Python tensorflow.python.ops.variable_scope.get_variable_scope() Examples

def testAtrousFullyConvolutionalValues(self):
    """Verify dense feature extraction with atrous convolution."""
    nominal_stride = 32
    for output_stride in [4, 8, 16, 32, None]:
      with arg_scope(resnet_utils.resnet_arg_scope(is_training=False)):
        with ops.Graph().as_default():
          with self.test_session() as sess:
            random_seed.set_random_seed(0)
            inputs = create_test_input(2, 81, 81, 3)
            # Dense feature extraction followed by subsampling.
            output, _ = self._resnet_small(
                inputs, None, global_pool=False, output_stride=output_stride)
            if output_stride is None:
              factor = 1
            else:
              factor = nominal_stride // output_stride
            output = resnet_utils.subsample(output, factor)
            # Make the two networks use the same weights.
            variable_scope.get_variable_scope().reuse_variables()
            # Feature extraction at the nominal network rate.
            expected, _ = self._resnet_small(inputs, None, global_pool=False)
            sess.run(variables.global_variables_initializer())
            self.assertAllClose(
                output.eval(), expected.eval(), atol=1e-4, rtol=1e-4) 

def call(self, inputs, state):
    """Run the cell on embedded inputs."""
    with ops.device("/cpu:0"):
      if self._initializer:
        initializer = self._initializer
      elif vs.get_variable_scope().initializer:
        initializer = vs.get_variable_scope().initializer
      else:
        # Default initializer for embeddings should have variance=1.
        sqrt3 = math.sqrt(3)  # Uniform(-sqrt(3), sqrt(3)) has variance=1.
        initializer = init_ops.random_uniform_initializer(-sqrt3, sqrt3)

      if isinstance(state, tuple):
        data_type = state[0].dtype
      else:
        data_type = state.dtype

      embedding = vs.get_variable(
          "embedding", [self._embedding_classes, self._embedding_size],
          initializer=initializer,
          dtype=data_type)
      embedded = embedding_ops.embedding_lookup(embedding,
                                                array_ops.reshape(inputs, [-1]))

      return self._cell(embedded, state) 

def _get_concat_variable(name, shape, dtype, num_shards):
  """Get a sharded variable concatenated into one tensor."""
  sharded_variable = _get_sharded_variable(name, shape, dtype, num_shards)
  if len(sharded_variable) == 1:
    return sharded_variable[0]

  concat_name = name + "/concat"
  concat_full_name = vs.get_variable_scope().name + "/" + concat_name + ":0"
  for value in ops.get_collection(ops.GraphKeys.CONCATENATED_VARIABLES):
    if value.name == concat_full_name:
      return value

  concat_variable = array_ops.concat(sharded_variable, 0, name=concat_name)
  ops.add_to_collection(ops.GraphKeys.CONCATENATED_VARIABLES,
                        concat_variable)
  return concat_variable 

def __call__(self, inputs, state, scope=None):
    """Run the cell on embedded inputs."""
    with vs.variable_scope(scope or type(self).__name__):  # "EmbeddingWrapper"
      with ops.device("/cpu:0"):
        if self._initializer:
          initializer = self._initializer
        elif vs.get_variable_scope().initializer:
          initializer = vs.get_variable_scope().initializer
        else:
          # Default initializer for embeddings should have variance=1.
          sqrt3 = math.sqrt(3)  # Uniform(-sqrt(3), sqrt(3)) has variance=1.
          initializer = init_ops.random_uniform_initializer(-sqrt3, sqrt3)

        if type(state) is tuple:
          data_type = state[0].dtype
        else:
          data_type = state.dtype

        embedding = vs.get_variable(
            "embedding", [self._embedding_classes, self._embedding_size],
            initializer=initializer,
            dtype=data_type)
        embedded = embedding_ops.embedding_lookup(
            embedding, array_ops.reshape(inputs, [-1]))
    return self._cell(embedded, state) 

def _get_concat_variable(name, shape, dtype, num_shards):
  """Get a sharded variable concatenated into one tensor."""
  sharded_variable = _get_sharded_variable(name, shape, dtype, num_shards)
  if len(sharded_variable) == 1:
    return sharded_variable[0]

  concat_name = name + "/concat"
  concat_full_name = vs.get_variable_scope().name + "/" + concat_name + ":0"
  for value in ops.get_collection(ops.GraphKeys.CONCATENATED_VARIABLES):
    if value.name == concat_full_name:
      return value

  concat_variable = array_ops.concat(0, sharded_variable, name=concat_name)
  ops.add_to_collection(ops.GraphKeys.CONCATENATED_VARIABLES,
                        concat_variable)
  return concat_variable 

def testTrainEvalWithReuse(self):
    train_batch_size = 2
    eval_batch_size = 1
    train_height, train_width = 231, 231
    eval_height, eval_width = 281, 281
    num_classes = 1000
    with self.cached_session():
      train_inputs = random_ops.random_uniform(
          (train_batch_size, train_height, train_width, 3))
      logits, _ = overfeat.overfeat(train_inputs)
      self.assertListEqual(logits.get_shape().as_list(),
                           [train_batch_size, num_classes])
      variable_scope.get_variable_scope().reuse_variables()
      eval_inputs = random_ops.random_uniform(
          (eval_batch_size, eval_height, eval_width, 3))
      logits, _ = overfeat.overfeat(
          eval_inputs, is_training=False, spatial_squeeze=False)
      self.assertListEqual(logits.get_shape().as_list(),
                           [eval_batch_size, 2, 2, num_classes])
      logits = math_ops.reduce_mean(logits, [1, 2])
      predictions = math_ops.argmax(logits, 1)
      self.assertEqual(predictions.get_shape().as_list(), [eval_batch_size]) 

def _get_concat_variable(name, shape, dtype, num_shards):
  """Get a sharded variable concatenated into one tensor."""
  sharded_variable = _get_sharded_variable(name, shape, dtype, num_shards)
  if len(sharded_variable) == 1:
    return sharded_variable[0]

  concat_name = name + "/concat"
  concat_full_name = vs.get_variable_scope().name + "/" + concat_name + ":0"
  for value in ops.get_collection(ops.GraphKeys.CONCATENATED_VARIABLES):
    if value.name == concat_full_name:
      return value

  concat_variable = array_ops.concat(sharded_variable, 0, name=concat_name)
  ops.add_to_collection(ops.GraphKeys.CONCATENATED_VARIABLES,
                        concat_variable)
  return concat_variable 

def call(self, inputs, state):
    """Run the cell on embedded inputs."""
    with ops.device("/cpu:0"):
      if self._initializer:
        initializer = self._initializer
      elif vs.get_variable_scope().initializer:
        initializer = vs.get_variable_scope().initializer
      else:
        # Default initializer for embeddings should have variance=1.
        sqrt3 = math.sqrt(3)  # Uniform(-sqrt(3), sqrt(3)) has variance=1.
        initializer = init_ops.random_uniform_initializer(-sqrt3, sqrt3)

      if isinstance(state, tuple):
        data_type = state[0].dtype
      else:
        data_type = state.dtype

      embedding = vs.get_variable(
          "embedding", [self._embedding_classes, self._embedding_size],
          initializer=initializer,
          dtype=data_type)
      embedded = embedding_ops.embedding_lookup(embedding,
                                                array_ops.reshape(inputs, [-1]))

      return self._cell(embedded, state) 

def testTrainEvalWithReuse(self):
    train_batch_size = 2
    eval_batch_size = 1
    train_height, train_width = 224, 224
    eval_height, eval_width = 256, 256
    num_classes = 1000
    with self.cached_session():
      train_inputs = random_ops.random_uniform(
          (train_batch_size, train_height, train_width, 3))
      logits, _ = vgg.vgg_a(train_inputs)
      self.assertListEqual(logits.get_shape().as_list(),
                           [train_batch_size, num_classes])
      variable_scope.get_variable_scope().reuse_variables()
      eval_inputs = random_ops.random_uniform(
          (eval_batch_size, eval_height, eval_width, 3))
      logits, _ = vgg.vgg_a(
          eval_inputs, is_training=False, spatial_squeeze=False)
      self.assertListEqual(logits.get_shape().as_list(),
                           [eval_batch_size, 2, 2, num_classes])
      logits = math_ops.reduce_mean(logits, [1, 2])
      predictions = math_ops.argmax(logits, 1)
      self.assertEqual(predictions.get_shape().as_list(), [eval_batch_size]) 

def call(self, inputs, forward=True):
    vs = variable_scope.get_variable_scope()
    vars_before = vs.global_variables()

    if forward:
      x1, x2 = inputs
      out = self._forward(x1, x2)
    else:
      y1, y2 = inputs
      out = self._backward(y1, y2)

    # Add any created variables to the Layer's variable stores
    new_vars = vs.global_variables()[len(vars_before):]
    train_vars = vs.trainable_variables()
    for new_var in new_vars:
      if new_var in train_vars:
        self._trainable_weights.append(new_var)
      else:
        self._non_trainable_weights.append(new_var)

    return out 

def __call__(self, inputs, state, scope=None):
    """Run the cell on embedded inputs."""
    with vs.variable_scope(scope or type(self).__name__):  # "EmbeddingWrapper"
      with ops.device("/cpu:0"):
        if self._initializer:
          initializer = self._initializer
        elif vs.get_variable_scope().initializer:
          initializer = vs.get_variable_scope().initializer
        else:
          # Default initializer for embeddings should have variance=1.
          sqrt3 = math.sqrt(3)  # Uniform(-sqrt(3), sqrt(3)) has variance=1.
          initializer = init_ops.random_uniform_initializer(-sqrt3, sqrt3)

        if type(state) is tuple:
          data_type = state[0].dtype
        else:
          data_type = state.dtype

        embedding = vs.get_variable(
            "embedding", [self._embedding_classes, self._embedding_size],
            initializer=initializer,
            dtype=data_type)
        embedded = embedding_ops.embedding_lookup(
            embedding, array_ops.reshape(inputs, [-1]))
    return self._cell(embedded, state) 

def call(self, inputs, forward=True):
    vs = variable_scope.get_variable_scope()
    vars_before = vs.global_variables()

    if forward:
      x1, x2 = inputs
      out = self._forward(x1, x2)
    else:
      y1, y2 = inputs
      out = self._backward(y1, y2)

    # Add any created variables to the Layer's variable stores
    new_vars = vs.global_variables()[len(vars_before):]
    train_vars = vs.trainable_variables()
    for new_var in new_vars:
      if new_var in train_vars:
        self._trainable_weights.append(new_var)
      else:
        self._non_trainable_weights.append(new_var)

    return out 

def _get_concat_variable(name, shape, dtype, num_shards):
  """Get a sharded variable concatenated into one tensor."""
  sharded_variable = _get_sharded_variable(name, shape, dtype, num_shards)
  if len(sharded_variable) == 1:
    return sharded_variable[0]

  concat_name = name + "/concat"
  concat_full_name = vs.get_variable_scope().name + "/" + concat_name + ":0"
  for value in ops.get_collection(ops.GraphKeys.CONCATENATED_VARIABLES):
    if value.name == concat_full_name:
      return value

  concat_variable = array_ops.concat(0, sharded_variable, name=concat_name)
  ops.add_to_collection(ops.GraphKeys.CONCATENATED_VARIABLES,
                        concat_variable)
  return concat_variable 

def _get_concat_variable(name, shape, dtype, num_shards):
  """Get a sharded variable concatenated into one tensor."""
  sharded_variable = _get_sharded_variable(name, shape, dtype, num_shards)
  if len(sharded_variable) == 1:
    return sharded_variable[0]

  concat_name = name + "/concat"
  concat_full_name = vs.get_variable_scope().name + "/" + concat_name + ":0"
  for value in ops.get_collection(ops.GraphKeys.CONCATENATED_VARIABLES):
    if value.name == concat_full_name:
      return value

  concat_variable = array_ops.concat(sharded_variable, 0, name=concat_name)
  ops.add_to_collection(ops.GraphKeys.CONCATENATED_VARIABLES,
                        concat_variable)
  return concat_variable 

def testTrainEvalWithReuse(self):
    train_batch_size = 2
    eval_batch_size = 1
    train_height, train_width = 224, 224
    eval_height, eval_width = 256, 256
    num_classes = 1000
    with self.test_session():
      train_inputs = random_ops.random_uniform(
          (train_batch_size, train_height, train_width, 3))
      logits, _ = vgg.vgg_19(train_inputs)
      self.assertListEqual(logits.get_shape().as_list(),
                           [train_batch_size, num_classes])
      variable_scope.get_variable_scope().reuse_variables()
      eval_inputs = random_ops.random_uniform(
          (eval_batch_size, eval_height, eval_width, 3))
      logits, _ = vgg.vgg_19(
          eval_inputs, is_training=False, spatial_squeeze=False)
      self.assertListEqual(logits.get_shape().as_list(),
                           [eval_batch_size, 2, 2, num_classes])
      logits = math_ops.reduce_mean(logits, [1, 2])
      predictions = math_ops.argmax(logits, 1)
      self.assertEquals(predictions.get_shape().as_list(), [eval_batch_size]) 

def testAtrousFullyConvolutionalValues(self):
    """Verify dense feature extraction with atrous convolution."""
    nominal_stride = 32
    for output_stride in [4, 8, 16, 32, None]:
      with arg_scope(resnet_utils.resnet_arg_scope(is_training=False)):
        with ops.Graph().as_default():
          with self.test_session() as sess:
            random_seed.set_random_seed(0)
            inputs = create_test_input(2, 81, 81, 3)
            # Dense feature extraction followed by subsampling.
            output, _ = self._resnet_small(
                inputs, None, global_pool=False, output_stride=output_stride)
            if output_stride is None:
              factor = 1
            else:
              factor = nominal_stride // output_stride
            output = resnet_utils.subsample(output, factor)
            # Make the two networks use the same weights.
            variable_scope.get_variable_scope().reuse_variables()
            # Feature extraction at the nominal network rate.
            expected, _ = self._resnet_small(inputs, None, global_pool=False)
            sess.run(variables.global_variables_initializer())
            self.assertAllClose(
                output.eval(), expected.eval(), atol=1e-4, rtol=1e-4) 

def testTrainEvalWithReuse(self):
    train_batch_size = 2
    eval_batch_size = 1
    train_height, train_width = 224, 224
    eval_height, eval_width = 300, 400
    num_classes = 1000
    with self.test_session():
      train_inputs = random_ops.random_uniform(
          (train_batch_size, train_height, train_width, 3))
      logits, _ = alexnet.alexnet_v2(train_inputs)
      self.assertListEqual(logits.get_shape().as_list(),
                           [train_batch_size, num_classes])
      variable_scope.get_variable_scope().reuse_variables()
      eval_inputs = random_ops.random_uniform(
          (eval_batch_size, eval_height, eval_width, 3))
      logits, _ = alexnet.alexnet_v2(
          eval_inputs, is_training=False, spatial_squeeze=False)
      self.assertListEqual(logits.get_shape().as_list(),
                           [eval_batch_size, 4, 7, num_classes])
      logits = math_ops.reduce_mean(logits, [1, 2])
      predictions = math_ops.argmax(logits, 1)
      self.assertEquals(predictions.get_shape().as_list(), [eval_batch_size]) 

def testTrainEvalWithReuse(self):
    train_batch_size = 2
    eval_batch_size = 1
    train_height, train_width = 224, 224
    eval_height, eval_width = 256, 256
    num_classes = 1000
    with self.test_session():
      train_inputs = random_ops.random_uniform(
          (train_batch_size, train_height, train_width, 3))
      logits, _ = vgg.vgg_16(train_inputs)
      self.assertListEqual(logits.get_shape().as_list(),
                           [train_batch_size, num_classes])
      variable_scope.get_variable_scope().reuse_variables()
      eval_inputs = random_ops.random_uniform(
          (eval_batch_size, eval_height, eval_width, 3))
      logits, _ = vgg.vgg_16(
          eval_inputs, is_training=False, spatial_squeeze=False)
      self.assertListEqual(logits.get_shape().as_list(),
                           [eval_batch_size, 2, 2, num_classes])
      logits = math_ops.reduce_mean(logits, [1, 2])
      predictions = math_ops.argmax(logits, 1)
      self.assertEquals(predictions.get_shape().as_list(), [eval_batch_size]) 

def testTrainEvalWithReuse(self):
    train_batch_size = 2
    eval_batch_size = 1
    train_height, train_width = 224, 224
    eval_height, eval_width = 256, 256
    num_classes = 1000
    with self.cached_session():
      train_inputs = random_ops.random_uniform(
          (train_batch_size, train_height, train_width, 3))
      logits, _ = vgg.vgg_19(train_inputs)
      self.assertListEqual(logits.get_shape().as_list(),
                           [train_batch_size, num_classes])
      variable_scope.get_variable_scope().reuse_variables()
      eval_inputs = random_ops.random_uniform(
          (eval_batch_size, eval_height, eval_width, 3))
      logits, _ = vgg.vgg_19(
          eval_inputs, is_training=False, spatial_squeeze=False)
      self.assertListEqual(logits.get_shape().as_list(),
                           [eval_batch_size, 2, 2, num_classes])
      logits = math_ops.reduce_mean(logits, [1, 2])
      predictions = math_ops.argmax(logits, 1)
      self.assertEqual(predictions.get_shape().as_list(), [eval_batch_size]) 

def rnn_decoder(decoder_inputs, initial_state, cell, scope=None):
  """RNN Decoder that creates training and sampling sub-graphs.

  Args:
    decoder_inputs: Inputs for decoder, list of tensors.
      This is used only in training sub-graph.
    initial_state: Initial state for the decoder.
    cell: RNN cell to use for decoder.
    scope: Scope to use, if None new will be produced.

  Returns:
    List of tensors for outputs and states for training and sampling sub-graphs.
  """
  with vs.variable_scope(scope or "dnn_decoder"):
    states, sampling_states = [initial_state], [initial_state]
    outputs, sampling_outputs = [], []
    with ops.name_scope("training", values=[decoder_inputs, initial_state]):
      for i, inp in enumerate(decoder_inputs):
        if i > 0:
          vs.get_variable_scope().reuse_variables()
        output, new_state = cell(inp, states[-1])
        outputs.append(output)
        states.append(new_state)
    with ops.name_scope("sampling", values=[initial_state]):
      for i, _ in enumerate(decoder_inputs):
        if i == 0:
          sampling_outputs.append(outputs[i])
          sampling_states.append(states[i])
        else:
          sampling_output, sampling_state = cell(sampling_outputs[-1],
                                                 sampling_states[-1])
          sampling_outputs.append(sampling_output)
          sampling_states.append(sampling_state)
  return outputs, states, sampling_outputs, sampling_states 

def _build_variable_getter(rename=None):
  """Build a model variable getter that respects scope getter and renames."""
  # Respect current getter, if one is set.
  current_custom_getter = variable_scope.get_variable_scope().custom_getter
  def layer_variable_getter(getter, *args, **kwargs):
    if current_custom_getter is not None:
      getter = functools.partial(current_custom_getter, getter)
    kwargs['rename'] = rename
    return _model_variable_getter(getter, *args, **kwargs)
  return layer_variable_getter 

def sequence_to_final(inputs, noutput, scope=None, name=None, reverse=False):
  """Run an LSTM across all steps and returns only the final state.

  Args:
    inputs: (length, batch_size, depth) tensor
    noutput: size of output vector
    scope: optional scope name
    name: optional name for output tensor
    reverse: run in reverse

  Returns:
    Batch of size (batch_size, noutput).
  """
  with variable_scope.variable_scope(scope, "SequenceToFinal", [inputs]):
    length, batch_size, _ = _shape(inputs)
    lstm = core_rnn_cell_impl.BasicLSTMCell(noutput, state_is_tuple=False)
    state = array_ops.zeros([batch_size, lstm.state_size])
    inputs_u = array_ops.unstack(inputs)
    if reverse:
      inputs_u = list(reversed(inputs_u))
    for i in xrange(length):
      if i > 0:
        variable_scope.get_variable_scope().reuse_variables()
      output, state = lstm(inputs_u[i], state)
    outputs = array_ops.reshape(output, [batch_size, noutput], name=name)
    return outputs 

def testConv2DSameEven(self):
    n, n2 = 4, 2

    # Input image.
    x = create_test_input(1, n, n, 1)

    # Convolution kernel.
    w = create_test_input(1, 3, 3, 1)
    w = array_ops.reshape(w, [3, 3, 1, 1])

    variable_scope.get_variable('Conv/weights', initializer=w)
    variable_scope.get_variable('Conv/biases', initializer=array_ops.zeros([1]))
    variable_scope.get_variable_scope().reuse_variables()

    y1 = layers.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
    y1_expected = math_ops.to_float([[14, 28, 43, 26], [28, 48, 66, 37],
                                     [43, 66, 84, 46], [26, 37, 46, 22]])
    y1_expected = array_ops.reshape(y1_expected, [1, n, n, 1])

    y2 = resnet_utils.subsample(y1, 2)
    y2_expected = math_ops.to_float([[14, 43], [43, 84]])
    y2_expected = array_ops.reshape(y2_expected, [1, n2, n2, 1])

    y3 = resnet_utils.conv2d_same(x, 1, 3, stride=2, scope='Conv')
    y3_expected = y2_expected

    y4 = layers.conv2d(x, 1, [3, 3], stride=2, scope='Conv')
    y4_expected = math_ops.to_float([[48, 37], [37, 22]])
    y4_expected = array_ops.reshape(y4_expected, [1, n2, n2, 1])

    with self.test_session() as sess:
      sess.run(variables.global_variables_initializer())
      self.assertAllClose(y1.eval(), y1_expected.eval())
      self.assertAllClose(y2.eval(), y2_expected.eval())
      self.assertAllClose(y3.eval(), y3_expected.eval())
      self.assertAllClose(y4.eval(), y4_expected.eval()) 

def testAtrousFullyConvolutionalValues(self):
    """Verify dense feature extraction with atrous convolution."""
    nominal_stride = 32
    for output_stride in [4, 8, 16, 32, None]:
      with arg_scope(resnet_utils.resnet_arg_scope()):
        with ops.Graph().as_default():
          with self.cached_session() as sess:
            random_seed.set_random_seed(0)
            inputs = create_input(2, 81, 81, 3)
            # Dense feature extraction followed by subsampling.
            output, _ = self._resnet_small(
                inputs,
                None,
                is_training=False,
                global_pool=False,
                output_stride=output_stride)
            if output_stride is None:
              factor = 1
            else:
              factor = nominal_stride // output_stride
            output = resnet_utils.subsample(output, factor)
            # Make the two networks use the same weights.
            variable_scope.get_variable_scope().reuse_variables()
            # Feature extraction at the nominal network rate.
            expected, _ = self._resnet_small(
                inputs, None, is_training=False, global_pool=False)
            sess.run(variables.global_variables_initializer())
            self.assertAllClose(
                output.eval(), expected.eval(), atol=2e-4, rtol=1e-4) 

def call(self, inputs, state):
    """Run one step of UGRNN.
    Args:
      inputs: input Tensor, 2D, batch x input size.
      state: state Tensor, 2D, batch x num units.
    Returns:
      new_output: batch x num units, Tensor representing the output of the UGRNN
        after reading `inputs` when previous state was `state`. Identical to
        `new_state`.
      new_state: batch x num units, Tensor representing the state of the UGRNN
        after reading `inputs` when previous state was `state`.
    Raises:
      ValueError: If input size cannot be inferred from inputs via
        static shape inference.
    """
    sigmoid = math_ops.sigmoid

    input_size = inputs.get_shape().with_rank(2)[1]
    if input_size.value is None:
      raise ValueError("Could not infer input size from inputs.get_shape()[-1]")

    with vs.variable_scope(vs.get_variable_scope(),
                           initializer=self._initializer):
      cell_inputs = array_ops.concat([inputs, state], 1)
      if self._linear is None:
        self._linear = _Linear(cell_inputs, 2 * self._num_units, True)
      rnn_matrix = self._linear(cell_inputs)

      [g_act, c_act] = array_ops.split(
          axis=1, num_or_size_splits=2, value=rnn_matrix)

      c = self._activation(c_act)
      g = sigmoid(g_act + self._forget_bias)
      new_state = g * state + (1.0 - g) * c
      new_output = new_state

    return new_output, new_state 

def ndlstm_base_unrolled(inputs, noutput, scope=None, reverse=False):
  """Run an LSTM, either forward or backward.

  This is a 1D LSTM implementation using unrolling and the TensorFlow
  LSTM op.

  Args:
    inputs: input sequence (length, batch_size, ninput)
    noutput: depth of output
    scope: optional scope name
    reverse: run LSTM in reverse

  Returns:
    Output sequence (length, batch_size, noutput)

  """
  with variable_scope.variable_scope(scope, "SeqLstmUnrolled", [inputs]):
    length, batch_size, _ = _shape(inputs)
    lstm_cell = core_rnn_cell_impl.BasicLSTMCell(noutput, state_is_tuple=False)
    state = array_ops.zeros([batch_size, lstm_cell.state_size])
    output_u = []
    inputs_u = array_ops.unstack(inputs)
    if reverse:
      inputs_u = list(reversed(inputs_u))
    for i in xrange(length):
      if i > 0:
        variable_scope.get_variable_scope().reuse_variables()
      output, state = lstm_cell(inputs_u[i], state)
      output_u += [output]
    if reverse:
      output_u = list(reversed(output_u))
    outputs = array_ops.stack(output_u)
    return outputs 

def tied_rnn_seq2seq(encoder_inputs, decoder_inputs, cell,
                     loop_function=None, dtype=dtypes.float32, scope=None):
    """RNN sequence-to-sequence model with tied encoder and decoder parameters.

    This model first runs an RNN to encode encoder_inputs into a state vector, and
    then runs decoder, initialized with the last encoder state, on decoder_inputs.
    Encoder and decoder use the same RNN cell and share parameters.

    Args:
      encoder_inputs: A list of 2D Tensors [batch_size x input_size].
      decoder_inputs: A list of 2D Tensors [batch_size x input_size].
      cell: rnn_cell.RNNCell defining the cell function and size.
      loop_function: If not None, this function will be applied to i-th output
        in order to generate i+1-th input, and decoder_inputs will be ignored,
        except for the first element ("GO" symbol), see rnn_decoder for details.
      dtype: The dtype of the initial state of the rnn cell (default: tf.float32).
      scope: VariableScope for the created subgraph; default: "tied_rnn_seq2seq".

    Returns:
      A tuple of the form (outputs, state), where:
        outputs: A list of the same length as decoder_inputs of 2D Tensors with
          shape [batch_size x output_size] containing the generated outputs.
        state: The state of each decoder cell in each time-step. This is a list
          with length len(decoder_inputs) -- one item for each time-step.
          It is a 2D Tensor of shape [batch_size x cell.state_size].
    """
    with variable_scope.variable_scope("combined_tied_rnn_seq2seq"):
        scope = scope or "tied_rnn_seq2seq"
        _, enc_state = rnn.rnn(
            cell, encoder_inputs, dtype=dtype, scope=scope)
        variable_scope.get_variable_scope().reuse_variables()
        return rnn_decoder(decoder_inputs, enc_state, cell,
                           loop_function=loop_function, scope=scope) 

def testConv2DSameEven(self):
    n, n2 = 4, 2

    # Input image.
    x = create_input(1, n, n, 1)

    # Convolution kernel.
    w = create_input(1, 3, 3, 1)
    w = array_ops.reshape(w, [3, 3, 1, 1])

    variable_scope.get_variable('Conv/weights', initializer=w)
    variable_scope.get_variable('Conv/biases', initializer=array_ops.zeros([1]))
    variable_scope.get_variable_scope().reuse_variables()

    y1 = layers.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
    y1_expected = math_ops.cast([[14, 28, 43, 26], [28, 48, 66, 37],
                                 [43, 66, 84, 46], [26, 37, 46, 22]],
                                dtypes.float32)
    y1_expected = array_ops.reshape(y1_expected, [1, n, n, 1])

    y2 = resnet_utils.subsample(y1, 2)
    y2_expected = math_ops.cast([[14, 43], [43, 84]], dtypes.float32)
    y2_expected = array_ops.reshape(y2_expected, [1, n2, n2, 1])

    y3 = resnet_utils.conv2d_same(x, 1, 3, stride=2, scope='Conv')
    y3_expected = y2_expected

    y4 = layers.conv2d(x, 1, [3, 3], stride=2, scope='Conv')
    y4_expected = math_ops.cast([[48, 37], [37, 22]], dtypes.float32)
    y4_expected = array_ops.reshape(y4_expected, [1, n2, n2, 1])

    with self.cached_session() as sess:
      sess.run(variables.global_variables_initializer())
      self.assertAllClose(y1.eval(), y1_expected.eval())
      self.assertAllClose(y2.eval(), y2_expected.eval())
      self.assertAllClose(y3.eval(), y3_expected.eval())
      self.assertAllClose(y4.eval(), y4_expected.eval()) 

def testConv2DSameEven(self):
    n, n2 = 4, 2

    # Input image.
    x = create_input(1, n, n, 1)

    # Convolution kernel.
    w = create_input(1, 3, 3, 1)
    w = array_ops.reshape(w, [3, 3, 1, 1])

    variable_scope.get_variable('Conv/weights', initializer=w)
    variable_scope.get_variable('Conv/biases', initializer=array_ops.zeros([1]))
    variable_scope.get_variable_scope().reuse_variables()

    y1 = layers.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
    y1_expected = math_ops.cast([[14, 28, 43, 26],
                                 [28, 48, 66, 37],
                                 [43, 66, 84, 46],
                                 [26, 37, 46, 22]],
                                dtypes.float32)
    y1_expected = array_ops.reshape(y1_expected, [1, n, n, 1])

    y2 = resnet_utils.subsample(y1, 2)
    y2_expected = math_ops.cast([[14, 43], [43, 84]], dtypes.float32)
    y2_expected = array_ops.reshape(y2_expected, [1, n2, n2, 1])

    y3 = resnet_utils.conv2d_same(x, 1, 3, stride=2, scope='Conv')
    y3_expected = y2_expected

    y4 = layers.conv2d(x, 1, [3, 3], stride=2, scope='Conv')
    y4_expected = math_ops.cast([[48, 37], [37, 22]], dtypes.float32)
    y4_expected = array_ops.reshape(y4_expected, [1, n2, n2, 1])

    with self.cached_session() as sess:
      sess.run(variables.global_variables_initializer())
      self.assertAllClose(y1.eval(), y1_expected.eval())
      self.assertAllClose(y2.eval(), y2_expected.eval())
      self.assertAllClose(y3.eval(), y3_expected.eval())
      self.assertAllClose(y4.eval(), y4_expected.eval()) 

def testAtrousFullyConvolutionalValues(self):
    """Verify dense feature extraction with atrous convolution."""
    nominal_stride = 32
    for output_stride in [4, 8, 16, 32, None]:
      with arg_scope(resnet_utils.resnet_arg_scope()):
        with ops.Graph().as_default():
          with self.cached_session() as sess:
            random_seed.set_random_seed(0)
            inputs = create_input(2, 81, 81, 3)
            # Dense feature extraction followed by subsampling.
            output, _ = self._resnet_small(
                inputs,
                None,
                is_training=False,
                global_pool=False,
                output_stride=output_stride)
            if output_stride is None:
              factor = 1
            else:
              factor = nominal_stride // output_stride
            output = resnet_utils.subsample(output, factor)
            # Make the two networks use the same weights.
            variable_scope.get_variable_scope().reuse_variables()
            # Feature extraction at the nominal network rate.
            expected, _ = self._resnet_small(
                inputs, None, is_training=False, global_pool=False)
            sess.run(variables.global_variables_initializer())
            self.assertAllClose(
                output.eval(), expected.eval(), atol=1e-4, rtol=1e-4)