ValueError: setting an array element with a sequence from BatchNormalization layer of Keras

Question

我正在实施一些东西，我发现批量规范化层抛出奇怪的值错误。

我用来生成错误的代码如下：

x = Input(shape=(25,14,19))
bn = BatchNormalization(
        momentum=0.1, 
        epsilon=0.00001, 
        gamma_regularizer=keras.initializers.ones(), 
        beta_constraint=keras.initializers.zeros())

y = bn(x)

堆栈跟踪是：

---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-20-e16165265878> in <module>()
      6         gamma_regularizer=keras.initializers.ones(),
      7         beta_constraint=keras.initializers.zeros())
----> 8 y = bn(x)
      9 

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/keras/engine/base_layer.pyc in __call__(self, inputs, **kwargs)
    430                                          '`layer.build(batch_input_shape)`')
    431                 if len(input_shapes) == 1:
--> 432                     self.build(input_shapes[0])
    433                 else:
    434                     self.build(input_shapes)

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/keras/layers/normalization.pyc in build(self, input_shape)
    105                                          initializer=self.gamma_initializer,
    106                                          regularizer=self.gamma_regularizer,
--> 107                                          constraint=self.gamma_constraint)
    108         else:
    109             self.gamma = None

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/keras/legacy/interfaces.pyc in wrapper(*args, **kwargs)
     89                 warnings.warn('Update your `' + object_name +
     90                               '` call to the Keras 2 API: ' + signature, stacklevel=2)
---> 91             return func(*args, **kwargs)
     92         wrapper._original_function = func
     93         return wrapper

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/keras/engine/base_layer.pyc in add_weight(self, name, shape, dtype, initializer, regularizer, trainable, constraint)
    253         if regularizer is not None:
    254             with K.name_scope('weight_regularizer'):
--> 255                 self.add_loss(regularizer(weight))
    256         if trainable:
    257             self._trainable_weights.append(weight)

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/keras/initializers.pyc in __call__(self, shape, dtype)
     44 
     45     def __call__(self, shape, dtype=None):
---> 46         return K.constant(1, shape=shape, dtype=dtype)
     47 
     48 

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/keras/backend/tensorflow_backend.pyc in constant(value, dtype, shape, name)
    425     if dtype is None:
    426         dtype = floatx()
--> 427     return tf.constant(value, dtype=dtype, shape=shape, name=name)
    428 
    429 

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/constant_op.pyc in constant(value, dtype, shape, name, verify_shape)
    212   tensor_value.tensor.CopyFrom(
    213       tensor_util.make_tensor_proto(
--> 214           value, dtype=dtype, shape=shape, verify_shape=verify_shape))
    215   dtype_value = attr_value_pb2.AttrValue(type=tensor_value.tensor.dtype)
    216   const_tensor = g.create_op(

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/tensorflow/python/framework/tensor_util.pyc in make_tensor_proto(values, dtype, shape, verify_shape)
    427     # If shape is None, numpy.prod returns None when dtype is not set, but raises
    428     # exception when dtype is set to np.int64
--> 429     if shape is not None and np.prod(shape, dtype=np.int64) == 0:
    430       nparray = np.empty(shape, dtype=np_dt)
    431     else:

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/numpy/core/fromnumeric.pyc in prod(a, axis, dtype, out, keepdims)
   2564 
   2565     return _methods._prod(a, axis=axis, dtype=dtype,
-> 2566                           out=out, **kwargs)
   2567 
   2568 

/Users/jaejunlee/anaconda2/lib/python2.7/site-packages/numpy/core/_methods.pyc in _prod(a, axis, dtype, out, keepdims)
     33 
     34 def _prod(a, axis=None, dtype=None, out=None, keepdims=False):
---> 35     return umr_prod(a, axis, dtype, out, keepdims)
     36 
     37 def _any(a, axis=None, dtype=None, out=None, keepdims=False):

ValueError: setting an array element with a sequence.

当输入形状的批量大小未知时，伽玛初始化似乎有问题？输入应该是由其他 Conv2D 生成的 2D (25 x 14)，因此它的通道大小（即特征）为 19.

谁能帮我解决这个问题？

Answer 1

我认为您错误地使用了 regularizer 和 constraint 参数而不是 initializer 参数:

bn = BatchNormalization(
        momentum=0.1, 
        epsilon=0.00001, 
        gamma_initializer=keras.initializers.ones(), 
        beta_initializer=keras.initializers.zeros())

ValueError: setting an array element with a sequence from BatchNormalization layer of Keras

ValueError: setting an array element with a sequence from BatchNormalization layer of Keras

python

keras

tensorflow

keras-layer

batch-normalization