python - AttributeError：“PrefetchDataset”对象没有属性“make_initializable_iterator”

Question

我收到：'AttributeError：'PrefetchDataset'对象没有属性'make_initializable_iterator'' 我正在使用：import tensorflow.compat.v1 as tf tf.disable_v2_behavior()

train_data, train_data_infer=_get_training_data(FLAGS)
iter_train = train_data.make_initializable_iterator()
iter_train_infer=train_data_infer.make_initializable_iterator()

谢谢

标签： python

Answer 1

您正在寻找的方法是tf.compat.v1.data.make_initializable_iterator(dataset)

只要你用tf.compat.v1，更换应该就够了

iter_train_infer=train_data_infer.make_initializable_iterator()

和

iter_train_infer = tf.data.make_initializable_iterator(train_data_infer)

如果你得到make_initializable_iterator is not supported when eager execution is enabled，然后tf.compat.v1.disable_eager_execution()在你的应用程序启动时添加。