Keras Tuner -调整多个相同的参数

Question

我已经构建了我的模型，如果我只调优一个变量，并且不在其他层上重复它，它就可以工作。例如，仅在一个层中调整unit。就像我下面的代码一样，我试图调优多个units，但它不起作用。

def model_builder(hp):
  model = keras.Sequential()

  # Tune the number of units in the first Dense layer
  hp_units_1 = hp.Int('units_1', min_value=32, max_value=512, step=32)
  model.add(keras.layers.Dense(
      units=hp_units_1, 
      input_dim=7,
      activation='relu'))


  hp_units_2 = hp.Int('units_2', min_value=32, max_value=512, step=32)
  model.add(keras.layers.Dense(units=hp_units_2 , activation='relu'))
  model.add(keras.layers.Dense(4, activation='sigmoid'))

  # Tune the learning rate for the optimizer
  # Choose an optimal value from 0.01, 0.001, or 0.0001
  hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])

  model.compile(optimizer=keras.optimizers.Adam(learning_rate=hp_learning_rate),
                loss=keras.losses.MeanSquaredError(),
                metrics=['mse'])

  return model

尽管我使用_1 so不重复键，但它抛出了一个错误。

Invalid model 0/5
Traceback (most recent call last):
  File "/usr/local/lib/python3.7/dist-packages/kerastuner/engine/hypermodel.py", line 104, in build
    model = self.hypermodel.build(hp)
  File "<ipython-input-19-f964baafcc40>", line 5, in model_builder
    hp_units = hp.Int('units_1', min_value=32, max_value=512, step=32)
  File "/usr/local/lib/python3.7/dist-packages/kerastuner/engine/hyperparameters.py", line 744, in Int
    return self._retrieve(hp)
  File "/usr/local/lib/python3.7/dist-packages/kerastuner/engine/hyperparameters.py", line 625, in _retrieve
    return self.values[hp.name]
KeyError: 'units_1'

Answer 1

最后，可能由于Colab保存运行时的方式，它在某种程度上使用了带有新关键字的旧关键字(虽然听起来很奇怪)。

因此，为了解决这个问题，我不得不在我的调谐器中添加参数overwrite=True：

tuner = kt.Hyperband(model_builder,
                     objective='val_loss',
                     max_epochs=10,
                     factor=3,
                     overwrite=True)

Answer 2

错误可能是由于在构建模型时定义的变量造成的。您可能可以在开始构建模型之前定义变量，如下所示并检查结果：

def model_builder(hp):
  hp_units_1 = hp.Int('units_1', min_value=32, max_value=512, step=32)
  hp_units_2 = hp.Int('units_2', min_value=32, max_value=512, step=32)
  # Tune the learning rate for the optimizer
  # Choose an optimal value from 0.01, 0.001, or 0.0001
  hp_learning_rate = hp.Choice('learning_rate', values=[1e-2, 1e-3, 1e-4])
  model = keras.Sequential()      
  model.add(keras.layers.Dense(units=hp_units_1, input_dim=7, activation='relu'))     
  model.add(keras.layers.Dense(units=hp_units_2 , activation='relu'))
  model.add(keras.layers.Dense(4, activation='sigmoid'))
  model.compile(optimizer=keras.optimizers.Adam(learning_rate=hp_learning_rate),
                loss=keras.losses.MeanSquaredError(),
                metrics=['mse'])

  return model