python类greater()的实例源码

def _mean(self):
    mean = self.mu * self._ones()
    if self.allow_nan_stats:
      nan = np.array(np.nan, dtype=self.dtype.as_numpy_dtype())
      return math_ops.select(
          math_ops.greater(self.df, self._ones()), mean,
          array_ops.fill(self.batch_shape(), nan, name="nan"))
    else:
      return control_flow_ops.with_dependencies([
          check_ops.assert_less(
              array_ops.ones((), dtype=self.dtype), self.df,
              message="mean not defined for components of df <= 1"),
      ], mean)

def safe_divide(numerator, denominator, name):
    """Divides two values, returning 0 if the denominator is <= 0.
    Args:
      numerator: A real `Tensor`.
      denominator: A real `Tensor`, with dtype matching `numerator`.
      name: Name for the returned op.
    Returns:
      0 if `denominator` <= 0, else `numerator` / `denominator`
    """
    return tf.where(
        math_ops.greater(denominator, 0),
        math_ops.divide(numerator, denominator),
        tf.zeros_like(numerator),
        name=name)

def _safe_div(numerator, denominator, name):
    """Divides two values, returning 0 if the denominator is <= 0.
    Args:
      numerator: A real `Tensor`.
      denominator: A real `Tensor`, with dtype matching `numerator`.
      name: Name for the returned op.
    Returns:
      0 if `denominator` <= 0, else `numerator` / `denominator`
    """
    return tf.where(
        math_ops.greater(denominator, 0),
        math_ops.divide(numerator, denominator),
        tf.zeros_like(numerator),
        name=name)

def safe_divide(numerator, denominator, name):
    """Divides two values, returning 0 if the denominator is <= 0.
    Args:
      numerator: A real `Tensor`.
      denominator: A real `Tensor`, with dtype matching `numerator`.
      name: Name for the returned op.
    Returns:
      0 if `denominator` <= 0, else `numerator` / `denominator`
    """
    return tf.where(
        math_ops.greater(denominator, 0),
        math_ops.divide(numerator, denominator),
        tf.zeros_like(numerator),
        name=name)

def safe_divide(numerator, denominator, name):
    """Divides two values, returning 0 if the denominator is <= 0.
    Args:
      numerator: A real `Tensor`.
      denominator: A real `Tensor`, with dtype matching `numerator`.
      name: Name for the returned op.
    Returns:
      0 if `denominator` <= 0, else `numerator` / `denominator`
    """
    return tf.where(
        math_ops.greater(denominator, 0),
        math_ops.divide(numerator, denominator),
        tf.zeros_like(numerator),
        name=name)

def _safe_div(numerator, denominator, name):
    """Divides two values, returning 0 if the denominator is <= 0.
    Args:
      numerator: A real `Tensor`.
      denominator: A real `Tensor`, with dtype matching `numerator`.
      name: Name for the returned op.
    Returns:
      0 if `denominator` <= 0, else `numerator` / `denominator`
    """
    return tf.where(
        math_ops.greater(denominator, 0),
        math_ops.divide(numerator, denominator),
        tf.zeros_like(numerator),
        name=name)

def _logits_to_predictions(self, logits):
    """See `_MultiClassHead`."""
    with ops.name_scope(None, "predictions", (logits,)):
      return {
          prediction_key.PredictionKey.LOGITS:
              logits,
          prediction_key.PredictionKey.PROBABILITIES:
              math_ops.sigmoid(
                  logits, name=prediction_key.PredictionKey.PROBABILITIES),
          prediction_key.PredictionKey.CLASSES:
              math_ops.to_int64(
                  math_ops.greater(logits, 0),
                  name=prediction_key.PredictionKey.CLASSES)
      }

def setUp(self):
    super(CoreBinaryOpsTest, self).setUp()

    self.x_probs_broadcast_tensor = array_ops.reshape(
        self.x_probs_lt.tensor, [self.x_size, 1, self.probs_size])

    self.channel_probs_broadcast_tensor = array_ops.reshape(
        self.channel_probs_lt.tensor, [1, self.channel_size, self.probs_size])

    # == and != are not element-wise for tf.Tensor, so they shouldn't be
    # elementwise for LabeledTensor, either.
    self.ops = [
        ('add', operator.add, math_ops.add, core.add),
        ('sub', operator.sub, math_ops.subtract, core.sub),
        ('mul', operator.mul, math_ops.multiply, core.mul),
        ('div', operator.truediv, math_ops.div, core.div),
        ('mod', operator.mod, math_ops.mod, core.mod),
        ('pow', operator.pow, math_ops.pow, core.pow_function),
        ('equal', None, math_ops.equal, core.equal),
        ('less', operator.lt, math_ops.less, core.less),
        ('less_equal', operator.le, math_ops.less_equal, core.less_equal),
        ('not_equal', None, math_ops.not_equal, core.not_equal),
        ('greater', operator.gt, math_ops.greater, core.greater),
        ('greater_equal', operator.ge, math_ops.greater_equal,
         core.greater_equal),
    ]
    self.test_lt_1 = self.x_probs_lt
    self.test_lt_2 = self.channel_probs_lt
    self.test_lt_1_broadcast = self.x_probs_broadcast_tensor
    self.test_lt_2_broadcast = self.channel_probs_broadcast_tensor
    self.broadcast_axes = [self.a0, self.a1, self.a3]

def test_reflexive(self):
    labeled_tensor = self.x_probs_lt + 1  # all elements must be >0 for division
    for op_name, infix_op, _, lt_op in self.ops:
      if infix_op is not None:
        expected_lt = lt_op(2, labeled_tensor)
        actual_lt = infix_op(2, labeled_tensor)
        # Python uses greater for the reflexive version of less (and vise-versa)
        if 'less' in op_name:
          op_name = op_name.replace('less', 'greater')
        elif 'greater' in op_name:
          op_name = op_name.replace('greater', 'less')
        self.assertIn(op_name, actual_lt.name)
        self.assertLabeledTensorsEqual(expected_lt, actual_lt)

def __gt__(self, other):
    return greater(self, other)

def _ndtr(x):
  """Implements ndtr core logic."""
  half_sqrt_2 = constant_op.constant(
      0.5 * math.sqrt(2.), dtype=x.dtype, name="half_sqrt_2")
  w = x * half_sqrt_2
  z = math_ops.abs(w)
  y = array_ops.where(math_ops.less(z, half_sqrt_2),
                      1. + math_ops.erf(w),
                      array_ops.where(math_ops.greater(w, 0.),
                                      2. - math_ops.erfc(z),
                                      math_ops.erfc(z)))
  return 0.5 * y

def gumbel_decoder_fn(encoder_state, embedding_matrix, output_projections, maximum_length,
                      start_of_sequence_id=2, end_of_sequence_id=3, temperature=1.0,
                      name=None):

    with tf.name_scope(name, "gumbel_decoder_fn", [
            encoder_state, embedding_matrix, output_projections, maximum_length,
            start_of_sequence_id, end_of_sequence_id, temperature]) as scope:
        batch_size = tf.shape(encoder_state)[0]
        W, b = output_projections
        start_of_sequence_id = ops.convert_to_tensor(start_of_sequence_id, tf.int32)
        end_of_sequence_id = ops.convert_to_tensor(end_of_sequence_id, tf.int32)
        temperature = ops.convert_to_tensor(temperature, tf.float32)
        maximum_length = ops.convert_to_tensor(maximum_length, tf.int32)

    def _decoder_fn(time, cell_state, cell_input, cell_output, context_state):
        with ops.name_scope(name, "gumbel_decoder_fn",
                            [time, cell_state, cell_input, cell_output, context_state]):
            if cell_input is not None:
                # -- not None if training
                raise ValueError("Expected cell_input to be None, but saw: %s" %
                                 cell_input)
            if cell_output is None:
                # -- initial values
                next_done = array_ops.zeros([batch_size, ], dtype=dtypes.bool)
                next_cell_state = encoder_state
                next_cell_input = tf.reshape(tf.tile(embedding_matrix[start_of_sequence_id], [batch_size]),
                                             shape=tf.shape(encoder_state))
                emit_output = cell_output
                next_context_state = context_state

            else:
                # -- transition function
                with ops.name_scope(name, "gumbel_output_fn", [W, b, cell_output, end_of_sequence_id, temperature]):
                    # -- output projection parameters usually used for output logits prior to softmax
                    output_logits = tf.add(tf.matmul(cell_output, W), b)  # [B, H] * [H, V] + [V] -> [B, V]

                    # -- stopping criterion if argmax is
                    output_argmax = tf.cast(tf.argmax(output_logits, axis=1), tf.int32)
                    next_done = tf.equal(output_argmax, end_of_sequence_id)

                    # -- sample from gumbel softmax (aka concrete) distribution, higher the temperature the spikier
                    output_probs = gumbel_softmax(output_logits, temperature=temperature, hard=False)

                    # soft embeddings for the next input
                    next_cell_input = tf.matmul(output_probs, embedding_matrix)  # [B, V] * [V, H] -> [B, H]

                next_cell_state = cell_state
                emit_output = cell_output
                next_context_state = context_state

            next_done = control_flow_ops.cond(math_ops.greater(time, maximum_length),
                                         lambda: array_ops.ones([batch_size, ], dtype=dtypes.bool),
                                         lambda: next_done)

        return next_done, next_cell_state, next_cell_input, emit_output, next_context_state

    return _decoder_fn

def softmax_cross_entropy(logits, onehot_labels, weight=1.0,
                          label_smoothing=0, scope=None):
  """Creates a cross-entropy loss using tf.nn.softmax_cross_entropy_with_logits.

  `weight` acts as a coefficient for the loss. If a scalar is provided,
  then the loss is simply scaled by the given value. If `weight` is a
  tensor of size [`batch_size`], then the loss weights apply to each
  corresponding sample.

  If `label_smoothing` is nonzero, smooth the labels towards 1/num_classes:
      new_onehot_labels = onehot_labels * (1 - label_smoothing)
                          + label_smoothing / num_classes

  Args:
    logits: [batch_size, num_classes] logits outputs of the network .
    onehot_labels: [batch_size, num_classes] target one_hot_encoded labels.
    weight: Coefficients for the loss. The tensor must be a scalar or a tensor
      of shape [batch_size].
    label_smoothing: If greater than 0 then smooth the labels.
    scope: the scope for the operations performed in computing the loss.

  Returns:
    A scalar `Tensor` representing the loss value.

  Raises:
    ValueError: If the shape of `logits` doesn't match that of `onehot_labels`
      or if the shape of `weight` is invalid or if `weight` is None.
  """
  with ops.name_scope(scope, "softmax_cross_entropy_loss",
                      [logits, onehot_labels]):
    logits.get_shape().assert_is_compatible_with(onehot_labels.get_shape())

    onehot_labels = math_ops.cast(onehot_labels, logits.dtype)

    if label_smoothing > 0:
      num_classes = math_ops.cast(
          array_ops.shape(onehot_labels)[1], logits.dtype)
      smooth_positives = 1.0 - label_smoothing
      smooth_negatives = label_smoothing / num_classes
      onehot_labels = onehot_labels * smooth_positives + smooth_negatives

    losses = nn.softmax_cross_entropy_with_logits(logits, onehot_labels,
                                                  name="xentropy")
    return compute_weighted_loss(losses, weight)

def sigmoid_cross_entropy(
    logits, multi_class_labels, weights=_WEIGHT_SENTINEL, label_smoothing=0,
    scope=None, weight=_WEIGHT_SENTINEL):
  """Creates a cross-entropy loss using tf.nn.sigmoid_cross_entropy_with_logits.

  `weight` acts as a coefficient for the loss. If a scalar is provided,
  then the loss is simply scaled by the given value. If `weight` is a
  tensor of size [`batch_size`], then the loss weights apply to each
  corresponding sample.

  If `label_smoothing` is nonzero, smooth the labels towards 1/2:

      new_multiclass_labels = multiclass_labels * (1 - label_smoothing)
                              + 0.5 * label_smoothing

  Args:
    logits: [batch_size, num_classes] logits outputs of the network .
    multi_class_labels: [batch_size, num_classes] target labels in (0, 1).
    weights: Coefficients for the loss. The tensor must be a scalar, a tensor of
      shape [batch_size] or shape [batch_size, num_classes].
    label_smoothing: If greater than 0 then smooth the labels.
    scope: The scope for the operations performed in computing the loss.
    weight: Deprecated alias for `weights`.

  Returns:
    A scalar `Tensor` representing the loss value.

  Raises:
    ValueError: If the shape of `logits` doesn't match that of
      `multi_class_labels` or if the shape of `weight` is invalid, or if
      `weight` is None.
  """
  weights = _weights(weights, weight)
  with ops.name_scope(scope, "sigmoid_cross_entropy_loss",
                      [logits, multi_class_labels, weights]):
    logits.get_shape().assert_is_compatible_with(multi_class_labels.get_shape())

    multi_class_labels = math_ops.cast(multi_class_labels, logits.dtype)

    if label_smoothing > 0:
      multi_class_labels = (multi_class_labels * (1 - label_smoothing) +
                            0.5 * label_smoothing)

    losses = nn.sigmoid_cross_entropy_with_logits(logits, multi_class_labels,
                                                  name="xentropy")
    return compute_weighted_loss(losses, weights)

def softmax_cross_entropy(
    logits, onehot_labels, weights=_WEIGHT_SENTINEL, label_smoothing=0,
    scope=None, weight=_WEIGHT_SENTINEL):
  """Creates a cross-entropy loss using tf.nn.softmax_cross_entropy_with_logits.

  `weight` acts as a coefficient for the loss. If a scalar is provided,
  then the loss is simply scaled by the given value. If `weight` is a
  tensor of size [`batch_size`], then the loss weights apply to each
  corresponding sample.

  If `label_smoothing` is nonzero, smooth the labels towards 1/num_classes:
      new_onehot_labels = onehot_labels * (1 - label_smoothing)
                          + label_smoothing / num_classes

  Args:
    logits: [batch_size, num_classes] logits outputs of the network .
    onehot_labels: [batch_size, num_classes] target one_hot_encoded labels.
    weights: Coefficients for the loss. The tensor must be a scalar or a tensor
      of shape [batch_size].
    label_smoothing: If greater than 0 then smooth the labels.
    scope: the scope for the operations performed in computing the loss.
    weight: Deprecated alias for `weights`.

  Returns:
    A scalar `Tensor` representing the loss value.

  Raises:
    ValueError: If the shape of `logits` doesn't match that of `onehot_labels`
      or if the shape of `weight` is invalid or if `weight` is None.
  """
  weights = _weights(weights, weight)
  with ops.name_scope(scope, "softmax_cross_entropy_loss",
                      [logits, onehot_labels, weights]):
    logits.get_shape().assert_is_compatible_with(onehot_labels.get_shape())

    onehot_labels = math_ops.cast(onehot_labels, logits.dtype)

    if label_smoothing > 0:
      num_classes = math_ops.cast(
          array_ops.shape(onehot_labels)[1], logits.dtype)
      smooth_positives = 1.0 - label_smoothing
      smooth_negatives = label_smoothing / num_classes
      onehot_labels = onehot_labels * smooth_positives + smooth_negatives

    losses = nn.softmax_cross_entropy_with_logits(logits, onehot_labels,
                                                  name="xentropy")
    return compute_weighted_loss(losses, weights)

def sigmoid_cross_entropy(
    logits, multi_class_labels, weights=1.0, label_smoothing=0, scope=None):
  """Creates a cross-entropy loss using tf.nn.sigmoid_cross_entropy_with_logits.

  `weights` acts as a coefficient for the loss. If a scalar is provided,
  then the loss is simply scaled by the given value. If `weights` is a
  tensor of size [`batch_size`], then the loss weights apply to each
  corresponding sample.

  If `label_smoothing` is nonzero, smooth the labels towards 1/2:

      new_multiclass_labels = multiclass_labels * (1 - label_smoothing)
                              + 0.5 * label_smoothing

  Args:
    logits: [batch_size, num_classes] logits outputs of the network .
    multi_class_labels: [batch_size, num_classes] labels in (0, 1).
    weights: Coefficients for the loss. The tensor must be a scalar, a tensor of
      shape [batch_size] or shape [batch_size, num_classes].
    label_smoothing: If greater than 0 then smooth the labels.
    scope: The scope for the operations performed in computing the loss.

  Returns:
    A scalar `Tensor` representing the loss value.

  Raises:
    ValueError: If the shape of `logits` doesn't match that of
      `multi_class_labels` or if the shape of `weights` is invalid, or if
      `weights` is None.
  """
  with ops.name_scope(scope, "sigmoid_cross_entropy_loss",
                      [logits, multi_class_labels, weights]) as scope:
    logits.get_shape().assert_is_compatible_with(multi_class_labels.get_shape())

    multi_class_labels = math_ops.cast(multi_class_labels, logits.dtype)

    if label_smoothing > 0:
      multi_class_labels = (multi_class_labels * (1 - label_smoothing) +
                            0.5 * label_smoothing)

    losses = nn.sigmoid_cross_entropy_with_logits(labels=multi_class_labels,
                                                  logits=logits,
                                                  name="xentropy")
    return compute_weighted_loss(losses, weights, scope=scope)

def softmax_cross_entropy(
    logits, onehot_labels, weights=1.0, label_smoothing=0, scope=None):
  """Creates a cross-entropy loss using tf.nn.softmax_cross_entropy_with_logits.

  `weights` acts as a coefficient for the loss. If a scalar is provided,
  then the loss is simply scaled by the given value. If `weights` is a
  tensor of size [`batch_size`], then the loss weights apply to each
  corresponding sample.

  If `label_smoothing` is nonzero, smooth the labels towards 1/num_classes:
      new_onehot_labels = onehot_labels * (1 - label_smoothing)
                          + label_smoothing / num_classes

  Args:
    logits: [batch_size, num_classes] logits outputs of the network .
    onehot_labels: [batch_size, num_classes] one-hot-encoded labels.
    weights: Coefficients for the loss. The tensor must be a scalar or a tensor
      of shape [batch_size].
    label_smoothing: If greater than 0 then smooth the labels.
    scope: the scope for the operations performed in computing the loss.

  Returns:
    A scalar `Tensor` representing the mean loss value.

  Raises:
    ValueError: If the shape of `logits` doesn't match that of `onehot_labels`
      or if the shape of `weights` is invalid or if `weights` is None.
  """
  with ops.name_scope(scope, "softmax_cross_entropy_loss",
                      [logits, onehot_labels, weights]) as scope:
    logits.get_shape().assert_is_compatible_with(onehot_labels.get_shape())

    onehot_labels = math_ops.cast(onehot_labels, logits.dtype)

    if label_smoothing > 0:
      num_classes = math_ops.cast(
          array_ops.shape(onehot_labels)[1], logits.dtype)
      smooth_positives = 1.0 - label_smoothing
      smooth_negatives = label_smoothing / num_classes
      onehot_labels = onehot_labels * smooth_positives + smooth_negatives

    losses = nn.softmax_cross_entropy_with_logits(labels=onehot_labels,
                                                  logits=logits,
                                                  name="xentropy")
    return compute_weighted_loss(losses, weights, scope=scope)

def __init__(self,
               n_classes,
               label_name=None,
               weight_column_name=None,
               enable_centered_bias=False,
               head_name=None,
               loss_fn=_softmax_cross_entropy_loss,
               thresholds=None,
               metric_class_ids=None):
    """_Head for classification.

    Args:
      n_classes: Number of classes, must be greater than 2 (for 2 classes, use
        `_BinaryLogisticHead`).
      label_name: String, name of the key in label dict. Can be null if label
        is a tensor (single headed models).
      weight_column_name: A string defining feature column name representing
        weights. It is used to down weight or boost examples during training. It
        will be multiplied by the loss of the example.
      enable_centered_bias: A bool. If True, estimator will learn a centered
        bias variable for each class. Rest of the model structure learns the
        residual after centered bias.
      head_name: name of the head. If provided, predictions, summary, metrics
        keys will be suffixed by `"/" + head_name` and the default variable
        scope will be `head_name`.
      loss_fn: Loss function.
      thresholds: thresholds for eval.
      metric_class_ids: List of class IDs for which we should report per-class
        metrics. Must all be in the range `[0, n_classes)`.

    Raises:
      ValueError: if `n_classes` or `metric_class_ids` is invalid.
    """
    super(_MultiClassHead, self).__init__(
        problem_type=constants.ProblemType.CLASSIFICATION,
        logits_dimension=n_classes,
        label_name=label_name,
        weight_column_name=weight_column_name,
        head_name=head_name)

    if (n_classes is None) or (n_classes <= 2):
      raise ValueError("n_classes must be > 2: %s." % n_classes)
    self._thresholds = thresholds if thresholds else (.5,)
    self._loss_fn = loss_fn
    self._enable_centered_bias = enable_centered_bias
    self._metric_class_ids = tuple([] if metric_class_ids is None else
                                   metric_class_ids)
    for class_id in self._metric_class_ids:
      if (class_id < 0) or (class_id >= n_classes):
        raise ValueError("Class ID %s not in [0, %s)." % (class_id, n_classes))

def get_logits_and_probs(logits=None,
                         probs=None,
                         multidimensional=False,
                         validate_args=False,
                         name="get_logits_and_probs"):
  """Converts logit to probabilities (or vice-versa), and returns both.

  Args:
    logits: Numeric `Tensor` representing log-odds.
    probs: Numeric `Tensor` representing probabilities.
    multidimensional: `Boolean`, default `False`.
      If `True`, represents whether the last dimension of `logits` or `probs`,
      a `[N1, N2, ... k]` dimensional tensor, representing the
      logit or probability of `shape[-1]` classes.
    validate_args: `Boolean`, default `False`.  When `True`, either assert `0 <=
      probs <= 1` (if not `multidimensional`) or that the last dimension of
      `probs` sums to one.
    name: A name for this operation (optional).

  Returns:
    logits, probs: Tuple of `Tensor`s. If `probs` has an entry that is `0` or
      `1`, then the corresponding entry in the returned logit will be `-Inf` and
      `Inf` respectively.

  Raises:
    ValueError: if neither `probs` nor `logits` were passed in, or both were.
  """
  with ops.name_scope(name, values=[probs, logits]):
    if (probs is None) == (logits is None):
      raise ValueError("Must pass probs or logits, but not both.")

    if probs is None:
      logits = ops.convert_to_tensor(logits, name="logits")
      if multidimensional:
        return logits, nn.softmax(logits, name="probs")
      return logits, math_ops.sigmoid(logits, name="probs")

    probs = ops.convert_to_tensor(probs, name="probs")
    if validate_args:
      with ops.name_scope("validate_probs"):
        one = constant_op.constant(1., probs.dtype)
        dependencies = [check_ops.assert_non_negative(probs)]
        if multidimensional:
          dependencies += [assert_close(math_ops.reduce_sum(probs, -1), one,
                                        message="probs does not sum to 1.")]
        else:
          dependencies += [check_ops.assert_less_equal(
              probs, one, message="probs has components greater than 1.")]
        probs = control_flow_ops.with_dependencies(dependencies, probs)

    with ops.name_scope("logits"):
      if multidimensional:
        # Here we don't compute the multidimensional case, in a manner
        # consistent with respect to the unidimensional case. We do so
        # following the TF convention. Typically, you might expect to see
        # logits = log(probs) - log(gather(probs, pivot)). A side-effect of
        # being consistent with the TF approach is that the unidimensional case
        # implicitly handles the second dimension but the multidimensional case
        # explicitly keeps the pivot dimension.
        return math_ops.log(probs), probs
      return math_ops.log(probs) - math_ops.log1p(-1. * probs), probs

def softplus_inverse(x, name=None):
  """Computes the inverse softplus, i.e., x = softplus_inverse(softplus(x)).

  Mathematically this op is equivalent to:

  ```none
  softplus_inverse = log(exp(x) - 1.)