Python tensorflow.python.ops.nn.bias_add() Examples

def call(self, inputs):
    shape = inputs.get_shape().as_list()
    input_dim = shape[-1]
    output_shape = shape[:-1] + [self.units]
    if len(output_shape) > 2:
      # Reshape the input to 2D.
      output_shape_tensors = array_ops.unstack(array_ops.shape(inputs))
      output_shape_tensors[-1] = self.units
      output_shape_tensor = array_ops.stack(output_shape_tensors)
      inputs = array_ops.reshape(inputs, [-1, input_dim])

    outputs = standard_ops.matmul(inputs, self.kernel)
    if self.use_bias:
      outputs = nn.bias_add(outputs, self.bias)

    if len(output_shape) > 2:
      # Reshape the output back to the original ndim of the input.
      outputs = array_ops.reshape(outputs, output_shape_tensor)
      outputs.set_shape(output_shape)

    if self.activation is not None:
      return self.activation(outputs)  # pylint: disable=not-callable
    return outputs 

def call(self, inputs):
    shape = inputs.get_shape().as_list()
    input_dim = shape[-1]
    output_shape = shape[:-1] + [self.units]
    if len(output_shape) > 2:
      # Reshape the input to 2D.
      output_shape_tensors = array_ops.unstack(array_ops.shape(inputs))
      output_shape_tensors[-1] = self.units
      output_shape_tensor = array_ops.stack(output_shape_tensors)
      inputs = array_ops.reshape(inputs, [-1, input_dim])

    outputs = standard_ops.matmul(inputs, self.kernel)
    if self.use_bias:
      outputs = nn.bias_add(outputs, self.bias)

    if len(output_shape) > 2:
      # Reshape the output back to the original ndim of the input.
      outputs = array_ops.reshape(outputs, output_shape_tensor)
      outputs.set_shape(output_shape)

    if self.activation is not None:
      return self.activation(outputs)  # pylint: disable=not-callable
    return outputs 

def testDeterministicGradients(self):
    with self.session(force_gpu=True):
      # There are problems with using force_gpu=True and cached_session with
      # both eager mode and graph mode in the same test. Using a non-cached
      # session and putting everything inside the same session context is
      # a compromise.
      for op_binding in (tf.nn.bias_add, nn.bias_add, nn_ops.bias_add):
        for data_layout in ('channels_first', 'channels_last'):
          # With the selected layer configuration, at least in TensorFlow
          # version 2.0, when data_layout='channels_last', bias_add operates
          # deterministically by default. I don't know if this is true for
          # all layer configurations. These cases are still being tested here,
          # for completeness.
          for data_rank in (1, 2, 3):
            for data_type in (dtypes.float16, dtypes.float32, dtypes.float64):
              self._testDeterministicGradientsCase(op_binding, data_layout,
                                                   data_rank, data_type) 

def call(self, inputs):
    # Apply the actual ops.
    if self.data_format == 'channels_last':
      strides = (1,) + self.strides + (1,)
    else:
      strides = (1, 1) + self.strides
    outputs = nn.separable_conv2d(
        inputs,
        self.depthwise_kernel,
        self.pointwise_kernel,
        strides=strides,
        padding=self.padding.upper(),
        rate=self.dilation_rate,
        data_format=utils.convert_data_format(self.data_format, ndim=4))

    if self.use_bias:
      outputs = nn.bias_add(
          outputs,
          self.bias,
          data_format=utils.convert_data_format(self.data_format, ndim=4))

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def call(self, inputs):
        w = self.compute_spectral_norm()
        inputs = ops.convert_to_tensor(inputs, dtype=self.dtype)
        rank = common_shapes.rank(inputs)
        if rank > 2:
            # Broadcasting is required for the inputs.
            outputs = standard_ops.tensordot(inputs, w, [[rank - 1], [0]])
            # Reshape the output back to the original ndim of the input.
            if not context.executing_eagerly():
                shape = inputs.get_shape().as_list()
                output_shape = shape[:-1] + [self.units]
                outputs.set_shape(output_shape)
        else:
            outputs = gen_math_ops.mat_mul(inputs, w)
        if self.use_bias:
            outputs = nn.bias_add(outputs, self.bias)
        if self.activation is not None:
            return self.activation(outputs)  # pylint: disable=not-callable
        return outputs 

def call(self, inputs, training=True):
        # BN if training
        if training:
            inputs = tf.keras.layers.BatchNormalization()(inputs)
        # dropout if training
        if training and self.dropout_rate > 0.0:
            inputs = dropout(inputs, self.dropout_rate, self.num_features_nonzero, self.is_sparse_inputs)
        if not training:
            print("gcn not training now")
        # convolve
        hidden_vectors = list()
        for i in range(len(self.adjs)):
            pre_sup = tf.matmul(inputs, self.kernels[i], a_is_sparse=self.is_sparse_inputs)
            hidden_vector = tf.sparse.sparse_dense_matmul(tf.cast(self.adjs[i], tf.float32), pre_sup)
            hidden_vectors.append(hidden_vector)
        outputs = tf.add_n(hidden_vectors)
        # bias
        if self.use_bias:
            outputs = nn.bias_add(outputs, self.bias)
        # activation
        if self.activation is not None:
            return self.activation(outputs)
        return outputs 

def call(self, inputs):
    inputs = ops.convert_to_tensor(inputs, dtype=self.dtype)
    shape = inputs.get_shape().as_list()
    if len(shape) > 2:
      # Broadcasting is required for the inputs.
      outputs = standard_ops.tensordot(inputs, self.kernel, [[len(shape) - 1],
                                                             [0]])
      # Reshape the output back to the original ndim of the input.
      if context.in_graph_mode():
        output_shape = shape[:-1] + [self.units]
        outputs.set_shape(output_shape)
    else:
      outputs = standard_ops.matmul(inputs, self.kernel)
    if self.use_bias:
      outputs = nn.bias_add(outputs, self.bias)
    if self.activation is not None:
      return self.activation(outputs)  # pylint: disable=not-callable
    return outputs 

def call(self, inputs):
    inputs = ops.convert_to_tensor(inputs, dtype=self.dtype)
    shape = inputs.get_shape().as_list()
    output_shape = shape[:-1] + [self.units]
    if len(output_shape) > 2:
      # Broadcasting is required for the inputs.
      outputs = standard_ops.tensordot(inputs, self.kernel, [[len(shape) - 1],
                                                             [0]])
      # Reshape the output back to the original ndim of the input.
      outputs.set_shape(output_shape)
    else:
      outputs = standard_ops.matmul(inputs, self.kernel)
    if self.use_bias:
      outputs = nn.bias_add(outputs, self.bias)
    if self.activation is not None:
      return self.activation(outputs)  # pylint: disable=not-callable
    return outputs 

def call(self, inputs):
        inputs = dropout(inputs, self.dropout_rate, self.num_features_nonzero, self.is_sparse_inputs)
        outputs = tf.matmul(inputs, self.kernel, a_is_sparse=self.is_sparse_inputs)
        if self.use_bias:
            outputs = nn.bias_add(outputs, self.bias)
        if self.activation is not None:
            return self.activation(outputs)
        return outputs 

def call(self, inputs):
    if self.data_format == 'channels_first':
      # Reshape to channels last
      inputs = array_ops.transpose(inputs, (0, 2, 3, 1))

    # Apply the actual ops.
    outputs = separable_conv2d_tf_nn(
        inputs,
        self.depthwise_kernel,
        self.pointwise_kernel,
        strides=(1,) + self.strides + (1,),
        padding=self.padding.upper(),
        rate=self.dilation_rate)

    if self.data_format == 'channels_first':
      # Reshape to channels first
      outputs = array_ops.transpose(outputs, (0, 3, 1, 2))

    if self.bias is not None:
      outputs = nn.bias_add(
          outputs,
          self.bias,
          data_format=utils.convert_data_format(self.data_format, ndim=4))

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def _testBias(self, np_inputs, np_bias, use_gpu=False):
    np_val = self._npBias(np_inputs, np_bias)
    with self.cached_session(use_gpu=use_gpu):
      tf_val = self.evaluate(nn_ops.bias_add(np_inputs, np_bias))
    self.assertAllCloseAccordingToType(np_val, tf_val) 

def call(self, inputs):
    inputs = ops.convert_to_tensor(inputs, dtype=self.dtype)
    ndim = self._input_rank

    if self.rectify:
      inputs = nn.relu(inputs)

    # Compute normalization pool.
    if ndim == 2:
      norm_pool = math_ops.matmul(math_ops.square(inputs), self.gamma)
      norm_pool = nn.bias_add(norm_pool, self.beta)
    elif self.data_format == "channels_last" and ndim <= 5:
      shape = self.gamma.shape.as_list()
      gamma = array_ops.reshape(self.gamma, (ndim - 2) * [1] + shape)
      norm_pool = nn.convolution(math_ops.square(inputs), gamma, "VALID")
      norm_pool = nn.bias_add(norm_pool, self.beta)
    else:  # generic implementation
      # This puts channels in the last dimension regardless of input.
      norm_pool = math_ops.tensordot(
          math_ops.square(inputs), self.gamma, [[self._channel_axis()], [0]])
      norm_pool += self.beta
      if self.data_format == "channels_first":
        # Return to channels_first format if necessary.
        axes = list(range(ndim - 1))
        axes.insert(1, ndim - 1)
        norm_pool = array_ops.transpose(norm_pool, axes)

    if self.inverse:
      norm_pool = math_ops.sqrt(norm_pool)
    else:
      norm_pool = math_ops.rsqrt(norm_pool)
    outputs = inputs * norm_pool

    if not context.executing_eagerly():
      outputs.set_shape(self.compute_output_shape(inputs.shape))
    return outputs 

def _eval_op(self, features, labels, logits=None, logits_input=None,
               name="eval_op"):
    labels = _check_labels(labels, self._label_name)
    if self._enable_centered_bias:
      logits = nn.bias_add(logits, _centered_bias(
          self.logits_dimension,
          self._centered_bias_weight_collection))
    loss_unweighted = self._eval_loss_fn(logits, labels)
    loss, _ = _loss(loss_unweighted,
                    _weight_tensor(features, self._weight_column_name),
                    name=name)

    predictions = self._logits_to_prediction(logits)

    return predictions, loss 

def _infer_op(self, logits=None, logits_input=None):
    if self._enable_centered_bias:
      logits = nn.bias_add(logits, _centered_bias(
          self.logits_dimension,
          self._centered_bias_weight_collection))
    return self._logits_to_prediction(logits) 

def _eval_op(self, features, labels, logits=None, logits_input=None,
               name="eval_op"):
    labels = _check_labels(labels, self._label_name)
    if self._enable_centered_bias:
      logits = nn.bias_add(logits, _centered_bias(
          self.logits_dimension,
          self._centered_bias_weight_collection))
    loss_unweighted = self._eval_loss_fn(logits, labels)
    loss, _ = _loss(loss_unweighted,
                    _weight_tensor(features, self._weight_column_name),
                    name=name)

    predictions = self._logits_to_prediction(logits)

    return predictions, loss 

def call(self, inputs):
    inputs = ops.convert_to_tensor(inputs, dtype=self.dtype)
    ndim = self._input_rank

    shape = self.gamma.get_shape().as_list()
    gamma = array_ops.reshape(self.gamma, (ndim - 2) * [1] + shape)

    # Compute normalization pool.
    if self.data_format == 'channels_first':
      norm_pool = nn.convolution(
          math_ops.square(inputs),
          gamma,
          'VALID',
          data_format='NC' + 'DHW' [-(ndim - 2):])
      if ndim == 3:
        norm_pool = array_ops.expand_dims(norm_pool, 2)
        norm_pool = nn.bias_add(norm_pool, self.beta, data_format='NCHW')
        norm_pool = array_ops.squeeze(norm_pool, [2])
      elif ndim == 5:
        shape = array_ops.shape(norm_pool)
        norm_pool = array_ops.reshape(norm_pool, shape[:3] + [-1])
        norm_pool = nn.bias_add(norm_pool, self.beta, data_format='NCHW')
        norm_pool = array_ops.reshape(norm_pool, shape)
      else:  # ndim == 4
        norm_pool = nn.bias_add(norm_pool, self.beta, data_format='NCHW')
    else:  # channels_last
      norm_pool = nn.convolution(math_ops.square(inputs), gamma, 'VALID')
      norm_pool = nn.bias_add(norm_pool, self.beta, data_format='NHWC')
    norm_pool = math_ops.sqrt(norm_pool)

    if self.inverse:
      outputs = inputs * norm_pool
    else:
      outputs = inputs / norm_pool
    outputs.set_shape(inputs.get_shape())
    return outputs 

def call(self, inputs):
    outputs = self._convolution_op(inputs, self.kernel)

    if self.use_bias:
      if self.data_format == 'channels_first':
        if self.rank == 1:
          # nn.bias_add does not accept a 1D input tensor.
          bias = array_ops.reshape(self.bias, (1, self.filters, 1))
          outputs += bias
        if self.rank == 2:
          outputs = nn.bias_add(outputs, self.bias, data_format='NCHW')
        if self.rank == 3:
          # As of Mar 2017, direct addition is significantly slower than
          # bias_add when computing gradients. To use bias_add, we collapse Z
          # and Y into a single dimension to obtain a 4D input tensor.
          outputs_shape = outputs.shape.as_list()
          outputs_4d = array_ops.reshape(outputs,
                                         [outputs_shape[0], outputs_shape[1],
                                          outputs_shape[2] * outputs_shape[3],
                                          outputs_shape[4]])
          outputs_4d = nn.bias_add(outputs_4d, self.bias, data_format='NCHW')
          outputs = array_ops.reshape(outputs_4d, outputs_shape)
      else:
        outputs = nn.bias_add(outputs, self.bias, data_format='NHWC')

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def _testBiasNCHW(self, np_inputs, np_bias, use_gpu):
    np_val = self._npBias(np_inputs, np_bias)
    np_inputs = self._NHWCToNCHW(np_inputs)
    with self.cached_session(use_gpu=use_gpu):
      tf_val = self.evaluate(nn_ops.bias_add(np_inputs, np_bias,
                                             data_format="NCHW"))
    tf_val = self._NCHWToNHWC(tf_val)
    self.assertAllCloseAccordingToType(self._AtLeast3d(np_val), tf_val) 

def _patch_bias_add():
  tf.nn.bias_add = _new_bias_add_1_14 # access via public API
  nn.bias_add = _new_bias_add_1_14 # called from tf.keras.layers.convolutional.Conv
  nn_ops.bias_add = _new_bias_add_1_14 # called from tests

# The original, pre-patched method can be viewed at
# https://github.com/tensorflow/tensorflow/blob/v1.14.0/tensorflow/python/ops/nn_ops.py#L2628 

def call(self, inputs):
    outputs = nn.convolution(
        input=inputs,
        filter=self.kernel,
        dilation_rate=self.dilation_rate,
        strides=self.strides,
        padding=self.padding.upper(),
        data_format=utils.convert_data_format(self.data_format, self.rank + 2))
    if self.bias is not None:
      if self.rank != 2 and self.data_format == 'channels_first':
        # bias_add does not support channels_first for non-4D inputs.
        if self.rank == 1:
          bias = array_ops.reshape(self.bias, (1, self.filters, 1))
        if self.rank == 3:
          bias = array_ops.reshape(self.bias, (1, self.filters, 1, 1))
        outputs += bias
      else:
        outputs = nn.bias_add(
            outputs,
            self.bias,
            data_format=utils.convert_data_format(self.data_format, 4))
        # Note that we passed rank=4 because bias_add will only accept
        # NHWC and NCWH even if the rank of the inputs is 3 or 5.

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def call(self, inputs):
    if self.data_format == 'channels_first':
      # Reshape to channels last
      inputs = array_ops.transpose(inputs, (0, 2, 3, 1))

    # Apply the actual ops.
    outputs = nn.separable_conv2d(
        inputs,
        self.depthwise_kernel,
        self.pointwise_kernel,
        strides=(1,) + self.strides + (1,),
        padding=self.padding.upper(),
        rate=self.dilation_rate)

    if self.data_format == 'channels_first':
      # Reshape to channels first
      outputs = array_ops.transpose(outputs, (0, 3, 1, 2))

    if self.bias:
      outputs = nn.bias_add(
          outputs,
          self.bias,
          data_format=utils.convert_data_format(self.data_format, ndim=4))

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def call(self, inputs):
    if self.data_format == 'channels_first':
      # Reshape to channels last
      inputs = array_ops.transpose(inputs, (0, 2, 3, 1))

    # Apply the actual ops.
    outputs = nn.separable_conv2d(
        inputs,
        self.depthwise_kernel,
        self.pointwise_kernel,
        strides=(1,) + self.strides + (1,),
        padding=self.padding.upper(),
        rate=self.dilation_rate)

    if self.data_format == 'channels_first':
      # Reshape to channels first
      outputs = array_ops.transpose(outputs, (0, 3, 1, 2))

    if self.bias:
      outputs = nn.bias_add(
          outputs,
          self.bias,
          data_format=utils.convert_data_format(self.data_format, ndim=4))

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def call(self, inputs):
    outputs = nn.convolution(
        input=inputs,
        filter=self.kernel,
        dilation_rate=self.dilation_rate,
        strides=self.strides,
        padding=self.padding.upper(),
        data_format=utils.convert_data_format(self.data_format, self.rank + 2))
    if self.bias is not None:
      if self.rank != 2 and self.data_format == 'channels_first':
        # bias_add does not support channels_first for non-4D inputs.
        if self.rank == 1:
          bias = array_ops.reshape(self.bias, (1, self.filters, 1))
        if self.rank == 3:
          bias = array_ops.reshape(self.bias, (1, self.filters, 1, 1))
        outputs += bias
      else:
        outputs = nn.bias_add(
            outputs,
            self.bias,
            data_format=utils.convert_data_format(self.data_format, 4))
        # Note that we passed rank=4 because bias_add will only accept
        # NHWC and NCWH even if the rank of the inputs is 3 or 5.

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def bias_add(x, bias, data_format=None):
  """Adds a bias vector to a tensor.

  Arguments:
      x: Tensor or variable.
      bias: Bias tensor to add.
      data_format: Data format for 3D, 4D or 5D tensors:
          one of "channels_first", "channels_last".

  Returns:
      Output tensor.

  Raises:
      ValueError: In case of invalid `data_format` argument.
  """
  if data_format is None:
    data_format = image_data_format()
  if data_format not in {'channels_first', 'channels_last'}:
    raise ValueError('Unknown data_format ' + str(data_format))
  if ndim(x) == 5:
    if data_format == 'channels_first':
      x += reshape(bias, (1, int_shape(bias)[0], 1, 1, 1))
    elif data_format == 'channels_last':
      x += reshape(bias, (1, 1, 1, 1, int_shape(bias)[0]))
  elif ndim(x) == 4:
    if data_format == 'channels_first':
      # No support yet for NCHW in bias_add.
      x += reshape(bias, (1, int_shape(bias)[0], 1, 1))
    elif data_format == 'channels_last':
      x = nn.bias_add(x, bias, data_format='NHWC')
  elif ndim(x) == 3:
    if data_format == 'channels_first':
      x += reshape(bias, (1, int_shape(bias)[0], 1))
    elif data_format == 'channels_last':
      x += reshape(bias, (1, 1, int_shape(bias)[0]))
  else:
    x = nn.bias_add(x, bias)
  return x


# RANDOMNESS 

def call(self, inputs):
    if self.data_format == 'channels_first':
      # Reshape to channels last
      inputs = array_ops.transpose(inputs, (0, 2, 3, 1))

    # Apply the actual ops.
    outputs = nn.separable_conv2d(
        inputs,
        self.depthwise_kernel,
        self.pointwise_kernel,
        strides=(1,) + self.strides + (1,),
        padding=self.padding.upper(),
        rate=self.dilation_rate)

    if self.data_format == 'channels_first':
      # Reshape to channels first
      outputs = array_ops.transpose(outputs, (0, 3, 1, 2))

    if self.bias is not None:
      outputs = nn.bias_add(
          outputs,
          self.bias,
          data_format=utils.convert_data_format(self.data_format, ndim=4))

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def call(self, inputs):
    outputs = nn.convolution(
        input=inputs,
        filter=self.kernel,
        dilation_rate=self.dilation_rate,
        strides=self.strides,
        padding=self.padding.upper(),
        data_format=utils.convert_data_format(self.data_format, self.rank + 2))

    if self.bias is not None:
      if self.data_format == 'channels_first':
        # bias_add only supports NHWC.
        # TODO(fchollet): remove this when `bias_add` is feature-complete.
        if self.rank == 1:
          bias = array_ops.reshape(self.bias, (1, self.filters, 1))
          outputs += bias
        if self.rank == 2:
          bias = array_ops.reshape(self.bias, (1, self.filters, 1, 1))
          outputs += bias
        if self.rank == 3:
          # As of Mar 2017, direct addition is significantly slower than
          # bias_add when computing gradients. To use bias_add, we collapse Z
          # and Y into a single dimension to obtain a 4D input tensor.
          outputs_shape = outputs.shape.as_list()
          outputs_4d = array_ops.reshape(outputs,
                                         [outputs_shape[0], outputs_shape[1],
                                          outputs_shape[2] * outputs_shape[3],
                                          outputs_shape[4]])
          outputs_4d = nn.bias_add(outputs_4d, self.bias, data_format='NCHW')
          outputs = array_ops.reshape(outputs_4d, outputs_shape)
      else:
        outputs = nn.bias_add(outputs, self.bias, data_format='NHWC')

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def call(self, inputs):
    inputs = ops.convert_to_tensor(inputs, dtype=self.dtype)
    ndim = self._input_rank

    shape = self.gamma.get_shape().as_list()
    gamma = array_ops.reshape(self.gamma, (ndim - 2) * [1] + shape)

    # Compute normalization pool.
    if self.data_format == 'channels_first':
      norm_pool = nn.convolution(
          math_ops.square(inputs),
          gamma,
          'VALID',
          data_format='NC' + 'DHW' [-(ndim - 2):])
      if ndim == 3:
        norm_pool = array_ops.expand_dims(norm_pool, 2)
        norm_pool = nn.bias_add(norm_pool, self.beta, data_format='NCHW')
        norm_pool = array_ops.squeeze(norm_pool, [2])
      elif ndim == 5:
        shape = array_ops.shape(norm_pool)
        norm_pool = array_ops.reshape(norm_pool, shape[:3] + [-1])
        norm_pool = nn.bias_add(norm_pool, self.beta, data_format='NCHW')
        norm_pool = array_ops.reshape(norm_pool, shape)
      else:  # ndim == 4
        norm_pool = nn.bias_add(norm_pool, self.beta, data_format='NCHW')
    else:  # channels_last
      norm_pool = nn.convolution(math_ops.square(inputs), gamma, 'VALID')
      norm_pool = nn.bias_add(norm_pool, self.beta, data_format='NHWC')
    norm_pool = math_ops.sqrt(norm_pool)

    if self.inverse:
      outputs = inputs * norm_pool
    else:
      outputs = inputs / norm_pool
    outputs.set_shape(inputs.get_shape())
    return outputs 

def head_ops(self,
               features,
               labels,
               mode,
               train_op_fn,
               logits=None,
               logits_input=None,
               scope=None):
    """See `_Head`."""
    _check_mode_valid(mode)
    _check_logits_input_not_supported(logits, logits_input)

    centered_bias = None
    if self._enable_centered_bias:
      centered_bias = _centered_bias(self._logits_dimension, self.head_name)
      logits = nn.bias_add(logits, centered_bias)

    predictions = self._logits_to_predictions(logits)
    loss = None
    train_op = None
    eval_metric_ops = None
    if (mode != model_fn.ModeKeys.INFER) and (labels is not None):
      labels_tensor = _to_labels_tensor(labels, self._label_name)
      loss = _training_loss(
          features,
          labels_tensor,
          logits,
          loss_fn=self._loss_fn,
          weight_column_name=self._weight_column_name,
          head_name=self.head_name)
      if (mode == model_fn.ModeKeys.TRAIN) and (train_op_fn is not None):
        train_op = _train_op(loss, labels_tensor, train_op_fn, centered_bias,
                             self._logits_dimension, self._loss_fn)
      eval_metric_ops = _eval_metric_ops(self._default_metrics(), features,
                                         labels, predictions)

    return model_fn.ModelFnOps(
        mode=mode,
        predictions=predictions,
        loss=loss,
        train_op=train_op,
        eval_metric_ops=eval_metric_ops,
        output_alternatives=self._create_output_alternatives(predictions)) 

def call(self, inputs):
    inputs_shape = array_ops.shape(inputs)
    batch_size = inputs_shape[0]
    if self.data_format == 'channels_first':
      c_axis, h_axis, w_axis = 1, 2, 3
    else:
      c_axis, h_axis, w_axis = 3, 1, 2

    height, width = inputs_shape[h_axis], inputs_shape[w_axis]
    kernel_h, kernel_w = self.kernel_size
    stride_h, stride_w = self.strides

    # Infer the dynamic output shape:
    out_height = utils.deconv_output_length(height,
                                            kernel_h,
                                            self.padding,
                                            stride_h)
    out_width = utils.deconv_output_length(width,
                                           kernel_w,
                                           self.padding,
                                           stride_w)
    if self.data_format == 'channels_first':
      output_shape = (batch_size, self.filters, out_height, out_width)
      strides = (1, 1, stride_h, stride_w)
    else:
      output_shape = (batch_size, out_height, out_width, self.filters)
      strides = (1, stride_h, stride_w, 1)

    output_shape_tensor = array_ops.stack(output_shape)
    outputs = nn.conv2d_transpose(
        inputs,
        self.kernel,
        output_shape_tensor,
        strides,
        padding=self.padding.upper(),
        data_format=utils.convert_data_format(self.data_format, ndim=4))

    if context.in_graph_mode():
      # Infer the static output shape:
      out_shape = inputs.get_shape().as_list()
      out_shape[c_axis] = self.filters
      out_shape[h_axis] = utils.deconv_output_length(out_shape[h_axis],
                                                     kernel_h,
                                                     self.padding,
                                                     stride_h)
      out_shape[w_axis] = utils.deconv_output_length(out_shape[w_axis],
                                                     kernel_w,
                                                     self.padding,
                                                     stride_w)
      outputs.set_shape(out_shape)

    if self.use_bias:
      outputs = nn.bias_add(
          outputs,
          self.bias,
          data_format=utils.convert_data_format(self.data_format, ndim=4))

    if self.activation is not None:
      return self.activation(outputs)
    return outputs 

def call(self, inputs):
    inputs_shape = array_ops.shape(inputs)
    batch_size = inputs_shape[0]
    if self.data_format == 'channels_first':
      c_axis, h_axis, w_axis = 1, 2, 3
    else:
      c_axis, h_axis, w_axis = 3, 1, 2

    height, width = inputs_shape[h_axis], inputs_shape[w_axis]
    kernel_h, kernel_w = self.kernel_size
    stride_h, stride_w = self.strides

    def get_deconv_dim(dim_size, stride_size, kernel_size, padding):
      if isinstance(dim_size, ops.Tensor):
        dim_size = math_ops.multiply(dim_size, stride_size)
      elif dim_size is not None:
        dim_size *= stride_size

      if padding == 'valid' and dim_size is not None:
        dim_size += max(kernel_size - stride_size, 0)
      return dim_size

    # Infer the dynamic output shape:
    out_height = get_deconv_dim(height, stride_h, kernel_h, self.padding)
    out_width = get_deconv_dim(width, stride_w, kernel_w, self.padding)

    if self.data_format == 'channels_first':
      output_shape = (batch_size, self.filters, out_height, out_width)
      strides = (1, 1, stride_h, stride_w)
    else:
      output_shape = (batch_size, out_height, out_width, self.filters)
      strides = (1, stride_h, stride_w, 1)

    output_shape_tensor = array_ops.stack(output_shape)
    outputs = nn.conv2d_transpose(
        inputs,
        self.kernel,
        output_shape_tensor,
        strides,
        padding=self.padding.upper(),
        data_format=utils.convert_data_format(self.data_format, ndim=4))

    # Infer the static output shape:
    out_shape = inputs.get_shape().as_list()
    out_shape[c_axis] = self.filters
    out_shape[h_axis] = get_deconv_dim(
        out_shape[h_axis], stride_h, kernel_h, self.padding)
    out_shape[w_axis] = get_deconv_dim(
        out_shape[w_axis], stride_w, kernel_w, self.padding)
    outputs.set_shape(out_shape)

    if self.bias:
      outputs = nn.bias_add(
          outputs,
          self.bias,
          data_format=utils.convert_data_format(self.data_format, ndim=4))

    if self.activation is not None:
      return self.activation(outputs)
    return outputs