data_iter(batch_size, features,

defdata_iter(batch_size,features,labels):

    num_examples=len(features)

    indices=list(range(num_examples))

    random.shuffle(indices)   

   for i in range(0,num_examples,batch_size):

            j=nd.array(indices[i:min(i+batch_size,num_examples)])

            yield features.take(j),labels.take(j)            # take函数根据索引返回对应元素