python - 如何在 tf.keras 自定义损失函数中触发 python 函数？

Question

在我的自定义损失函数中，我需要调用一个纯 Python 函数，传入计算出的 TD 误差和一些索引。该函数不需要返回任何内容或进行区分。这是我要调用的函数:

def update_priorities(self, traces_idxs, td_errors):
    """Updates the priorities of the traces with specified indexes."""
    self.priorities[traces_idxs] = td_errors + eps

我试过使用 tf.py_function 来调用包装函数，但只有当它嵌入到图中时才会被调用，即如果它有输入和输出并且使用了输出。因此，我试图通过一些张量而不对它们执行任何操作，现在调用了该函数。这是我的整个自定义损失函数:

def masked_q_loss(data, y_pred):
    """Computes the MSE between the Q-values of the actions that were taken and the cumulative
    discounted rewards obtained after taking those actions. Updates trace priorities.
    """
    action_batch, target_qvals, traces_idxs = data[:,0], data[:,1], data[:,2]
    seq = tf.cast(tf.range(0, tf.shape(action_batch)[0]), tf.int32)
    action_idxs = tf.transpose(tf.stack([seq, tf.cast(action_batch, tf.int32)]))
    qvals = tf.gather_nd(y_pred, action_idxs)

    def update_priorities(_qvals, _target_qvals, _traces_idxs):
        """Computes the TD error and updates memory priorities."""
        td_error = _target_qvals - _qvals
        _traces_idxs = tf.cast(_traces_idxs, tf.int32)
        mem.update_priorities(_traces_idxs, td_error)
        return _qvals

    qvals = tf.py_function(func=update_priorities, inp=[qvals, target_qvals, traces_idxs], Tout=[tf.float32])
    return tf.keras.losses.mse(qvals, target_qvals)

但是，由于调用 mem.update_priorities(_traces_idxs, td_error)

，我收到以下错误

ValueError: An operation has `None` for gradient. Please make sure that all of your ops have a gradient defined (i.e. are differentiable). Common ops without gradient: K.argmax, K.round, K.eval.

我不需要为 update_priorities 计算梯度，我只想在图形计算的特定点调用它，然后忘记它。我该怎么做？

Answer 1

在包装函数内的张量上使用 .numpy() 解决了这个问题:

def update_priorities(_qvals, _target_qvals, _traces_idxs):
    """Computes the TD error and updates memory priorities."""
    td_error = np.abs((_target_qvals - _qvals).numpy())
    _traces_idxs = (tf.cast(_traces_idxs, tf.int32)).numpy()
    mem.update_priorities(_traces_idxs, td_error)
    return _qvals