我想制作优化 Gpu 和 Cpu 的管道。 数据集

回归问题大约有 10000 个数据点和 4 个描述变量。

df = pd.read_csv("dataset")
X_train, X_test, y_train, y_test = 
train_test_split(df.iloc[:, :-1].values, df.iloc[:, -1].values)
scaler = MinMaxScaler()
scaler.fit(X_train)
X_train_scaled = scaler.transform(X_train)
batch_size = 64
with tf.Session() as sess:
  dataset = tf.data.Dataset.from_tensor_slices((X_train_scaled, y_train))
  dataset = dataset.cache()
  dataset = dataset.shuffle(len(X_train_scaled))
  dataset = dataset.repeat()
  dataset = dataset.batch(batch_size)
  dataset = dataset.prefetch(batch_size*10)
  iterator = dataset.make_one_shot_iterator()
  print(sess.run(iterator.get_next()[0]),sess.run(iterator.get_next()[1]))
  his = model.fit(dataset, epochs=300, steps_per_epoch=1000, verbose=0)

[[0.54192635 0.36815166 0.37738184 0.13592493] [0.31898017 0.33687204 0.59490225 0.59597855] [0.2733711 0.26047393 0.42761693 0.99986595] [0.77025496 0.98919431 0.45632269 0.66447721] [0.64305949 0.50236967 0.53823311 0.56313673]
[429.66 460.53 428.49 446.62 456.84]

在拟合模型中，出现以下错误

AttributeError: 'PrefetchDataset' object has no attribute 'ndim'

我看到了这个问题的一些问题。

但这对我不起作用。

软件版本： Keras:2.2.4 Tensorflow:1.12.0