Python variables.model_variable函数代码示例

def sequence_softmax(inputs, noutput, scope=None, name=None, linear_name=None):
  """Run a softmax layer over all the time steps of an input sequence.

  Args:
    inputs: (length, batch_size, depth) tensor
    noutput: output depth
    scope: optional scope name
    name: optional name for output tensor
    linear_name: name for linear (pre-softmax) output

  Returns:
    A tensor of size (length, batch_size, noutput).

  """
  length, _, ninputs = _shape(inputs)
  inputs_u = array_ops.unstack(inputs)
  output_u = []
  with variable_scope.variable_scope(scope, "SequenceSoftmax", [inputs]):
    initial_w = random_ops.truncated_normal([0 + ninputs, noutput], stddev=0.1)
    initial_b = constant_op.constant(0.1, shape=[noutput])
    w = variables.model_variable("weights", initializer=initial_w)
    b = variables.model_variable("biases", initializer=initial_b)
    for i in xrange(length):
      with variable_scope.variable_scope(scope, "SequenceSoftmaxStep",
                                         [inputs_u[i]]):
        # TODO(tmb) consider using slim.fully_connected(...,
        # activation_fn=tf.nn.softmax)
        linear = nn_ops.xw_plus_b(inputs_u[i], w, b, name=linear_name)
        output = nn_ops.softmax(linear)
        output_u += [output]
    outputs = array_ops.stack(output_u, name=name)
  return outputs

 def testGetLocalVariables(self):
   with self.test_session():
     with variable_scope.variable_scope('A'):
       _ = variables_lib2.model_variable('a', [5])
     with variable_scope.variable_scope('B'):
       _ = variables_lib2.model_variable('a', [5])
     self.assertEquals([], variables_lib2.get_local_variables('A'))
     self.assertEquals([], variables_lib2.get_local_variables('B'))

 def testGetModelVariables(self):
   with self.test_session():
     with variable_scope.variable_scope('A'):
       a = variables_lib2.model_variable('a', [5])
     with variable_scope.variable_scope('B'):
       b = variables_lib2.model_variable('a', [5])
     self.assertEquals([a], variables_lib2.get_model_variables('A'))
     self.assertEquals([b], variables_lib2.get_model_variables('B'))

  def testVariableWithVariableDeviceChooser(self):

    with ops.Graph().as_default():
      device_fn = variables_lib2.VariableDeviceChooser()
      with arg_scope([variables_lib2.model_variable], device=device_fn):
        a = variables_lib2.model_variable('a', [5])
        b = variables_lib2.model_variable('b', [20])
        self.assertDeviceEqual(a.device, 'cpu:0')
        self.assertEqual(a.initial_value.op.colocation_groups(),
                         a.op.colocation_groups())
        self.assertDeviceEqual(b.device, 'cpu:0')
        self.assertEqual(a.initial_value.op.colocation_groups(),
                         a.op.colocation_groups())

 def testNotInLocalVariables(self):
   with self.test_session():
     with variable_scope.variable_scope('A'):
       a = variables_lib2.model_variable('a', [5])
       self.assertTrue(a in variables_lib.global_variables())
       self.assertTrue(a in ops.get_collection(ops.GraphKeys.MODEL_VARIABLES))
       self.assertFalse(a in variables_lib.local_variables())

 def testNameAndShape(self):
   with self.test_session():
     with variable_scope.variable_scope('A'):
       a = variables_lib2.model_variable('a', [5])
       self.assertEquals(a.op.name, 'A/a')
       self.assertListEqual(a.get_shape().as_list(), [5])
       self.assertListEqual([a], variables_lib2.get_model_variables('A'))

def _model_variable_getter(getter,
                           name,
                           shape=None,
                           dtype=None,
                           initializer=None,
                           regularizer=None,
                           trainable=True,
                           collections=None,
                           caching_device=None,
                           partitioner=None,
                           rename=None,
                           use_resource=None,
                           **_):
  """Getter that uses model_variable for compatibility with core layers."""
  short_name = name.split('/')[-1]
  if rename and short_name in rename:
    name_components = name.split('/')
    name_components[-1] = rename[short_name]
    name = '/'.join(name_components)
  return variables.model_variable(
      name,
      shape=shape,
      dtype=dtype,
      initializer=initializer,
      regularizer=regularizer,
      collections=collections,
      trainable=trainable,
      caching_device=caching_device,
      partitioner=partitioner,
      custom_getter=getter,
      use_resource=use_resource)

def bow_encoder(ids,
                vocab_size,
                embed_dim,
                sparse_lookup=True,
                initializer=None,
                regularizer=None,
                trainable=True,
                scope=None,
                reuse=None):
  """Maps a sequence of symbols to a vector per example by averaging embeddings.

  Args:
    ids: `[batch_size, doc_length]` `Tensor` or `SparseTensor` of type
      `int32` or `int64` with symbol ids.
    vocab_size: Integer number of symbols in vocabulary.
    embed_dim: Integer number of dimensions for embedding matrix.
    sparse_lookup: `bool`, if `True`, converts ids to a `SparseTensor`
        and performs a sparse embedding lookup. This is usually faster,
        but not desirable if padding tokens should have an embedding. Empty rows
        are assigned a special embedding.
    initializer: An initializer for the embeddings, if `None` default for
        current scope is used.
    regularizer: Optional regularizer for the embeddings.
    trainable: If `True` also add variables to the graph collection
      `GraphKeys.TRAINABLE_VARIABLES` (see tf.Variable).
    scope: Optional string specifying the variable scope for the op, required
        if `reuse=True`.
    reuse: If `True`, variables inside the op will be reused.

  Returns:
    Encoding `Tensor` `[batch_size, embed_dim]` produced by
    averaging embeddings.

  Raises:
    ValueError: If `embed_dim` or `vocab_size` are not specified.
  """
  if not vocab_size or not embed_dim:
    raise ValueError('Must specify vocab size and embedding dimension')
  with variable_scope.variable_scope(
      scope, 'bow_encoder', [ids], reuse=reuse):
    embeddings = variables.model_variable(
        'embeddings', shape=[vocab_size, embed_dim],
        initializer=initializer, regularizer=regularizer,
        trainable=trainable)
    if sparse_lookup:
      if isinstance(ids, sparse_tensor.SparseTensor):
        sparse_ids = ids
      else:
        sparse_ids = sparse_ops.dense_to_sparse_tensor(ids)
      return contrib_embedding_ops.safe_embedding_lookup_sparse(
          [embeddings], sparse_ids, combiner='mean', default_id=0)
    else:
      if isinstance(ids, sparse_tensor.SparseTensor):
        raise TypeError('ids are expected to be dense Tensor, got: %s', ids)
      return math_ops.reduce_mean(
          embedding_ops.embedding_lookup(embeddings, ids),
          reduction_indices=1)

def l2_normalization(
        inputs,
        scaling=False,
        scale_initializer=init_ops.ones_initializer(),
        reuse=None,
        variables_collections=None,
        outputs_collections=None,
        trainable=True,
        scope=None):
    """Implement L2 normalization on every feature (i.e. spatial normalization).

    Should be extended in some near future to other dimensions, providing a more
    flexible normalization framework.

    inputs: a 4-D tensor with dimensions [batch_size, height, width, channels].
    scaling: whether or not to add a post scaling operation along the dimensions
      which have been normalized.
    scale_initializer: An initializer for the weights.
    reuse: whether or not the layer and its variables should be reused. To be
      able to reuse the layer scope must be given.
    variables_collections: optional list of collections for all the variables or
      a dictionary containing a different list of collection per variable.
    outputs_collections: collection to add the outputs.
    trainable: If `True` also add variables to the graph collection
      `GraphKeys.TRAINABLE_VARIABLES` (see tf.Variable).
    scope: Optional scope for `variable_scope`.
    Returns:
      A `Tensor` representing the output of the operation.
    """

    with variable_scope.variable_scope(
            scope, 'L2Normalization', [inputs], reuse=reuse) as sc:

        inputs_shape = inputs.get_shape()
        inputs_rank = inputs_shape.ndims
        params_shape = inputs_shape[-1:]
        dtype = inputs.dtype.base_dtype

        # Normalize along spatial dimensions.
        norm_dim = tf.range(1, inputs_rank-1)
        outputs = nn.l2_normalize(inputs, norm_dim, epsilon=1e-12)
        # Additional scaling.
        if scaling:
            scale_collections = utils.get_variable_collections(
                variables_collections, 'scale')
            scale = variables.model_variable('gamma',
                                             shape=params_shape,
                                             dtype=dtype,
                                             initializer=scale_initializer,
                                             collections=scale_collections,
                                             trainable=trainable)
            outputs = tf.multiply(outputs, scale)
        return utils.collect_named_outputs(outputs_collections,
                                           sc.original_name_scope, outputs)

  def testDeviceFn(self):

    class DevFn(object):

      def __init__(self):
        self.counter = -1

      def __call__(self, op):
        self.counter += 1
        return '/cpu:%d' % self.counter

    with ops.Graph().as_default():
      with arg_scope([variables_lib2.model_variable], device=DevFn()):
        a = variables_lib2.model_variable('a', [5])
        b = variables_lib2.model_variable('b', [20])
        self.assertDeviceEqual(a.device, '/cpu:0')
        self.assertEqual(a.initial_value.op.colocation_groups(),
                         a.op.colocation_groups())
        self.assertDeviceEqual(b.device, '/cpu:1')
        self.assertEqual(b.initial_value.op.colocation_groups(),
                         b.op.colocation_groups())

def embed_sequence(ids,
                   vocab_size=None,
                   embed_dim=None,
                   unique=False,
                   initializer=None,
                   regularizer=None,
                   trainable=True,
                   scope=None,
                   reuse=None):
  """Maps a sequence of symbols to a sequence of embeddings.

  Typical use case would be reusing embeddings between an encoder and decoder.

  Args:
    ids: `[batch_size, doc_length]` `Tensor` of type `int32` or `int64`
      with symbol ids.
    vocab_size: Integer number of symbols in vocabulary.
    embed_dim: Integer number of dimensions for embedding matrix.
    unique: If `True`, will first compute the unique set of indices, and then
         lookup each embedding once, repeating them in the output as needed.
    initializer: An initializer for the embeddings, if `None` default for
        current scope is used.
    regularizer: Optional regularizer for the embeddings.
    trainable: If `True` also add variables to the graph collection
      `GraphKeys.TRAINABLE_VARIABLES` (see `tf.Variable`).
    scope: Optional string specifying the variable scope for the op, required
        if `reuse=True`.
    reuse: If `True`, variables inside the op will be reused.

  Returns:
    `Tensor` of `[batch_size, doc_length, embed_dim]` with embedded sequences.

  Raises:
    ValueError: if `embed_dim` or `vocab_size` are not specified when 
      `reuse` is `None` or `False`.
  """
  if not (reuse or (vocab_size and embed_dim)):
    raise ValueError('Must specify vocab size and embedding dimension when not'
                     'reusing. Got vocab_size=%s and embed_dim=%s' % (
                         vocab_size, embed_dim))
  with variable_scope.variable_scope(
      scope, 'EmbedSequence', [ids], reuse=reuse):
    shape = [vocab_size, embed_dim]
    if reuse and vocab_size is None or embed_dim is None:
      shape = None
    embeddings = variables.model_variable(
        'embeddings', shape=shape,
        initializer=initializer, regularizer=regularizer,
        trainable=trainable)
    if unique:
      return contrib_embedding_ops.embedding_lookup_unique(embeddings, ids)
    return embedding_ops.embedding_lookup(embeddings, ids)

def bias_add(inputs,
             activation_fn=None,
             initializer=init_ops.zeros_initializer,
             regularizer=None,
             reuse=None,
             variables_collections=None,
             outputs_collections=None,
             trainable=True,
             scope=None):
  """Adds a bias to the inputs.

  Can be used as a normalizer function for conv2d and fully_connected.

  Args:
    inputs: a tensor of with at least rank 2 and value for the last dimension,
      e.g. `[batch_size, depth]`, `[None, None, None, depth]`.
    activation_fn: Optional activation function.
    initializer: An initializer for the bias, defaults to 0.
    regularizer: A regularizer like the result of
      `l1_regularizer` or `l2_regularizer`.
    reuse: whether or not the layer and its variables should be reused. To be
      able to reuse the layer scope must be given.
    variables_collections: optional collections for the variables.
    outputs_collections: collections to add the outputs.
    trainable: If `True` also add variables to the graph collection
      `GraphKeys.TRAINABLE_VARIABLES` (see tf.Variable).
    scope: Optional scope for variable_op_scope.

  Returns:
    a tensor representing the result of adding biases to the inputs.
  """
  with variable_scope.variable_op_scope([inputs],
                                        scope, 'BiasAdd', reuse=reuse) as sc:
    inputs = ops.convert_to_tensor(inputs)
    dtype = inputs.dtype.base_dtype
    num_features = utils.last_dimension(inputs.get_shape(), min_rank=2)
    biases_collections = utils.get_variable_collections(variables_collections,
                                                        'biases')
    biases = variables.model_variable('biases',
                                      shape=[num_features,],
                                      dtype=dtype,
                                      initializer=initializer,
                                      regularizer=regularizer,
                                      collections=biases_collections,
                                      trainable=trainable)
    outputs = nn.bias_add(inputs, biases)
    if activation_fn:
      outputs = activation_fn(outputs)
    return utils.collect_named_outputs(outputs_collections, sc.name, outputs)

def _create_joint_embedding_lookup(columns_to_tensors,
                                   embedding_lookup_arguments,
                                   num_outputs,
                                   trainable,
                                   weight_collections):
  """Creates an embedding lookup for all columns sharing a single weight."""
  for arg in embedding_lookup_arguments:
    assert arg.weight_tensor is None, (
        'Joint sums for weighted sparse columns are not supported. '
        'Please use weighted_sum_from_feature_columns instead.')
    assert arg.combiner == 'sum', (
        'Combiners other than sum are not supported for joint sums. '
        'Please use weighted_sum_from_feature_columns instead.')
  assert len(embedding_lookup_arguments) >= 1, (
      'At least one column must be in the model.')
  prev_size = 0
  sparse_tensors = []
  for a in embedding_lookup_arguments:
    t = a.input_tensor
    values = t.values + prev_size
    prev_size += a.vocab_size
    sparse_tensors.append(
        ops.SparseTensor(t.indices,
                         values,
                         t.shape))
  sparse_tensor = sparse_ops.sparse_concat(1, sparse_tensors)
  with variable_scope.variable_scope(
      None, default_name='linear_weights', values=columns_to_tensors.values()):
    variable = contrib_variables.model_variable(
        name='weights',
        shape=[prev_size, num_outputs],
        dtype=dtypes.float32,
        initializer=init_ops.zeros_initializer,
        trainable=trainable,
        collections=weight_collections)
    if isinstance(variable, variables.Variable):
      variable = [variable]
    else:
      variable = variable._get_variable_list()  # pylint: disable=protected-access
    predictions = embedding_ops.safe_embedding_lookup_sparse(
        variable,
        sparse_tensor,
        sparse_weights=None,
        default_id=0,
        combiner='sum',
        name='_weights')
    return variable, predictions

def _create_embedding_lookup(column,
                             columns_to_tensors,
                             embedding_lookup_arguments,
                             num_outputs,
                             trainable,
                             weight_collections):
  """Creates variables and returns predictions for linear weights in a model.

  Args:
   column: the column we're working on.
   columns_to_tensors: a map from column name to tensors.
   embedding_lookup_arguments: arguments for embedding lookup.
   num_outputs: how many outputs.
   trainable: whether the variable we create is trainable.
   weight_collections: weights will be placed here.

  Returns:
  variables: the created embeddings.
  predictions: the computed predictions.
  """
  with variable_scope.variable_scope(
      None, default_name=column.name, values=columns_to_tensors.values()):
    variable = contrib_variables.model_variable(
        name='weights',
        shape=[embedding_lookup_arguments.vocab_size, num_outputs],
        dtype=dtypes.float32,
        initializer=embedding_lookup_arguments.initializer,
        trainable=trainable,
        collections=weight_collections)
    if isinstance(variable, variables.Variable):
      variable = [variable]
    else:
      variable = variable._get_variable_list()  # pylint: disable=protected-access
    predictions = embedding_ops.safe_embedding_lookup_sparse(
        variable,
        embedding_lookup_arguments.input_tensor,
        sparse_weights=embedding_lookup_arguments.weight_tensor,
        default_id=0,
        combiner=embedding_lookup_arguments.combiner,
        name=column.name + '_weights')
    return variable, predictions

  def init_state(self, state_name, batch_size, dtype, learned_state=False):
    """Creates an initial state compatible with this cell.

    Args:
      state_name: name of the state tensor
      batch_size: model batch size
      dtype: dtype for the tensor values i.e. tf.float32
      learned_state: whether the initial state should be learnable. If false,
        the initial state is set to all 0's

    Returns:
      The created initial state.
    """
    state_size = (
        self.state_size_flat if self._flattened_state else self.state_size)
    # list of 2 zero tensors or variables tensors, depending on if
    # learned_state is true
    ret_flat = [(variables.model_variable(
        state_name + str(i),
        shape=s,
        dtype=dtype,
        initializer=tf.truncated_normal_initializer(stddev=0.03))
                 if learned_state else tf.zeros(
                     [batch_size] + s, dtype=dtype, name=state_name))
                for i, s in enumerate(state_size)]

    # duplicates initial state across the batch axis if it's learned
    if learned_state:
      ret_flat = [
          tf.stack([tensor
                    for i in range(int(batch_size))])
          for tensor in ret_flat
      ]
    for s, r in zip(state_size, ret_flat):
      r.set_shape([None] + s)
    return tf.contrib.framework.nest.pack_sequence_as(
        structure=[1, 1], flat_sequence=ret_flat)