Python tensorflow.Tensor方法代码示例

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def _clip_gradients(self, grad):
    """Clips gradients if the hyperparameter `gradient_clip_norm` requires it.

    Sparse tensors, in the form of IndexedSlices returned for the
    gradients of embeddings, require special handling.

    Args:
      grad: Gradient Tensor, IndexedSlices, or None.

    Returns:
      Optionally clipped gradient.
    """
    if grad is not None and self.hyperparams.gradient_clip_norm > 0:
      logging.info('Clipping gradient %s', grad)
      if isinstance(grad, tf.IndexedSlices):
        tmp = tf.clip_by_norm(grad.values, self.hyperparams.gradient_clip_norm)
        return tf.IndexedSlices(tmp, grad.indices, grad.dense_shape)
      else:
        return tf.clip_by_norm(grad, self.hyperparams.gradient_clip_norm)
    else:
      return grad 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def dropout(input_tensor, dropout_prob):
    """Perform dropout.

    Args:
      input_tensor: float Tensor.
      dropout_prob: Python float. The probability of dropping out a value (NOT of
        *keeping* a dimension as in `tf.nn.dropout`).

    Returns:
      A version of `input_tensor` with dropout applied.
    """
    if dropout_prob is None or dropout_prob == 0.0:
        return input_tensor

    output = tf.nn.dropout(input_tensor, 1.0 - dropout_prob)
    return output 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def _create_learning_rate(hyperparams, step_var):
  """Creates learning rate var, with decay and switching for CompositeOptimizer.

  Args:
    hyperparams: a GridPoint proto containing optimizer spec, particularly
      learning_method to determine optimizer class to use.
    step_var: tf.Variable, global training step.

  Returns:
    a scalar `Tensor`, the learning rate based on current step and hyperparams.
  """
  if hyperparams.learning_method != 'composite':
    base_rate = hyperparams.learning_rate
  else:
    spec = hyperparams.composite_optimizer_spec
    switch = tf.less(step_var, spec.switch_after_steps)
    base_rate = tf.cond(switch, lambda: tf.constant(spec.method1.learning_rate),
                        lambda: tf.constant(spec.method2.learning_rate))
  return tf.train.exponential_decay(
      base_rate,
      step_var,
      hyperparams.decay_steps,
      hyperparams.decay_base,
      staircase=hyperparams.decay_staircase) 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def testBuildManualStepLearningRate(self):
    learning_rate_text_proto = """
      manual_step_learning_rate {
        schedule {
          step: 0
          learning_rate: 0.006
        }
        schedule {
          step: 90000
          learning_rate: 0.00006
        }
      }
    """
    global_summaries = set([])
    learning_rate_proto = optimizer_pb2.LearningRate()
    text_format.Merge(learning_rate_text_proto, learning_rate_proto)
    learning_rate = optimizer_builder._create_learning_rate(
        learning_rate_proto, global_summaries)
    self.assertTrue(isinstance(learning_rate, tf.Tensor)) 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def compute_exponential_averages(variables, decay):
    """Given a list of tensorflow scalar variables
    create ops corresponding to their exponential
    averages
    Parameters
    ----------
    variables: [tf.Tensor]
        List of scalar tensors.
    Returns
    -------
    averages: [tf.Tensor]
        List of scalar tensors corresponding to averages
        of al the `variables` (in order)
    apply_op: tf.runnable
        Op to be run to update the averages with current value
        of variables.
    """
    averager = tf.train.ExponentialMovingAverage(decay=decay)
    apply_op = averager.apply(variables)
    return [averager.average(v) for v in variables], apply_op 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def _normal_distribution_cdf(x, stddev):
  """Evaluates the CDF of the normal distribution.

  Normal distribution with mean 0 and standard deviation stddev,
  evaluated at x=x.

  input and output `Tensor`s have matching shapes.

  Args:
    x: a `Tensor`
    stddev: a `Tensor` with the same shape as `x`.

  Returns:
    a `Tensor` with the same shape as `x`.

  """
  return 0.5 * (1.0 + tf.erf(x / (math.sqrt(2) * stddev + 1e-20))) 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def cv_squared(x):
  """The squared coefficient of variation of a sample.

  Useful as a loss to encourage a positive distribution to be more uniform.
  Epsilons added for numerical stability.
  Returns 0 for an empty Tensor.

  Args:
    x: a `Tensor`.

  Returns:
    a `Scalar`.
  """
  epsilon = 1e-10
  float_size = tf.to_float(tf.size(x)) + epsilon
  mean = tf.reduce_sum(x) / float_size
  variance = tf.reduce_sum(tf.square(x - mean)) / float_size
  return variance / (tf.square(mean) + epsilon) 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def __init__(self, pad_mask):
    """Compute and store the location of the padding.

    Args:
      pad_mask (tf.Tensor): Reference padding tensor of shape
        [batch_size,length] or [dim_origin] (dim_origin=batch_size*length)
        containing non-zeros positive values to indicate padding location.
    """
    self.nonpad_ids = None
    self.dim_origin = None

    with tf.name_scope("pad_reduce/get_ids"):
      pad_mask = tf.reshape(pad_mask, [-1])  # Flatten the batch
      # nonpad_ids contains coordinates of zeros rows (as pad_mask is
      # float32, checking zero equality is done with |x| < epsilon, with
      # epsilon=1e-9 as standard, here pad_mask only contains positive values
      # so tf.abs would be redundant)
      self.nonpad_ids = tf.to_int32(tf.where(pad_mask < 1e-9))
      self.dim_origin = tf.shape(pad_mask)[:1] 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def remove(self, x):
    """Remove padding from the given tensor.

    Args:
      x (tf.Tensor): of shape [dim_origin,...]

    Returns:
      a tensor of shape [dim_compressed,...] with dim_compressed <= dim_origin
    """
    with tf.name_scope("pad_reduce/remove"):
      x_shape = x.get_shape().as_list()
      x = tf.gather_nd(
          x,
          indices=self.nonpad_ids,
      )
      if not tf.contrib.eager.in_eager_mode():
        # This is a hack but for some reason, gather_nd return a tensor of
        # undefined shape, so the shape is set up manually
        x.set_shape([None] + x_shape[1:])
    return x 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def restore(self, x):
    """Add padding back to the given tensor.

    Args:
      x (tf.Tensor): of shape [dim_compressed,...]

    Returns:
      a tensor of shape [dim_origin,...] with dim_compressed >= dim_origin. The
      dim is restored from the original reference tensor
    """
    with tf.name_scope("pad_reduce/restore"):
      x = tf.scatter_nd(
          indices=self.nonpad_ids,
          updates=x,
          shape=tf.concat([self.dim_origin, tf.shape(x)[1:]], axis=0),
      )
    return x 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def combine(self, expert_out, multiply_by_gates=True):
    """Sum together the expert output, weighted by the gates.

    The slice corresponding to a particular batch element `b` is computed
    as the sum over all experts `i` of the expert output, weighted by the
    corresponding gate values.  If `multiply_by_gates` is set to False, the
    gate values are ignored.

    Args:
      expert_out: a list of `num_experts` `Tensor`s, each with shape
        `[expert_batch_size_i, <extra_output_dims>]`.
      multiply_by_gates: a boolean

    Returns:
      a `Tensor` with shape `[batch_size, <extra_output_dims>]`.
    """
    # see comments on convert_gradient_to_tensor
    stitched = common_layers.convert_gradient_to_tensor(
        tf.concat(expert_out, 0))
    if multiply_by_gates:
      stitched *= tf.expand_dims(self._nonzero_gates, 1)
    combined = tf.unsorted_segment_sum(stitched, self._batch_index,
                                       tf.shape(self._gates)[0])
    return combined 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def dispatch(self, inp):
    """Create one input Tensor for each expert.

    Args:
      inp: a list of length num_datashards `Tensor`s with shapes
        `[batch_size[d], <extra_input_dims>]`.
    Returns:
      a list of `num_experts` `Tensor`s with shapes
        `[num_examples[i], <extra_input_dims>]`.
    """
    dispatched = self._dp(lambda a, b: a.dispatch(b), self._dispatchers, inp)
    ret = self._ep(tf.concat, transpose_list_of_lists(dispatched), 0)
    if ret[0].dtype == tf.float32:
      # see comments on common_layers.convert_gradient_to_tensor
      ret = self._ep(common_layers.convert_gradient_to_tensor, ret)
    return ret 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def body(self, features):
    """Most models will override this function.

    Compute label logits for one shard as a function of the transformed
    features.

    Args:
      features: A dictionary of key to Tensor.  Each Tensor has shape
         [batch_size, ?, ?, hidden_size].

    Returns:
      output: tensor of logits with shape [batch_size, O, P, body_output_size.
      losses: either single loss as a scalar, a list, a tensor (to be averaged)
              or a dictionary of losses.
    """
    raise NotImplementedError("Abstract Method") 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def eval_autoregressive(self, features=None, decode_length=50):
    """Autoregressive eval.

    Quadratic time in decode_length.

    Args:
      features: an map of string to `Tensor`
      decode_length: an integer.  How many additional timesteps to decode.

    Returns:
      logits: `Tensor`
      losses: a dictionary: {loss-name (string): floating point `Scalar`}.
          Contains a single key "training".
    """
    results = self._slow_greedy_infer(features, decode_length=decode_length)
    return results["logits"], results["losses"] 

# 需要导入模块: import tensorflow [as 别名]
# 或者: from tensorflow import Tensor [as 别名]
def _beam_decode(self, features, decode_length, beam_size, top_beams, alpha):
    """Beam search decoding.

    Models should ideally implement a more efficient version of this function.

    Args:
      features: an map of string to `Tensor`
      decode_length: an integer.  How many additional timesteps to decode.
      beam_size: number of beams.
      top_beams: an integer. How many of the beams to return.
      alpha: Float that controls the length penalty. larger the alpha, stronger
        the preference for longer translations.

    Returns:
       samples: an integer `Tensor`. Top samples from the beam search
    """
    return self._beam_decode_slow(features, decode_length, beam_size, top_beams,
                                  alpha)