#include "tensorflow/core/util/tensor_slice_reader.h"

namespace tensorflow {
namespace checkpoint {

class TensorSliceReader;

// A wrapper around BundleReader (for V2 checkpoints) and
// checkpoint::TensorSliceReader (for V1), that is more easily SWIG wrapped for
// other languages.
//
// The class currently only interacts with single-slice (i.e., non-partitioned)
// variables.
class CheckpointReader {
 public:

using Model = std::function<Status(AbstractContext*,
                                   absl::Span<AbstractTensorHandle* const>,
                                   absl::Span<AbstractTensorHandle*>)>;

// Runs `model` maybe wrapped in a function call op. This can be thought as
// being equivalent to the following python code.
//
// if use_function:
//   outputs = tf.function(model)(inputs)
// else:
//   outputs = model(inputs)

      // and hence expect the input to be forwarded to the return tensor.
      EXPECT_EQ(orig_ptr, TF_TensorData(t));
    } else {
      // In sync mode, forwarding can't really happen since the client code will
      // have a reference count on the input tensor while the kernel is being
      // executed and thus it cannot be re-used for the return tensor.
      EXPECT_NE(orig_ptr, TF_TensorData(t));

# Make sure code inside the TensorFlow codebase can use tf2.enabled() at import.
_os.environ["TF2_BEHAVIOR"] = "1"
from tensorflow.python import tf2 as _tf2
_tf2.enable()

# API IMPORTS PLACEHOLDER

# WRAPPER_PLACEHOLDER

# Make sure directory containing top level submodules is in
# the __path__ so that "from tensorflow.foo import bar" works.
# We're using bitwise, but there's nothing special about that.

from tensorflow.python.util.lazy_loader import LazyLoader as _LazyLoader
from tensorflow.python.util.lazy_loader import KerasLazyLoader as _KerasLazyLoader

# API IMPORTS PLACEHOLDER

# WRAPPER_PLACEHOLDER

if "dev" in __version__:   # pylint: disable=undefined-variable
  _logging.warning("""

  TensorFlow's `tf-nightly` package will soon be updated to TensorFlow 2.0.

  opts.prefix = prefix;

  for (int i = 0; i < dst_inputs.size(); ++i) {
    opts.input_map[ToTensorId(src_inputs[i])] =
        TensorId(dst_inputs[i].node()->name(), dst_inputs[i].index());
  }
  opts.skip_mapped_nodes = true;

  for (const tensorflow::Operation& op : control_deps) {
    opts.control_dependencies.push_back(op.node()->name());
  }

  for (int i = 0; i < nreturn_nodes; ++i) {

// which is passed to the functions referenced in the TFE_CustomDevice struct
// `device` (execute, delete_device, etc.). It can for example contain the
// names of wrapped devices.
//
// There are currently no graph semantics implemented for registered custom
// devices, so executing tf.functions which contain operations placed on the
// custom devices will fail.
//

            * `FreeBufferHandle`

* `tf.train.CheckpointOptions` and `tf.saved_model.SaveOptions`
    * These now take in a new argument called `experimental_sharding_callback`.
      This is a callback function wrapper that will be executed to determine how
      tensors will be split into shards when the saver writes the checkpoint
      shards to disk. `tf.train.experimental.ShardByTaskPolicy` is the default

  AbstractFunctionKind getKind() const { return kind_; }

  // Returns the AbstractFunction as a FunctionDef.
  virtual Status GetFunctionDef(const FunctionDef**) = 0;

  // Returns a shared reference to the wrapped function.
  virtual absl::StatusOr<core::RefCountPtr<FunctionRecord>>
  GetFunctionRecord() = 0;

 private:
  const AbstractFunctionKind kind_;
};

using AbstractFunctionPtr =

      description: If compiling from source
  - type: textarea
    id: what-happened
    attributes:
      label: Current behavior?
      description: Also tell us, what did you expect to happen?
      placeholder: Tell us what you see!
    validations:
      required: true
  - type: textarea
    id: code-to-reproduce
    attributes:
      label: Standalone code to reproduce the issue