VariableInfo::VariableInfo(
    int index, absl::string_view name, Var* var,
    const std::optional<ManagedStackTrace>& definition_stack_trace)
    : index_(index),
      name_(name),
      var_(var),
      definition_stack_trace_(definition_stack_trace) {}

VariableInfo::VariableInfo(VariableInfo&& other)
    : index_(other.index_),
      var_(other.var_),
      definition_stack_trace_(other.definition_stack_trace_),

      grads = gt.gradient(y, vars_)  # uses registered gradient function.

    # compute with composition
    with backprop.GradientTape() as gt:
      for var_ in vars_:
        gt.watch(var_)
      re_y = compute_composite(**kwargs)  # uses composite function.
      re_grads = gt.gradient(re_y, vars_)  # uses gradients compposite function.

    for v, re_v in zip(y, re_y):
      self.assertAllClose(v, re_v)

  int index() const { return index_; }

  // A pointer to the resource variable.  May be null if this VariableInfo is
  // "empty", i.e. it does not track a resource variable.
  Var* var() const { return var_; }

  // Returns the variable name.
  absl::string_view name() const { return name_; }

  // Returns true if the resource variable lock was successfully acquired by
  // this thread.

    func.return %4 : tensor<100x50xf32>
  }
  // CHECK: "tf.VarHandleOp"
  // CHECK-SAME: shared_name = "var1"
  // CHECK-SAME: _is_initialized = true
  // CHECK: "tf.VarHandleOp"
  // CHECK-SAME: shared_name = "var2"
  // CHECK-SAME: _is_initialized = true
  // CHECK: "tf.VarHandleOp"

  // Create 2 resources and initialize them with dummy values.
  TF_CHECK_OK(device->resource_manager()->Create(
      container, "var1", new tensorflow::Var(tensorflow::DataType::DT_FLOAT)));
  TF_CHECK_OK(device->resource_manager()->Create(
      container, "var2", new tensorflow::Var(tensorflow::DataType::DT_FLOAT)));
}

Status FakeSession::Create(const tensorflow::GraphDef& graph) {

    - five
    - six
    - seven
    - eight
    - nine
    env:
    - name: VAR_3
      valueFrom:
        secretKeyRef:
          key: some-other-key
          name: some-oher-name
    - name: VAR_2
      valueFrom:
        secretKeyRef:
          key: other-key
          name: other-name
    - name: VAR_1
      valueFrom:
        secretKeyRef:
          key: some-key
          name: some-name

    - five
    - six
    - seven
    - eight
    - nine
    env:
    - name: VAR_3
      valueFrom:
        secretKeyRef:
          key: some-other-key
          name: some-oher-name
    - name: VAR_2
      valueFrom:
        secretKeyRef:
          key: other-key
          name: other-name
    - name: VAR_1
      valueFrom:
        secretKeyRef:
          key: some-key
          name: some-name

    %var_0 = "tf.VarHandleOp"() {shared_name = "var_0"} : () -> tensor<!tf_type.resource<tensor<2xf32>>>
    "tf.AssignVariableOp"(%var_0, %cst_0) : (tensor<!tf_type.resource<tensor<2xf32>>>, tensor<2xf32>) -> ()
    %cst_1 = "tf.Const"() {value = dense<2.000000e+00> : tensor<3xf32>} : () -> tensor<3xf32>
    %var_1 = "tf.VarHandleOp"() {shared_name = "var_1"} : () -> tensor<!tf_type.resource<tensor<3xf32>>>

    // CHECK: %[[VAR0:.*]] = "tf.VarHandleOp"()
    %0 = "tf.VarHandleOp"() {container = "", shared_name = "var0"} : () -> tensor<!tf_type.resource<tensor<!tf_type.string>>>

    // CHECK: "tf.AssignVariableOp"(%[[VAR0]], %[[ASSET0]])
    "tf.AssignVariableOp"(%0, %arg0) : (tensor<!tf_type.resource<tensor<!tf_type.string>>>, tensor<!tf_type.string>) -> ()

    // CHECK: %[[VAR1:.*]] = "tf.VarHandleOp"()

    %var_0 = "tf.VarHandleOp"() {shared_name = "var_0"} : () -> tensor<!tf_type.resource<tensor<2xf32>>>
    "tf.AssignVariableOp"(%var_0, %cst_0) : (tensor<!tf_type.resource<tensor<2xf32>>>, tensor<2xf32>) -> ()
    %add_0 = "tf.Identity"(%cst_0) : (tensor<2xf32>) -> tensor<2xf32>
    %var_1 = "tf.VarHandleOp"() {shared_name = "var_1"} : () -> tensor<!tf_type.resource<tensor<2xf32>>>