// (`output_tensors`).
  //
  // If provided, a non-null `forward_function` will be used instead of the
  // backward function (`backward_function_getter` /
  // `backward_function_deleter`) to compute jvps for this operation. If
  // `forward_function` is null, a GradientTape is used on the backward function
  // to compute the jvp, which will waste computation when executing eagerly.
  //

}

// Runs `model` maybe wrapped in a function.
Status RunModel(Model model, AbstractContext* ctx,
                absl::Span<AbstractTensorHandle* const> inputs,
                absl::Span<AbstractTensorHandle*> outputs, bool use_function) {
  if (use_function) {
    const char* fn_name = "test_fn";
    core::RefCountPtr<AbstractFunction> scoped_func;
    // Returning null tensors from a tf.function is not supported, so we keep

                               TF_Buffer* buf, TF_Status* status) {
  auto* function_def = tensorflow::unwrap(ctx)->FindFunctionDef(function_name);
  if (function_def == nullptr) {
    status->status = tensorflow::errors::NotFound(
        "Unable to find FunctionDef with name: ", function_name);
    return;
  }
  string str = function_def->SerializeAsString();

        "@com_google_absl//absl/strings",
    ],
)

cc_library(
    name = "graph_function",
    srcs = ["graph_function.cc"],
    hdrs = [
        "graph_function.h",
    ],
    visibility = [
        "//tensorflow:internal",
    ],
    deps = [
        ":abstract_function",
        "//tensorflow/core:framework",
        "//tensorflow/core/platform:refcount",

                                                     const char* function_name,
                                                     TF_Buffer* buf,
                                                     TF_Status* status);

// Get GraphDebugInfo containing stack traces mapping to node names
TF_CAPI_EXPORT extern void TFE_ContextGetGraphDebugInfo(
    TFE_Context* ctx, const char* function_name, TF_Buffer* buf,
    TF_Status* status);

    x.reset(x_raw);
  }

  Status s = RunModel(TestScalarShape, ctx.get(),
                      /*inputs=*/{x.get()},
                      /*outputs=*/{},
                      /*use_function=*/UseFunction());
  ASSERT_EQ(errors::OK, s.code()) << s.message();
}

// Checks that inputs[0] is a matrix with shape 2x4.
Status TestTensorShape2x4(AbstractContext* ctx,

                       /*tfrt*/ ::testing::Values(false),
                       /*use_function*/ ::testing::Values(true, false)));
#else
INSTANTIATE_TEST_SUITE_P(
    UnifiedCAPI, CppGradients,
    ::testing::Combine(::testing::Values("graphdef", "mlir"),
                       /*tfrt*/ ::testing::Values(false),
                       /*use_function*/ ::testing::Values(true, false)));
#endif
}  // namespace
}  // namespace internal

"""
Top-level module of TensorFlow. By convention, we refer to this module as
`tf` instead of `tensorflow`, following the common practice of importing
TensorFlow via the command `import tensorflow as tf`.

The primary function of this module is to import all of the public TensorFlow
interfaces into a single place. The interfaces themselves are located in
sub-modules, as described below.

using tensorflow::tracing::TracingTensorHandle;

void TF_SetTracingImplementation(const char* name, TF_Status* s) {
  tsl::Set_TF_Status_from_Status(s, SetDefaultTracingEngine(name));
}

// Creates a new TensorFlow function, it is an execution context attached to a
// given tracing context.
TF_ExecutionContext* TF_CreateFunction(const char* fn_name, TF_Status* s) {
  return wrap(CreateTracingExecutionContext(fn_name, s));
}

        closed over by local functions. Previously, such variables were
        sometimes incorrectly ignored.
    *   functions returned by the `get_concrete_function` method of
        `tf.function` objects can now be called with arguments consistent with
        the original arguments or type specs passed to `get_concrete_function`.
        This calling convention is now the preferred way to use concrete