Python 3.x 保存时保存tf.keras.Model断言错误：未跟踪对象

我有一个基于

tf.keras.model

构建的模型，当我尝试使用

model.save（/path/to/model）

保存模型时，我得到了错误

AssertionError: Tried to export a function which references untracked object Tensor("Network/truediv_3:0", shape=(4, 3), dtype=float32).TensorFlow objects (e.g. tf.Variable) captured by functions must be tracked by assigning them to an attribute of a tracked object or assigned to an attribute of the main object directly.

它所指的张量计算如下

def gmm(self, gamma, z):
    # some code omitted

    # here is the tensor:
    self.mu = tf.divide(
                tf.matmul(gamma, z, transpose_a=True), sums_exp_dims
            )
    # more code omitted

我需要将这个张量作为训练例程的类属性。我尝试“转储”参数：

def dump(self):
    out = {'mu':self.mu}
    self.mu = None
    return out

但这导致了计算图中更多的错误。有没有办法保存模型以避免断言错误