Python keras.backend.backend() Examples

def set_img_format():
    try:
        if K.backend() == 'theano':
            K.set_image_data_format('channels_first')
        else:
            K.set_image_data_format('channels_last')
    except AttributeError:
        if K._BACKEND == 'theano':
            K.set_image_dim_ordering('th')
        else:
            K.set_image_dim_ordering('tf') 

def classifier(base_layers, input_rois, num_rois, nb_classes=21, trainable=False):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround

    if K.backend() == 'tensorflow':
        pooling_regions = 14
        # Changed the input shape to 1088 from 1024 because of nn_base's output being 1088. Not sure if this is correct
        input_shape = (num_rois, 14, 14, 1088)
    elif K.backend() == 'theano':
        pooling_regions = 7
        input_shape = (num_rois, 1024, 7, 7)

    out_roi_pool = RoiPoolingConv(pooling_regions, num_rois)([base_layers, input_rois])
    out = classifier_layers(out_roi_pool, input_shape=input_shape, trainable=True)

    out = TimeDistributed(Flatten())(out)

    out_class = TimeDistributed(Dense(nb_classes, activation='softmax', kernel_initializer='zero'), name='dense_class_{}'.format(nb_classes))(out)
    # note: no regression target for bg class
    out_regr = TimeDistributed(Dense(4 * (nb_classes-1), activation='linear', kernel_initializer='zero'), name='dense_regress_{}'.format(nb_classes))(out)
    return [out_class, out_regr] 

def test_save_load_all(self):
        for backend in self.list_backends():
            try:
                set_keras_backend(backend)
            except ModuleNotFoundError:
                continue
            K.set_learning_phase(0)  # test
            for use_attn_mask in [True, False]:
                model = self.create_small_model(use_attn_mask)
                path = '/tmp/{}.model'.format(uuid.uuid4())
                try:
                    model.save(path)
                    new_model = keras.models.load_model(path, custom_objects={'MultiHeadAttention': MultiHeadAttention,
                                                                              'LayerNormalization': LayerNormalization,
                                                                              'Gelu': Gelu})
                    TestTransformer.compare_two_models(model, new_model)
                except Exception as e:
                    raise e
                finally:
                    if os.path.exists(path):
                        os.remove(path) 

def test_save_load_weights(self):
        for backend in self.list_backends():
            try:
                set_keras_backend(backend)
            except ModuleNotFoundError:
                continue
            K.set_learning_phase(0)  # test
            for use_attn_mask in [True, False]:
                model = self.create_small_model(use_attn_mask)
                path = '/tmp/{}.model'.format(uuid.uuid4())
                try:
                    model.save_weights(path)
                    model.load_weights(path)
                except Exception as e:
                    raise e
                finally:
                    if os.path.exists(path):
                        os.remove(path) 

def swish(x,
          name="swish"):
    """
    Swish activation function from 'Searching for Activation Functions,' https://arxiv.org/abs/1710.05941.

    Parameters:
    ----------
    x : keras.backend tensor/variable/symbol
        Input tensor/variable/symbol.
    name : str, default 'swish'
        Block name.

    Returns
    -------
    keras.backend tensor/variable/symbol
        Resulted tensor/variable/symbol.
    """
    w = nn.Activation("sigmoid", name=name + "/sigmoid")(x)
    x = nn.multiply([x, w], name=name + "/mul")
    return x 

def classifier(base_layers, input_rois, num_rois, nb_classes=21, trainable=False):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround

    if K.backend() == 'tensorflow':
        pooling_regions = 14
        input_shape = (num_rois, 14, 14, 1024)
    elif K.backend() == 'theano':
        pooling_regions = 7
        input_shape = (num_rois, 1024, 7, 7)

    out_roi_pool = RoiPoolingConv(pooling_regions, num_rois)([base_layers, input_rois])
    out = classifier_layers(out_roi_pool, input_shape=input_shape, trainable=True)

    out = TimeDistributed(Flatten())(out)

    out_class = TimeDistributed(Dense(nb_classes, activation='softmax', kernel_initializer='zero'), name='dense_class_{}'.format(nb_classes))(out)
    # note: no regression target for bg class
    out_regr = TimeDistributed(Dense(4 * (nb_classes-1), activation='linear', kernel_initializer='zero'), name='dense_regress_{}'.format(nb_classes))(out)
    return [out_class, out_regr] 

def classifier(base_layers, input_rois, num_rois, nb_classes = 21, trainable=False):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround

    if K.backend() == 'tensorflow':
        pooling_regions = 14
        input_shape = (num_rois,14,14,1024)
    elif K.backend() == 'theano':
        pooling_regions = 7
        input_shape = (num_rois,1024,7,7)

    out_roi_pool = RoiPoolingConv(pooling_regions, num_rois)([base_layers, input_rois])
    out = classifier_layers(out_roi_pool, input_shape=input_shape, trainable=True)

    out = TimeDistributed(Flatten())(out)

    out_class = TimeDistributed(Dense(nb_classes, activation='softmax', kernel_initializer='zero'), name='dense_class_{}'.format(nb_classes))(out)
    # note: no regression target for bg class
    out_regr = TimeDistributed(Dense(4 * (nb_classes-1), activation='linear', kernel_initializer='zero'), name='dense_regress_{}'.format(nb_classes))(out)
    return [out_class, out_regr] 

def classifier(base_layers, input_rois, num_rois, nb_classes = 21, trainable=False):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround

    if K.backend() == 'tensorflow':
        pooling_regions = 7
        input_shape = (num_rois,7,7,512)
    elif K.backend() == 'theano':
        pooling_regions = 7
        input_shape = (num_rois,512,7,7)

    out_roi_pool = RoiPoolingConv(pooling_regions, num_rois)([base_layers, input_rois])

    out = TimeDistributed(Flatten(name='flatten'))(out_roi_pool)
    out = TimeDistributed(Dense(4096, activation='relu', name='fc1'))(out)
    out = TimeDistributed(Dropout(0.5))(out)
    out = TimeDistributed(Dense(4096, activation='relu', name='fc2'))(out)
    out = TimeDistributed(Dropout(0.5))(out)

    out_class = TimeDistributed(Dense(nb_classes, activation='softmax', kernel_initializer='zero'), name='dense_class_{}'.format(nb_classes))(out)
    # note: no regression target for bg class
    out_regr = TimeDistributed(Dense(4 * (nb_classes-1), activation='linear', kernel_initializer='zero'), name='dense_regress_{}'.format(nb_classes))(out)

    return [out_class, out_regr] 

def classifier(base_layers, input_rois, num_rois, nb_classes = 21, trainable=False):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround

    if K.backend() == 'tensorflow':
        pooling_regions = 7
        input_shape = (num_rois, 7, 7, 512)
    elif K.backend() == 'theano':
        pooling_regions = 7
        input_shape = (num_rois, 512, 7, 7)

    out_roi_pool = RoiPoolingConv(pooling_regions, num_rois)([base_layers, input_rois])

    out = TimeDistributed(Flatten(name='flatten'))(out_roi_pool)
    out = TimeDistributed(Dense(4096, activation='relu', name='fc1'))(out)
    out = TimeDistributed(Dense(4096, activation='relu', name='fc2'))(out)

    out_class = TimeDistributed(Dense(nb_classes, activation='softmax', kernel_initializer='zero'), name='dense_class_{}'.format(nb_classes))(out)
    # note: no regression target for bg class
    out_regr = TimeDistributed(Dense(4 * (nb_classes-1), activation='linear', kernel_initializer='zero'), name='dense_regress_{}'.format(nb_classes))(out)

    return [out_class, out_regr] 

def _handle_broken_model(self, model, error):
        del model

        n = self.genome_handler.n_classes
        loss = log_loss(np.concatenate(([1], np.zeros(n - 1))), np.ones(n) / n)
        accuracy = 1 / n
        gc.collect()

        if K.backend() == 'tensorflow':
            K.clear_session()
            tf.reset_default_graph()

        print('An error occurred and the model could not train:')
        print(error)
        print(('Model assigned poor score. Please ensure that your model'
               'constraints live within your computational resources.'))
        return loss, accuracy 

def contain_tf_gpu_mem_usage():
    """
    By default TensorFlow may try to reserve all available GPU memory
    making it impossible to train multiple networks at once.
    This function will disable such behaviour in TensorFlow.
    """
    from keras import backend
    if backend.backend() != 'tensorflow':
        return
    try:
        # noinspection PyPackageRequirements
        import tensorflow as tf
    except ImportError:
        pass
    else:
        from keras.backend.tensorflow_backend import set_session
        config = tf.ConfigProto()
        config.gpu_options.allow_growth = True  # dynamically grow the memory
        sess = tf.Session(config=config)
        set_session(sess) 

def test_save_load_weights(self):
        for backend in self.list_backends():
            try:
                set_keras_backend(backend)
            except ModuleNotFoundError:
                continue
            K.set_learning_phase(0)  # test
            for use_attn_mask in [True, False]:
                model = self.create_small_model(use_attn_mask)
                path = '/tmp/{}.model'.format(uuid.uuid4())
                try:
                    model.save_weights(path)
                    model.load_weights(path)
                except Exception as e:
                    raise e
                finally:
                    if os.path.exists(path):
                        os.remove(path) 

def test_save_load_all(self):
        for backend in self.list_backends():
            try:
                set_keras_backend(backend)
            except ModuleNotFoundError:
                continue
            K.set_learning_phase(0)  # test
            for use_attn_mask in [True, False]:
                model = self.create_small_model(use_attn_mask)
                path = '/tmp/{}.model'.format(uuid.uuid4())
                try:
                    model.save(path)
                    new_model = keras.models.load_model(path, custom_objects={'MultiHeadAttention': MultiHeadAttention,
                                                                              'LayerNormalization': LayerNormalization,
                                                                              'Gelu': Gelu})
                    TestTransformer.compare_two_models(model, new_model)
                except Exception as e:
                    raise e
                finally:
                    if os.path.exists(path):
                        os.remove(path) 

def classifier(base_layers, input_rois, num_rois, nb_classes = 21, trainable=True):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround

    if K.backend() == 'tensorflow':
        pooling_regions = 14
        input_shape = (num_rois,14,14,1024)
    elif K.backend() == 'theano':
        raise ValueError("Theano backend not supported")

    out_roi_pool = RoiPoolingConv(pooling_regions, num_rois,trainable=trainable)([base_layers, input_rois])
    out = classifier_layers(out_roi_pool, input_shape=input_shape, trainable=True)

    out = TimeDistributed(Flatten())(out)

    out_class = TimeDistributed(Dense(nb_classes, activation='softmax', kernel_initializer='zero',trainable=trainable), name='dense_class_{}'.format(nb_classes),trainable=trainable)(out)
    # note: no regression target for bg class
    out_regr = TimeDistributed(Dense(4 * (nb_classes-1), activation='linear', kernel_initializer='zero',trainable=trainable), name='dense_regress_{}'.format(nb_classes),trainable=trainable)(out)
    return [out_class, out_regr] 

def __init__(self, log_dir='./logs',
                 histogram_freq=0,
                 batch_size=32,
                 write_graph=True,
                 write_grads=False,
                 write_images=False,
                 embeddings_freq=0,
                 embeddings_layer_names=None,
                 embeddings_metadata=None):
        super(TensorBoard, self).__init__()
        if K.backend() != 'tensorflow':
            raise RuntimeError('TensorBoard callback only works '
                               'with the TensorFlow backend.')
        self.log_dir = log_dir
        self.histogram_freq = histogram_freq
        self.merged = None
        self.write_graph = write_graph
        self.write_grads = write_grads
        self.write_images = write_images
        self.embeddings_freq = embeddings_freq
        self.embeddings_layer_names = embeddings_layer_names
        self.embeddings_metadata = embeddings_metadata or {}
        self.batch_size = batch_size 

def clip_norm(g, c, n):
    if c > 0:
        if K.backend() == 'tensorflow':
            import tensorflow as tf
            import copy
            condition = n >= c
            then_expression = tf.scalar_mul(c / n, g)
            else_expression = g

            if hasattr(then_expression, 'get_shape'):
                g_shape = copy.copy(then_expression.get_shape())
            elif hasattr(then_expression, 'dense_shape'):
                g_shape = copy.copy(then_expression.dense_shape)
            if condition.dtype != tf.bool:
                condition = tf.cast(condition, 'bool')
            g = K.tensorflow_backend.control_flow_ops.cond(
                condition, lambda: then_expression, lambda: else_expression)
            if hasattr(then_expression, 'get_shape'):
                g.set_shape(g_shape)
            elif hasattr(then_expression, 'dense_shape'):
                g._dense_shape = g_shape
        else:
            g = K.switch(n >= c, g * c / n, g)
    return g 

def classifier(base_layers, input_rois, num_rois, nb_classes = 21, trainable=False):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround

    if K.backend() == 'tensorflow':
        pooling_regions = 14
        input_shape = (num_rois,14,14,1024)
    elif K.backend() == 'theano':
        pooling_regions = 7
        input_shape = (num_rois,1024,7,7)

    out_roi_pool = RoiPoolingConv(pooling_regions, num_rois)([base_layers, input_rois])
    out = classifier_layers(out_roi_pool, input_shape=input_shape, trainable=True)

    out = TimeDistributed(Flatten())(out)

    out_class = TimeDistributed(Dense(nb_classes, activation='softmax', kernel_initializer='zero'), name='dense_class_{}'.format(nb_classes))(out)
    # note: no regression target for bg class
    out_regr = TimeDistributed(Dense(4 * (nb_classes-1), activation='linear', kernel_initializer='zero'), name='dense_regress_{}'.format(nb_classes))(out)
    return [out_class, out_regr] 

def dot_product(x, kernel):
    """
    Wrapper for dot product operation, in order to be compatible with both
    Theano and Tensorflow
    Args:
        x (): input
        kernel (): weights
    Returns:
    """
    if K.backend() == 'tensorflow':
        return K.squeeze(K.dot(x, K.expand_dims(kernel)), axis=-1)
    else:
        return K.dot(x, kernel) 

def classifier_layers(x, input_shape, trainable=False):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround
    # (hence a smaller stride in the region that follows the ROI pool)
    if K.backend() == 'tensorflow':
        x = conv_block_td(x, 3, [512, 512, 2048], stage=5, block='a', input_shape=input_shape, strides=(2, 2), trainable=trainable)
    elif K.backend() == 'theano':
        x = conv_block_td(x, 3, [512, 512, 2048], stage=5, block='a', input_shape=input_shape, strides=(1, 1), trainable=trainable)

    x = identity_block_td(x, 3, [512, 512, 2048], stage=5, block='b', trainable=trainable)
    x = identity_block_td(x, 3, [512, 512, 2048], stage=5, block='c', trainable=trainable)
    x = TimeDistributed(AveragePooling2D((7, 7)), name='avg_pool')(x)

    return x 

def set_keras_backend(backend):
    global K
    if K.backend() != backend:
        os.environ['KERAS_BACKEND'] = backend
        reload(K)
        assert K.backend() == backend 

def __init__(self, output_shape, method="bilinear", *kwargs):
        """Initializes a new Resize2D layer.
        
        Arguments:
            output_shape {(int, int)} -- 2D tuple with the (height, width) of the new image
        
        Keyword Arguments:
            method {str} -- The method to use to resize the image. (default: {"bilinear"})
                            Possible values are: bilinear, nearest_neighbor, bicubic, area
        
        Raises:
            ValueError -- ValueError will be raised if arguments are invalid.
        """

        if not isinstance(output_shape, (list, tuple)) or len(output_shape) != 2:
            raise ValueError("`output_shape` must be a tuple or list of length 2.")

        if K.backend() != "tensorflow":
            raise ValueError("This layer is only supported using the tensorflow backend.")

        global tf
        import tensorflow as tf
        allowed_methods = {
            "bilinear": tf.image.ResizeMethod.BILINEAR,
            "nearest_neighbor": tf.image.ResizeMethod.NEAREST_NEIGHBOR,
            "bicubic": tf.image.ResizeMethod.BICUBIC,
            "area": tf.image.ResizeMethod.AREA
        }

        if method not in allowed_methods:
            raise ValueError("`mode´ has to be one of the values: {0}".format(allowed_methods.keys()))


        self._kwargs = kwargs
        self._output_shape = output_shape
        self._method = allowed_methods[method]
        super().__init__(*kwargs) 

def dot_product(x, kernel):
    """
    Wrapper for dot product operation, in order to be compatible with both
    Theano and Tensorflow
    Args:
        x (): input
        kernel (): weights
    Returns:
    """
    if K.backend() == 'tensorflow':
        # todo: check that this is correct
        return K.squeeze(K.dot(x, K.expand_dims(kernel)), axis=-1)
    else:
        return K.dot(x, kernel) 

def test_switchnorm_mode_twice():
    # This is a regression test for issue #4881 with the old
    # switch normalization functions in the Theano backend.
    model = Sequential()
    model.add(SwitchNormalization(input_shape=(10, 5, 5), axis=1))
    model.add(SwitchNormalization(input_shape=(10, 5, 5), axis=1))
    model.compile(loss='mse', optimizer='sgd')

    x = np.random.normal(loc=5.0, scale=10.0, size=(20, 10, 5, 5))
    model.fit(x, x, epochs=1, verbose=0)
    model.predict(x) 

def test_basic_switchnorm():
    layer_test(SwitchNormalization,
               kwargs={'momentum': 0.9,
                       'epsilon': 0.1,
                       'gamma_regularizer': regularizers.l2(0.01),
                       'beta_regularizer': regularizers.l2(0.01)},
               input_shape=(3, 4, 2))
    layer_test(SwitchNormalization,
               kwargs={'momentum': 0.9,
                       'epsilon': 0.1,
                       'axis': 1},
               input_shape=(3, 4, 2))
    layer_test(SwitchNormalization,
               kwargs={'gamma_initializer': 'ones',
                       'beta_initializer': 'ones',
                       'moving_mean_initializer': 'zeros',
                       'moving_variance_initializer': 'ones'},
               input_shape=(3, 4, 2, 4))
    if K.backend() != 'theano':
        layer_test(SwitchNormalization,
                   kwargs={'momentum': 0.9,
                           'epsilon': 0.1,
                           'axis': 1,
                           'scale': False,
                           'center': False},
                   input_shape=(3, 4, 2, 4)) 

def dot_product(x, kernel):
    """
    Wrapper for dot product operation, in order to be compatible with both
    Theano and Tensorflow
    Args:
        x (): input
        kernel (): weights
    Returns:
    """
    if K.backend() == 'tensorflow':
        return K.squeeze(K.dot(x, K.expand_dims(kernel)), axis=-1)
    else:
        return K.dot(x, kernel) 

def classifier_layers(x, input_shape, trainable=False):

    # compile times on theano tend to be very high, so we use smaller ROI pooling regions to workaround
    # (hence a smaller stride in the region that follows the ROI pool)
    if K.backend() == 'tensorflow':
        x = conv_block_td(x, 3, [512, 512, 2048], stage=5, block='a', input_shape=input_shape, strides=(2, 2), trainable=trainable)
    elif K.backend() == 'theano':
        x = conv_block_td(x, 3, [512, 512, 2048], stage=5, block='a', input_shape=input_shape, strides=(1, 1), trainable=trainable)

    x = identity_block_td(x, 3, [512, 512, 2048], stage=5, block='b', trainable=trainable)
    x = identity_block_td(x, 3, [512, 512, 2048], stage=5, block='c', trainable=trainable)
    x = TimeDistributed(AveragePooling2D((7, 7)), name='avg_pool')(x)

    return x 

def dot_product(x, kernel):
    """
    Wrapper for dot product operation, in order to be compatible with both
    Theano and Tensorflow
    Args:
        x (): input
        kernel (): weights
    Returns:
    """
    if K.backend() == 'tensorflow':
        return K.squeeze(K.dot(x, K.expand_dims(kernel)), axis=-1)
    else:
        return K.dot(x, kernel) 

def get_weight_path():
    if K.image_dim_ordering() == 'th':
        print('pretrained weights not available for VGG with theano backend')
        return
    else:
        return 'vgg16_weights_tf_dim_ordering_tf_kernels.h5' 

def dot_product(x, kernel):
    if K.backend() == 'tensorflow':
        return K.squeeze(K.dot(x, K.expand_dims(kernel)), axis=-1)
    else:
        return K.dot(x, kernel) 

def dot_product(x, kernel):
    """
    Wrapper for dot product operation, in order to be compatible with both
    Theano and Tensorflow
    Args:
        x (): input
        kernel (): weights
    Returns:
    """
    if K.backend() == 'tensorflow':
        return K.squeeze(K.dot(x, K.expand_dims(kernel)), axis=-1)
    else:
        return K.dot(x, kernel)