AttributeError：“float”对象在自定义 Keras 模型中没有属性“pop”，并覆盖 train_step

Question

我收到此错误 AttributeError: 'float' object has no attribute 'pop' in my Encoder-Decoder custom Keras model while using model.fit with custom train_step.我不明白这个错误发生在哪里。当我们不知道错误的位置时，我需要有关如何解决此类错误的帮助？

Epoch 1/10
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-561-573fadf7e010> in <module>()
----> 1 model.fit(dataset,epochs=10)

5 frames
/usr/local/lib/python3.6/dist-packages/tensorflow/python/keras/callbacks.py in _batch_update_progbar(self, logs)
    921 
    922     logs = copy.copy(logs) if logs else {}
--> 923     batch_size = logs.pop('size', 0)
    924     num_steps = logs.pop('num_steps', 1)  # DistStrat can run >1 steps.
    925     logs.pop('batch', None)

AttributeError: 'float' object has no attribute 'pop'

链接到我的代码：https://github.com/abhishek203/E-Abhishek/blob/master/Untitled4.ipynb

Answer 1

您的训练步骤的返回语句应该类似于 {'loss':loss} 而不是 loss