Python ops.convert_to_tensor_or_indexed_slices方法代码示例

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def _NextIteration(data, name=None):
  data = ops.convert_to_tensor_or_indexed_slices(data, as_ref=True)
  if isinstance(data, ops.Tensor):
    if data.dtype.is_ref_dtype:
      return ref_next_iteration(data, name=name)
    else:
      return next_iteration(data, name=name)
  else:
    if not isinstance(data, (ops.IndexedSlices, sparse_tensor.SparseTensor)):
      raise TypeError("Type %s not supported" % type(data))
    values = _NextIteration(data.values, name=name)
    indices = next_iteration(data.indices, name="indices")
    if isinstance(data, ops.IndexedSlices):
      dense_shape = data.dense_shape
      if dense_shape is not None:
        dense_shape = next_iteration(dense_shape, name="dense_shape")
      return ops.IndexedSlices(values, indices, dense_shape)
    else:
      dense_shape = next_iteration(data.shape, name="dense_shape")
      return sparse_tensor.SparseTensor(indices, values, dense_shape) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def _AsTensorList(x, p):
  """Return x as a list of Tensors or IndexedSlices.

  For entries of `x` that are Operations, this returns an Identity of `p`
  with a dependency on the operation.

  Args:
    x: A Tensor/IndexedSlices/Operation or a list or tuple of them.
    p: A Tensor to return for entries in `x` that are Operations.

  Returns:
    A list of Tensors or IndexedSlices.
  """
  if not isinstance(x, (list, _basetuple)):
    x = [x]

  l = []
  for v in x:
    if isinstance(v, ops.Operation):
      v = with_dependencies([v], p)
    v = ops.convert_to_tensor_or_indexed_slices(v)
    if isinstance(v, ops.Tensor):
      l.append(array_ops.identity(v))
    else:
      l.append(ops.IndexedSlices(array_ops.identity(v.values),
                                 array_ops.identity(v.indices)))
  return l 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def with_dependencies(dependencies, output_tensor, name=None):
  """Produces the content of `output_tensor` only after `dependencies`.

  In some cases, a user may want the output of an operation to be
  consumed externally only after some other dependencies have run
  first. This function ensures returns `output_tensor`, but only after all
  operations in `dependencies` have run. Note that this means that there is
  no guarantee that `output_tensor` will be evaluated after any `dependencies`
  have run.

  See also @{tf.tuple$tuple} and @{tf.group$group}.

  Args:
    dependencies: Iterable of operations to run before this op finishes.
    output_tensor: A `Tensor` or `IndexedSlices` that will be returned.
    name: (Optional) A name for this operation.

  Returns:
    Same as `output_tensor`.

  Raises:
    TypeError: if `output_tensor` is not a `Tensor` or `IndexedSlices`.
  """
  with ops.name_scope(name, "control_dependency",
                      list(dependencies) + [output_tensor]) as name:
    with ops.colocate_with(output_tensor):
      with ops.control_dependencies(dependencies):
        output_tensor = ops.convert_to_tensor_or_indexed_slices(output_tensor)
        if isinstance(output_tensor, ops.Tensor):
          return _Identity(output_tensor, name=name)
        else:
          return ops.IndexedSlices(_Identity(output_tensor.values, name=name),
                                   output_tensor.indices,
                                   output_tensor.dense_shape) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def with_dependencies(dependencies, output_tensor, name=None):
  """Produces the content of `output_tensor` only after `dependencies`.

  In some cases, a user may want the output of an operation to be
  consumed externally only after some other dependencies have run
  first. This function ensures returns `output_tensor`, but only after all
  operations in `dependencies` have run. Note that this means that there is
  no guarantee that `output_tensor` will be evaluated after any `dependencies`
  have run.

  See also `tuple` and `group`.

  Args:
    dependencies: Iterable of operations to run before this op finishes.
    output_tensor: A `Tensor` or `IndexedSlices` that will be returned.
    name: (Optional) A name for this operation.

  Returns:
    Same as `output_tensor`.

  Raises:
    TypeError: if `output_tensor` is not a `Tensor` or `IndexedSlices`.
  """
  with ops.name_scope(name, "control_dependency",
                      list(dependencies) + [output_tensor]) as name:
    with ops.colocate_with(output_tensor):
      with ops.control_dependencies(dependencies):
        output_tensor = ops.convert_to_tensor_or_indexed_slices(output_tensor)
        if isinstance(output_tensor, ops.Tensor):
          return _Identity(output_tensor, name=name)
        else:
          return ops.IndexedSlices(_Identity(output_tensor.values, name=name),
                                   output_tensor.indices,
                                   output_tensor.dense_shape) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def _Identity(data, name=None):
  """Return a tensor with the same shape and contents as the input tensor.

  Args:
    data: A Tensor.
    name: A name for this operation (optional).

  Returns:
    A Tensor with the same type and value as the input Tensor.
  """
  data = ops.convert_to_tensor_or_indexed_slices(data, as_ref=True)
  if isinstance(data, ops.Tensor):
    if data.dtype.is_ref_dtype:
      return gen_array_ops._ref_identity(data, name=name)
    else:
      return array_ops.identity(data, name=name)
  else:
    if not isinstance(data, (ops.IndexedSlices, sparse_tensor.SparseTensor)):
      raise TypeError("Type %s not supported" % type(data))
    values = _Identity(data.values, name=name)
    indices = array_ops.identity(data.indices, name="indices")
    if isinstance(data, ops.IndexedSlices):
      dense_shape = data.dense_shape
      if dense_shape is not None:
        dense_shape = array_ops.identity(dense_shape, name="dense_shape")
      return ops.IndexedSlices(values, indices, dense_shape)
    else:
      dense_shape = array_ops.identity(data.shape, name="dense_shape")
      return sparse_tensor.SparseTensor(indices, values, dense_shape) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def exit(data, name=None):
  """Exits the current frame to its parent frame.

  Exit makes its input `data` available to the parent frame.

  Args:
    data: The tensor to be made available to the parent frame.
    name: A name for this operation (optional).

  Returns:
    The same tensor as `data`.
  """
  data = ops.convert_to_tensor_or_indexed_slices(data, as_ref=True)
  if isinstance(data, ops.Tensor):
    if data.dtype.is_ref_dtype:
      return gen_control_flow_ops._ref_exit(data, name)
    else:
      return gen_control_flow_ops._exit(data, name)
  else:
    if not isinstance(data, (ops.IndexedSlices, sparse_tensor.SparseTensor)):
      raise TypeError("Type %s not supported" % type(data))
    values = exit(data.values, name=name)
    indices = gen_control_flow_ops._exit(data.indices, name="indices")
    if isinstance(data, ops.IndexedSlices):
      dense_shape = data.dense_shape
      if dense_shape is not None:
        dense_shape = gen_control_flow_ops._exit(dense_shape, name)
      return ops.IndexedSlices(values, indices, dense_shape)
    else:
      dense_shape = gen_control_flow_ops._exit(data.shape, name)
      return sparse_tensor.SparseTensor(indices, values, dense_shape) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def with_dependencies(dependencies, output_tensor, name=None):
  """Produces the content of `output_tensor` only after `dependencies`.

  In some cases, a user may want the output of an operation to be
  consumed externally only after some other dependencies have run
  first. This function ensures returns `output_tensor`, but only after all
  operations in `dependencies` have run. Note that this means that there is
  no guarantee that `output_tensor` will be evaluated after any `dependencies`
  have run.

  See also `tuple` and `group`.

  Args:
    dependencies: A list of operations to run before this op finishes.
    output_tensor: A `Tensor` or `IndexedSlices` that will be returned.
    name: (Optional) A name for this operation.

  Returns:
    Same as `output_tensor`.

  Raises:
    TypeError: if `output_tensor` is not a `Tensor` or `IndexedSlices`.
  """
  with ops.name_scope(name, "control_dependency",
                      dependencies + [output_tensor]) as name:
    with ops.colocate_with(output_tensor):
      with ops.control_dependencies(dependencies):
        output_tensor = ops.convert_to_tensor_or_indexed_slices(output_tensor)
        if isinstance(output_tensor, ops.Tensor):
          return _Identity(output_tensor, name=name)
        else:
          return ops.IndexedSlices(_Identity(output_tensor.values, name=name),
                                   output_tensor.indices,
                                   output_tensor.dense_shape) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def with_dependencies(dependencies, output_tensor, name=None):
  """Produces the content of `output_tensor` only after `dependencies`.

  In some cases, a user may want the output of an operation to be
  consumed externally only after some other dependencies have run
  first. This function ensures returns `output_tensor`, but only after all
  operations in `dependencies` have run. Note that this means that there is
  no guarantee that `output_tensor` will be evaluated after any `dependencies`
  have run.

  See also @{tf.tuple$tuple} and @{tf.group$group}.

  Args:
    dependencies: Iterable of operations to run before this op finishes.
    output_tensor: A `Tensor` or `IndexedSlices` that will be returned.
    name: (Optional) A name for this operation.

  Returns:
    Same as `output_tensor`.

  Raises:
    TypeError: if `output_tensor` is not a `Tensor` or `IndexedSlices`.
  """
  if context.in_eager_mode():
    return output_tensor
  with ops.name_scope(name, "control_dependency",
                      list(dependencies) + [output_tensor]) as name:
    with ops.colocate_with(output_tensor):
      with ops.control_dependencies(dependencies):
        output_tensor = ops.convert_to_tensor_or_indexed_slices(output_tensor)
        if isinstance(output_tensor, ops.Tensor):
          return _Identity(output_tensor, name=name)
        else:
          return ops.IndexedSlices(_Identity(output_tensor.values, name=name),
                                   output_tensor.indices,
                                   output_tensor.dense_shape) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def _SwitchRefOrTensor(data, pred, name="Switch"):
  """Forwards `data` to an output determined by `pred`.

  If `pred` is false, the `data` input is forwared to the first output.
  Otherwise, the data goes to the second output.

  This op handles `Tensor`s and `IndexedSlices`.

  Args:
    data: The tensor to be forwarded to the appropriate output.
    pred: A scalar that specifies which output port will receive data.
    name: A name for this operation (optional).

  Returns:
    `(output_false, output_true)`: If `pred` is true, data will be forwarded to
    `output_true`, otherwise it goes to `output_false`.

  Raises:
    TypeError: if data is not a Tensor or IndexedSlices
  """
  data = ops.convert_to_tensor_or_indexed_slices(data, name="data")
  # NOTE(vrv): ops.colocate_with(data, ignore_existing=True) below
  # addresses the following scenario.
  #
  # Assume you execute Optimizer.apply_gradients() in a branch of a cond().
  #
  # 1. The update op is created inside a `with ops.colocate(var):` block
  #
  # 2. Some tensor `data` is captured and a switch is created in a
  #    `with ops.colocate_with(data):` block.
  #
  # with ops.colocate_with(var):
  #  with ops.colocate_with(data):
  #    op = ...
  #
  # var and data may be pinned to different devices, so we want to ops
  # created within ops.colocate_with(data) to ignore the existing stack.
  with ops.colocate_with(data, ignore_existing=True):
    if isinstance(data, ops.Tensor):
      if data.dtype._is_ref_dtype:  # pylint: disable=protected-access
        return ref_switch(data, pred, name=name)
    return switch(data, pred, name=name) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def _SwitchRefOrTensor(data, pred, name="Switch"):
  """Forwards `data` to an output determined by `pred`.

  If `pred` is true, the `data` input is forwared to the first output.
  Otherwise, the data goes to the second output.

  This op handles `Tensor`s and `IndexedSlices`.

  Args:
    data: The tensor to be forwarded to the appropriate output.
    pred: A scalar that specifies which output port will receive data.
    name: A name for this operation (optional).

  Returns:
    `(output_false, output_false)`: If `pred` is true, data will be forwarded to
    `output_true`, otherwise it goes to `output_false`.

  Raises:
    TypeError: if data is not a Tensor or IndexedSlices
  """
  data = ops.convert_to_tensor_or_indexed_slices(data, name="data")
  # NOTE(vrv): ops.colocate_with(data, ignore_existing=True) below
  # addresses the following scenario.
  #
  # Assume you execute Optimizer.apply_gradients() in a branch of a cond().
  #
  # 1. The update op is created inside a `with ops.colocate(var):` block
  #
  # 2. Some tensor `data` is captured and a switch is created in a
  #    `with ops.colocate_with(data):` block.
  #
  # with ops.colocate_with(var):
  #  with ops.colocate_with(data):
  #    op = ...
  #
  # var and data may be pinned to different devices, so we want to ops
  # created within ops.colocate_with(data) to ignore the existing stack.
  with ops.colocate_with(data, ignore_existing=True):
    if isinstance(data, ops.Tensor):
      if data.dtype._is_ref_dtype:  # pylint: disable=protected-access
        return ref_switch(data, pred, name=name)
    return switch(data, pred, name=name) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def _Enter(data, frame_name, is_constant=False, parallel_iterations=10,
           use_ref=True, use_input_shape=True, name=None):
  """Creates or finds a child frame, and makes `data` available to it.

  The unique `frame_name` is used by the `Executor` to identify frames. If
  `is_constant` is true, `data` is a constant in the child frame; otherwise
  it may be changed in the child frame. At most `parallel_iterations`
  iterations are run in parallel in the child frame.

  Args:
    data: The tensor to be made available to the child frame.
    frame_name: The name of the child frame.
    is_constant: If true, the output is constant within the child frame.
    parallel_iterations: The number of iterations allowed to run in parallel.
    use_ref: If true, use ref_enter if data is of ref type.
    name: A name for this operation (optional).

  Returns:
    The same tensor as `data`.
  """
  data = ops.convert_to_tensor_or_indexed_slices(data, as_ref=True)
  if isinstance(data, ops.Tensor):
    if data.dtype.is_ref_dtype and use_ref:
      result = ref_enter(data, frame_name, is_constant, parallel_iterations,
                         name=name)
    else:
      result = enter(data, frame_name, is_constant, parallel_iterations,
                     name=name)
    if use_input_shape:
      result.set_shape(data.get_shape())
    return result
  else:
    if not isinstance(data, (ops.IndexedSlices, sparse_tensor.SparseTensor)):
      raise TypeError("Type %s not supported" % type(data))
    values = _Enter(data.values, frame_name, is_constant,
                    parallel_iterations=parallel_iterations,
                    use_input_shape=use_input_shape, name=name)
    indices = enter(data.indices, frame_name, is_constant,
                    parallel_iterations, name="indices")
    if use_input_shape:
      indices.set_shape(data.indices.get_shape())
    if isinstance(data, ops.IndexedSlices):
      dense_shape = data.dense_shape
      if dense_shape is not None:
        dense_shape = enter(dense_shape, frame_name, is_constant,
                            parallel_iterations, name="dense_shape")
        if use_input_shape:
          dense_shape.set_shape(data.dense_shape.get_shape())
      return ops.IndexedSlices(values, indices, dense_shape)
    else:
      dense_shape = enter(data.shape, frame_name, is_constant,
                          parallel_iterations, name="dense_shape")
      if use_input_shape:
        dense_shape.set_shape(data.shape.get_shape())
      return sparse_tensor.SparseTensor(indices, values, dense_shape) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def switch(data, pred, dtype=None, name=None):
  """Forwards `data` to an output determined by `pred`.

  If `pred` is true, the `data` input is forwared to the first output.
  Otherwise, the data goes to the second output.

  This op handles `Tensor`s and `IndexedSlices`.

  Args:
    data: The tensor to be forwarded to the appropriate output.
    pred: A scalar that specifies which output port will receive data.
    dtype: Optional element type for the returned tensor. If missing,
           the type is inferred from the type of `value`.
    name: A name for this operation (optional).

  Returns:
    `(output_false, output_true)`: If `pred` is true, data will be forwarded
    to `output_true`, otherwise it goes to `output_false`.
  """
  with ops.name_scope(name, "Switch", [data, pred]) as name:
    data = ops.convert_to_tensor_or_indexed_slices(data, dtype=dtype,
                                                   name="data", as_ref=True)
    pred = ops.convert_to_tensor(pred, name="pred")
    if isinstance(data, ops.Tensor):
      return gen_control_flow_ops._switch(data, pred, name=name)
    else:
      if not isinstance(data, (ops.IndexedSlices, sparse_tensor.SparseTensor)):
        raise TypeError("Type %s not supported" % type(data))
      val, ind = data.values, data.indices
      val_f, val_t = gen_control_flow_ops._switch(val, pred, name=name)
      ind_f, ind_t = gen_control_flow_ops._switch(ind, pred, name="indices")
      if isinstance(data, ops.IndexedSlices):
        dense_shape = data.dense_shape
        if dense_shape is not None:
          dense_shape_f, dense_shape_t = gen_control_flow_ops._switch(
              dense_shape, pred, name="dense_shape")
        else:
          dense_shape_f, dense_shape_t = None, None
        return (ops.IndexedSlices(val_f, ind_f, dense_shape_f),
                ops.IndexedSlices(val_t, ind_t, dense_shape_t))
      else:
        dense_shape = data.shape
        dense_shape_f, dense_shape_t = gen_control_flow_ops._switch(
            data.shape, pred, name="dense_shape")
        return (sparse_tensor.SparseTensor(ind_f, val_f, dense_shape_f),
                sparse_tensor.SparseTensor(ind_t, val_t, dense_shape_t)) 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def merge(inputs, name=None):
  """Returns the value of an available element of `inputs`.

  This op tests each of the tensors in `inputs` in turn to determine if any of
  them is available. If it finds an available tensor, it returns it and its
  index in `inputs`.

  It is an error if more than one tensor in `inputs` is available. If no tensor
  in `inputs` is available, the returned tensor and index are not set.

  This op handles both `Tensor`s and `IndexedSlices`. If inputs has a mix of
  `Tensor`s and `IndexedSlices`, all inputs are converted to IndexedSlices
  before merging.

  Args:
    inputs: The input tensors, at most one of which is available.
    name: A name for this operation (optional).

  Returns:
    A tuple containing the chosen input tensor and its index in `inputs`.

  Raises:
    ValueError: If any of the inputs is None, or inputs are IndexedSlices and
      some but not all have a dense_shape property.
  """
  if any([inp is None for inp in inputs]):
    raise ValueError("At least one of the merge inputs is None: %s" % inputs)
  with ops.name_scope(name, "Merge", inputs) as name:
    inputs = [ops.convert_to_tensor_or_indexed_slices(inp, as_ref=True)
              for inp in inputs]
    if all([isinstance(v, ops.Tensor) for v in inputs]):
      if all([v.dtype.is_ref_dtype for v in inputs]):
        return gen_control_flow_ops._ref_merge(inputs, name)
      else:
        return gen_control_flow_ops._merge(inputs, name)
    elif all([isinstance(v, sparse_tensor.SparseTensor) for v in inputs]):
      # Only handle the case when all inputs are SparseTensor.
      values, _ = merge([inp.values for inp in inputs], name=name)
      indices, chosen_index = gen_control_flow_ops._merge(
          [inp.indices for inp in inputs], name="indices")
      dense_shape, _ = gen_control_flow_ops._merge(
          [inp.shape for inp in inputs], name="dense_shape")
      return (sparse_tensor.SparseTensor(indices, values, dense_shape),
              chosen_index)
    else:
      # For now convert all the inputs as IndexedSlices.
      inputs = math_ops._as_indexed_slices_list(inputs, optimize=False)
      values, _ = merge([inp.values for inp in inputs], name=name)
      indices, chosen_index = gen_control_flow_ops._merge(
          [inp.indices for inp in inputs], name="indices")
      if any(inp.dense_shape is not None for inp in inputs):
        if any(inp.dense_shape is None for inp in inputs):
          raise ValueError("Either all merged IndexedSlices must have a "
                           "dense_shape, or none must have a dense_shape.")
        dense_shape, _ = gen_control_flow_ops._merge(
            [inp.dense_shape for inp in inputs], name="dense_shape")
      else:
        dense_shape = None
      return ops.IndexedSlices(values, indices, dense_shape), chosen_index
# pylint: enable=protected-access 

# 需要导入模块: from tensorflow.python.framework import ops [as 别名]
# 或者: from tensorflow.python.framework.ops import convert_to_tensor_or_indexed_slices [as 别名]
def _SwitchRefOrTensor(data, pred, name="Switch"):
  """Forwards `data` to an output determined by `pred`.

  If `pred` is false, the `data` input is forwarded to the first output.
  Otherwise, the data goes to the second output.

  This op handles `Tensor`s and `IndexedSlices`.

  Args:
    data: The tensor to be forwarded to the appropriate output.
    pred: A scalar that specifies which output port will receive data.
    name: A name for this operation (optional).

  Returns:
    `(output_false, output_true)`: If `pred` is true, data will be forwarded to
    `output_true`, otherwise it goes to `output_false`.

  Raises:
    TypeError: if data is not a Tensor or IndexedSlices
  """
  data = ops.convert_to_tensor_or_indexed_slices(data, name="data")
  # NOTE(vrv): ops.colocate_with(data, ignore_existing=True) below
  # addresses the following scenario.
  #
  # Assume you execute Optimizer.apply_gradients() in a branch of a cond().
  #
  # 1. The update op is created inside a `with ops.colocate(var):` block
  #
  # 2. Some tensor `data` is captured and a switch is created in a
  #    `with ops.colocate_with(data):` block.
  #
  # with ops.colocate_with(var):
  #  with ops.colocate_with(data):
  #    op = ...
  #
  # var and data may be pinned to different devices, so we want to ops
  # created within ops.colocate_with(data) to ignore the existing stack.
  with ops.colocate_with(data, ignore_existing=True):
    if isinstance(data, ops.Tensor):
      if data.dtype._is_ref_dtype:  # pylint: disable=protected-access
        return ref_switch(data, pred, name=name)
    return switch(data, pred, name=name)