Python (Scikeras) - ValueError : Invalid parameter layers for estimator KerasClassifier

Question

我正在尝试使用 GridSearchCV 和 scikeras 包装器创建一个卷积神经网络，但我收到了一个无法找出原因的错误。

错误的核心是这样的:

ValueError: Invalid parameter layers for estimator KerasClassifier. This issue can likely be resolved by setting this parameter in the KerasClassifier constructor: KerasClassifier(layers=[128]) Check the list of available parameters with estimator.get_params().keys()

请在代码后面找到完整的错误。我尝试更改几行，或添加不同的参数，但是似乎没有什么可以改变我收到的错误。这是代码:

# first model using the chosen parameters

# Part 1: Create the model
def cnn_model0(layers):
  cnn = tf.keras.models.Sequential() # initialising the CNN
  
  # model layers
  cnn.add( # Step 1 - Convolution
      tf.keras.layers.Conv2D(filters=32, kernel_size=3, padding="same", activation="relu", input_shape=[50, 50, 3]))
  cnn.add( # Step 2 - Pooling
      tf.keras.layers.MaxPool2D(pool_size=2, strides=2, padding='valid'))
  cnn.add( # Second convolutional layer
      tf.keras.layers.Conv2D(filters=32, kernel_size=3, padding="same", activation="relu"))
  cnn.add( # Second pooling layer 
      tf.keras.layers.MaxPool2D(pool_size=2, strides=2, padding='valid'))
  cnn.add( # Step 3 - Flattening
      tf.keras.layers.Flatten())

  # Step 4 - Full connection (FC)
  for i, nodes in enumerate(layers):
      cnn.add(tf.keras.layers.Dense(units = nodes, activation = 'relu'))
  cnn.add(tf.keras.layers.Dense(units = 43, activation = 'softmax'))

  # Compiling the CNN
  cnn.compile(optimizer = 'Adam', loss = 'categorical_crossentropy', metrics = ['accuracy']) 
  return cnn


# Part 2: Fitting the CNN model
model = KerasClassifier(build_fn = cnn_model0, verbose = 1)

# establish the grid parameters
layers = [[128], (256, 128), (200, 150, 120)]
param_grid = dict(layers = layers)

# fit GridSearchCV
grid = GridSearchCV(estimator = model, param_grid = param_grid, verbose = 1)
grid_results = grid.fit(X_train, y_train, validation_data = (X_val, y_val))

# Part 3: Printing the results
print("Best: {0}, using {1}".format(grid_results.best_score_, grid_results.best_params_))

# result values
means = grid_results.cv_results_['mean_test_score']
stds = grid_results.cv_results_['std_test_score']
params = grid_results.cv_results_['params']
for mean, stdev, param in zip(means, stds, params):
  print('{0} ({1}) with: {2}'.format(mean, stdev, param))

导致错误的主要部分是:

model = KerasClassifier(build_fn = cnn_model0, verbose = 1)
layers = [[128], (256, 128), (200, 150, 120)]
param_grid = dict(layers = layers)
grid = GridSearchCV(estimator = model, param_grid = param_grid, verbose = 1)
grid_results = grid.fit(X_train, y_train, validation_data = (X_val, y_val))

从我收到的消息来看，我认为模型和正在建立的层中有一些不正确的地方。任何有助于缩小问题范围的帮助将不胜感激。我对很多机器学习仍然不熟悉。

提前致谢。

完整错误消息:

Fitting 5 folds for each of 3 candidates, totalling 15 fits

---------------------------------------------------------------------------

ValueError                                Traceback (most recent call last)

<ipython-input-33-b6a6389b51ee> in <module>()
     35 # fit GridSearchCV
     36 grid = GridSearchCV(estimator = model, param_grid = param_grid, verbose = 1)
---> 37 grid_results = grid.fit(X_train, y_train, validation_data = (X_val, y_val))
     38 
     39 # Part 3: Printing the results

12 frames

/usr/local/lib/python3.7/dist-packages/scikeras/wrappers.py in set_params(self, **params)
   1153                         "\nCheck the list of available parameters with"
   1154                         " `estimator.get_params().keys()`"
-> 1155                     ) from None
   1156         return self
   1157 

ValueError: Invalid parameter layers for estimator KerasClassifier.
This issue can likely be resolved by setting this parameter in the KerasClassifier constructor:
`KerasClassifier(layers=[128])`
Check the list of available parameters with `estimator.get_params().keys()`

Answer 1

在进行错误中所述的更改后，它现在可以工作了。代码更改自:

# Part 2: Fitting the CNN model
model = KerasClassifier(build_fn = cnn_model0, verbose = 1)

# establish the grid parameters
layers = [[128], (256, 128), (200, 150, 120)]
param_grid = dict(layers = layers)

至

# Part 2: Fitting the CNN model
model = KerasClassifier(build_fn = cnn_model0, verbose = 1, layers = [[128], (256, 128), (200, 150, 120)])

# establish the grid parameters
param_grid = dict(layers = layers)

虽然现在还有另一个问题，“param_grid = dict(layers =layers)”不再定义“layers”，但模型仍然会产生结果。