

我也尝试过使用 x[i] 和 x[:,i,i] （正如我的教授所建议的那样），但无济于事。我很困惑，只是在寻找关于我是否在这里做正确的事情的指导。我的教授建议，由于我不断收到错误，我应该在 jupyter notebook 中重新启动内核并重新运行代码。我有，我收到同样的错误......

如果您需要其他上下文，请告诉我。

class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(RNN, self).__init__()
        self.hidden_size = hidden_size

        self.rnn_cell = nn.RNNCell(input_size, hidden_size)
        self.fc = nn.Linear(hidden_size, output_size)
        self.softmax = nn.LogSoftmax(dim=1)

    def forward(self, x):
        """
        x: size [seq_length, 1, input_size]
        """
        h = torch.zeros(x.size(1), self.hidden_size)

        for i in range(x.size(0)):
            ### START YOUR CODE ###
            h = self.rnn_cell(x[:,:,i], h)
            ### END YOUR CODE ###

        ### START YOUR CODE ###
        # Hint: first call fc, then call softmax
        out = self.softmax(self.fc(self.hidden_size, h.size(0)))
        ### END YOUR CODE ###

        return out 
IndexError: Dimension out of range (expected to be in range of [-1, 0], but got 1)

Answer 1

我不是 RNN 方面的专家，但可以尝试一下。

class RNN(nn.Module):
def __init__(self, input_size, hidden_size, output_size):
    super(RNN, self).__init__()
    self.hidden_size = hidden_size

    self.rnn_cell = nn.RNN(input_size, hidden_size)
    self.fc = nn.Linear(hidden_size, output_size)


def forward(self, x):
    """
    x: size [seq_length, 1, input_size]
    """
    h = torch.zeros(num_layers(hidden), x.size(0), self.hidden_size)


        ### START YOUR CODE ###
    out,hidden = self.rnn_cell(x, h)
        ### END YOUR CODE ###

    ### START YOUR CODE ###
    # Hint: first call fc, then call softmax
    out = out.contiguous().view(-1, self.hidden_dim) #You need to reshape the output to fit the FC layer
    out = self.fc(out)
    return F.softmax(out)
    ### END YOUR CODE ###

    return out 

请尝试运行它，如果出现错误或任何疑问，请告诉我。 （不能向您询问详细信息，因为我无法评论 rn。）

如果您对我的回答有任何想法，请支持。

Answer 2

class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(RNN, self).__init__()
        self.hidden_size = hidden_size

        self.rnn_cell = nn.RNNCell(input_size, hidden_size)
        self.fc = nn.Linear(hidden_size, output_size)
        self.softmax = nn.LogSoftmax(dim=1)

    def forward(self, x):
        """
        x: size [seq_length, 1, input_size]
        """
        h = torch.zeros(x.size(1), self.hidden_size)

        for i in range(x.size(0)):
            ### START YOUR CODE ###
            h = self.rnn_cell(x[i,:,:], h)
            ### END YOUR CODE ###

        ### START YOUR CODE ###
        # Hint: first call fc, then call softmax
        out = self.fc(h)
        out = self.softmax(out)
        ### END YOUR CODE ###

        return out

Answer 3

class RNN(nn.Module):
    def __init__(self, input_size, hidden_size, output_size):
        super(RNN, self).__init__()
        self.hidden_size = hidden_size

        self.rnn_cell = nn.RNNCell(input_size, hidden_size)
        self.fc = nn.Linear(hidden_size, output_size)
        self.softmax = nn.LogSoftmax(dim=1)

    def forward(self, x):
        """
        x: size [seq_length, 1, input_size]
        """
        h = torch.zeros(x.size(1), self.hidden_size)

        for i in range(x.size(0)):
            ### START YOUR CODE ###
            h = self.rnn_cell(x[i,:,:], h)
            ### END YOUR CODE ###

        ### START YOUR CODE ###
        # Hint: first call fc, then call softmax
        out = self.softmax(self.fc(h))
        ### END YOUR CODE ###

        return out