var _ = F2 /* ERROR "impossible type assertion: F2().(*X2)\n\t*X2 does not implement T2 (missing method M)" */ ().(*X2)

type X2 struct{}

func _() {
	switch F2().(type) {
	case * /* ERROR "impossible type switch case: *X2\n\tF2() (value of type T2) cannot have dynamic type *X2 (missing method M)" */ X2:
	}

namespace experimental {
namespace cc {

// FunctionMetadata stores additional function information, including
// optional signaturedef feeds/fetches (for TF1-based ConcreteFunctions),
// a valid function path (for TF2-based ConcreteFunctions), and
// the types + number of inputs and outputs.
class FunctionMetadata final {
  // TODO(bmzhao): Add getters here as necessary.
 private:
  friend class ConcreteFunction;

  // to populate `MetaInfoDef`s `function_aliases` when the quantized model is
  // exported to the saved model. This field is usually only populated for the
  // TF2 models.
  map<string, string> function_aliases = 6;

  // Holds information about the asset files used for the model. It essentially
  // associates asset file names with the tensors to which the asset file names

}

void CreateReplicatedClusteringPipeline(OpPassManager &pm,
                                        llvm::StringRef module_name) {
  // Since the internal bridge clustering passes are shared among TF1/TF2
  // TF2-only passes should go here. However, this should be very rare and
  // new passes generally should go into the internal
  // AddReplicatedBridgeClusteringPipelinePasses.

# Tensorflow C SavedModel API

## Overview

These are the new experimental C SavedModel APIs for loading and running
SavedModels in a TF2-idiomatic fashion. See
[RFC 207](https://github.com/tensorflow/community/pull/207) for additional
context.

The directory structure is as follows:

```none
saved_model/

  public/

  internal/

  core/

```

## saved_model/public

namespace tensorflow {
namespace internal {

// Executes a VarHandleOp using `ctx`, and fills `handle` with the DT_RESOURCE
// TensorHandle associated with the variable. This is equivalent to creating an
// unitialized TF2 tf.Variable.
// https://github.com/tensorflow/tensorflow/blob/516608035f85cec8b126712b0ff8407220206b22/tensorflow/python/ops/resource_variable_ops.py#L1867-L1872
Status CreateUninitializedResourceVariable(ImmediateExecutionContext* ctx,

  return Status();
}

Status FlatTensorFunction::MakeCallOp(
    absl::Span<AbstractTensorHandle* const> inputs, ImmediateOpPtr* out) const {
  out->reset(ctx_->CreateOperation());
  // In eager mode, TF2 python executes functions by constructing an op with
  // the name of the functiondef:
  // https://github.com/tensorflow/tensorflow/blob/66668ec0ca432e2f38a575b814f45b6d299d01ed/tensorflow/python/eager/function.py#L545

#include "tensorflow/core/framework/function.pb.h"
#include "tensorflow/core/protobuf/saved_object_graph.pb.h"

namespace tensorflow {

// FlatTensorFunction models a TF2 eager runtime view of a callable function,
// taking + returning flat lists of tensors, including any captures.
// Effectively, it is a thin wrapper around a FunctionDef owned by the

// enabled or disabled, it will decide whether or not to run the bridge.
//
// The config_proto param is a required input for all TF1 graphs but it is
// redundant for TF2 graphs.
// If getting rollout policy involves graph analysis, `record_stats` is used
// to decide whether to emit metrics on unsupported features of the graph.
MlirBridgeRolloutPolicy GetMlirBridgeRolloutPolicy(

"""Serves as a common "main" function for all the SavedModel tests.

There is a fair amount of setup needed to initialize tensorflow and get it
into a proper TF2 execution mode. This hides that boilerplate.
"""

import tempfile
from absl import app
from absl import flags
from absl import logging
import tensorflow.compat.v1 as tf