Question

我有一个nlp数据集，根据Pytorch官方教程，我将数据集更改为word_to_idx和tag_to_idx，例如：

word_to_idx =  {'I': 0, 'have': 1, 'used': 2, 'transfers': 3, 'on': 4, 'three': 5, 'occasions': 6, 'now': 7, 'and': 8, 'each': 9, 'time': 10}
tag_to_idx =  {'PRON': 0, 'VERB': 1, 'NOUN': 2, 'ADP': 3, 'NUM': 4, 'ADV': 5, 'CONJ': 6, 'DET': 7, 'ADJ': 8, 'PRT': 9, '.': 10, 'X': 11}

我想用BiLSTM完成POS标记任务。这是我的BiLSTM代码：

class LSTMTagger(nn.Module):

    def __init__(self, embedding_dim, hidden_dim, vocab_size, tagset_size):
        super(LSTMTagger, self).__init__()
        self.hidden_dim = hidden_dim
        self.word_embeddings = nn.Embedding(vocab_size, tagset_size)

        # The LSTM takes word embeddings as inputs, and outputs hidden states
        self.lstm = nn.LSTM(embedding_dim, hidden_dim, bidirectional=True)

        # The linear layer that maps from hidden state space to tag space
        self.hidden2tag = nn.Linear(in_features=hidden_dim * 2, out_features=tagset_size)


    def forward(self, sentence):
        embeds = self.word_embeddings(sentence)
        lstm_out, _ = self.lstm(embeds.view(len(sentence), 1, -1))
        tag_space = self.hidden2tag(lstm_out.view(len(sentence), -1))

        # tag_scores = F.softmax(tag_space, dim=1)
        tag_scores = F.log_softmax(tag_space, dim=1)
        return tag_scores

然后我在Pycharm中运行训练代码，例如：

EMBEDDING_DIM = 6
HIDDEN_DIM = 6
NUM_EPOCHS = 3

model = LSTMTagger(embedding_dim=EMBEDDING_DIM,
                   hidden_dim=HIDDEN_DIM,
                   vocab_size=len(word_to_idx),
                   tagset_size=len(tag_to_idx))

loss_function = nn.NLLLoss()
optimizer = optim.SGD(model.parameters(), lr=0.1)

# See what the scores are before training
with torch.no_grad():
    inputs = prepare_sequence(training_data[0][0], word_to_idx)
    tag_scores = model(inputs)
    print(tag_scores)
    print(tag_scores.size())

但是，它显示行tag_scores = model(inputs)和行lstm_out, _ = self.lstm(embeds.view(len(sentence), 1, -1))的错误。错误是：

Traceback (most recent call last):
  line 140, in <module>
    tag_scores = model(inputs)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in __call__
    result = self.forward(*input, **kwargs)
  line 115, in forward
    lstm_out, _ = self.lstm(embeds.view(len(sentence), 1, -1))
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/nn/modules/module.py", line 493, in __call__
    result = self.forward(*input, **kwargs)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/nn/modules/rnn.py", line 559, in forward
    return self.forward_tensor(input, hx)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/nn/modules/rnn.py", line 539, in forward_tensor
    output, hidden = self.forward_impl(input, hx, batch_sizes, max_batch_size, sorted_indices)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/nn/modules/rnn.py", line 519, in forward_impl
    self.check_forward_args(input, hx, batch_sizes)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/nn/modules/rnn.py", line 490, in check_forward_args
    self.check_input(input, batch_sizes)
  File "/Library/Frameworks/Python.framework/Versions/3.7/lib/python3.7/site-packages/torch/nn/modules/rnn.py", line 153, in check_input
    self.input_size, input.size(-1)))
RuntimeError: input.size(-1) must be equal to input_size. Expected 6, got 12

我不知道如何使用它进行调试。有人可以帮我解决这个问题吗？预先感谢！

Answer 1

错误在这里：

self.word_embeddings = nn.Embedding(vocab_size, tagset_size)

您使用的标签数量不是12，而不是LSTM层期望的6，而不是使用嵌入维度。

Pytorch BiLSTM POS标记问题：RuntimeError：input.size（-1）必须等于input_size。预期为6，得到12

1 个答案: