Created by: 131250208

使用的自定义数据集，依照官方文档的数据定义如下：

# 自定义的生成器
def my_reader(features, labels):
    def gen():
        for ind, f in enumerate(features):
            yield f, labels[ind]
    return gen

BATCH_SIZE = 20

train_reader = paddle.batch(
    paddle.reader.shuffle(
        my_reader(features_train, labels_train), buf_size=500),
        batch_size=BATCH_SIZE)

test_reader = paddle.batch(
    paddle.reader.shuffle(
        my_reader(features_dev, labels_dev), buf_size=500),
        batch_size=BATCH_SIZE)

reader的输出：

next(train_reader())[0]
# (array([0.8967, 0.7465, 0.8835]), 2)

我尝试了一下官方案例的手写数字识别数据的输出，输出的格式是一致的，不知道为什么会报错。 这是我的variable定义：

feed_order = ['confs', 'label']
confs = fluid.data(name = feed_order[0], shape = [None, 3], dtype = "float32")
label = fluid.data(name = feed_order[1], shape = [None, 1], dtype = "int64")

报错代码的行:

for batch_id, data in enumerate(train_reader()):
            metrics = exe.run(main_program, 
                              feed = feeder.feed(data),
                              fetch_list = [avg_loss, accuracy])