import theano.tensor as T
x = T.imatrix('x')
shared_words = theano.shared(value=np.asarray(all_words, dtype=theano.config.floatX),
                             name='shared_words', borrow=True)
layer0_input = shared_words[x.flatten()].reshape((x.shape[0], 1, x.shape[1], shared_words.shape[1]))

all_words是词向量矩阵，这个layer0_input是将shared_words降为1维后进行reshape,可是x.shape[0]，与x.shape[1]都是多少啊