Python tensorflow.get_variable() Examples

def build_permutation(self):

        with tf.variable_scope("encoder"):
            
            with tf.variable_scope("embedding"):
                # Embed input sequence
                W_embed =tf.get_variable("weights", [1,self.input_dimension+2, self.input_embed], initializer=self.initializer) # +2 for TW feat. here too
                embedded_input = tf.nn.conv1d(self.input_, W_embed, 1, "VALID", name="embedded_input")
                # Batch Normalization
                embedded_input = tf.layers.batch_normalization(embedded_input, axis=2, training=self.is_training, name='layer_norm', reuse=None)

            with tf.variable_scope("dynamic_rnn"):
                # Encode input sequence
                cell1 = LSTMCell(self.num_neurons, initializer=self.initializer)  # BNLSTMCell(self.num_neurons, self.training) or cell1 = DropoutWrapper(cell1, output_keep_prob=0.9)
                # Return the output activations [Batch size, Sequence Length, Num_neurons] and last hidden state as tensors.
                encoder_output, encoder_state = tf.nn.dynamic_rnn(cell1, embedded_input, dtype=tf.float32)

        with tf.variable_scope('decoder'):
            # Ptr-net returns permutations (self.positions), with their log-probability for backprop
            self.ptr = Pointer_decoder(encoder_output, self.config)
            self.positions, self.log_softmax, self.attending, self.pointing = self.ptr.loop_decode(encoder_state)
            variable_summaries('log_softmax',self.log_softmax, with_max_min = True) 

def set_input_shape(self, input_shape):
        batch_size, rows, cols, input_channels = input_shape
        kernel_shape = tuple(self.kernel_shape) + (input_channels,
                                                   self.output_channels)
        assert len(kernel_shape) == 4
        assert all(isinstance(e, int) for e in kernel_shape), kernel_shape
        with tf.variable_scope(self.name):
            init = tf.truncated_normal(kernel_shape, stddev=0.1)
            self.kernels = self.get_variable(self.w_name, init)
            self.b = self.get_variable(
                'b', .1 + np.zeros((self.output_channels,)).astype('float32'))
        input_shape = list(input_shape)
        self.input_shape = input_shape
        input_shape[0] = 1
        dummy_batch = tf.zeros(input_shape)
        dummy_output = self.fprop(dummy_batch)
        output_shape = [int(e) for e in dummy_output.get_shape()]
        output_shape[0] = 1
        self.output_shape = tuple(output_shape) 

def setUp(self):
        super(TestRunnerMultiGPU, self).setUp()
        self.sess = tf.Session()

        inputs = []
        outputs = []
        self.niter = 10
        niter = self.niter
        # A Simple graph with `niter` sub-graphs.
        with tf.variable_scope(None, 'runner'):
            for i in range(niter):
                v = tf.get_variable('v%d' % i, shape=(100, 10))
                w = tf.get_variable('w%d' % i, shape=(100, 1))

                inputs += [{'v': v, 'w': w}]
                outputs += [{'v': v, 'w': w}]

        self.runner = RunnerMultiGPU(inputs, outputs, sess=self.sess) 

def variable_on_cpu(name, shape, initializer, trainable=True):
    """Helper to create a Variable stored on CPU memory.

    Args:
            name: name of the variable
            shape: list of ints
            initializer: initializer for Variable
            trainable: boolean defining if the variable is for training
    Returns:
            Variable Tensor
    """
    var = tf.get_variable(
        name, shape, initializer=initializer, trainable=trainable)
    return var


# layers 

def _apply_with_captured_variables(self, function):
    """Applies a function using previously-captured variables.

    Args:
      function: Function to apply using captured variables.  The function
          should take one argument, its enclosing variable scope.

    Returns:
      Results of function application.
    """

    def _custom_getter(getter, *args, **kwargs):
      """Retrieves the normal or moving-average variables."""
      return self._component.get_variable(var_params=getter(*args, **kwargs))

    with tf.variable_scope(
        'cell', reuse=True, custom_getter=_custom_getter) as scope:
      return function(scope) 

def __call__(self, inputs, state, scope=None):
    """GRU cell with layer normalization."""
    input_dim = inputs.get_shape().as_list()[1]
    num_units = self._num_units

    with tf.variable_scope(scope or "gru_cell"):
      with tf.variable_scope("gates"):
        w_h = tf.get_variable(
            "w_h", [num_units, 2 * num_units],
            initializer=self._w_h_initializer())
        w_x = tf.get_variable(
            "w_x", [input_dim, 2 * num_units],
            initializer=self._w_x_initializer(input_dim))
        z_and_r = (_layer_norm(tf.matmul(state, w_h), scope="layer_norm/w_h") +
                   _layer_norm(tf.matmul(inputs, w_x), scope="layer_norm/w_x"))
        z, r = tf.split(tf.sigmoid(z_and_r), 2, 1)
      with tf.variable_scope("candidate"):
        w = tf.get_variable(
            "w", [input_dim, num_units], initializer=self._w_initializer)
        u = tf.get_variable(
            "u", [num_units, num_units], initializer=self._u_initializer)
        h_hat = (r * _layer_norm(tf.matmul(state, u), scope="layer_norm/u") +
                 _layer_norm(tf.matmul(inputs, w), scope="layer_norm/w"))
      new_h = (1 - z) * state + z * self._activation(h_hat)
    return new_h, new_h 

def wrap_variable(self, var):
        """wrap layer.w into variables"""
        val = self.lay.w.get(var, None)
        if val is None:
            shape = self.lay.wshape[var]
            args = [0., 1e-2, shape]
            if 'moving_mean' in var:
                val = np.zeros(shape)
            elif 'moving_variance' in var:
                val = np.ones(shape)
            else:
                val = np.random.normal(*args)
            self.lay.w[var] = val.astype(np.float32)
            self.act = 'Init '
        if not self.var: return

        val = self.lay.w[var]
        self.lay.w[var] = tf.constant_initializer(val)
        if var in self._SLIM: return
        with tf.variable_scope(self.scope):
            self.lay.w[var] = tf.get_variable(var,
                shape = self.lay.wshape[var],
                dtype = tf.float32,
                initializer = self.lay.w[var]) 

def _create_loss(self):
        """ Step 4: define the loss function """
        with tf.name_scope('loss'):
            # construct variables for NCE loss
            nce_weight = tf.get_variable('nce_weight',
                                         shape=[self.vocab_size, self.embed_size],
                                         initializer=tf.truncated_normal_initializer(
                                             stddev=1.0 / (self.embed_size ** 0.5)))
            nce_bias = tf.get_variable('nce_bias', initializer=tf.zeros([VOCAB_SIZE]))

            # define loss function to be NCE loss function
            self.loss = tf.reduce_mean(tf.nn.nce_loss(weights=nce_weight,
                                                      biases=nce_bias,
                                                      labels=self.target_words,
                                                      inputs=self.embed,
                                                      num_sampled=self.num_sampled,
                                                      num_classes=self.vocab_size), name='loss') 

def encode(self, inputs):

        # Tensor blocks holding the input sequences [Batch Size, Sequence Length, Features]
        #self.input_ = tf.placeholder(tf.float32, [self.batch_size, self.max_length, self.input_dimension], name="input_raw")

        with tf.variable_scope("embedding"):
          # Embed input sequence
          W_embed =tf.get_variable("weights",[1,self.input_dimension, self.input_embed], initializer=self.initializer)
          self.embedded_input = tf.nn.conv1d(inputs, W_embed, 1, "VALID", name="embedded_input")
          # Batch Normalization
          self.enc = tf.layers.batch_normalization(self.embedded_input, axis=2, training=self.is_training, name='layer_norm', reuse=None)
        
        with tf.variable_scope("stack"):
          # Blocks
          for i in range(self.num_stacks): # num blocks
              with tf.variable_scope("block_{}".format(i)):
                  ### Multihead Attention
                  self.enc = multihead_attention(self.enc, num_units=self.input_embed, num_heads=self.num_heads, dropout_rate=0.1, is_training=self.is_training)
                  
                  ### Feed Forward
                  self.enc = feedforward(self.enc, num_units=[4*self.input_embed, self.input_embed], is_training=self.is_training)

          # Return the output activations [Batch size, Sequence Length, Num_neurons] as tensors.
          self.encoder_output = self.enc ### NOTE: encoder_output is the ref for attention ###
          return self.encoder_output 

def conv_linear(args, kw, kh, nin, nout, rate, do_bias, bias_start, prefix):
  """Convolutional linear map."""
  if not isinstance(args, (list, tuple)):
    args = [args]
  with tf.variable_scope(prefix):
    with tf.device("/cpu:0"):
      k = tf.get_variable("CvK", [kw, kh, nin, nout])
    if len(args) == 1:
      arg = args[0]
    else:
      arg = tf.concat(axis=3, values=args)
    res = tf.nn.convolution(arg, k, dilation_rate=(rate, 1), padding="SAME")
    if not do_bias: return res
    with tf.device("/cpu:0"):
      bias_term = tf.get_variable(
          "CvB", [nout], initializer=tf.constant_initializer(bias_start))
    bias_term = tf.reshape(bias_term, [1, 1, 1, nout])
    return res + bias_term 

def test_adam(self):
        with self.test_session() as sess:
            w = tf.get_variable(
                "w",
                shape=[3],
                initializer=tf.constant_initializer([0.1, -0.2, -0.1]))
            x = tf.constant([0.4, 0.2, -0.5])
            loss = tf.reduce_mean(tf.square(x - w))
            tvars = tf.trainable_variables()
            grads = tf.gradients(loss, tvars)
            global_step = tf.train.get_or_create_global_step()
            optimizer = optimization.AdamWeightDecayOptimizer(learning_rate=0.2)
            train_op = optimizer.apply_gradients(zip(grads, tvars), global_step)
            init_op = tf.group(tf.global_variables_initializer(),
                               tf.local_variables_initializer())
            sess.run(init_op)
            for _ in range(100):
                sess.run(train_op)
            w_np = sess.run(w)
            self.assertAllClose(w_np.flat, [0.4, 0.2, -0.5], rtol=1e-2, atol=1e-2) 

def __init__(self, key_dim, memory_size, vocab_size,
               choose_k=256, alpha=0.1, correct_in_top=1, age_noise=8.0,
               var_cache_device='', nn_device=''):
    self.key_dim = key_dim
    self.memory_size = memory_size
    self.vocab_size = vocab_size
    self.choose_k = min(choose_k, memory_size)
    self.alpha = alpha
    self.correct_in_top = correct_in_top
    self.age_noise = age_noise
    self.var_cache_device = var_cache_device  # Variables are cached here.
    self.nn_device = nn_device  # Device to perform nearest neighbour matmul.

    caching_device = var_cache_device if var_cache_device else None
    self.update_memory = tf.constant(True)  # Can be fed "false" if needed.
    self.mem_keys = tf.get_variable(
        'memkeys', [self.memory_size, self.key_dim], trainable=False,
        initializer=tf.random_uniform_initializer(-0.0, 0.0),
        caching_device=caching_device)
    self.mem_vals = tf.get_variable(
        'memvals', [self.memory_size], dtype=tf.int32, trainable=False,
        initializer=tf.constant_initializer(0, tf.int32),
        caching_device=caching_device)
    self.mem_age = tf.get_variable(
        'memage', [self.memory_size], dtype=tf.float32, trainable=False,
        initializer=tf.constant_initializer(0.0), caching_device=caching_device)
    self.recent_idx = tf.get_variable(
        'recent_idx', [self.vocab_size], dtype=tf.int32, trainable=False,
        initializer=tf.constant_initializer(0, tf.int32))

    # variable for projecting query vector into memory key
    self.query_proj = tf.get_variable(
        'memory_query_proj', [self.key_dim, self.key_dim], dtype=tf.float32,
        initializer=tf.truncated_normal_initializer(0, 0.01),
        caching_device=caching_device) 

def _append_base_layers(self, hidden_layers):
    """Appends layers defined by the base class to the |hidden_layers|."""
    last_layer = hidden_layers[-1]

    logits = tf.nn.xw_plus_b(last_layer,
                             self._component.get_variable('weights_softmax'),
                             self._component.get_variable('bias_softmax'))
    return hidden_layers + [last_layer, logits] 

def feat_norm(input, dimZ):
    beta = tf.get_variable('beta', shape=(dimZ,), initializer=tf.constant_initializer(value=0.0))
    gamma = tf.get_variable('gamma', shape=(dimZ,), initializer=tf.constant_initializer(value=1.0))
    output,_, _ = tf.nn.fused_batch_norm(input, gamma, beta)
    return output 

def create(self,
             fixed_embeddings,
             linked_embeddings,
             context_tensor_arrays,
             attention_tensor,
             during_training,
             stride=None):
    """Requires |stride|; otherwise see base class."""
    # TODO(googleuser): Normalize the arguments to create(). 'stride'
    # is unused by the recurrent network units, while 'context_tensor_arrays'
    # and 'attenion_tensor_array' is unused by bulk network units. b/33587044
    if stride is None:
      raise ValueError("PairwiseConvNetwork needs 'stride'")

    input_tensor = get_input_tensor_with_stride(fixed_embeddings,
                                                linked_embeddings, stride)

    # TODO(googleuser): Add dropout.
    del context_tensor_arrays, attention_tensor, during_training  # Unused.

    num_steps = tf.shape(input_tensor)[1]
    arg1 = tf.expand_dims(input_tensor, 1)
    arg1 = tf.tile(arg1, tf.stack([1, num_steps, 1, 1]))
    arg2 = tf.expand_dims(input_tensor, 2)
    arg2 = tf.tile(arg2, tf.stack([1, 1, num_steps, 1]))
    conv = tf.concat([arg1, arg2], 3)
    for i in xrange(self._num_layers):
      with tf.variable_scope('conv%d' % i, reuse=True) as scope:
        conv = tf.nn.conv2d(
            conv,
            self._component.get_variable('weights'), [1, 1, 1, 1],
            padding='SAME')
        conv = tf.nn.bias_add(conv, self._component.get_variable('biases'))
        if i in self._relu_layers:
          conv = tf.nn.relu(conv, name=scope.name)
    return [tf.reshape(conv, [-1, num_steps], name='reshape_activations')] 

def get_variable(self, var_name=None, var_params=None):
    """Returns either the original or averaged version of a given variable.

    If the master.read_from_avg flag is set to True, and the
    ExponentialMovingAverage (EMA) object has been attached, then this will ask
    the EMA object for the given variable.

    This is to allow executing inference from the averaged version of
    parameters.

    Arguments:
      var_name: Name of the variable.
      var_params: tf.Variable for which to retrieve an average.

    Only one of |var_name| or |var_params| needs to be provided.  If both are
    provided, |var_params| takes precedence.

    Returns:
      tf.Variable object corresponding to original or averaged version.
    """
    if var_params:
      var_name = var_params.name
    else:
      check.NotNone(var_name, 'specify at least one of var_name or var_params')
      var_params = tf.get_variable(var_name)

    if self.moving_average and self.master.read_from_avg:
      logging.info('Retrieving average for: %s', var_name)
      var_params = self.moving_average.average(var_params)
      assert var_params
    logging.info('Returning: %s', var_params.name)
    return var_params 

def _AddVariable(self, shape, dtype, name, initializer=None):
    if name in self.variables:
      return self.variables[name]
    self.variables[name] = tf.get_variable(name, shape, dtype, initializer)
    if initializer is not None:
      self.inits[name] = state_ops.init_variable(self.variables[name],
                                                 initializer)
    return self.variables[name] 

def _variable_on_cpu(name, shape, initializer):
  """Helper to create a Variable stored on CPU memory.

  Args:
    name: name of the variable
    shape: list of ints
    initializer: initializer for Variable

  Returns:
    Variable Tensor
  """
  with tf.device('/cpu:0'):
    var = tf.get_variable(name, shape, initializer=initializer)
  return var 

def __call__(self, op):
    device_string = ''
    if self._num_ps > 0:
      task_id = self._next_task_id
      self._next_task_id = (self._next_task_id + 1) % self._num_ps
      device_string = '%s/task:%d' % (self._ps_device, task_id)
    device_string += '/%s' % self._placement
    return device_string


# TODO(sguada) Remove once get_variable is able to colocate op.devices. 

def _assert_variance_in_range(self, initializer, shape, variance,
                                tol=1e-2):
    with tf.Graph().as_default() as g:
      with self.test_session(graph=g) as sess:
        var = tf.get_variable(
            name='test',
            shape=shape,
            dtype=tf.float32,
            initializer=initializer)
        sess.run(tf.global_variables_initializer())
        values = sess.run(var)
        self.assertAllClose(np.var(values), variance, tol, tol) 

def layer_norm(x, nmaps, prefix, epsilon=1e-5):
  """Layer normalize the 4D tensor x, averaging over the last dimension."""
  with tf.variable_scope(prefix):
    scale = tf.get_variable("layer_norm_scale", [nmaps],
                            initializer=tf.ones_initializer())
    bias = tf.get_variable("layer_norm_bias", [nmaps],
                           initializer=tf.zeros_initializer())
    mean, variance = tf.nn.moments(x, [3], keep_dims=True)
    norm_x = (x - mean) / tf.sqrt(variance + epsilon)
    return norm_x * scale + bias 

def create(self,
             fixed_embeddings,
             linked_embeddings,
             context_tensor_arrays,
             attention_tensor,
             during_training,
             stride=None):
    """Requires |stride|; otherwise see base class."""
    if stride is None:
      raise RuntimeError("ConvNetwork needs 'stride' and must be called in the "
                         "bulk feature extractor component.")
    input_tensor = get_input_tensor_with_stride(fixed_embeddings,
                                                linked_embeddings, stride)

    # TODO(googleuser): Add context and attention.
    del context_tensor_arrays, attention_tensor

    # On CPU, add a dimension so that the 'image' has shape
    # [stride, 1, num_steps, D].
    conv = tf.expand_dims(input_tensor, 1)
    for i in range(len(self._depths) - 1):
      with tf.variable_scope('conv%d' % i, reuse=True) as scope:
        if during_training:
          conv.set_shape([None, 1, None, self._depths[i]])
          conv = self._maybe_apply_dropout(conv, stride)
        conv = tf.nn.conv2d(
            conv,
            self._component.get_variable('weights'), [1, 1, 1, 1],
            padding='SAME')
        conv = tf.nn.bias_add(conv, self._component.get_variable('biases'))
        if i < (len(self._weights) - 1) or not self._output_dim:
          conv = self._nonlinearity(conv, name=scope.name)
    return [
        tf.reshape(
            conv, [-1, self._depths[-1]], name='reshape_activations')
    ] 

def _fully_connected(self, x, out_dim):
    """FullyConnected layer for final output."""
    x = tf.reshape(x, [self.hps.batch_size, -1])
    w = tf.get_variable(
        'DW', [x.get_shape()[1], out_dim],
        initializer=tf.uniform_unit_scaling_initializer(factor=1.0))
    b = tf.get_variable('biases', [out_dim],
                        initializer=tf.constant_initializer())
    return tf.nn.xw_plus_b(x, w, b) 

def _variable_on_cpu(name, shape, initializer):
  """Helper to create a Variable stored on CPU memory.

  Args:
    name: name of the variable
    shape: list of ints
    initializer: initializer for Variable

  Returns:
    Variable Tensor
  """
  with tf.device('/cpu:0'):
    dtype = tf.float16 if FLAGS.use_fp16 else tf.float32
    var = tf.get_variable(name, shape, initializer=initializer, dtype=dtype)
  return var 

def predict_rewards(self,input_):

        with tf.variable_scope("encoder"):

            Encoder = Attentive_encoder(self.config)
            encoder_output = Encoder.encode(input_)
            frame = tf.reduce_mean(encoder_output, 1) # [Batch size, Sequence Length, Num_neurons] to [Batch size, Num_neurons]

        with tf.variable_scope("ffn"):
            # ffn 1
            h0 = tf.layers.dense(frame, self.num_neurons, activation=tf.nn.relu, kernel_initializer=self.initializer)
            # ffn 2
            w1 =tf.get_variable("w1", [self.num_neurons, 1], initializer=self.initializer)
            b1 = tf.Variable(self.init_baseline, name="b1")
            self.predictions = tf.squeeze(tf.matmul(h0, w1)+b1) 

def deconv2d_bn_lrn_drop(scope_or_name, inputs, kernel_shape, out_shape, subS=2, activation=tf.nn.relu,
                       use_bn=False,
                       use_mvn=False,
                       is_training=True,
                       use_lrn=False,
                       keep_prob=1.0,
                       dropout_maps=False,
                       initOpt=0):
    with tf.variable_scope(scope_or_name):
        if initOpt == 0:
            stddev = np.sqrt(2.0 / (kernel_shape[0] * kernel_shape[1] * kernel_shape[2] + kernel_shape[3]))
        if initOpt == 1:
            stddev = 5e-2
        if initOpt == 2:
            stddev = min(np.sqrt(2.0 / (kernel_shape[0] * kernel_shape[1] * kernel_shape[2])),5e-2)
        kernel = tf.get_variable("weights", kernel_shape,
                                 initializer=tf.random_normal_initializer(stddev=stddev))
        bias = tf.get_variable("bias", kernel_shape[2],
                               initializer=tf.constant_initializer(value=0.1))
        conv=tf.nn.conv2d_transpose(inputs, kernel, out_shape, strides=[1, subS, subS, 1], padding='SAME', name='conv')
        outputs = tf.nn.bias_add(conv, bias, name='preActivation')
        if use_bn:
            # outputs = tf.layers.batch_normalization(outputs, axis=3, training=is_training, name="batchNorm")
            outputs = batch_norm(outputs, is_training=is_training, scale=True, fused=True, scope="batchNorm")
        if use_mvn:
            outputs = feat_norm(outputs, kernel_shape[3])
        if activation:
            outputs = activation(outputs, name='activation')
        if use_lrn:
            outputs = tf.nn.local_response_normalization(outputs, name='localResponseNorm')
        if dropout_maps:
            conv_shape = tf.shape(outputs)
            n_shape = tf.stack([conv_shape[0], 1, 1, conv_shape[3]])
            outputs = tf.nn.dropout(outputs, keep_prob, noise_shape=n_shape)
        else:
            outputs = tf.nn.dropout(outputs, keep_prob)
        return outputs 

def testConv2DSameOdd(self):
    n, n2 = 5, 3

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

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

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

    y1 = slim.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
    y1_expected = tf.to_float([[14, 28, 43, 58, 34],
                               [28, 48, 66, 84, 46],
                               [43, 66, 84, 102, 55],
                               [58, 84, 102, 120, 64],
                               [34, 46, 55, 64, 30]])
    y1_expected = tf.reshape(y1_expected, [1, n, n, 1])

    y2 = resnet_utils.subsample(y1, 2)
    y2_expected = tf.to_float([[14, 43, 34],
                               [43, 84, 55],
                               [34, 55, 30]])
    y2_expected = tf.reshape(y2_expected, [1, n2, n2, 1])

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

    y4 = slim.conv2d(x, 1, [3, 3], stride=2, scope='Conv')
    y4_expected = y2_expected

    with self.test_session() as sess:
      sess.run(tf.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_test_input(1, n, n, 1)

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

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

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

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

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

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

    with self.test_session() as sess:
      sess.run(tf.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 testConv2DSameOdd(self):
    n, n2 = 5, 3

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

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

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

    y1 = slim.conv2d(x, 1, [3, 3], stride=1, scope='Conv')
    y1_expected = tf.to_float([[14, 28, 43, 58, 34],
                               [28, 48, 66, 84, 46],
                               [43, 66, 84, 102, 55],
                               [58, 84, 102, 120, 64],
                               [34, 46, 55, 64, 30]])
    y1_expected = tf.reshape(y1_expected, [1, n, n, 1])

    y2 = resnet_utils.subsample(y1, 2)
    y2_expected = tf.to_float([[14, 43, 34],
                               [43, 84, 55],
                               [34, 55, 30]])
    y2_expected = tf.reshape(y2_expected, [1, n2, n2, 1])

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

    y4 = slim.conv2d(x, 1, [3, 3], stride=2, scope='Conv')
    y4_expected = y2_expected

    with self.test_session() as sess:
      sess.run(tf.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_test_input(1, n, n, 1)

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

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

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

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

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

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

    with self.test_session() as sess:
      sess.run(tf.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())