python - 使用 tf.data 创建 TensorFlow 输入管道时出错

Question

我有以下代码：

with tf.device('/cpu:0'):



train_data = (train_data.map(to_feature_map,
                               num_parallel_calls=tf.data.experimental.AUTOTUNE
                               )
  .shuffle(1000)
  .batch(32, drop_remainder=True)
  .prefetch(tf.data.experimental.AUTOTUNE)
  )

我收到以下错误：

TypeError: Tensors in list passed to 'input' of 'EagerPyFunc' Op have types [<NOT CONVERTIBLE TO TENSOR>, int32] that are invalid. Tensors: [{'input_word_ids': <tf.Tensor 'args_2:0' shape=(32, 128) dtype=int32>, 'input_mask': <tf.Tensor 'args_0:0' shape=(32, 128) dtype=int32>, 'input_type_ids': <tf.Tensor 'args_1:0' shape=(32, 128) dtype=int32>}, <tf.Tensor 'args_3:0' shape=(32,) dtype=int32>]

我该如何解决这个问题