python报错TypeError: Dimension value must be integer or None or have an __index__ method, got 3.0

Question

我在运行lanenet程序的时候，训练数据集出现的这个错误

  File "C:\Users\dell\anaconda3\envs\tensorflow\lib\site-packages\tensorflow\python\framework\tensor_shape.py", line 197, in __init__
    six.raise_from(
  File "<string>", line 3, in raise_from
TypeError: Dimension value must be integer or None or have an __index__ method, got 3.0

```python


@tf_export(v1=["Dimension"])
class Dimension(object):
  """Represents the value of one dimension in a TensorShape."""

  __slots__ = ["_value"]

  def __init__(self, value):
    """Creates a new Dimension with the given value."""
    if value is None:
      self._value = None
    elif isinstance(value, Dimension):
      self._value = value
    else:
      try:
        # int(...) compensates for the int/long dichotomy on Python 2.X.
        # TODO(b/143206389): Remove once we fully migrate to 3.X.
        self._value = int(value.__index__())
      except AttributeError:
        six.raise_from(
            TypeError("Dimension value must be integer or None or have "
                      "an __index__ method, got {!r}".format(value)),
            None)
      if self._value < 0:
        raise ValueError("Dimension %d must be >= 0" % self._value)

Answer 1

1.原因：这个错误通常发生在使用tensorflow或keras的时候，当定义一个层或一个张量的维度时，传入了一个浮点数而不是一个整数或者None。例如，如果使用tf.keras.Input (shape = (3.0, ))，就会报这个错误，因为3.0不是一个合法的维度值。
2.解决方法：检查代码中定义层或张量的地方，确保维度值是整数或者None，或者有一个__index__方法可以返回一个整数。例如，可以使用tf.keras.Input (shape = (3, ))或者tf.keras.Input (shape = (None, ))来避免这个错误。