topology.py

# -*- coding: utf-8 -*-
from __future__ import print_function
from __future__ import absolute_import
from __future__ import division

import numpy as np

import types as python_types
import warnings
import copy
import os
import inspect
from six.moves import zip

from .. import backend as K
from .. import initializations
from ..utils.io_utils import ask_to_proceed_with_overwrite
from ..utils.generic_utils import func_dump, func_load


def to_list(x):
    '''This normalizes a list/tensor into a list.

    If a tensor is passed, we return
    a list of size 1 containing the tensor.
    '''
    if isinstance(x, list):
        return x
    return [x]


def object_list_uid(object_list):
    object_list = to_list(object_list)
    return ', '.join([str(abs(id(x))) for x in object_list])


class InputSpec(object):
    '''This specifies the ndim, dtype and shape of every input to a layer.
    Every layer should expose (if appropriate) an `input_spec` attribute:
    a list of instances of InputSpec (one per input tensor).

    A None entry in a shape is compatible with any dimension,
    a None shape is compatible with any shape.
    '''
    def __init__(self, dtype=None, shape=None, ndim=None):
        if isinstance(ndim, str):
            if '+' not in ndim:
                raise ValueError('When passing a str "ndim", '
                                 'it should have the form "2+", "3+", etc.')
            int_ndim = ndim[:ndim.find('+')]
            if not int_ndim.isdigit():
                raise ValueError('When passing a str "ndim", '
                                 'it should have the form "2+", "3+", etc.')
        if shape is not None:
            self.ndim = len(shape)
        else:
            self.ndim = ndim
        self.dtype = dtype
        self.shape = shape


class Node(object):
    '''A `Node` describes the connectivity between two layers.

    Each time a layer is connected to some new input,
    a node is added to `layer.inbound_nodes`.
    Each time the output of a layer is used by another layer,
    a node is added to `layer.outbound_nodes`.

    # Attributes
        outbound_layer: the layer that takes
            `input_tensors` and turns them into `output_tensors`.
        inbound_layers: a list of layers, the same length as `input_tensors`,
            the layers from where `input_tensors` originate.
        node_indices: a list of integers, the same length as `inbound_layers`.
            `node_indices[i]` is the origin node of `input_tensors[i]`
            (necessary since each inbound layer might have several nodes,
            e.g. if the layer is being shared with a different data stream).
        tensor_indices: a list of integers,
            the same length as `inbound_layers`.
            `tensor_indices[i]` is the index of `input_tensors[i]` within the
            output of the inbound layer
            (necessary since each inbound layer might
            have multiple tensor outputs, with each one being
            independently manipulable).
        input_tensors: list of input tensors.
        output_tensors: list of output tensors.
        input_masks: list of input masks (a mask can be a tensor, or None).
        output_masks: list of output masks (a mask can be a tensor, or None).
        input_shapes: list of input shape tuples.
        output_shapes: list of output shape tuples.

    `node_indices` and `tensor_indices` are basically fine-grained coordinates
    describing the origin of the `input_tensors`, verifying the following:

    `input_tensors[i] == inbound_layers[i].inbound_nodes[node_indices[i]].output_tensors[tensor_indices[i]]`

    A node from layer A to layer B is added to:
        A.outbound_nodes
        B.inbound_nodes
    '''
    def __init__(self, outbound_layer,
                 inbound_layers, node_indices, tensor_indices,
                 input_tensors, output_tensors,
                 input_masks, output_masks,
                 input_shapes, output_shapes):
        # Layer instance (NOT a list).
        # this is the layer that takes a list of input tensors
        # and turns them into a list of output tensors.
        # the current node will be added to
        # the inbound_nodes of outbound_layer.
        self.outbound_layer = outbound_layer

        # The following 3 properties describe where
        # the input tensors come from: which layers,
        # and for each layer, which node and which
        # tensor output of each node.

        self.inbound_layers = inbound_layers  # List of layer instances
        self.node_indices = node_indices  # List of integers, 1:1 mapping with inbound_layers.
        self.tensor_indices = tensor_indices  # List of integers, 1:1 mapping with inbound_layers.

        # Tensor inputs and outputs of outbound_layer.
        self.input_tensors = input_tensors  # List of tensors. 1:1 mapping with inbound_layers.
        self.output_tensors = output_tensors  # List of tensors, created by outbound_layer.call().

        # input and output masks
        self.input_masks = input_masks  # List of tensors, 1:1 mapping with input_tensor.
        self.output_masks = output_masks  # List of tensors, created by outbound_layer.compute_mask().

        # input and output shapes
        self.input_shapes = input_shapes  # List of shape tuples, shapes of input_tensors.
        self.output_shapes = output_shapes  # List of shape tuples, shapes of output_tensors.

        # Add nodes to all layers involved.
        for layer in inbound_layers:
            if layer is not None:
                layer.outbound_nodes.append(self)
        outbound_layer.inbound_nodes.append(self)

    @classmethod
    def create_node(cls, outbound_layer,
                    inbound_layers, node_indices=None, tensor_indices=None):
        if not node_indices:
            node_indices = [0 for _ in range(len(inbound_layers))]
        else:
            assert len(node_indices) == len(inbound_layers)
        if not tensor_indices:
            tensor_indices = [0 for _ in range(len(inbound_layers))]

        input_tensors = []
        input_masks = []
        input_shapes = []

        for inbound_layer, node_index, tensor_index in zip(inbound_layers, node_indices, tensor_indices):
            inbound_node = inbound_layer.inbound_nodes[node_index]
            input_tensors.append(inbound_node.output_tensors[tensor_index])
            input_masks.append(inbound_node.output_masks[tensor_index])
            input_shapes.append(inbound_node.output_shapes[tensor_index])

        assert len(input_shapes) == len(input_tensors) == len(input_masks)

        if len(input_tensors) == 1:
            output_tensors = to_list(outbound_layer.call(input_tensors[0], mask=input_masks[0]))
            output_masks = to_list(outbound_layer.compute_mask(input_tensors[0], input_masks[0]))
            # TODO: try to auto-infer shape
            # if exception is raised by get_output_shape_for.
            output_shapes = to_list(outbound_layer.get_output_shape_for(input_shapes[0]))
        else:
            output_tensors = to_list(outbound_layer.call(input_tensors, mask=input_masks))
            output_masks = to_list(outbound_layer.compute_mask(input_tensors, input_masks))
            output_shapes = to_list(outbound_layer.get_output_shape_for(input_shapes))

        if not output_tensors or output_tensors[0] is None:
            raise TypeError('The `call` method of layer "' +
                            outbound_layer.name +
                            '" should return a tensor. Found: ' +
                            str(output_tensors[0]))
        if len(output_tensors) != len(output_shapes):
            raise ValueError('The `get_output_shape_for` method of layer "' +
                             outbound_layer.name +
                             '"" should return one shape tuple per '
                             'output tensor of the layer. Found: ' +
                             str(output_shapes))
        if len(output_tensors) != len(output_masks):
            raise ValueError('The `compute_mask` method of layer "' +
                             outbound_layer.name +
                             '" should return one mask tensor per '
                             'output tensor of the layer. Found: ' +
                             str(output_masks))

        for i in range(len(output_tensors)):
            output_tensors[i]._keras_shape = output_shapes[i]
            output_tensors[i]._uses_learning_phase = any([x._uses_learning_phase for x in input_tensors]) or outbound_layer.uses_learning_phase
            output_tensors[i]._keras_history = (outbound_layer, len(outbound_layer.inbound_nodes), i)

        return cls(outbound_layer,
                   inbound_layers, node_indices, tensor_indices,
                   input_tensors, output_tensors,
                   input_masks, output_masks,
                   input_shapes, output_shapes)

    def get_config(self):
        inbound_names = []
        for layer in self.inbound_layers:
            if layer:
                inbound_names.append(layer.name)
            else:
                inbound_names.append(None)
        return {'outbound_layer': self.outbound_layer.name if self.outbound_layer else None,
                'inbound_layers': inbound_names,
                'node_indices': self.node_indices,
                'tensor_indices': self.tensor_indices}


class Layer(object):
    '''Abstract base layer class.

    # Properties
        name: String, must be unique within a model.
        input_spec: List of InputSpec class instances
            each entry describes one required input:
                - ndim
                - dtype
            A layer with `n` input tensors must have
            an `input_spec` of length `n`.
        trainable: Boolean, whether the layer weights
            will be updated during training.
        uses_learning_phase: Whether any operation
            of the layer uses `K.in_training_phase()`
            or `K.in_test_phase()`.
        input_shape: Shape tuple. Provided for convenience,
            but note that there may be cases in which this
            attribute is ill-defined (e.g. a shared layer
            with multiple input shapes), in which case
            requesting `input_shape` will raise an Exception.
            Prefer using `layer.get_input_shape_for(input_shape)`,
            or `layer.get_input_shape_at(node_index)`.
        output_shape: Shape tuple. See above.
        inbound_nodes: List of nodes.
        outbound_nodes: List of nodes.
        supports_masking: Boolean.
        input, output: Input/output tensor(s). Note that if the layer is used
            more than once (shared layer), this is ill-defined
            and will raise an exception. In such cases, use
            `layer.get_input_at(node_index)`.
        input_mask, output_mask: Same as above, for masks.
        trainable_weights: List of variables.
        non_trainable_weights: List of variables.
        weights: The concatenation of the lists trainable_weights and
            non_trainable_weights (in this order).
        constraints: Dict mapping weights to constraints.

    # Methods
        call(x, mask=None): Where the layer's logic lives.
        __call__(x, mask=None): Wrapper around the layer logic (`call`).
            If x is a Keras tensor:
                - Connect current layer with last layer from tensor:
                    `self.add_inbound_node(last_layer)`
                - Add layer to tensor history
            If layer is not built:
                - Build from x._keras_shape
        get_weights()
        set_weights(weights)
        get_config()
        count_params()
        get_output_shape_for(input_shape)
        compute_mask(x, mask)
        get_input_at(node_index)
        get_output_at(node_index)
        get_input_shape_at(node_index)
        get_output_shape_at(node_index)
        get_input_mask_at(node_index)
        get_output_mask_at(node_index)

    # Class Methods
        from_config(config)

    # Internal methods:
        build(input_shape)
        add_inbound_node(layer, index=0)
        create_input_layer()
        assert_input_compatibility()
    '''
    def __init__(self, **kwargs):
        # These properties should have been set
        # by the child class, as appropriate.
        if not hasattr(self, 'input_spec'):
            self.input_spec = None
        if not hasattr(self, 'supports_masking'):
            self.supports_masking = False
        if not hasattr(self, 'uses_learning_phase'):
            self.uses_learning_phase = False

        # These lists will be filled via successive calls
        # to self.add_inbound_node().
        self.inbound_nodes = []
        self.outbound_nodes = []

        # These properties will be set upon call of self.build(),
        # which itself will be called upon self.add_inbound_node if necessary.
        if not hasattr(self, '_trainable_weights'):
            self._trainable_weights = []
        if not hasattr(self, '_non_trainable_weights'):
            self._non_trainable_weights = []
        if not hasattr(self, 'losses'):
            self.losses = []
        if not hasattr(self, 'constraints'):
            self.constraints = {}  # dict {tensor: constraint instance}
        self.built = False

        # These properties should be set by the user via keyword arguments.
        # note that 'input_dtype', 'input_shape' and 'batch_input_shape'
        # are only applicable to input layers: do not pass these keywords
        # to non-input layers.
        allowed_kwargs = {'input_shape',
                          'batch_input_shape',
                          'input_dtype',
                          'name',
                          'trainable'}
        for kwarg in kwargs.keys():
            if kwarg not in allowed_kwargs:
                raise TypeError('Keyword argument not understood:', kwarg)
        name = kwargs.get('name')
        if not name:
            prefix = self.__class__.__name__.lower()
            name = prefix + '_' + str(K.get_uid(prefix))
        self.name = name

        self.trainable = kwargs.get('trainable', True)
        if 'batch_input_shape' in kwargs or 'input_shape' in kwargs:
            # In this case we will create an input layer
            # to insert before the current layer
            if 'batch_input_shape' in kwargs:
                batch_input_shape = tuple(kwargs['batch_input_shape'])
            elif 'input_shape' in kwargs:
                batch_input_shape = (None,) + tuple(kwargs['input_shape'])
            self.batch_input_shape = batch_input_shape
            input_dtype = kwargs.get('input_dtype', K.floatx())
            self.input_dtype = input_dtype

    @property
    def trainable_weights(self):
        trainable = getattr(self, 'trainable', True)
        if trainable:
            return self._trainable_weights
        else:
            return []

    @trainable_weights.setter
    def trainable_weights(self, weights):
        self._trainable_weights = weights

    @property
    def non_trainable_weights(self):
        trainable = getattr(self, 'trainable', True)
        if not trainable:
            return self._trainable_weights + self._non_trainable_weights
        else:
            return self._non_trainable_weights

    @non_trainable_weights.setter
    def non_trainable_weights(self, weights):
        self._non_trainable_weights = weights

    @property
    def regularizers(self):
        warnings.warn('The `regularizers` property of '
                      'layers/models is deprecated. '
                      'Regularization losses are now managed via the `losses` '
                      'layer/model property.')
        return []

    @regularizers.setter
    def regularizers(self, _):
        warnings.warn('The `regularizers` property of layers/models '
                      'is deprecated. '
                      'Regularization losses are now managed via the `losses` '
                      'layer/model property.')

    def create_input_layer(self, batch_input_shape,
                           input_dtype=None, name=None):
        if not name:
            prefix = self.__class__.__name__.lower() + '_input_'
            name = prefix + str(K.get_uid(prefix))
        if not input_dtype:
            input_dtype = K.floatx()

        self.batch_input_shape = batch_input_shape
        self.input_dtype = input_dtype

        # Instantiate the input layer.
        x = Input(batch_shape=batch_input_shape,
                  dtype=input_dtype, name=name)
        # This will build the current layer
        # and create the node connecting the current layer
        # to the input layer we just created.
        self(x)

    def add_weight(self, shape, initializer, name=None,
                   trainable=True,
                   regularizer=None,
                   constraint=None):
        '''Adds a weight variable to the layer.

        # Arguments:
            shape: The shape tuple of the weight.
            initializer: An Initializer instance (callable).
            trainable: A boolean, whether the weight should
                be trained via backprop or not (assuming
                that the layer itself is also trainable).
            regularizer: An optional Regularizer instance.
        '''
        initializer = initializations.get(initializer)
        weight = initializer(shape, name=name)
        if regularizer is not None:
            self.add_loss(regularizer(weight))
        if constraint is not None:
            self.constraints[weight] = constraint
        if trainable:
            self._trainable_weights.append(weight)
        else:
            self._non_trainable_weights.append(weight)
        return weight

    def assert_input_compatibility(self, input):
        '''This checks that the tensor(s) `input`
        verify the input assumptions of the layer
        (if any). If not, exceptions are raised.
        '''
        if not self.input_spec:
            return True
        if not isinstance(self.input_spec, list):
            raise TypeError('input_spec must be a list of '
                            'InputSpec instances. Found: ' +
                            str(self.input_spec))
        inputs = to_list(input)
        if len(self.input_spec) > 1:
            if len(inputs) != len(self.input_spec):
                raise ValueError('Layer ' + self.name + ' expects ' +
                                 str(len(self.input_spec)) + ' inputs, '
                                 'but it received ' + str(len(inputs)) +
                                 ' input tensors. Input received: ' +
                                 str(input))
        for input_index, (x, spec) in enumerate(zip(inputs, self.input_spec)):
            if spec is None:
                continue

            # Check ndim.
            if spec.ndim is not None:
                if isinstance(spec.ndim, str):
                    int_ndim = spec.ndim[:spec.ndim.find('+')]
                    ndim = int(int_ndim)
                    if K.ndim(x) < ndim:
                        raise ValueError('Input ' + str(input_index) +
                                         ' is incompatible with layer ' +
                                         self.name + ': expected ndim >= ' +
                                         str(ndim) + ', found ndim=' +
                                         str(K.ndim(x)))
                else:
                    if K.ndim(x) != spec.ndim:
                        raise ValueError('Input ' + str(input_index) +
                                         ' is incompatible with layer ' +
                                         self.name + ': expected ndim=' +
                                         str(spec.ndim) + ', found ndim=' +
                                         str(K.ndim(x)))
            if spec.dtype is not None:
                if K.dtype(x) != spec.dtype:
                    raise ValueError('Input ' + str(input_index) +
                                     ' is incompatible with layer ' +
                                     self.name + ': expected dtype=' +
                                     str(spec.dtype) + ', found dtype=' +
                                     str(K.dtype(x)))
            if spec.shape is not None:
                if hasattr(x, '_keras_shape'):
                    x_shape = x._keras_shape
                elif hasattr(K, 'int_shape'):
                    # Tensorflow shape inference.
                    x_shape = K.int_shape(x)
                else:
                    continue
                for spec_dim, dim in zip(spec.shape, x_shape):
                    if spec_dim is not None:
                        if spec_dim != dim:
                            raise ValueError(
                                'Input ' + str(input_index) +
                                ' is incompatible with layer ' +
                                self.name + ': expected shape=' +
                                str(spec.shape) + ', found shape=' +
                                str(x_shape))

    def call(self, x, mask=None):
        '''This is where the layer's logic lives.

        # Arguments
            x: input tensor, or list/tuple of input tensors.
            mask: a masking tensor (or list of tensors). Used mainly in RNNs.

        # Returns:
            A tensor or list/tuple of tensors.
        '''
        return x

    def __call__(self, x, mask=None):
        '''Wrapper around self.call(), for handling
        internal Keras references.

        If a Keras tensor is passed:
            - We call self.add_inbound_node().
            - If necessary, we `build` the layer to match
                the _keras_shape of the input(s).
            - We update the _keras_shape of every input tensor with
                its new shape (obtained via self.get_output_shape_for).
                This is done as part of add_inbound_node().
            - We update the _keras_history of the output tensor(s)
                with the current layer.
                This is done as part of add_inbound_node().

        # Arguments
            x: Can be a tensor or list/tuple of tensors.
            mask: Tensor or list/tuple of tensors.
        '''
        if not self.built:
            # Raise exceptions in case the input is not compatible
            # with the input_spec specified in the layer constructor.
            self.assert_input_compatibility(x)

            # Collect input shapes to build layer.
            input_shapes = []
            for x_elem in to_list(x):
                if hasattr(x_elem, '_keras_shape'):
                    input_shapes.append(x_elem._keras_shape)
                elif hasattr(K, 'int_shape'):
                    input_shapes.append(K.int_shape(x_elem))
                else:
                    raise ValueError('You tried to call layer "' + self.name +
                                     '". This layer has no information'
                                     ' about its expected input shape, '
                                     'and thus cannot be built. '
                                     'You can build it manually via: '
                                     '`layer.build(batch_input_shape)`')
            if len(input_shapes) == 1:
                self.build(input_shapes[0])
            else:
                self.build(input_shapes)
            self.built = True

        # Raise exceptions in case the input is not compatible
        # with the input_spec set at build time.
        self.assert_input_compatibility(x)

        input_tensors = to_list(x)
        inbound_layers = []
        node_indices = []
        tensor_indices = []
        for input_tensor in input_tensors:
            if hasattr(input_tensor, '_keras_history') and input_tensor._keras_history:
                # This is a Keras tensor.
                previous_layer, node_index, tensor_index = input_tensor._keras_history
                inbound_layers.append(previous_layer)
                node_indices.append(node_index)
                tensor_indices.append(tensor_index)
            else:
                inbound_layers = None
                break

        if inbound_layers:
            # This will call layer.build() if necessary.
            self.add_inbound_node(inbound_layers, node_indices, tensor_indices)
            # Outputs were already computed when calling self.add_inbound_node.
            outputs = self.inbound_nodes[-1].output_tensors
        else:
            # This case appears if the input was not a Keras tensor.
            outputs = to_list(self.call(x, mask))

        # Apply activity regularizer if any:
        if hasattr(self, 'activity_regularizer') and self.activity_regularizer is not None:
            regularization_losses = [self.activity_regularizer(x) for x in outputs]
            self.add_loss(regularization_losses, input_tensors)

        # If single output tensor: return it,
        # else return a list (at least 2 elements).
        if len(outputs) == 1:
            return outputs[0]
        else:
            return outputs

    def add_inbound_node(self, inbound_layers,
                         node_indices=None, tensor_indices=None):
        '''
        # Arguments
            inbound_layers: Can be a layer instance
                or a list/tuple of layer instances.
            node_indices: Integer (or list of integers).
                The input layer might have a number of
                parallel output streams;
                this is the index of the stream (in the input layer)
                where to connect the current layer.
            tensor_indices: Integer or list of integers.
                The output of the inbound node might be a list/tuple
                of tensor, and we might only be interested in
                one specific entry.
                This index allows you to specify the index of
                the entry in the output list
                (if applicable). "None" means that we take all outputs
                (as a list).
        '''
        inbound_layers = to_list(inbound_layers)
        if not node_indices:
            node_indices = [0 for _ in range(len(inbound_layers))]
        else:
            node_indices = to_list(node_indices)
            assert len(node_indices) == len(inbound_layers)
        if not tensor_indices:
            tensor_indices = [0 for _ in range(len(inbound_layers))]
        else:
            tensor_indices = to_list(tensor_indices)

        if not self.built:
            # collect input_shapes for call to build()
            input_shapes = []
            for layer, node_index, tensor_index in zip(inbound_layers, node_indices, tensor_indices):
                input_shapes.append(layer.inbound_nodes[node_index].output_shapes[tensor_index])
            # call build()
            if len(input_shapes) == 1:
                self.build(input_shape=input_shapes[0])
            else:
                self.build(input_shape=input_shapes)
            self.built = True
        # creating the node automatically updates self.inbound_nodes
        # as well as outbound_nodes on inbound layers.
        Node.create_node(self, inbound_layers, node_indices, tensor_indices)

    def get_output_shape_for(self, input_shape):
        '''Computes the output shape of the layer given
        an input shape (assumes that the layer will be built
        to match that input shape).

        # Arguments
            input_shape: Shape tuple (tuple of integers)
                or list of shape tuples (one per output tensor of the layer).
                Shape tuples can include None for free dimensions,
                instead of an integer.
        '''
        return input_shape

    def compute_mask(self, input, input_mask=None):
        '''Computes an output masking tensor, given an input tensor
        (or list thereof) and an input mask (or list thereof).

        # Arguments
            input: Tensor or list of tensors.
            input_mask: Tensor or list of tensors.

        # Returns
            None or a tensor (or list of tensors,
                one per output tensor of the layer).
        '''
        if not hasattr(self, 'supports_masking') or not self.supports_masking:
            if input_mask is not None:
                if isinstance(input_mask, list):
                    if any(input_mask):
                        raise ValueError('Layer ' + self.name +
                                         ' does not support masking, '
                                         'but was passed an input_mask: ' +
                                         str(input_mask))
                else:
                    raise ValueError('Layer ' + self.name +
                                     ' does not support masking, '
                                     'but was passed an input_mask: ' +
                                     str(input_mask))
            # masking not explicitly supported: return None as mask
            return None
        # if masking is explictly supported, by default
        # carry over the input mask
        return input_mask

    def build(self, input_shape):
        '''Creates the layer weights.
        Must be implemented on all layers that have weights.

        # Arguments
            input_shape: Keras tensor (future input to layer)
                or list/tuple of Keras tensors to reference
                for weight shape computations.
        '''
        self.built = True

    def _get_node_attribute_at_index(self, node_index, attr, attr_name):
        '''Retrieves an attribute (e.g. input_tensors) from a node.

        # Arguments
            node_index: Integer index of the node from which
                to retrieve the attribute.
            attr: Exact node attribute name.
            attr_name: Human-readable attribute name, for error messages.
        '''
        if not self.inbound_nodes:
            raise RuntimeError('The layer has never been called '
                               'and thus has no defined ' + attr_name + '.')
        if not len(self.inbound_nodes) > node_index:
            raise ValueError('Asked to get ' + attr_name +
                             ' at node ' + str(node_index) +
                             ', but the layer has only ' +
                             str(len(self.inbound_nodes)) + ' inbound nodes.')
        values = getattr(self.inbound_nodes[node_index], attr)
        if len(values) == 1:
            return values[0]
        else:
            return values

    def get_input_shape_at(self, node_index):
        '''Retrieves the input shape(s) of a layer at a given node.
        '''
        return self._get_node_attribute_at_index(node_index,
                                                 'input_shapes',
                                                 'input shape')

    def get_output_shape_at(self, node_index):
        '''Retrieves the output shape(s) of a layer at a given node.
        '''
        return self._get_node_attribute_at_index(node_index,
                                                 'output_shapes',
                                                 'output shape')

    def get_input_at(self, node_index):
        '''Retrieves the input tensor(s) of a layer at a given node.
        '''
        return self._get_node_attribute_at_index(node_index,
                                                 'input_tensors',
                                                 'input')

    def get_output_at(self, node_index):
        '''Retrieves the output tensor(s) of a layer at a given node.
        '''
        return self._get_node_attribute_at_index(node_index,
                                                 'output_tensors',
                                                 'output')

    def get_input_mask_at(self, node_index):
        '''Retrieves the input mask tensor(s) of a layer at a given node.
        '''
        return self._get_node_attribute_at_index(node_index,
                                                 'input_masks',
                                                 'input mask')

    def get_output_mask_at(self, node_index):
        '''Retrieves the output mask tensor(s) of a layer at a given node.
        '''
        return self._get_node_attribute_at_index(node_index,
                                                 'output_masks',
                                                 'output mask')

    @property
    def input(self):
        '''Retrieves the input tensor(s) of a layer (only applicable if
        the layer has exactly one inbound node, i.e. if it is connected
        to one incoming layer).
        '''
        if len(self.inbound_nodes) > 1:
            raise AttributeError('Layer ' + self.name +
                                 ' has multiple inbound nodes, '
                                 'hence the notion of "layer input" '
                                 'is ill-defined. '
                                 'Use `get_input_at(node_index)` instead.')
        elif not self.inbound_nodes:
            raise AttributeError('Layer ' + self.name +
                                 ' is not connected, no input to return.')
        return self._get_node_attribute_at_index(0, 'input_tensors',
                                                 'input')

    @property
    def output(self):
        '''Retrieves the output tensor(s) of a layer (only applicable if
        the layer has exactly one inbound node, i.e. if it is connected
        to one incoming layer).
        '''
        if len(self.inbound_nodes) == 0:
            raise AttributeError('Layer ' + self.name +
                                 ' has no inbound nodes.')
        if len(self.inbound_nodes) > 1:
            raise AttributeError('Layer ' + self.name +
                                 ' has multiple inbound nodes, '
                                 'hence the notion of "layer output" '
                                 'is ill-defined. '
                                 'Use `get_output_at(node_index)` instead.')
        return self._get_node_attribute_at_index(0, 'output_tensors',
                                                 'output')

    @property
    def input_mask(self):
        '''Retrieves the input mask tensor(s) of a layer (only applicable if
        the layer has exactly one inbound node, i.e. if it is connected
        to one incoming layer).
        '''
        if len(self.inbound_nodes) != 1:
            raise AttributeError('Layer ' + self.name +
                                 ' has multiple inbound nodes, ' +
                                 'hence the notion of "layer input mask" '
                                 'is ill-defined. '
                                 'Use `get_input_mask_at(node_index)` instead.')
        return self._get_node_attribute_at_index(0, 'input_masks',
                                                 'input mask')

    @property
    def output_mask(self):
        '''Retrieves the output mask tensor(s) of a layer (only applicable if
        the layer has exactly one inbound node, i.e. if it is connected
        to one incoming layer).
        '''
        if len(self.inbound_nodes) != 1:
            raise AttributeError('Layer ' + self.name +
                                 ' has multiple inbound nodes, '
                                 'hence the notion of "layer output mask" '
                                 'is ill-defined. '
                                 'Use `get_output_mask_at(node_index)` '
                                 'instead.')
        return self._get_node_attribute_at_index(0, 'output_masks',
                                                 'output mask')

    @property
    def input_shape(self):
        '''Retrieves the input shape tuple(s) of a layer. Only applicable
        if the layer has one inbound node,
        or if all inbound nodes have the same input shape.
        '''
        if not self.inbound_nodes:
            raise AttributeError('The layer has never been called '
                                 'and thus has no defined input shape.')
        all_input_shapes = set([str(node.input_shapes) for node in self.inbound_nodes])
        if len(all_input_shapes) == 1:
            input_shapes = self.inbound_nodes[0].input_shapes
            if len(input_shapes) == 1:
                return input_shapes[0]
            else:
                return input_shapes
        else:
            raise AttributeError('The layer "' + str(self.name) +
                                 ' has multiple inbound nodes, '
                                 'with different input shapes. Hence '
                                 'the notion of "input shape" is '
                                 'ill-defined for the layer. '
                                 'Use `get_input_shape_at(node_index)` '
                                 'instead.')

    @property
    def output_shape(self):
        '''Retrieves the output shape tuple(s) of a layer. Only applicable
        if the layer has one inbound node,
        or if all inbound nodes have the same output shape.
        '''
        if not self.inbound_nodes:
            raise AttributeError('The layer has never been called '
                                 'and thus has no defined output shape.')
        all_output_shapes = set([str(node.output_shapes) for node in self.inbound_nodes])
        if len(all_output_shapes) == 1:
            output_shapes = self.inbound_nodes[0].output_shapes
            if len(output_shapes) == 1:
                return output_shapes[0]
            else:
                return output_shapes
        else:
            raise AttributeError('The layer "' + str(self.name) +
                                 ' has multiple inbound nodes, '
                                 'with different output shapes. Hence '
                                 'the notion of "output shape" is '
                                 'ill-defined for the layer. '
                                 'Use `get_output_shape_at(node_index)` '
                                 'instead.')

    def add_loss(self, losses, inputs=None):
        if losses is None:
            return
        # Update self.losses
        losses = to_list(losses)
        if not hasattr(self, 'losses'):
            self.losses = []
        try:
            self.losses += losses
        except AttributeError:
            # In case self.losses isn't settable
            # (i.e. it's a getter method).
            # In that case the `losses` property is
            # auto-computed and shouldn't be set.
            pass
        # Update self._per_input_updates
        if not hasattr(self, '_per_input_losses'):
            self._per_input_losses = {}
        if inputs is not None:
            inputs_hash = object_list_uid(inputs)
        else:
            # Updates indexed by None are unconditional
            # rather than input-dependent
            inputs_hash = None
        if inputs_hash not in self._per_input_losses:
            self._per_input_losses[inputs_hash] = []
        self._per_input_losses[inputs_hash] += losses

    def add_update(self, updates, inputs=None):
        if updates is None:
            return
        # Update self.updates
        updates = to_list(updates)
        if not hasattr(self, 'updates'):
            self.updates = []
        try:
            self.updates += updates
        except AttributeError:
            # In case self.updates isn't settable
            # (i.e. it's a getter method).
            # In that case the `updates` property is
            # auto-computed and shouldn't be set.
            pass
        # Update self._per_input_updates
        if not hasattr(self, '_per_input_updates'):
            self._per_input_updates = {}
        if inputs is not None:
            inputs_hash = object_list_uid(inputs)
        else:
            # Updates indexed by None are unconditional
            # rather than input-dependent
            inputs_hash = None
        if inputs_hash not in self._per_input_updates:
            self._per_input_updates[inputs_hash] = []
        self._per_input_updates[inputs_hash] += updates

    def get_updates_for(self, inputs):
        if not hasattr(self, '_per_input_updates'):
            return []
        if inputs is not None:
            inputs_hash = object_list_uid(inputs)
        else:
            inputs_hash = None
        if inputs_hash in self._per_input_updates:
            return self._per_input_updates[inputs_hash]
        return []

    def get_losses_for(self, inputs):
        if not hasattr(self, '_per_input_losses'):
            return []
        if inputs is not None:
            inputs_hash = object_list_uid(inputs)
        else:
            inputs_hash = None
        if inputs_hash in self._per_input_losses:
            return self._per_input_losses[inputs_hash]
        return []

    @property
    def weights(self):
        return self.trainable_weights + self.non_trainable_weights

    def set_weights(self, weights):
        '''Sets the weights of the layer, from Numpy arrays.

        # Arguments
            weights: a list of Numpy arrays. The number
                of arrays and their shape must match
                number of the dimensions of the weights
                of the layer (i.e. it should match the
                output of `get_weights`).
        '''
        params = self.weights
        if len(params) != len(weights):
            raise ValueError('You called `set_weights(weights)` on layer "' +
                             self.name +
                             '" with a  weight list of length ' +
                             str(len(weights)) +
                             ', but the layer was expecting ' +
                             str(len(params)) +
                             ' weights. Provided weights: ' +
                             str(weights)[:50] + '...')
        if not params:
            return
        weight_value_tuples = []
        param_values = K.batch_get_value(params)
        for pv, p, w in zip(param_values, params, weights):
            if pv.shape != w.shape:
                raise ValueError('Layer weight shape ' +
                                 str(pv.shape) +
                                 ' not compatible with '
                                 'provided weight shape ' + str(w.shape))
            weight_value_tuples.append((p, w))
        K.batch_set_value(weight_value_tuples)

    def get_weights(self):
        '''Returns the current weights of the layer,
        as a list of numpy arrays.
        '''
        params = self.weights
        return K.batch_get_value(params)

    def get_config(self):
        '''Returns a Python dictionary (serializable)
        containing the configuration of a layer.
        The same layer can be reinstantiated later
        (without its trained weights) from this configuration.

        The config of a layer does not include connectivity
        information, nor the layer class name. These are handled
        by Container (one layer of abstraction above).
        '''
        config = {'name': self.name,
                  'trainable': self.trainable}
        if hasattr(self, 'batch_input_shape'):
            config['batch_input_shape'] = self.batch_input_shape
        if hasattr(self, 'input_dtype'):
            config['input_dtype'] = self.input_dtype
        return config

    @classmethod
    def from_config(cls, config):
        '''This method is the reverse of get_config,
        capable of instantiating the same layer from the config
        dictionary. It does not handle layer connectivity
        (handled by Container), nor weights (handled by `set_weights`).

        # Arguments
            config: A Python dictionary, typically the
                output of get_config.
        '''
        return cls(**config)

    def count_params(self):
        '''Returns the total number of floats (or ints)
        composing the weights of the layer.
        '''
        if not self.built:
            if self.__class__.__name__ == 'Sequential':
                self.build()
            else:
                raise RuntimeError('You tried to call `count_params` on ' +
                                   self.name + ', but the layer isn\'t built. '
                                   'You can build it manually via: `' +
                                   self.name + '.build(batch_input_shape)`.')
        return sum([K.count_params(p) for p in self.weights])


class InputLayer(Layer):
    '''Layer to be used as an entry point into a graph.
    It can either wrap an existing tensor (pass an `input_tensor` argument)
    or create its a placeholder tensor (pass arguments `input_shape`
    or `batch_input_shape` as well as `input_dtype`).

    # Arguments
        input_shape: Shape tuple, not including the batch axis.
        batch_input_shape: Shape tuple, including the batch axis.
        input_dtype: Datatype of the input.
        input_tensor: Optional tensor to use as layer input
            instead of creating a placeholder.
        sparse: Boolean, whether the placeholder created
            is meant to be sparse.
        name: Name of the layer (string).
    '''
    def __init__(self, input_shape=None, batch_input_shape=None,
                 input_dtype=None, input_tensor=None, sparse=False, name=None):
        self.input_spec = None
        self.supports_masking = False
        self.uses_learning_phase = False
        self.trainable = False
        self.built = True
        self._trainable_weights = []
        self._non_trainable_weights = []
        self.inbound_nodes = []
        self.outbound_nodes = []
        self.constraints = {}
        self.sparse = sparse

        if not name:
            prefix = 'input'
            name = prefix + '_' + str(K.get_uid(prefix))
        self.name = name

        if input_shape and batch_input_shape:
            raise ValueError('Only provide the input_shape OR '
                             'batch_input_shape argument to '
                             'InputLayer, not both at the same time.')
        if input_tensor is not None:
            # Attempt automatic input shape inference.
            try:
                batch_input_shape = K.int_shape(input_tensor)
            except:
                if not input_shape and not batch_input_shape:
                    raise ValueError('InputLayer was provided '
                                     'an input_tensor argument, '
                                     'but its input shape cannot be '
                                     'automatically inferred. '
                                     'You should pass an input_shape or '
                                     'batch_input_shape argument.')
        if not batch_input_shape:
            if not input_shape:
                raise ValueError('An Input layer should be passed either '
                                 'a `batch_input_shape` or an `input_shape`.')
            else:
                batch_input_shape = (None,) + tuple(input_shape)
        else:
            batch_input_shape = tuple(batch_input_shape)

        if not input_dtype:
            if input_tensor is None:
                input_dtype = K.floatx()
            else:
                input_dtype = K.dtype(input_tensor)

        self.batch_input_shape = batch_input_shape
        self.input_dtype = input_dtype

        if input_tensor is None:
            input_tensor = K.placeholder(shape=batch_input_shape,
                                         dtype=input_dtype,
                                         sparse=self.sparse,
                                         name=self.name)
        else:
            input_tensor._keras_shape = batch_input_shape
        # Create an input node to add to self.outbound_node
        # and set output_tensors' _keras_history.
        input_tensor._uses_learning_phase = False
        input_tensor._keras_history = (self, 0, 0)
        Node(self,
             inbound_layers=[],
             node_indices=[],
             tensor_indices=[],
             input_tensors=[input_tensor],
             output_tensors=[input_tensor],
             input_masks=[None],
             output_masks=[None],
             input_shapes=[batch_input_shape],
             output_shapes=[batch_input_shape])

    def get_config(self):
        config = {'batch_input_shape': self.batch_input_shape,
                  'input_dtype': self.input_dtype,
                  'sparse': self.sparse,
                  'name': self.name}
        return config


def Input(shape=None, batch_shape=None,
          name=None, dtype=K.floatx(), sparse=False,
          tensor=None):
    '''`Input()` is used to instantiate a Keras tensor.
    A Keras tensor is a tensor object from the underlying backend
    (Theano or TensorFlow), which we augment with certain
    attributes that allow us to build a Keras model
    just by knowing the inputs and outputs of the model.

    For instance, if a, b and c and Keras tensors,
    it becomes possible to do:
    `model = Model(input=[a, b], output=c)`

    The added Keras attributes are:
        ._keras_shape: Integer shape tuple propagated
            via Keras-side shape inference.
        ._keras_history: Last layer applied to the tensor.
            the entire layer graph is retrievable from that layer,
            recursively.

    # Arguments
        shape: A shape tuple (integer), not including the batch size.
            For instance, `shape=(32,)` indicates that the expected input
            will be batches of 32-dimensional vectors.
        batch_shape: A shape tuple (integer), including the batch size.
            For instance, `batch_shape=(10, 32)` indicates that
            the expected input will be batches of 10 32-dimensional vectors.
            `batch_shape=(None, 32)` indicates batches of an arbitrary number
            of 32-dimensional vectors.
        name: An optional name string for the layer.
            Should be unique in a model (do not reuse the same name twice).
            It will be autogenerated if it isn't provided.
        dtype: The data type expected by the input, as a string
            (`float32`, `float64`, `int32`...)
        sparse: A boolean specifying whether the placeholder
            to be created is sparse.

    # Example

        ```python
        # this is a logistic regression in Keras
        a = Input(shape=(32,))
        b = Dense(16, activation='softmax')(a)
        model = Model(input=a, output=b)
        ```
    '''
    if not batch_shape and tensor is None:
        assert shape, ('Please provide to Input either a `shape`'
                       ' or a `batch_shape` argument. Note that '
                       '`shape` does not include the batch '
                       'dimension.')
    if shape and not batch_shape:
        batch_shape = (None,) + tuple(shape)
    input_layer = InputLayer(batch_input_shape=batch_shape,
                             name=name, input_dtype=dtype,
                             sparse=sparse,
                             input_tensor=tensor)
    # Return tensor including _keras_shape and _keras_history.
    # Note that in this case train_output and test_output are the same pointer.
    outputs = input_layer.inbound_nodes[0].output_tensors
    if len(outputs) == 1:
        return outputs[0]
    else:
        return outputs


class Merge(Layer):
    '''A `Merge` layer can be used to merge a list of tensors
    into a single tensor, following some merge `mode`.

    # Example

    ```python
    model1 = Sequential()
    model1.add(Dense(32, input_dim=32))

    model2 = Sequential()
    model2.add(Dense(32, input_dim=32))

    merged_model = Sequential()
    merged_model.add(Merge([model1, model2], mode='concat', concat_axis=1))
    ```

    # Arguments
        layers: Can be a list of Keras tensors or
            a list of layer instances. Must be more
            than one layer/tensor.
        mode: String or lambda/function. If string, must be one
            of: 'sum', 'mul', 'concat', 'ave', 'cos', 'dot', 'max'.
            If lambda/function, it should take as input a list of tensors
            and return a single tensor.
        concat_axis: Integer, axis to use in mode `concat`.
        dot_axes: Integer or tuple of integers,
            axes to use in mode `dot` or `cos`.
        output_shape: Either a shape tuple (tuple of integers),
            or a lambda/function
            to compute `output_shape`
            (only if merge mode is a lambda/function).
            If the argument is a tuple,
            it should be expected output shape, *not* including the batch size
            (same convention as the `input_shape` argument in layers).
            If the argument is callable,
            it should take as input a list of shape tuples
            (1:1 mapping to input tensors)
            and return a single shape tuple, including the
            batch size (same convention as the
            `get_output_shape_for` method of layers).
        node_indices: Optional list of integers containing
            the output node index for each input layer
            (in case some input layers have multiple output nodes).
            will default to an array of 0s if not provided.
        tensor_indices: Optional list of indices of output tensors
            to consider for merging
            (in case some input layer node returns multiple tensors).
        output_mask: Mask or lambda/function to compute the output mask (only
            if merge mode is a lambda/function). If the latter case, it should
            take as input a list of masks and return a single mask.
    '''
    def __init__(self, layers=None, mode='sum', concat_axis=-1,
                 dot_axes=-1, output_shape=None, output_mask=None,
                 arguments=None, node_indices=None, tensor_indices=None,
                 name=None):
        self.layers = layers
        self.mode = mode
        self.concat_axis = concat_axis
        self.dot_axes = dot_axes
        self._output_shape = output_shape
        self.node_indices = node_indices
        self._output_mask = output_mask
        self.arguments = arguments if arguments else {}

        # Layer parameters.
        self.inbound_nodes = []
        self.outbound_nodes = []
        self.constraints = {}
        self._trainable_weights = []
        self._non_trainable_weights = []
        self.supports_masking = True
        self.uses_learning_phase = False
        self.input_spec = None  # Compatible with anything.
        if not name:
            prefix = self.__class__.__name__.lower()
            name = prefix + '_' + str(K.get_uid(prefix))
        self.name = name

        if layers:
            # This exists for backwards compatibility.
            # equivalent to:
            # merge = Merge(layers=None)
            # output = merge([input_tensor_1, input_tensor_2])
            if not node_indices:
                # By default we connect to
                # the 1st output stream in the input layer.
                node_indices = [0 for _ in range(len(layers))]
            self._arguments_validation(layers, mode,
                                       concat_axis, dot_axes,
                                       node_indices, tensor_indices)
            self.built = True
            self.add_inbound_node(layers, node_indices, tensor_indices)
        else:
            self.built = False

    def _arguments_validation(self, layers, mode, concat_axis, dot_axes,
                              node_indices, tensor_indices):
        '''Validates user-passed arguments and raises exceptions
        as appropriate.
        '''
        if not callable(mode):
            if mode not in {'sum', 'mul', 'concat', 'ave', 'cos', 'dot', 'max'}:
                raise ValueError('Invalid merge mode: ' + str(mode))
        if not isinstance(layers, (list, tuple)) or len(layers) < 2:
            raise TypeError('A Merge should only be applied to a list of '
                            'layers with at least 2 elements. Found: ' +
                            str(layers))

        if tensor_indices is None:
            tensor_indices = [None for _ in range(len(layers))]

        input_shapes = []
        for i, layer in enumerate(layers):
            layer_output_shape = layer.get_output_shape_at(node_indices[i])
            if isinstance(layer_output_shape, list):
                # Case: the layer has multiple output tensors
                # and we only need a specific one.
                layer_output_shape = layer_output_shape[tensor_indices[i]]
            input_shapes.append(layer_output_shape)

        if mode in {'sum', 'mul', 'ave', 'cos', 'max'}:
            input_shapes_set = set(input_shapes)
            if len(input_shapes_set) > 1:
                raise ValueError('Only layers of same output shape can '
                                 'be merged using ' + mode + ' mode. ' +
                                 'Layer shapes: %s' % input_shapes)
        if mode in {'cos', 'dot'}:
            if len(layers) > 2:
                raise ValueError(mode + ' merge takes exactly 2 layers')
            shape1 = input_shapes[0]
            shape2 = input_shapes[1]
            n1 = len(shape1)
            n2 = len(shape2)
            if isinstance(dot_axes, int):
                if dot_axes < 0:
                    self.dot_axes = [dot_axes % n1, dot_axes % n2]
                else:
                    self.dot_axes = [dot_axes, ] * 2
            if not isinstance(self.dot_axes, (list, tuple)):
                raise TypeError('Invalid type for dot_axes - '
                                'should be a list.')
            if len(self.dot_axes) != 2:
                raise ValueError('Invalid format for dot_axes - '
                                 'should contain two elements.')
            if not isinstance(self.dot_axes[0], int) or not isinstance(self.dot_axes[1], int):
                raise ValueError('Invalid format for dot_axes - '
                                 'list elements should be "int".')
            if shape1[self.dot_axes[0]] != shape2[self.dot_axes[1]]:
                raise ValueError('Dimension incompatibility using dot mode: '
                                 '%s != %s. ' % (shape1[self.dot_axes[0]], shape2[self.dot_axes[1]]) +
                                 'Layer shapes: %s, %s' % (shape1, shape2))
        elif mode == 'concat':
            reduced_inputs_shapes = [list(shape) for shape in input_shapes]
            shape_set = set()
            for i in range(len(reduced_inputs_shapes)):
                del reduced_inputs_shapes[i][self.concat_axis]
                shape_set.add(tuple(reduced_inputs_shapes[i]))
            if len(shape_set) > 1:
                raise ValueError('"concat" mode can only merge '
                                 'layers with matching '
                                 'output shapes except for the concat axis. '
                                 'Layer shapes: %s' % (input_shapes))

    def call(self, inputs, mask=None):
        if not isinstance(inputs, list) or len(inputs) <= 1:
            raise TypeError('Merge must be called on a list of tensors '
                            '(at least 2). Got: ' + str(inputs))
        # Case: "mode" is a lambda or function.
        if callable(self.mode):
            arguments = self.arguments
            arg_spec = inspect.getargspec(self.mode)
            if 'mask' in arg_spec.args:
                arguments['mask'] = mask
            return self.mode(inputs, **arguments)

        if self.mode == 'sum' or self.mode == 'ave':
            s = inputs[0]
            for i in range(1, len(inputs)):
                s += inputs[i]
            if self.mode == 'ave':
                s /= len(inputs)
            return s

        elif self.mode == 'concat':
            return K.concatenate(inputs, axis=self.concat_axis)

        elif self.mode == 'mul':
            s = inputs[0]
            for i in range(1, len(inputs)):
                s *= inputs[i]
            return s
        elif self.mode == 'max':
            s = inputs[0]
            for i in range(1, len(inputs)):
                s = K.maximum(s, inputs[i])
            return s
        elif self.mode == 'dot':
            l1 = inputs[0]
            l2 = inputs[1]
            output = K.batch_dot(l1, l2, self.dot_axes)
            return output

        elif self.mode == 'cos':
            l1 = inputs[0]
            l2 = inputs[1]
            denominator = K.sqrt(K.batch_dot(l1, l1, self.dot_axes) *
                                 K.batch_dot(l2, l2, self.dot_axes))
            denominator = K.maximum(denominator, K.epsilon())
            output = K.batch_dot(l1, l2, self.dot_axes) / denominator
            output = K.expand_dims(output, 1)
            return output
        else:
            raise ValueError('Unknown merge mode.')

    def __call__(self, inputs, mask=None):
        '''We disable successive calls to __call__ for Merge layers.
        Although there is no technical obstacle to
        making it possible to __call__ a Merge instance many times
        (it is just a layer), it would make for a rather inelegant API.
        '''
        if not isinstance(inputs, list):
            raise TypeError('Merge can only be called on a list of tensors, '
                            'not a single tensor. Received: ' + str(inputs))
        if self.built:
            raise RuntimeError('A Merge layer cannot be used more than once, '
                               'please use '
                               'the "merge" function instead: '
                               '`merged_tensor = merge([tensor_1, tensor2])`.')

        all_keras_tensors = True
        for x in inputs:
            if not hasattr(x, '_keras_history'):
                all_keras_tensors = False
                break

        if all_keras_tensors:
            layers = []
            node_indices = []
            tensor_indices = []
            for x in inputs:
                layer, node_index, tensor_index = x._keras_history
                layers.append(layer)
                node_indices.append(node_index)
                tensor_indices.append(tensor_index)
            self._arguments_validation(layers, self.mode,
                                       self.concat_axis, self.dot_axes,
                                       node_indices, tensor_indices)
            self.built = True
            self.add_inbound_node(layers, node_indices, tensor_indices)

            outputs = self.inbound_nodes[-1].output_tensors
            return outputs[0]  # Merge only returns a single tensor.
        else:
            return self.call(inputs, mask)

    def get_output_shape_for(self, input_shape):
        # Must have multiple input shape tuples.
        assert isinstance(input_shape, list)
        # Case: callable self._output_shape.
        if callable(self.mode):
            if callable(self._output_shape):
                output_shape = self._output_shape(input_shape)
                return output_shape
            elif self._output_shape is not None:
                return (input_shape[0][0],) + tuple(self._output_shape)
            else:
                # TODO: consider shape auto-inference with TF.
                raise ValueError('The Merge layer ' + self.name +
                                 ' has a callable `mode` argument, '
                                 'and we cannot infer its output shape '
                                 'because no `output_shape` '
                                 'argument was provided. '
                                 'Make sure to pass a shape tuple '
                                 '(or callable) '
                                 '`output_shape` to Merge.')
        # Pre-defined merge modes.
        input_shapes = input_shape
        if self.mode in ['sum', 'mul', 'ave', 'max']:
            # All tuples in input_shapes should be the same.
            return input_shapes[0]
        elif self.mode == 'concat':
            output_shape = list(input_shapes[0])
            for shape in input_shapes[1:]:
                if output_shape[self.concat_axis] is None or shape[self.concat_axis] is None:
                    output_shape[self.concat_axis] = None
                    break
                output_shape[self.concat_axis] += shape[self.concat_axis]
            return tuple(output_shape)
        elif self.mode in ['dot', 'cos']:
            shape1 = list(input_shapes[0])
            shape2 = list(input_shapes[1])
            shape1.pop(self.dot_axes[0])
            shape2.pop(self.dot_axes[1])
            shape2.pop(0)
            output_shape = shape1 + shape2
            if len(output_shape) == 1:
                output_shape += [1]
            return tuple(output_shape)

    def compute_mask(self, inputs, mask=None):
        if mask is None or all([m is None for m in mask]):
            return None

        assert hasattr(mask, '__len__') and len(mask) == len(inputs)

        if self.mode in ['sum', 'mul', 'ave']:
            masks = [K.expand_dims(m, 0) for m in mask if m is not None]
            return K.all(K.concatenate(masks, axis=0), axis=0, keepdims=False)
        elif self.mode == 'concat':
            # Make a list of masks while making sure
            # the dimensionality of each mask
            # is the same as the corresponding input.
            masks = []
            for input_i, mask_i in zip(inputs, mask):
                if mask_i is None:
                    # Input is unmasked. Append all 1s to masks,
                    # but cast it to uint8 first
                    masks.append(K.cast(K.ones_like(input_i), 'uint8'))
                elif K.ndim(mask_i) < K.ndim(input_i):
                    # Mask is smaller than the input, expand it
                    masks.append(K.expand_dims(mask_i))
                else:
                    masks.append(mask_i)
            concatenated = K.concatenate(masks, axis=self.concat_axis)
            return K.all(concatenated, axis=-1, keepdims=False)
        elif self.mode in ['cos', 'dot']:
            return None
        elif callable(self.mode):
            if callable(self._output_mask):
                return self._output_mask(mask)
            else:
                return self._output_mask
        else:
            # This should have been caught earlier.
            raise ValueError('Invalid merge mode: {}'.format(self.mode))

    def get_config(self):
        if isinstance(self.mode, python_types.LambdaType):
            mode = func_dump(self.mode)
            mode_type = 'lambda'
        elif callable(self.mode):
            mode = self.mode.__name__
            mode_type = 'function'
        else:
            mode = self.mode
            mode_type = 'raw'

        if isinstance(self._output_shape, python_types.LambdaType):
            output_shape = func_dump(self._output_shape)
            output_shape_type = 'lambda'
        elif callable(self._output_shape):
            output_shape = self._output_shape.__name__
            output_shape_type = 'function'
        else:
            output_shape = self._output_shape
            output_shape_type = 'raw'

        if isinstance(self._output_mask, python_types.LambdaType):
            output_mask = func_dump(self._output_mask)
            output_mask_type = 'lambda'
        elif callable(self._output_mask):
            output_mask = self._output_mask.__name__
            output_mask_type = 'function'
        else:
            output_mask = self._output_mask
            output_mask_type = 'raw'

        return {'name': self.name,
                'mode': mode,
                'mode_type': mode_type,
                'concat_axis': self.concat_axis,
                'dot_axes': self.dot_axes,
                'output_shape': output_shape,
                'output_shape_type': output_shape_type,
                'output_mask': output_mask,
                'output_mask_type': output_mask_type,
                'arguments': self.arguments}

    @classmethod
    def from_config(cls, config):
        mode_type = config.pop('mode_type')
        if mode_type == 'function':
            mode = globals()[config['mode']]
        elif mode_type == 'lambda':
            mode = func_load(config['mode'], globs=globals())
        else:
            mode = config['mode']

        output_shape_type = config.pop('output_shape_type', None)
        if output_shape_type == 'function':
            output_shape = globals()[config['output_shape']]
        elif output_shape_type == 'lambda':
            output_shape = func_load(config['output_shape'],
                                     globs=globals())
        else:
            output_shape = config.get('output_shape')

        output_mask_type = config.pop('output_mask_type', None)
        if output_mask_type == 'function':
            output_mask = globals()[config['output_mask']]
        elif output_mask_type == 'lambda':
            output_mask = func_load(config['output_mask'],
                                    globs=globals())
        else:
            output_mask = config.get('output_mask')

        config['mode'] = mode
        config['output_shape'] = output_shape
        config['output_mask'] = output_mask
        return super(Merge, cls).from_config(config)


def merge(inputs, mode='sum', concat_axis=-1,
          dot_axes=-1, output_shape=None, output_mask=None,
          arguments=None, name=None):
    '''Functional merge, to apply to Keras tensors (NOT layers).
    Returns a Keras tensor.

    # Example

    ```python
    tensor_a = Input(shape=(32,))
    tensor_b = Input(shape=(32,))
    merged_tensor = merge([tensor_a, tensor_b], mode='concat', concat_axis=1)
    ```

    # Arguments
        mode: String or lambda/function. If string, must be one
            of: 'sum', 'mul', 'concat', 'ave', 'cos', 'dot'.
            If lambda/function, it should take as input a list of tensors
            and return a single tensor.
        concat_axis: Integer, axis to use in mode `concat`.
        dot_axes: Integer or tuple of integers,
            axes to use in mode `dot` or `cos`.
        output_shape: Shape tuple (tuple of integers), or lambda/function
            to compute output_shape (only if merge mode is a lambda/function).
            If the latter case, it should take as input a list of shape tuples
            (1:1 mapping to input tensors) and return a single shape tuple,
            including the batch size
            (same convention as the `get_output_shape_for` method of layers).
        node_indices: Optional list of integers containing
            the output node index for each input layer
            (in case some input layers have multiple output nodes).
            will default to an array of 0s if not provided.
        tensor_indices: Optional list of indices of output tensors
            to consider for merging
            (in case some input layer node returns multiple tensors).
    '''
    all_keras_tensors = True
    for x in inputs:
        if not hasattr(x, '_keras_history'):
            all_keras_tensors = False
            break
    if all_keras_tensors:
        input_layers = []
        node_indices = []
        tensor_indices = []
        for x in inputs:
            input_layer, node_index, tensor_index = x._keras_history
            input_layers.append(input_layer)
            node_indices.append(node_index)
            tensor_indices.append(tensor_index)
        merge_layer = Merge(input_layers, mode=mode,
                            concat_axis=concat_axis,
                            dot_axes=dot_axes,
                            output_shape=output_shape,
                            output_mask=output_mask,
                            arguments=arguments,
                            node_indices=node_indices,
                            tensor_indices=tensor_indices,
                            name=name)
        return merge_layer.inbound_nodes[0].output_tensors[0]
    else:
        merge_layer = Merge(mode=mode,
                            concat_axis=concat_axis,
                            dot_axes=dot_axes,
                            output_shape=output_shape,
                            output_mask=output_mask,
                            arguments=arguments,
                            name=name)
        return merge_layer(inputs)


class Container(Layer):
    '''A Container is a directed acyclic graph of layers.

    It is the topological form of a "model". A Model
    is simply a Container with added training routines.

    # Properties
        name
        inputs
        outputs
        input_layers
        output_layers
        input_spec (list of class instances)
            each entry describes one required input:
                - ndim
                - dtype
        trainable (boolean)
        input_shape
        output_shape
        inbound_nodes: list of nodes
        outbound_nodes: list of nodes
        supports_masking (boolean)
        trainable_weights (list of variables)
        non_trainable_weights (list of variables)
        constraints (list of tuples (weight, constraint))

    # Methods
        summary
        get_layer
        get_weights
        set_weights
        get_config
        get_output_shape_for

    # Class Methods
        from_config
    '''
    def __init__(self, input, output, name=None):
        # Handle name argument.
        if not name:
            prefix = self.__class__.__name__.lower()
            name = prefix + '_' + str(K.get_uid(prefix))
        self.name = name

        # Whether container weights are trainable.
        self.trainable = True

        # Container-specific properties.
        if isinstance(input, (list, tuple)):
            self.inputs = list(input)  # Tensor or list of tensors.
        else:
            self.inputs = [input]
        if isinstance(output, (list, tuple)):
            self.outputs = list(output)
        else:
            self.outputs = [output]

        # Check for redundancy in inputs.
        inputs_set = set(self.inputs)
        if len(inputs_set) != len(self.inputs):
            raise ValueError('The list of inputs passed to the model '
                             'is redundant. '
                             'All inputs should only appear once.'
                             ' Found: ' + str(self.inputs))

        # List of initial layers (1 to 1 mapping with self.inputs,
        # hence the same layer might appear twice)
        self.input_layers = []
        # TODO: probably useless because input layers must be Input layers
        # (node_indices = [0], tensor_indices = [0])
        self.input_layers_node_indices = []
        self.input_layers_tensor_indices = []
        # list of layers (1 to 1 mapping with self.inputs,
        # hence the same layer might appear twice)
        self.output_layers = []
        # TODO: probably useless
        self.output_layers_node_indices = []
        self.output_layers_tensor_indices = []
        # all layers in order of horizontal graph traversal.
        # Entries are unique. Includes input and output layers.
        self.layers = []

        # this is for performance optimization
        # when calling the Container on new inputs.
        # every time the Container is called on a set on input tensors,
        # we compute the output tensors,
        # output masks and output shapes in one pass,
        # then cache them here. When of of these output is queried later,
        # we retrieve it from there instead of recomputing it.
        self._output_mask_cache = {}
        self._output_tensor_cache = {}
        self._output_shape_cache = {}

        # Arguments validation.
        for x in self.inputs:
            # Check that x is a Keras tensor.
            if not hasattr(x, '_keras_history'):
                cls_name = self.__class__.__name__
                raise TypeError('Input tensors to a ' + cls_name + ' ' +
                                'must be Keras tensors. Found: ' + str(x) +
                                ' (missing Keras metadata).')
            # Check that x is an input tensor.
            layer, node_index, tensor_index = x._keras_history
            if len(layer.inbound_nodes) > 1 or (layer.inbound_nodes and layer.inbound_nodes[0].inbound_layers):
                cls_name = self.__class__.__name__
                warnings.warn(cls_name + ' inputs must come from '
                              'a Keras Input layer, '
                              'they cannot be the output of '
                              'a previous non-Input layer. '
                              'Here, a tensor specified as '
                              'input to "' + self.name +
                              '" was not an Input tensor, '
                              'it was generated by layer ' +
                              layer.name + '.\n'
                              'Note that input tensors are '
                              'instantiated via `tensor = Input(shape)`.\n'
                              'The tensor that caused the issue was: ' +
                              str(x.name))
        for x in self.outputs:
            if not hasattr(x, '_keras_history'):
                cls_name = self.__class__.__name__
                raise TypeError('Output tensors to a ' + cls_name + ' must be '
                                'Keras tensors. Found: ' + str(x))
        # Build self.output_layers:
        for x in self.outputs:
            layer, node_index, tensor_index = x._keras_history
            self.output_layers.append(layer)
            self.output_layers_node_indices.append(node_index)
            self.output_layers_tensor_indices.append(tensor_index)

        # Fill in the output mask cache.
        masks = []
        for x in self.inputs:
            layer, node_index, tensor_index = x._keras_history
            node = layer.inbound_nodes[node_index]
            mask = node.output_masks[tensor_index]
            masks.append(mask)
        mask_cache_key = ','.join([str(id(x)) for x in self.inputs])
        mask_cache_key += '_' + ','.join([str(id(x)) for x in masks])
        masks = []
        for x in self.outputs:
            layer, node_index, tensor_index = x._keras_history
            node = layer.inbound_nodes[node_index]
            mask = node.output_masks[tensor_index]
            masks.append(mask)
        if len(masks) == 1:
            mask = masks[0]
        else:
            mask = masks
        self._output_mask_cache[mask_cache_key] = mask

        # Build self.input_layers:
        for x in self.inputs:
            layer, node_index, tensor_index = x._keras_history
            # It's supposed to be an input layer, so only one node
            # and one tensor output.
            assert node_index == 0
            assert tensor_index == 0
            self.input_layers.append(layer)
            self.input_layers_node_indices.append(node_index)
            self.input_layers_tensor_indices.append(tensor_index)

        # Build self.input_names and self.output_names.
        self.input_names = []
        self.output_names = []
        for layer in self.input_layers:
            self.input_names.append(layer.name)
        for layer in self.output_layers:
            self.output_names.append(layer.name)

        self.internal_input_shapes = [x._keras_shape for x in self.inputs]
        self.internal_output_shapes = [x._keras_shape for x in self.outputs]

        # Container_nodes: set of nodes included in the graph
        # (not all nodes included in the layers
        # are relevant to the current graph).
        container_nodes = set()  # ids of all nodes relevant to the Container
        nodes_depths = {}  # dict {node: depth value}
        layers_depths = {}  # dict {layer: depth value}
        layer_indices = {}  # dict {layer: index in traversal}

        def make_node_marker(node, depth):
            return str(id(node)) + '-' + str(depth)

        def build_map_of_graph(tensor, seen_nodes=set(), depth=0,
                               layer=None, node_index=None, tensor_index=None):
            '''This recursively updates the maps nodes_depths,
            layers_depths and the set container_nodes.
            Does not try to detect cycles in graph (TODO?)

            # Arguments
                tensor: Some tensor in a graph.
                seen_nodes: Set of node ids ("{layer.name}_ib-{node_index}")
                    of nodes seen so far. Useful to prevent infinite loops.
                depth: Current depth in the graph (0 = last output).
                layer: Layer from which `tensor` comes from. If not provided,
                    will be obtained from `tensor._keras_history`.
                node_index: Node index from which `tensor` comes from.
                tensor_index: Tensor_index from which `tensor` comes from.
            '''
            if not layer or node_index is None or tensor_index is None:
                layer, node_index, tensor_index = tensor._keras_history
            node = layer.inbound_nodes[node_index]

            # Prevent cycles.
            seen_nodes.add(make_node_marker(node, depth))

            node_key = layer.name + '_ib-' + str(node_index)
            # Update container_nodes.
            container_nodes.add(node_key)
            # Update nodes_depths.
            node_depth = nodes_depths.get(node)
            if node_depth is None:
                nodes_depths[node] = depth
            else:
                nodes_depths[node] = max(depth, node_depth)
            # Update layers_depths.
            previously_seen_depth = layers_depths.get(layer)
            if previously_seen_depth is None:
                current_depth = depth
            else:
                current_depth = max(depth, previously_seen_depth)
            layers_depths[layer] = current_depth
            if layer not in layer_indices:
                layer_indices[layer] = len(layer_indices)

            # Propagate to all previous tensors connected to this node.
            for i in range(len(node.inbound_layers)):
                x = node.input_tensors[i]
                layer = node.inbound_layers[i]
                node_index = node.node_indices[i]
                tensor_index = node.tensor_indices[i]
                next_node = layer.inbound_nodes[node_index]
                # use node_marker to prevent cycles
                node_marker = make_node_marker(next_node, current_depth + 1)
                if node_marker not in seen_nodes:
                    build_map_of_graph(x, seen_nodes, current_depth + 1,
                                       layer, node_index, tensor_index)

        for x in self.outputs:
            seen_nodes = set()
            build_map_of_graph(x, seen_nodes, depth=0)

        # Build a dict {depth: list of nodes with this depth}
        nodes_by_depth = {}
        for node, depth in nodes_depths.items():
            if depth not in nodes_by_depth:
                nodes_by_depth[depth] = []
            nodes_by_depth[depth].append(node)

        # Build a dict {depth: list of layers with this depth}
        layers_by_depth = {}
        for layer, depth in layers_depths.items():
            if depth not in layers_by_depth:
                layers_by_depth[depth] = []
            layers_by_depth[depth].append(layer)

        # Get sorted list of layer depths.
        depth_keys = list(layers_by_depth.keys())
        depth_keys.sort(reverse=True)

        # Set self.layers and self.layers_by_depth.
        layers = []
        for depth in depth_keys:
            layers_for_depth = layers_by_depth[depth]
            # Container.layers needs to have a deterministic order:
            # here we order them by traversal order.
            if K.legacy_weight_ordering():
                layers_for_depth.sort(key=lambda x: x.name)
            else:
                layers_for_depth.sort(key=lambda x: layer_indices[x])
            for layer in layers_for_depth:
                layers.append(layer)
        self.layers = layers
        self.layers_by_depth = layers_by_depth

        # Get sorted list of node depths.
        depth_keys = list(nodes_by_depth.keys())
        depth_keys.sort(reverse=True)

        # Check that all tensors required are computable.
        # computable_tensors: all tensors in the graph
        # that can be computed from the inputs provided.
        computable_tensors = []
        for x in self.inputs:
            computable_tensors.append(x)

        layers_with_complete_input = []  # To provide a better error msg.
        for depth in depth_keys:
            for node in nodes_by_depth[depth]:
                layer = node.outbound_layer
                if layer:
                    for x in node.input_tensors:
                        if x not in computable_tensors:
                            raise RuntimeError(
                                'Graph disconnected: '
                                'cannot obtain value for tensor ' +
                                str(x) + ' at layer "' + layer.name + '". '
                                'The following previous layers '
                                'were accessed without issue: ' +
                                str(layers_with_complete_input))
                    for x in node.output_tensors:
                        computable_tensors.append(x)
                    layers_with_complete_input.append(layer.name)

        # Set self.nodes and self.nodes_by_depth.
        self.container_nodes = container_nodes
        self.nodes_by_depth = nodes_by_depth

        # Ensure name unicity, which will be crucial for serialization
        # (since serialized nodes refer to layers by their name).
        all_names = [layer.name for layer in self.layers]
        for name in all_names:
            if all_names.count(name) != 1:
                raise RuntimeError('The name "' + name + '" is used ' +
                                   str(all_names.count(name)) +
                                   ' times in the model. '
                                   'All layer names should be unique.')

        # Layer parameters.
        # The new container starts with a single inbound node
        # for its inputs, and no outbound nodes.
        self.outbound_nodes = []  # Will be appended to by future calls to __call__
        self.inbound_nodes = []  # Will be appended to below, and by future calls to __call__
        # Create the node linking internal inputs to internal outputs.
        Node(outbound_layer=self,
             inbound_layers=[],
             node_indices=[],
             tensor_indices=[],
             input_tensors=self.inputs,
             output_tensors=self.outputs,
             # No container-level masking for now.
             input_masks=[None for _ in self.inputs],
             output_masks=[None for _ in self.outputs],
             input_shapes=[x._keras_shape for x in self.inputs],
             output_shapes=[x._keras_shape for x in self.outputs])
        self.built = True
        self.supports_masking = False
        # The following are implemented as property functions:
        # self.constraints
        # self.trainable_weights
        # self.non_trainable_weights
        # self.input_spec

    def get_layer(self, name=None, index=None):
        '''Returns a layer based on either its name (unique)
        or its index in the graph. Indices are based on
        order of horizontal graph traversal (bottom-up).

        # Arguments
            name: String, name of layer.
            index: Integer, index of layer.

        # Returns
            A layer instance.
        '''
        # It would be unreliable to build a dictionary
        # based on layer names, because names can potentially
        # be changed at any point by the user
        # without the container being notified of it.
        if index is not None:
            if len(self.layers) <= index:
                raise ValueError('Was asked to retrieve layer at index ' +
                                 str(index) + ' but model only has ' +
                                 str(len(self.layers)) + ' layers.')
            else:
                return self.layers[index]
        else:
            assert name, 'Provide either a layer name or layer index.'
        layer = None
        for layer in self.layers:
            if layer.name == name:
                return layer
        if not layer:
            raise ValueError('No such layer: ' + name)

    @property
    def updates(self):
        updates = []
        for layer in self.layers:
            if hasattr(layer, 'updates'):
                if len(layer.inbound_nodes) == 1:
                    updates += layer.updates
                else:
                    # Collect updates that are dependent on inputs
                    # that are part of the model.
                    for node_index, node in enumerate(layer.inbound_nodes):
                        node_key = layer.name + '_ib-' + str(node_index)
                        if node_key in self.container_nodes:
                            # The model owns this layer node.
                            inputs = node.input_tensors
                            updates += layer.get_updates_for(inputs)
                    # Collect unconditional updates.
                    updates += layer.get_updates_for(None)
        return updates

    @property
    def losses(self):
        losses = []
        for layer in self.layers:
            if hasattr(layer, 'losses'):
                if len(layer.inbound_nodes) == 1:
                    losses += layer.losses
                else:
                    # Collect losses that are dependent on inputs
                    # that are part of the model.
                    for node_index, node in enumerate(layer.inbound_nodes):
                        node_key = layer.name + '_ib-' + str(node_index)
                        if node_key in self.container_nodes:
                            # The model owns this layer node.
                            inputs = node.input_tensors
                            losses += layer.get_losses_for(inputs)
                    # Collect unconditional losses.
                    losses += layer.get_losses_for(None)
        return losses

    @property
    def stateful(self):
        return any([(hasattr(layer, 'stateful') and layer.stateful) for layer in self.layers])

    def reset_states(self):
        for layer in self.layers:
            if hasattr(layer, 'reset_states') and getattr(layer, 'stateful', False):
                layer.reset_states()

    @property
    def state_updates(self):
        '''Returns the `updates` from all layers that are
        stateful.  This is useful for separating training updates and
        state updates, e.g. when we need to update a layer's internal state
        during prediction.
        '''
        state_updates = []
        for layer in self.layers:
            if getattr(layer, 'stateful', False):
                if hasattr(layer, 'updates'):
                    state_updates += layer.updates
        return state_updates

    @property
    def constraints(self):
        cons = {}
        for layer in self.layers:
            for key, value in layer.constraints.items():
                if key in cons and cons[key] != value:
                    raise ValueError('Received multiple constraints '
                                     'for one weight tensor: ' + str(key))
                cons[key] = value
        return cons

    @property
    def regularizers(self):
        warnings.warn('The `regularizers` attribute of layers/models '
                      'is deprecated. '
                      'Regularization losses are now managed via the `losses` '
                      'layer/model property.\n'
                      'The `regularizers` attribute will be removed '
                      'after 06/2017.')
        return []

    @property
    def trainable_weights(self):
        if not self.trainable:
            return []
        weights = []
        for layer in self.layers:
            weights += layer.trainable_weights
        return weights

    @property
    def non_trainable_weights(self):
        weights = []
        for layer in self.layers:
            weights += layer.non_trainable_weights
        if not self.trainable:
            trainable_weights = []
            for layer in self.layers:
                trainable_weights += layer.trainable_weights
            return trainable_weights + weights
        return weights

    def get_weights(self):
        '''Returns the weights of the model,
        as a flat list of Numpy arrays.
        '''
        weights = []
        for layer in self.layers:
            weights += layer.weights
        return K.batch_get_value(weights)

    def set_weights(self, weights):
        '''Sets the weights of the model.
        The `weights` argument should be a list
        of Numpy arrays with shapes and types matching
        the output of `model.get_weights()`.
        '''
        tuples = []
        for layer in self.layers:
            nb_param = len(layer.weights)
            layer_weights = weights[:nb_param]
            for sw, w in zip(layer.weights, layer_weights):
                tuples.append((sw, w))
            weights = weights[nb_param:]
        K.batch_set_value(tuples)

    @property
    def input_spec(self):
        specs = []
        for layer in getattr(self, 'input_layers', []):
            if layer.input_spec is None:
                specs.append(None)
            else:
                if not isinstance(layer.input_spec, list):
                    raise TypeError('Layer ' + layer.name +
                                    ' has an input_spec attribute that '
                                    'is not a list. We expect a list. '
                                    'Found input_spec = ' +
                                    str(layer.input_spec))
                specs += layer.input_spec
        return specs

    @property
    def uses_learning_phase(self):
        '''True if any layer in the graph uses it.
        '''
        layers_learning_phase = any([layer.uses_learning_phase for layer in self.layers])
        return layers_learning_phase

    def call(self, input, mask=None):
        '''`call` just reapplies all ops in the graph to the new inputs
        (e.g. build a new computational graph from the provided inputs).

        It is callable on non-Keras tensors.

        # Arguments
            input: A tensor or list of tensors.
            mask: A mask or list of masks. A mask can be
                either a tensor or None (no mask).

        # Returns
            A tensor if there is a single output, or
            a list of tensors if there are more than one outputs.
        '''
        inputs = to_list(input)
        if mask is None:
            masks = [None for _ in range(len(inputs))]
        else:
            masks = to_list(mask)
        cache_key = ','.join([str(id(x)) for x in inputs])
        cache_key += '_' + ','.join([str(id(x)) for x in masks])
        if cache_key in self._output_tensor_cache:
            return self._output_tensor_cache[cache_key]
        else:
            output_tensors, output_masks, output_shapes = self.run_internal_graph(inputs, masks)
            return output_tensors

    def compute_mask(self, input, mask):
        inputs = to_list(input)
        if mask is None:
            masks = [None for _ in range(len(inputs))]
        else:
            masks = to_list(mask)
        cache_key = ','.join([str(id(x)) for x in inputs])
        cache_key += '_' + ','.join([str(id(x)) for x in masks])
        if cache_key in self._output_mask_cache:
            return self._output_mask_cache[cache_key]
        else:
            output_tensors, output_masks, output_shapes = self.run_internal_graph(inputs, masks)
            return output_masks

    def get_output_shape_for(self, input_shape):
        input_shapes = to_list(input_shape)
        if len(input_shapes) != len(self.input_layers):
            raise ValueError('Invalid input_shape argument ' +
                             str(input_shape) + ': model has ' +
                             str(len(self.input_layers)) + ' tensor inputs.')

        cache_key = ','.join([str(x) for x in input_shapes])
        if cache_key in self._output_shape_cache:
            output_shapes = self._output_shape_cache[cache_key]
            if isinstance(output_shapes, list) and len(output_shapes) == 1:
                return output_shapes[0]
            return output_shapes
        else:
            # Bad luck, we have to run the graph manually.
            layers_to_output_shapes = {}
            for i in range(len(input_shapes)):
                layer = self.input_layers[i]
                input_shape = input_shapes[i]
                # It's an input layer: get_output_shape_for is identity,
                # and there is only one node and one tensor output.
                shape_key = layer.name + '_0_0'
                layers_to_output_shapes[shape_key] = input_shape

            depth_keys = list(self.nodes_by_depth.keys())
            depth_keys.sort(reverse=True)
            # Iterate over nodes, by depth level.
            if len(depth_keys) > 1:
                for depth in depth_keys:
                    nodes = self.nodes_by_depth[depth]
                    for node in nodes:
                        # This is always a single layer, never a list.
                        layer = node.outbound_layer
                        if layer in self.input_layers:
                            # We've already covered the input layers
                            # a few lines above.
                            continue
                        # Potentially redundant list,
                        # same size of node.input_tensors.
                        input_shapes = []
                        for j in range(len(node.inbound_layers)):
                            inbound_layer = node.inbound_layers[j]
                            node_index = node.node_indices[j]
                            tensor_index = node.tensor_indices[j]
                            shape_key = inbound_layer.name + '_%s_%s' % (node_index, tensor_index)
                            input_shape = layers_to_output_shapes[shape_key]
                            input_shapes.append(input_shape)

                        if len(input_shapes) == 1:
                            output_shape = layer.get_output_shape_for(input_shapes[0])
                        else:
                            output_shape = layer.get_output_shape_for(input_shapes)

                        output_shapes = to_list(output_shape)
                        node_index = layer.inbound_nodes.index(node)
                        for j in range(len(output_shapes)):
                            shape_key = layer.name + '_%s_%s' % (node_index, j)
                            layers_to_output_shapes[shape_key] = output_shapes[j]

            # Read final output shapes from layers_to_output_shapes.
            output_shapes = []
            output_shape_keys = []
            for i in range(len(self.output_layers)):
                layer = self.output_layers[i]
                node_index = self.output_layers_node_indices[i]
                tensor_index = self.output_layers_tensor_indices[i]
                shape_key = layer.name + '_%s_%s' % (node_index, tensor_index)
                output_shape_keys.append(shape_key)

            for i, key in enumerate(output_shape_keys):
                assert key in layers_to_output_shapes
                output_shapes.append(layers_to_output_shapes[key])
            # Store in cache.
            self._output_shape_cache[cache_key] = output_shapes
            if isinstance(output_shapes, list) and len(output_shapes) == 1:
                return output_shapes[0]
            return output_shapes

    def run_internal_graph(self, inputs, masks=None):
        '''Computes output tensors for new inputs.

        # Note:
            - Expects `inputs` to be a list (potentially with 1 element).
            - Can be run on non-Keras tensors.

        # Arguments
            inputs: List of tensors
            masks: List of masks (tensors or None).

        # Returns
            Three lists: output_tensors, output_masks, output_shapes
        '''
        if masks is None:
            masks = [None for _ in range(len(inputs))]

        # Dictionary mapping reference tensors to tuples
        # (computed tensor, compute mask)
        # we assume a 1:1 mapping from tensor to mask
        # TODO: raise exception when a `.compute_mask()` call
        # does not return a list the same size as `call`
        tensor_map = {}
        for x, y, mask in zip(self.inputs, inputs, masks):
            tensor_map[str(id(x))] = (y, mask)

        depth_keys = list(self.nodes_by_depth.keys())
        depth_keys.sort(reverse=True)
        for depth in depth_keys:
            nodes = self.nodes_by_depth[depth]
            for node in nodes:
                # This is always a single layer, never a list.
                layer = node.outbound_layer

                reference_input_tensors = node.input_tensors
                reference_output_tensors = node.output_tensors

                # If all previous input tensors are available in tensor_map,
                # then call node.inbound_layer on them.
                computed_data = []  # List of tuples (input, mask).
                for x in reference_input_tensors:
                    if str(id(x)) in tensor_map:
                        computed_data.append(tensor_map[str(id(x))])
                if len(computed_data) == len(reference_input_tensors):
                    # call layer
                    if len(computed_data) == 1:
                        computed_tensor, computed_mask = computed_data[0]
                        output_tensors = to_list(layer.call(computed_tensor,
                                                            computed_mask))
                        output_masks = to_list(layer.compute_mask(computed_tensor,
                                                                  computed_mask))
                        computed_tensors = [computed_tensor]
                        computed_masks = [computed_mask]
                    else:
                        computed_tensors = [x[0] for x in computed_data]
                        computed_masks = [x[1] for x in computed_data]
                        output_tensors = to_list(layer.call(computed_tensors,
                                                            computed_masks))
                        output_masks = to_list(layer.compute_mask(computed_tensors,
                                                                  computed_masks))

                    # Update model updates and losses:
                    layer_inputs = [x[0] for x in computed_data]
                    # Keep track of updates that depend on the inputs
                    # (e.g. BN updates).
                    self.add_update(layer.get_updates_for(layer_inputs), inputs)
                    # Keep track of unconditional updates (e.g. a counter).
                    self.add_update(layer.get_updates_for(None), None)
                    # Keep track of losses that depend on the inputs
                    # (e.g. activity regularizers).
                    self.add_loss(layer.get_losses_for(layer_inputs), inputs)
                    # Keep track of unconditional losses
                    # (e.g. weight regularizers).
                    self.add_loss(layer.get_losses_for(None), None)

                    # Update _keras_shape.
                    if all([hasattr(x, '_keras_shape') for x in computed_tensors]):
                        if len(computed_tensors) == 1:
                            shapes = to_list(layer.get_output_shape_for(computed_tensors[0]._keras_shape))
                            uses_learning_phase = computed_tensors[0]._uses_learning_phase or layer.uses_learning_phase
                        else:
                            shapes = to_list(layer.get_output_shape_for([x._keras_shape for x in computed_tensors]))
                            uses_learning_phase = any([x._uses_learning_phase for x in computed_tensors]) or layer.uses_learning_phase
                        for x, s in zip(output_tensors, shapes):
                            x._keras_shape = s
                            x._uses_learning_phase = uses_learning_phase

                    # Update tensor_map.
                    for x, y, mask in zip(reference_output_tensors, output_tensors, output_masks):
                        tensor_map[str(id(x))] = (y, mask)

        output_tensors = []
        output_masks = []
        output_shapes = []
        for x in self.outputs:
            # TODO: Better error message.
            assert str(id(x)) in tensor_map, 'Could not compute output ' + str(x)
            tensor, mask = tensor_map[str(id(x))]
            if hasattr(tensor, '_keras_shape') and output_shapes is not None:
                shape = tensor._keras_shape
                output_shapes.append(shape)
            else:
                output_shapes = None
            output_tensors.append(tensor)
            output_masks.append(mask)

        # Update cache;
        # keys are based on ids on input tensors and inputs masks.
        cache_key = ','.join([str(id(x)) for x in inputs])
        cache_key += '_' + ','.join([str(id(x)) for x in masks])

        if len(output_tensors) == 1:
            output_tensors = output_tensors[0]
            self._output_tensor_cache[cache_key] = output_tensors
        else:
            self._output_tensor_cache[cache_key] = output_tensors

        if len(output_masks) == 1:
            output_masks = output_masks[0]
            self._output_mask_cache[cache_key] = output_masks
        else:
            self._output_mask_cache[cache_key] = output_masks

        if output_shapes is not None:
            input_shapes = [x._keras_shape for x in inputs]
            cache_key = ','.join([str(x) for x in input_shapes])
            if len(output_shapes) == 1:
                output_shapes = output_shapes[0]
                self._output_shape_cache[cache_key] = output_shapes
            else:
                self._output_shape_cache[cache_key] = output_shapes
        return output_tensors, output_masks, output_shapes

    def get_config(self):
        config = {
            'name': self.name,
        }
        node_conversion_map = {}
        for layer in self.layers:
            if issubclass(layer.__class__, Container):
                # Containers start with a pre-existing node
                # linking their input to output.
                kept_nodes = 1
            else:
                kept_nodes = 0
            for original_node_index, node in enumerate(layer.inbound_nodes):
                node_key = layer.name + '_ib-' + str(original_node_index)
                if node_key in self.container_nodes:
                    node_conversion_map[node_key] = kept_nodes
                    kept_nodes += 1
        layer_configs = []
        for layer in self.layers:  # From the earliest layers on.
            layer_class_name = layer.__class__.__name__
            layer_config = layer.get_config()
            filtered_inbound_nodes = []
            for original_node_index, node in enumerate(layer.inbound_nodes):
                node_key = layer.name + '_ib-' + str(original_node_index)
                if node_key in self.container_nodes:
                    # The node is relevant to the model:
                    # add to filtered_inbound_nodes.
                    if node.inbound_layers:
                        node_data = []
                        for i in range(len(node.inbound_layers)):
                            inbound_layer = node.inbound_layers[i]
                            node_index = node.node_indices[i]
                            tensor_index = node.tensor_indices[i]
                            node_key = inbound_layer.name + '_ib-' + str(node_index)
                            new_node_index = node_conversion_map.get(node_key, 0)
                            node_data.append([inbound_layer.name,
                                              new_node_index,
                                              tensor_index])
                        filtered_inbound_nodes.append(node_data)
            layer_configs.append({
                'name': layer.name,
                'class_name': layer_class_name,
                'config': layer_config,
                'inbound_nodes': filtered_inbound_nodes,
            })
        config['layers'] = layer_configs

        # Gather info about inputs and outputs.
        model_inputs = []
        for i in range(len(self.input_layers)):
            layer = self.input_layers[i]
            node_index = self.input_layers_node_indices[i]
            node_key = layer.name + '_ib-' + str(node_index)
            new_node_index = node_conversion_map[node_key]
            tensor_index = self.input_layers_tensor_indices[i]
            model_inputs.append([layer.name, new_node_index, tensor_index])
        config['input_layers'] = model_inputs
        model_outputs = []
        for i in range(len(self.output_layers)):
            layer = self.output_layers[i]
            node_index = self.output_layers_node_indices[i]
            node_key = layer.name + '_ib-' + str(node_index)
            new_node_index = node_conversion_map[node_key]
            tensor_index = self.output_layers_tensor_indices[i]
            model_outputs.append([layer.name, new_node_index, tensor_index])
        config['output_layers'] = model_outputs
        return copy.deepcopy(config)

    @classmethod
    def from_config(cls, config, custom_objects=None):
        '''Instantiates a Model from its config (output of `get_config()`).
        '''
        from keras.utils.layer_utils import layer_from_config

        # layer instances created during
        # the graph reconstruction process
        created_layers = {}

        def process_layer(layer_data):
            # Iterate over saved layers, instantiate them,
            # then call them on appropriate inputs to create graph nodes
            layer_name = layer_data['name']

            # Instantiate layer.
            layer = layer_from_config(layer_data,
                                      custom_objects=custom_objects)
            created_layers[layer_name] = layer

            # Gather layer inputs.
            inbound_nodes_data = layer_data['inbound_nodes']
            for node_data in inbound_nodes_data:
                input_tensors = []
                for input_data in node_data:
                    inbound_layer_name, inbound_node_index, inbound_tensor_index = input_data
                    assert inbound_layer_name in created_layers, 'Missing layer: %s' % inbound_layer_name
                    inbound_layer = created_layers[inbound_layer_name]
                    inbound_node = inbound_layer.inbound_nodes[inbound_node_index]
                    input_tensors.append(inbound_node.output_tensors[inbound_tensor_index])
                # Call layer on its inputs, thus creating the node
                # and building the layer if needed.
                if input_tensors:
                    if len(input_tensors) == 1:
                        layer(input_tensors[0])
                    else:
                        layer(input_tensors)

        for layer_data in config['layers']:
            process_layer(layer_data)

        name = config.get('name')
        input_tensors = []
        output_tensors = []
        for layer_data in config['input_layers']:
            layer_name, node_index, tensor_index = layer_data
            assert layer_name in created_layers
            layer = created_layers[layer_name]
            layer_output_tensors = layer.inbound_nodes[node_index].output_tensors
            input_tensors.append(layer_output_tensors[tensor_index])
        for layer_data in config['output_layers']:
            layer_name, node_index, tensor_index = layer_data
            assert layer_name in created_layers
            layer = created_layers[layer_name]
            layer_output_tensors = layer.inbound_nodes[node_index].output_tensors
            output_tensors.append(layer_output_tensors[tensor_index])
        return cls(input=input_tensors, output=output_tensors, name=name)

    def save(self, filepath, overwrite=True):
        '''Save into a single HDF5 file:
            - The model architecture, allowing to re-instantiate the model.
            - The model weights.
            - The state of the optimizer, allowing to resume training
                exactly where you left off.

        This allows you to save the entirety of the state of a model
        in a single file.

        Saved models can be reinstantiated via `keras.models.load_model`.
        The model returned by `load_model`
        is a compiled model ready to be used (unless the saved model
        was never compiled in the first place).

        # Example

        ```python
        from keras.models import load_model

        model.save('my_model.h5')  # creates a HDF5 file 'my_model.h5'
        del model  # deletes the existing model

        # returns a compiled model
        # identical to the previous one
        model = load_model('my_model.h5')
        ```
        '''
        from ..models import save_model
        save_model(self, filepath, overwrite)

    def save_weights(self, filepath, overwrite=True):
        '''Dumps all layer weights to a HDF5 file.

        The weight file has:
            - `layer_names` (attribute), a list of strings
                (ordered names of model layers).
            - For every layer, a `group` named `layer.name`
                - For every such layer group, a group attribute `weight_names`,
                    a list of strings
                    (ordered names of weights tensor of the layer).
                - For every weight in the layer, a dataset
                    storing the weight value, named after the weight tensor.
        '''
        import h5py
        # If file exists and should not be overwritten:
        if not overwrite and os.path.isfile(filepath):
            proceed = ask_to_proceed_with_overwrite(filepath)
            if not proceed:
                return
        f = h5py.File(filepath, 'w')
        self.save_weights_to_hdf5_group(f)
        f.flush()
        f.close()

    def save_weights_to_hdf5_group(self, f):
        if hasattr(self, 'flattened_layers'):
            # Support for legacy Sequential/Merge behavior.
            flattened_layers = self.flattened_layers
        else:
            flattened_layers = self.layers

        f.attrs['layer_names'] = [layer.name.encode('utf8') for layer in flattened_layers]

        for layer in flattened_layers:
            g = f.create_group(layer.name)
            symbolic_weights = layer.weights
            weight_values = K.batch_get_value(symbolic_weights)
            weight_names = []
            for i, (w, val) in enumerate(zip(symbolic_weights, weight_values)):
                if hasattr(w, 'name') and w.name:
                    name = str(w.name)
                else:
                    name = 'param_' + str(i)
                weight_names.append(name.encode('utf8'))
            g.attrs['weight_names'] = weight_names
            for name, val in zip(weight_names, weight_values):
                param_dset = g.create_dataset(name, val.shape,
                                              dtype=val.dtype)
                if not val.shape:
                    # scalar
                    param_dset[()] = val
                else:
                    param_dset[:] = val

    def load_weights(self, filepath, by_name=False):
        '''Loads all layer weights from a HDF5 save file.

        If `by_name` is False (default) weights are loaded
        based on the network's topology, meaning the architecture
        should be the same as when the weights were saved.
        Note that layers that don't have weights are not taken
        into account in the topological ordering, so adding or
        removing layers is fine as long as they don't have weights.

        If `by_name` is True, weights are loaded into layers
        only if they share the same name. This is useful
        for fine-tuning or transfer-learning models where
        some of the layers have changed.
        '''
        import h5py
        f = h5py.File(filepath, mode='r')
        if 'layer_names' not in f.attrs and 'model_weights' in f:
            f = f['model_weights']
        if by_name:
            self.load_weights_from_hdf5_group_by_name(f)
        else:
            self.load_weights_from_hdf5_group(f)

        if hasattr(f, 'close'):
            f.close()

    def load_weights_from_hdf5_group(self, f):
        '''Weight loading is based on layer order in a list
        (matching model.flattened_layers for Sequential models,
        and model.layers for Model class instances), not
        on layer names.
        Layers that have no weights are skipped.
        '''
        if hasattr(self, 'flattened_layers'):
            # Support for legacy Sequential/Merge behavior.
            flattened_layers = self.flattened_layers
        else:
            flattened_layers = self.layers

        if 'nb_layers' in f.attrs:
            # Legacy format.
            nb_layers = f.attrs['nb_layers']
            if nb_layers != len(flattened_layers):
                raise ValueError('You are trying to load a weight file '
                                 'containing ' + str(nb_layers) +
                                 ' layers into a model with ' +
                                 str(len(flattened_layers)) + ' layers.')

            for k in range(nb_layers):
                g = f['layer_{}'.format(k)]
                weights = [g['param_{}'.format(p)] for p in range(g.attrs['nb_params'])]
                flattened_layers[k].set_weights(weights)
        else:
            # New file format.
            filtered_layers = []
            for layer in flattened_layers:
                weights = layer.weights
                if weights:
                    filtered_layers.append(layer)
            flattened_layers = filtered_layers

            layer_names = [n.decode('utf8') for n in f.attrs['layer_names']]
            filtered_layer_names = []
            for name in layer_names:
                g = f[name]
                weight_names = [n.decode('utf8') for n in g.attrs['weight_names']]
                if len(weight_names):
                    filtered_layer_names.append(name)
            layer_names = filtered_layer_names
            if len(layer_names) != len(flattened_layers):
                raise ValueError('You are trying to load a weight file '
                                 'containing ' + str(len(layer_names)) +
                                 ' layers into a model with ' +
                                 str(len(flattened_layers)) + ' layers.')

            # We batch weight value assignments in a single backend call
            # which provides a speedup in TensorFlow.
            weight_value_tuples = []
            for k, name in enumerate(layer_names):
                g = f[name]
                weight_names = [n.decode('utf8') for n in g.attrs['weight_names']]
                weight_values = [g[weight_name] for weight_name in weight_names]
                layer = flattened_layers[k]
                symbolic_weights = layer.weights
                if len(weight_values) != len(symbolic_weights):
                    raise ValueError('Layer #' + str(k) +
                                     ' (named "' + layer.name +
                                     '" in the current model) was found to '
                                     'correspond to layer ' + name +
                                     ' in the save file. '
                                     'However the new layer ' + layer.name +
                                     ' expects ' + str(len(symbolic_weights)) +
                                     ' weights, but the saved weights have ' +
                                     str(len(weight_values)) +
                                     ' elements.')
                if layer.__class__.__name__ == 'Convolution1D':
                    # This is for backwards compatibility with
                    # the old Conv1D weights format.
                    w = weight_values[0]
                    shape = w.shape
                    if shape[:2] != (layer.filter_length, 1) or shape[3] != layer.nb_filter:
                        # Legacy shape:
                        # (self.nb_filter, input_dim, self.filter_length, 1)
                        assert shape[0] == layer.nb_filter and shape[2:] == (layer.filter_length, 1)
                        w = np.transpose(w, (2, 3, 1, 0))
                        weight_values[0] = w
                weight_value_tuples += zip(symbolic_weights, weight_values)
            K.batch_set_value(weight_value_tuples)

    def load_weights_from_hdf5_group_by_name(self, f):
        ''' Name-based weight loading
        (instead of topological weight loading).
        Layers that have no matching name are skipped.
        '''
        if hasattr(self, 'flattened_layers'):
            # Support for legacy Sequential/Merge behavior.
            flattened_layers = self.flattened_layers
        else:
            flattened_layers = self.layers

        if 'nb_layers' in f.attrs:
                raise ValueError('The weight file you are trying to load is'
                                 ' in a legacy format that does not support'
                                 ' name-based weight loading.')
        else:
            # New file format.
            layer_names = [n.decode('utf8') for n in f.attrs['layer_names']]

            # Reverse index of layer name to list of layers with name.
            index = {}
            for layer in flattened_layers:
                if layer.name:
                    index.setdefault(layer.name, []).append(layer)

            # We batch weight value assignments in a single backend call
            # which provides a speedup in TensorFlow.
            weight_value_tuples = []
            for k, name in enumerate(layer_names):
                g = f[name]
                weight_names = [n.decode('utf8') for n in g.attrs['weight_names']]
                weight_values = [g[weight_name] for weight_name in weight_names]

                for layer in index.get(name, []):
                    symbolic_weights = layer.weights
                    if len(weight_values) != len(symbolic_weights):
                        raise ValueError('Layer #' + str(k) +
                                         ' (named "' + layer.name +
                                         '") expects ' +
                                         str(len(symbolic_weights)) +
                                         ' weight(s), but the saved weights' +
                                         ' have ' + str(len(weight_values)) +
                                         ' element(s).')
                    # Set values.
                    for i in range(len(weight_values)):
                        weight_value_tuples.append((symbolic_weights[i],
                                                    weight_values[i]))
            K.batch_set_value(weight_value_tuples)

    def _updated_config(self):
        '''Shared between different serialization methods.'''
        from keras import __version__ as keras_version

        config = self.get_config()
        model_config = {
            'class_name': self.__class__.__name__,
            'config': config,
            'keras_version': keras_version
        }
        return model_config

    def to_json(self, **kwargs):
        '''Returns a JSON string containing the network configuration.

        To load a network from a JSON save file, use
        `keras.models.model_from_json(json_string, custom_objects={})`.
        '''
        import json

        def get_json_type(obj):
            # If obj is any numpy type
            if type(obj).__module__ == np.__name__:
                return obj.item()

            # If obj is a python 'type'
            if type(obj).__name__ == type.__name__:
                return obj.__name__

            raise TypeError('Not JSON Serializable:', obj)

        model_config = self._updated_config()
        return json.dumps(model_config, default=get_json_type, **kwargs)

    def to_yaml(self, **kwargs):
        '''Returns a yaml string containing the network configuration.

        To load a network from a yaml save file, use
        `keras.models.model_from_yaml(yaml_string, custom_objects={})`.

        `custom_objects` should be a dictionary mapping
        the names of custom losses / layers / etc to the corresponding
        functions / classes.
        '''
        import yaml
        return yaml.dump(self._updated_config(), **kwargs)

    def summary(self, line_length=100, positions=[.33, .55, .67, 1.]):
        from keras.utils.layer_utils import print_summary

        if hasattr(self, 'flattened_layers'):
            # Support for legacy Sequential/Merge behavior.
            flattened_layers = self.flattened_layers
        else:
            flattened_layers = self.layers
        print_summary(flattened_layers,
                      getattr(self, 'container_nodes', None),
                      line_length=line_length,
                      positions=positions)


def get_source_inputs(tensor, layer=None, node_index=None):
    '''Returns the list of input tensors
    necessary to compute `tensor`.

    Output will always be a list of tensors
    (potentially with 1 element).

    # Arguments
        tensor: The tensor to start from.
        layer: Origin layer of the tensor. Will be
            determined via tensor._keras_history if not provided.
        node_index: Origin node index of the tensor.
    '''
    if not hasattr(tensor, '_keras_history'):
        return tensor

    if layer is None or node_index:
        layer, node_index, _ = tensor._keras_history
    if not layer.inbound_nodes:
        return [tensor]
    else:
        node = layer.inbound_nodes[node_index]
        if not node.inbound_layers:
            # Reached an Input layer, stop recursion.
            return node.input_tensors
        else:
            source_tensors = []
            for i in range(len(node.inbound_layers)):
                x = node.input_tensors[i]
                layer = node.inbound_layers[i]
                node_index = node.node_indices[i]
                previous_sources = get_source_inputs(x,
                                                     layer,
                                                     node_index)
                # Avoid input redundancy.
                for x in previous_sources:
                    if x not in source_tensors:
                        source_tensors.append(x)
            return source_tensors