Question

从embedding_matrix调用的函数。有人可以解释一下我为什么收到此错误以及如何解决该错误。

def getEmbeddingMatrix (word_index, vectorSource):
        wordVecSources = {'fasttext' : './vectors/crawl-300d-2M-subword.vec', 'custom-fasttext' : 
           './vectors/' + '20news-fasttext.json' }
        f = open (wordVecSources[vectorSource])
        allWv = {}
        if (vectorSource == 'custom-fasttext'):
            allWv = json.loads(f.read())
        elif (vectorSource == 'fasttext'):
            errorCount = 0
            for line in f:
                values = line.split()
                word = values[0].strip()
                try:
                    wv = np.asarray(values[1:], dtype='float32')
                    if (len(wv) != wvLength):
                        errorCount = errorCount + 1
                        continue
                except:
                    errorCount = errorCount + 1
                    continue
                allWv[word] = wv
            print ("# Bad Word Vectors:", errorCount)
        f.close()
        embedding_matrix = np.zeros((len(word_index)+1, wvLength))  # +1 for the masked 0
        for word, i in word_index.items():
            if word in allWv:
                embedding_matrix[i] = allWv[word]
        return embedding_matrix

early_stop = keras.callbacks.EarlyStopping(monitor='val_loss', min_delta=0, patience=5, verbose=2, mode='auto', restore_best_weights=False)
model = keras.models.Sequential()
if (vectorSource != 'none'):
    embedding_matrix = getEmbeddingMatrix (kTokenizer.word_index, vectorSource)
    embedding = keras.layers.embeddings.Embedding(input_dim=len(kTokenizer.word_index)+1, output_dim=wvLength, weights=[embedding_matrix], input_length=sequenceLength, trainable=False, mask_zero=True)
else:
    embedding = keras.layers.embeddings.Embedding(input_dim=len(kTokenizer.word_index)+1, output_dim=wvLength, input_length=sequenceLength, trainable=True, mask_zero=True)
model.add(embedding)
model.add(keras.layers.LSTM(units=150, dropout=0.2, recurrent_dropout=0.2, return_sequences=False))
model.add(keras.layers.Dense(numClasses, activation='softmax'))
model.compile(optimizer='adam', loss='categorical_crossentropy', metrics=['acc'])
print(model.summary())

调用getEmbeddingMatrix时，即使两个文件都位于同一位置，它也会给出KeyError'-f'

KeyError                                  Traceback (most recent call last)
<ipython-input-45-ca6718845b1d> in <module>
      2 model = keras.models.Sequential()
      3 if (vectorSource != 'none'):
----> 4     embedding_matrix = getEmbeddingMatrix (kTokenizer.word_index, vectorSource)
      5     embedding = keras.layers.embeddings.Embedding(input_dim=len(kTokenizer.word_index)+1, output_dim=wvLength, weights=[embedding_matrix], input_length=sequenceLength, trainable=False, mask_zero=True)
      6 else:

<ipython-input-40-b29c56e927a3> in getEmbeddingMatrix(word_index, vectorSource)
      1 def getEmbeddingMatrix (word_index, vectorSource):
      2     wordVecSources = {'fasttext' : './vectors/crawl-300d-2M-subword.vec', 'custom-fasttext' : './vectors/' + '20news-fasttext.json' }
----> 3     f = open (wordVecSources[vectorSource])
      4     allWv = {}
      5     if (vectorSource == 'custom-fasttext'):

KeyError: '-f'

Answer 1

字典wordVecSources的密钥可以是'fasttext'或'custom-fasttext'，但是它收到的密钥是'-f'。

这意味着您正在以某种方式为vectorSource分配'-f'值'none'，'fasttext'或'custom-fasttext'来防止{{1 }}。

在代码中检查与KeyError相关的shell命令行参数。例如，在vectorSource中说您有your_file.py，那么您应该运行类似vectorSource = str(sys.argv[1])的东西，而不是运行$python your_file.py fasttext，因为这里是第一个命令行参数（sys.argv [1]）在运行时为其$python your_file.py -f赋值。

LSTM模型的KeyError'-f'Python

1 个答案: