NotImplementedError: Learning rate schedule must override get_config

Question

我使用 tf.keras 创建了自定义计划，但在保存模型时遇到此错误：

NotImplementedError: Learning rate schedule must override get_config

class 看起来像这样：

class CustomSchedule(tf.keras.optimizers.schedules.LearningRateSchedule):

    def __init__(self, d_model, warmup_steps=4000):
        super(CustomSchedule, self).__init__()

        self.d_model = d_model
        self.d_model = tf.cast(self.d_model, tf.float32)

        self.warmup_steps = warmup_steps

    def __call__(self, step):
        arg1 = tf.math.rsqrt(step)
        arg2 = step * (self.warmup_steps**-1.5)

        return tf.math.rsqrt(self.d_model) * tf.math.minimum(arg1, arg2)

    def get_config(self):
        config = {
            'd_model':self.d_model,
            'warmup_steps':self.warmup_steps

        }
        base_config = super(CustomSchedule, self).get_config()
        return dict(list(base_config.items()) + list(config.items()))

Answer 1

当您使用自定义子类模型时，保存模型架构有点棘手。相反，使用 Model.save_weights() 仅保存权重更容易。

如果您将代码更改为此，您将不会看到该错误：

  def get_config(self):
    config = {
    'd_model': self.d_model,
    'warmup_steps': self.warmup_steps,

     }
    return config

NotImplementedError: Learning rate schedule must override get_config

NotImplementedError: Learning rate schedule must override get_config

python

machine-learning

transformer

keras

tensorflow2.0