if (!e->IsControlEdge() && e->src()->IsConstant()) {
      const_input = e->src();
      break;
    }
  }

  if (!const_input) {
    return false;
  }

  const TensorProto* proto = nullptr;
  if (!TryGetNodeAttr(const_input->def(), "value", &proto)) {
    return false;
  }

  return TensorShapeUtils::IsScalar(proto->tensor_shape());
}

                     const GraphDefBuilder::Options& opts) {
  if (opts.HaveError()) return nullptr;
  NodeBuilder node_builder(opts.GetNameForOp("Const"), "Const",
                           opts.op_registry());
  TensorProto value;
  value.set_dtype(dtype);
  for (int dim : shape) {
    value.mutable_tensor_shape()->add_dim()->set_size(dim);
  }
  return opts.WithAttr("value", value)
      .WithAttr("dtype", dtype)

// TF_Version returns a string describing version information of the
// TensorFlow library. TensorFlow uses semantic versioning.
TF_CAPI_EXPORT extern const char* TF_Version(void);

// Parsing a serialized TensorProto into a TF_Tensor.
TF_CAPI_EXPORT extern void TF_TensorFromProto(const TF_Buffer* from,
                                              TF_Tensor* to, TF_Status* status);

      ->set_disable_optimize_for_static_graph(true);
}

void TF_TensorFromProto(const TF_Buffer* from, TF_Tensor* to,
                        TF_Status* status) {
  TF_SetStatus(status, TF_OK, "");
  tensorflow::TensorProto from_tensor_proto;
  status->status = BufferToMessage(from, &from_tensor_proto);
  if (!status->status.ok()) {
    return;
  }
  status->status =

  TF_DeleteTensor(a);
  TF_DeleteTensor(b);
}

TEST(CAPI, TestFromProto) {
  Tensor t_cc(DT_FLOAT, TensorShape({2, 3}));
  t_cc.flat<float>().setConstant(1.0);
  tensorflow::TensorProto t_proto;
  t_cc.AsProtoField(&t_proto);

  TF_Buffer* t_buffer = TF_NewBuffer();
  TF_CHECK_OK(MessageToBuffer(t_proto, t_buffer));

  const int num_bytes = 6 * sizeof(float);
  float* values =

  func.return %0 : tensor<4xui64>
}

// CHECK-LABEL: func @tensor_proto
func.func @tensor_proto() -> tensor<!tf_type.quint8> {
  // tfrt_fallback_async.const_tensor_proto accepts a serialized tensor proto.
  // CHECK: tfrt_fallback_async.const_tensor_proto "\08\0C\12\00\22\01@"

  constexpr absl::string_view kAddOpExpr =
      R"mlir(
      %0 = "tf.Const"() {value = #tf_type<tensor_proto : "0x746674656"> : tensor<1x3x2x2x!tf_type.qint32>} : () -> tensor<1x3x2x2x!tf_type.qint32>
      %1 = "tf.Const"() {value = #tf_type<tensor_proto : "0x746674656"> : tensor<2x!tf_type.qint32>} : () -> tensor<2x!tf_type.qint32>
      %2 = "tf.Const"() {value = dense<1.0> : tensor<f32>} : () -> tensor<f32>

// CHECK: "tf.TensorProtoFloatTensor"() {bar = #tf_type<tensor_proto : "0x68656C6C6F"> : tensor<2x1x4xf32>} : () -> ()
  "tf.TensorProtoFloatTensor"(){bar = #tf_type<tensor_proto : "0x68656C6C6F"> : tensor<2x1x4xf32>} : () -> ()
// CHECK: "tf.TensorProtoStringTensor"() {bar = #tf_type<tensor_proto : "0x68656C6C6F"> : tensor<2x1x4x!tf_type.string>} : () -> ()
  "tf.TensorProtoStringTensor"(){bar = #tf_type<tensor_proto : "0x68656C6C6F"> : tensor<2x1x4x!tf_type.string>} : () -> ()

  %input_scales = "tf.Const"() { value = dense<2.0> : tensor<f32> } : () -> tensor<f32>
  %input_zps = "tf.Const"() { value = dense<4> : tensor<i32> } : () -> tensor<i32>
  // tensor_proto that points to dense<127> of type !tf_type.qint32.

// CHECK-DAG: %[[bias:.*]] = "tf.Const"() <{value = dense<0.000000e+00> : tensor<3xf32>}> : () -> tensor<3xf32>
// CHECK-DAG: %[[q_w1:.*]] = "tf.Const"() <{value = #tf_type<tensor_proto : "0x746674{{.*}}-> tensor<2x3x1x3x!tf_type.qint8>
// CHECK-DAG: %[[q_w2:.*]] = "tf.Const"() <{value = #tf_type<tensor_proto : "0x746674{{.*}}-> tensor<2x3x1x6x!tf_type.qint8>
// CHECK-DAG: %[[w_scale:.*]] = "tf.Const"() <{value = dense<0.0236220472> : tensor<f32>}> : () -> tensor<f32>