NotImplementedError: numpy() is only available when eager execution is enabled; while using TF2.4.1

Question

使用 tensorflow 2.4.1，我在 keras 中覆盖 SimpleRNNCell.call，在此处找到：

https://github.com/tensorflow/tensorflow/blob/85c8b2a817f95a3e979ecd1ed95bff1dc1335cff/tensorflow/python/keras/layers/recurrent.py#L1362

具体偏置部分如下：

    if self.bias is not None:
      bias_inv = np.arctanh(self.bias)
      h = K.bias_add(h, bias_inv)

我收到以下错误：

NotImplementedError: in user code:

    <ipython-input-12-a2655e34a197>:72 call  *
        inputs, mask=mask, training=training, initial_state=initial_state)
    <ipython-input-55-87c0b5bbed00>:23 call  *
        bias_inv = np.arctanh(self.bias)
    /usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/resource_variable_ops.py:483 __array__  **
        return np.asarray(self.numpy())
    /usr/local/lib/python3.7/dist-packages/tensorflow/python/ops/resource_variable_ops.py:620 numpy
        "numpy() is only available when eager execution is enabled.")

    NotImplementedError: numpy() is only available when eager execution is enabled.

我确定 eager execution 已启用，可能是什么问题？

Answer 1

@Ramy Hanna
AFAIK，Keras 在执行时将所有层和模型转换为图形。因此，即使开启了 Eager 模式，您也可能会遇到此类错误。您可以通过以下任一方式避免它们：

将图层用作函数（以测试您所做的更改）
设置 dynamic=True 标志（在文档中检查一次）

Answer 2

您需要使用 tensorflow 函数来实现您的层，而不是 numpy 函数。在这种情况下，您应该将 np.arctanh 替换为 tf.math.atanh:

bias_inv = tf.math.atanh(self.bias)

NotImplementedError: numpy() is only available when eager execution is enabled; while using TF2.4.1

NotImplementedError: numpy() is only available when eager execution is enabled; while using TF2.4.1

python

numpy

keras

tensorflow

eager-execution