训练量增加导致Keras错误

Question

我有一个词嵌入神经网络，它在我的语料库的小集上工作得很好，但是当我将训练集增加到成千上万条记录（不更改任何其他内容）时，出现以下错误：

ValueError: Error when checking input: expected documents to have shape (46,) but got array with shape (1,)

模型摘要如下：

__________________________________________________________________________________________________
Layer (type)                    Output Shape         Param #     Connected to
==================================================================================================
ingredients (InputLayer)        (None, 46)           0
__________________________________________________________________________________________________
documents (InputLayer)          (None, 46)           0
__________________________________________________________________________________________________
ingredients_embedding (Embeddin (None, 46, 50)       8709200     ingredients[0][0]
__________________________________________________________________________________________________
documents_embedding (Embedding) (None, 46, 50)       8709200     documents[0][0]
__________________________________________________________________________________________________
lambda_1 (Lambda)               (None, 50)           0           ingredients_embedding[0][0]
__________________________________________________________________________________________________
lambda_2 (Lambda)               (None, 50)           0           documents_embedding[0][0]
__________________________________________________________________________________________________
dot_product (Dot)               (None, 1)            0           lambda_1[0][0]
                                                                 lambda_2[0][0]
__________________________________________________________________________________________________
reshape_1 (Reshape)             (None, 1)            0           dot_product[0][0]
==================================================================================================
Total params: 17,418,400
Trainable params: 17,418,400
Non-trainable params: 0

这是模型的入口，生成器为训练生成数据：

negative_ratio, n_positive = 2, 100
t = Trainer()
training_data_pairs = t.index_and_encode()
training_size, embedding_size, input_size = len(training_data_pairs), 50, 46

batch = t.generate_batch(n_positive, negative_ratio=negative_ratio)
model = model(training_size, embedding_size, input_size)
h = model.fit_generator(
    batch,
    epochs=10,
    steps_per_epoch=int(training_size/n_positive),
    verbose=2
)