Keras backends

What is a "backend"?

Keras is a model-level library, providing high-level building blocks for developing deep learning models. It does not handle itself low-level operations such as tensor products, convolutions and so on. Instead, it relies on a specialized, well-optimized tensor manipulation library to do so, serving as the "backend engine" of Keras. Rather than picking one single tensor library and making the implementation of Keras tied to that library, Keras handles the problem in a modular way, and several different backend engines can be plugged seamlessly into Keras.

At this time, Keras has two backend implementations available: the TensorFlow backend and the Theano backend.

  • TensorFlow is an open-source symbolic tensor manipulation framework developed by Google, Inc.
  • Theano is an open-source symbolic tensor manipulation framework developed by LISA/MILA Lab at Université de Montréal.

In the future, we are likely to add more backend options. Go ask Microsoft about how their CNTK backend project is doing.

Switching from one backend to another

If you have run Keras at least once, you will find the Keras configuration file at:

~/.keras/keras.json

If it isn't there, you can create it.

The default configuration file looks like this:

{
    "image_data_format": "channels_last",
    "epsilon": 1e-07,
    "floatx": "float32",
    "backend": "tensorflow"
}

Simply change the field backend to either "theano" or "tensorflow", and Keras will use the new configuration next time you run any Keras code.

You can also define the environment variable KERAS_BACKEND and this will override what is defined in your config file :

KERAS_BACKEND=tensorflow python -c "from keras import backend"
Using TensorFlow backend.

keras.json details

{
    "image_data_format": "channels_last",
    "epsilon": 1e-07,
    "floatx": "float32",
    "backend": "tensorflow"
}

You can change these settings by editing ~/.keras/keras.json.

  • image_data_format: string, either "channels_last" or "channels_first". It specifies which data format convention Keras will follow. (keras.backend.image_data_format() returns it.)
  • For 2D data (e.g. image), "channels_last" assumes (rows, cols, channels) while "channels_first" assumes (channels, rows, cols).
  • For 3D data, "channels_last" assumes (conv_dim1, conv_dim2, conv_dim3, channels) while "channels_first" assumes (channels, conv_dim1, conv_dim2, conv_dim3).
  • epsilon: float, a numeric fuzzing constant used to avoid dividing by zero in some operations.
  • floatx: string, "float16", "float32", or "float64". Default float precision.
  • backend: string, "tensorflow" or "theano".

Using the abstract Keras backend to write new code

If you want the Keras modules you write to be compatible with both Theano and TensorFlow, you have to write them via the abstract Keras backend API. Here's an intro.

You can import the backend module via:

from keras import backend as K

The code below instantiates an input placeholder. It's equivalent to tf.placeholder() or T.matrix(), T.tensor3(), etc.

input = K.placeholder(shape=(2, 4, 5))
# also works:
input = K.placeholder(shape=(None, 4, 5))
# also works:
input = K.placeholder(ndim=3)

The code below instantiates a shared variable. It's equivalent to tf.variable() or theano.shared().

val = np.random.random((3, 4, 5))
var = K.variable(value=val)

# all-zeros variable:
var = K.zeros(shape=(3, 4, 5))
# all-ones:
var = K.ones(shape=(3, 4, 5))

Most tensor operations you will need can be done as you would in TensorFlow or Theano:

a = b + c * K.abs(d)
c = K.dot(a, K.transpose(b))
a = K.sum(b, axis=2)
a = K.softmax(b)
a = concatenate([b, c], axis=-1)
# etc...

Backend functions

epsilon

epsilon()

Returns the value of the fuzz factor used in numeric expressions.

Returns

A float.

Example

>>> keras.backend.epsilon()
1e-08

set_epsilon

set_epsilon(e)

Sets the value of the fuzz factor used in numeric expressions.

Arguments

  • e: float. New value of epsilon.

Example

>>> from keras import backend as K
>>> K.epsilon()
1e-08
>>> K.set_epsilon(1e-05)
>>> K.epsilon()
1e-05

floatx

floatx()

Returns the default float type, as a string (e.g. 'float16', 'float32', 'float64').

Returns

String, the current default float type.

Example

>>> keras.backend.floatx()
'float32'

set_floatx

set_floatx(floatx)

Sets the default float type.

Arguments

  • String: 'float16', 'float32', or 'float64'.

Example

>>> from keras import backend as K
>>> K.floatx()
'float32'
>>> K.set_floatx('float16')
>>> K.floatx()
'float16'

cast_to_floatx

cast_to_floatx(x)

Cast a Numpy array to the default Keras float type.

Arguments

  • x: Numpy array.

Returns

The same Numpy array, cast to its new type.

Example

>>> from keras import backend as K
>>> K.floatx()
'float32'
>>> arr = numpy.array([1.0, 2.0], dtype='float64')
>>> arr.dtype
dtype('float64')
>>> new_arr = K.cast_to_floatx(arr)
>>> new_arr
array([ 1.,  2.], dtype=float32)
>>> new_arr.dtype
dtype('float32')

image_data_format

image_data_format()

Returns the default image data format convention ('channels_first' or 'channels_last').

Returns

A string, either 'channels_first' or 'channels_last'

Example

>>> keras.backend.image_data_format()
'channels_first'

set_image_data_format

set_image_data_format(data_format)

Sets the value of the data format convention.

Arguments

  • data_format: string. 'channels_first' or 'channels_last'.

Example

>>> from keras import backend as K
>>> K.image_data_format()
'channels_first'
>>> K.set_image_data_format('channels_last')
>>> K.image_data_format()
'channels_last'

is_keras_tensor

is_keras_tensor(x)

Returns whether x is a Keras tensor.

Arguments

  • x: a potential tensor.

Returns

A boolean: whether the argument is a Keras tensor.

Examples

>>> from keras import backend as K
>>> np_var = numpy.array([1, 2])
>>> K.is_keras_tensor(np_var)
False
>>> keras_var = K.variable(np_var)
>>> K.is_keras_tensor(keras_var)  # A variable is not a Tensor.
False
>>> keras_placeholder = K.placeholder(shape=(2, 4, 5))
>>> K.is_keras_tensor(keras_placeholder)  # A placeholder is a Tensor.
True

set_image_dim_ordering

set_image_dim_ordering(dim_ordering)

Sets the value of the image data format.

Arguments

  • data_format: string. 'channels_first' or 'channels_last'.

Example

>>> from keras import backend as K
>>> K.image_data_format()
'channels_first'
>>> K.set_image_data_format('channels_last')
>>> K.image_data_format()
'channels_last'

image_dim_ordering

image_dim_ordering()

Legacy getter for data format.

learning_phase

learning_phase()

set_learning_phase

set_learning_phase(value)

get_uid

get_uid(prefix='')

Provides a unique UID given a string prefix.

Arguments

  • prefix: string.

Returns

An integer.

Example

>>> keras.backend.get_uid('dense')
>>> 1
>>> keras.backend.get_uid('dense')
>>> 2

reset_uids

reset_uids()

is_sparse

is_sparse(tensor)

to_dense

to_dense(tensor)

name_scope

name_scope()

variable

variable(value, dtype=None, name=None)

Instantiates a variable and returns it.

Arguments

  • value: Numpy array, initial value of the tensor.
  • dtype: Tensor type.
  • name: Optional name string for the tensor.

Returns

A variable instance (with Keras metadata included).

constant

constant(value, dtype=None, shape=None, name=None)

placeholder

placeholder(shape=None, ndim=None, dtype=None, sparse=False, name=None)

Instantiate an input data placeholder variable.

shape

shape(x)

Returns the shape of a tensor.

  • Warning: type returned will be different for Theano backend (Theano tensor type) and TF backend (TF TensorShape).

int_shape

int_shape(x)

Returns the shape of a Keras tensor or a Keras variable as a tuple of integers or None entries.

Arguments

  • x: Tensor or variable.

Returns

A tuple of integers (or None entries).

ndim

ndim(x)

dtype

dtype(x)

eval

eval(x)

Returns the value of a tensor.

zeros

zeros(shape, dtype=None, name=None)

Instantiates an all-zeros variable.

ones

ones(shape, dtype=None, name=None)

Instantiates an all-ones variable.

eye

eye(size, dtype=None, name=None)

Instantiates an identity matrix.

ones_like

ones_like(x, dtype=None, name=None)

zeros_like

zeros_like(x, dtype=None, name=None)

random_uniform_variable

random_uniform_variable(shape, low, high, dtype=None, name=None)

random_normal_variable

random_normal_variable(shape, mean, scale, dtype=None, name=None)

count_params

count_params(x)

Returns the number of scalars in a tensor.

  • Return: numpy integer.

cast

cast(x, dtype)

update

update(x, new_x)

update_add

update_add(x, increment)

update_sub

update_sub(x, decrement)

moving_average_update

moving_average_update(variable, value, momentum)

dot

dot(x, y)

batch_dot

batch_dot(x, y, axes=None)

Batchwise dot product.

batch_dot results in a tensor with less dimensions than the input. If the number of dimensions is reduced to 1, we use expand_dims to make sure that ndim is at least 2.

Arguments

x, y: tensors with ndim >= 2 - axes: list (or single) int with target dimensions

Returns

A tensor with shape equal to the concatenation of x's shape (less the dimension that was summed over) and y's shape (less the batch dimension and the dimension that was summed over). If the final rank is 1, we reshape it to (batch_size, 1).

Examples

Assume x = [[1, 2], [3, 4]] and y = [[5, 6], [7, 8]] batch_dot(x, y, axes=1) = [[17, 53]] which is the main diagonal of x.dot(y.T), although we never have to calculate the off-diagonal elements.

Shape inference: Let x's shape be (100, 20) and y's shape be (100, 30, 20). If dot_axes is (1, 2), to find the output shape of resultant tensor, loop through each dimension in x's shape and y's shape: x.shape[0] : 100 : append to output shape x.shape[1] : 20 : do not append to output shape, dimension 1 of x has been summed over. (dot_axes[0] = 1) y.shape[0] : 100 : do not append to output shape, always ignore first dimension of y y.shape[1] : 30 : append to output shape y.shape[2] : 20 : do not append to output shape, dimension 2 of y has been summed over. (dot_axes[1] = 2)

output_shape = (100, 30)

transpose

transpose(x)

gather

gather(reference, indices)

reference: a tensor. - indices: an int tensor of indices.

  • Return: a tensor of same type as reference.

max

max(x, axis=None, keepdims=False)

min

min(x, axis=None, keepdims=False)

sum

sum(x, axis=None, keepdims=False)

Sum of the values in a tensor, alongside the specified axis.

prod

prod(x, axis=None, keepdims=False)

Multiply the values in a tensor, alongside the specified axis.

mean

mean(x, axis=None, keepdims=False)

Mean of a tensor, alongside the specified axis.

std

std(x, axis=None, keepdims=False)

var

var(x, axis=None, keepdims=False)

any

any(x, axis=None, keepdims=False)

Bitwise reduction (logical OR).

all

all(x, axis=None, keepdims=False)

Bitwise reduction (logical AND).

argmax

argmax(x, axis=-1)

argmin

argmin(x, axis=-1)

square

square(x)

abs

abs(x)

sqrt

sqrt(x)

exp

exp(x)

log

log(x)

round

round(x)

sign

sign(x)

pow

pow(x, a)

clip

clip(x, min_value, max_value)

equal

equal(x, y)

not_equal

not_equal(x, y)

greater

greater(x, y)

greater_equal

greater_equal(x, y)

less

less(x, y)

less_equal

less_equal(x, y)

maximum

maximum(x, y)

minimum

minimum(x, y)

sin

sin(x)

cos

cos(x)

normalize_batch_in_training

normalize_batch_in_training(x, gamma, beta, reduction_axes, epsilon=0.001)

Computes mean and std for batch then apply batch_normalization on batch.

batch_normalization

batch_normalization(x, mean, var, beta, gamma, epsilon=0.001)

Apply batch normalization on x given mean, var, beta and gamma.

concatenate

concatenate(tensors, axis=-1)

reshape

reshape(x, shape)

permute_dimensions

permute_dimensions(x, pattern)

Transpose dimensions.

pattern should be a tuple or list of dimension indices, e.g. [0, 2, 1].

repeat_elements

repeat_elements(x, rep, axis)

Repeat the elements of a tensor along an axis, like np.repeat.

If x has shape (s1, s2, s3) and axis=1, the output will have shape (s1, s2 * rep, s3).

resize_images

resize_images(X, height_factor, width_factor, data_format)

Resize the images contained in a 4D tensor of shape - [batch, channels, height, width] (for 'channels_first' data_format) - [batch, height, width, channels] (for 'channels_last' data_format) by a factor of (height_factor, width_factor). Both factors should be positive integers.

resize_volumes

resize_volumes(X, depth_factor, height_factor, width_factor, data_format)

Resize the volume contained in a 5D tensor of shape - [batch, channels, depth, height, width] (for 'channels_first' data_format) - [batch, depth, height, width, channels] (for 'channels_last' data_format) by a factor of (depth_factor, height_factor, width_factor). Both factors should be positive integers.

repeat

repeat(x, n)

Repeat a 2D tensor.

If x has shape (samples, dim) and n=2, the output will have shape (samples, 2, dim).

arange

arange(start, stop=None, step=1, dtype='int32')

Creates a 1-D tensor containing a sequence of integers.

The function arguments use the same convention as Theano's arange: if only one argument is provided, it is in fact the "stop" argument.

The default type of the returned tensor is 'int32' to match TensorFlow's default.

tile

tile(x, n)

flatten

flatten(x)

batch_flatten

batch_flatten(x)

Turn a n-D tensor into a 2D tensor where the first dimension is conserved.

expand_dims

expand_dims(x, axis=-1)

Add a 1-sized dimension at index "dim".

squeeze

squeeze(x, axis)

Remove a 1-dimension from the tensor at index "axis".

temporal_padding

temporal_padding(x, padding=(1, 1))

Pad the middle dimension of a 3D tensor with "padding" zeros left and right.

Apologies for the inane API, but Theano makes this really hard.

spatial_2d_padding

spatial_2d_padding(x, padding=((1, 1), (1, 1)), data_format=None)

Pad the 2nd and 3rd dimensions of a 4D tensor with "padding[0]" and "padding[1]" (resp.) zeros left and right.

spatial_3d_padding

spatial_3d_padding(x, padding=((1, 1), (1, 1), (1, 1)), data_format=None)

Pad the 2nd, 3rd and 4th dimensions of a 5D tensor with "padding[0]", "padding[1]" and "padding[2]" (resp.) zeros left and right.

stack

stack(x, axis=0)

one_hot

one_hot(indices, num_classes)

Input: nD integer tensor of shape (batch_size, dim1, dim2, ... dim(n-1)) - Output: (n + 1)D one hot representation of the input with shape (batch_size, dim1, dim2, ... dim(n-1), num_classes)

reverse

reverse(x, axes)

Reverse a tensor along the the specified axes

pattern_broadcast

pattern_broadcast(x, broatcastable)

get_value

get_value(x)

batch_get_value

batch_get_value(xs)

Returns the value of more than one tensor variable, as a list of Numpy arrays.

set_value

set_value(x, value)

batch_set_value

batch_set_value(tuples)

get_variable_shape

get_variable_shape(x)

print_tensor(x, message='')

Print the message and the tensor when evaluated and return the same tensor.

function

function(inputs, outputs, updates=[])

gradients

gradients(loss, variables)

stop_gradient

stop_gradient(variables)

Returns variables but with zero gradient with respect to every other variables.

rnn

rnn(step_function, inputs, initial_states, go_backwards=False, mask=None, constants=None, unroll=False, input_length=None)

Iterates over the time dimension of a tensor.

Arguments

  • inputs: tensor of temporal data of shape (samples, time, ...) (at least 3D).
  • step_function:
  • Parameters:
    • input: tensor with shape (samples, ...) (no time dimension), representing input for the batch of samples at a certain time step.
    • states: list of tensors.
  • Returns:
    • output: tensor with shape (samples, ...) (no time dimension),
    • new_states: list of tensors, same length and shapes as 'states'.
  • initial_states: tensor with shape (samples, ...) (no time dimension), containing the initial values for the states used in the step function.
  • go_backwards: boolean. If True, do the iteration over the time dimension in reverse order.
  • mask: binary tensor with shape (samples, time), with a zero for every element that is masked.
  • constants: a list of constant values passed at each step.
  • unroll: whether to unroll the RNN or to use a symbolic loop (while_loop or scan depending on backend).
  • input_length: must be specified if using unroll.

Returns

A tuple (last_output, outputs, new_states). - last_output: the latest output of the rnn, of shape (samples, ...) - outputs: tensor with shape (samples, time, ...) where each entry outputs[s, t] is the output of the step function at time t for sample s. - new_states: list of tensors, latest states returned by the step function, of shape (samples, ...).

switch

switch(condition, then_expression, else_expression)

condition: scalar tensor.

in_train_phase

in_train_phase(x, alt, training=None)

Selects x in train phase, and alt otherwise.

Note that alt should have the same shape as x.

Returns

Either x or alt based on the training flag. the training flag defaults to K.learning_phase().

in_test_phase

in_test_phase(x, alt, training=None)

Selects x in test phase, and alt otherwise. Note that alt should have the same shape as x.

Returns

Either x or alt based on K.learning_phase.

elu

elu(x, alpha=1.0)

Exponential linear unit

Arguments

  • x: Tensor to compute the activation function for.
  • alpha: scalar

relu

relu(x, alpha=0.0, max_value=None)

softmax

softmax(x)

softplus

softplus(x)

softsign

softsign(x)

categorical_crossentropy

categorical_crossentropy(output, target, from_logits=False)

sparse_categorical_crossentropy

sparse_categorical_crossentropy(output, target, from_logits=False)

binary_crossentropy

binary_crossentropy(output, target, from_logits=False)

sigmoid

sigmoid(x)

hard_sigmoid

hard_sigmoid(x)

tanh

tanh(x)

dropout

dropout(x, level, noise_shape=None, seed=None)

Sets entries in x to zero at random, while scaling the entire tensor.

Arguments

  • x: tensor
  • level: fraction of the entries in the tensor that will be set to 0.
  • noise_shape: shape for randomly generated keep/drop flags, must be broadcastable to the shape of x
  • seed: random seed to ensure determinism.

l2_normalize

l2_normalize(x, axis)

in_top_k

in_top_k(predictions, targets, k)

Returns whether the targets are in the top k predictions

Arguments

  • predictions: A tensor of shape batch_size x classess and type float32.
  • targets: A tensor of shape batch_size and type int32 or int64.
  • k: An int, number of top elements to consider.

Returns

A tensor of shape batch_size and type int. output_i is 1 if targets_i is within top-k values of predictions_i

conv1d

conv1d(x, kernel, strides=1, padding='valid', data_format=None, dilation_rate=1)

1D convolution.

Arguments

  • kernel: kernel tensor.
  • strides: stride integer.
  • padding: string, "same", "causal" or "valid".
  • data_format: string, one of "channels_last", "channels_first"
  • dilation_rate: integer.

conv2d

conv2d(x, kernel, strides=(1, 1), padding='valid', data_format=None, dilation_rate=(1, 1))

2D convolution.

Arguments

  • kernel: kernel tensor.
  • strides: strides tuple.
  • padding: string, "same" or "valid".
  • data_format: "channels_last" or "channels_first". Whether to use Theano or TensorFlow data format in inputs/kernels/ouputs.

conv2d_transpose

conv2d_transpose(x, kernel, output_shape, strides=(1, 1), padding='valid', data_format=None)

2D deconvolution (transposed convolution).

Arguments

  • kernel: kernel tensor.
  • output_shape: desired dimensions of output.
  • strides: strides tuple.
  • padding: string, "same" or "valid".
  • data_format: "channels_last" or "channels_first". Whether to use Theano or TensorFlow data format in inputs/kernels/ouputs.

separable_conv2d

separable_conv2d(x, depthwise_kernel, pointwise_kernel, strides=(1, 1), padding='valid', data_format=None, dilation_rate=(1, 1))

conv3d

conv3d(x, kernel, strides=(1, 1, 1), padding='valid', data_format=None, dilation_rate=(1, 1, 1))

3D convolution.

Arguments

  • kernel: kernel tensor.
  • strides: strides tuple.
  • padding: string, "same" or "valid".
  • data_format: "channels_last" or "channels_first". Whether to use Theano or TensorFlow data format in inputs/kernels/ouputs.

pool2d

pool2d(x, pool_size, strides=(1, 1), padding='valid', data_format=None, pool_mode='max')

pool3d

pool3d(x, pool_size, strides=(1, 1, 1), padding='valid', data_format=None, pool_mode='max')

bias_add

bias_add(x, bias, data_format=None)

random_normal

random_normal(shape, mean=0.0, stddev=1.0, dtype=None, seed=None)

random_uniform

random_uniform(shape, minval=0.0, maxval=1.0, dtype=None, seed=None)

random_binomial

random_binomial(shape, p=0.0, dtype=None, seed=None)

truncated_normal

truncated_normal(shape, mean=0.0, stddev=1.0, dtype=None, seed=None)

ctc_interleave_blanks

ctc_interleave_blanks(Y)

ctc_create_skip_idxs

ctc_create_skip_idxs(Y)

ctc_update_log_p

ctc_update_log_p(skip_idxs, zeros, active, log_p_curr, log_p_prev)

ctc_path_probs

ctc_path_probs(predict, Y, alpha=0.0001)

ctc_cost

ctc_cost(predict, Y)

ctc_batch_cost

ctc_batch_cost(y_true, y_pred, input_length, label_length)

Runs CTC loss algorithm on each batch element.

Arguments

  • y_true: tensor (samples, max_string_length) containing the truth labels
  • y_pred: tensor (samples, time_steps, num_categories) containing the prediction, or output of the softmax
  • input_length: tensor (samples,1) containing the sequence length for each batch item in y_pred
  • label_length: tensor (samples,1) containing the sequence length for each batch item in y_true

Returns

Tensor with shape (samples,1) containing the CTC loss of each element

map_fn

map_fn(fn, elems, name=None)

Map the function fn over the elements elems and return the outputs.

Arguments

  • fn: Callable that will be called upon each element in elems
  • elems: tensor, at least 2 dimensional
  • name: A string name for the map node in the graph

Returns

Tensor with first dimension equal to the elems and second depending on fn

foldl

foldl(fn, elems, initializer=None, name=None)

Reduce elems using fn to combine them from left to right.

Arguments

  • fn: Callable that will be called upon each element in elems and an accumulator, for instance lambda acc, x: acc + x
  • elems: tensor
  • initializer: The first value used (elems[0] in case of None)
  • name: A string name for the foldl node in the graph

Returns

Same type and shape as initializer

foldr

foldr(fn, elems, initializer=None, name=None)

Reduce elems using fn to combine them from right to left.

Arguments

  • fn: Callable that will be called upon each element in elems and an accumulator, for instance lambda acc, x: acc + x
  • elems: tensor
  • initializer: The first value used (elems[-1] in case of None)
  • name: A string name for the foldr node in the graph

Returns

Same type and shape as initializer

backend

backend()

Publicly accessible method for determining the current backend.