reinterpret_cast<const uint8_t*>(custom_options.data());
    const size_t opt_size = custom_options.size();
    if (!flexbuffers::VerifyBuffer(opt_data, opt_size)) {
      return emitError(loc, "invalid custom options");
    }

    const flexbuffers::Vector& v =
        flexbuffers::GetRoot(opt_data, opt_size).AsVector();

    op_name = v[0].AsString().str();

// RUN: tf-opt -tf-optimize %s -o %t && FileCheck %s < %t

// CHECK-LABEL: convbiasaddmul
func.func @convbiasaddmul(%arg: tensor<256x32x32x3xf32>) -> tensor<256x8x7x16xf32> {
  %filter = arith.constant dense<2.0> : tensor<3x3x3x16xf32>
  %bias = arith.constant dense<3.0> : tensor<16xf32>
  %value = arith.constant dense<4.0> : tensor<16xf32>

 public:
  OptimizeIntGraph() = default;
  OptimizeIntGraph(const OptimizeIntGraph &) = default;
  void runOnOperation() override;
};

#include "tensorflow/compiler/mlir/quantization/stablehlo/passes/bridge/optimize.inc"

void OptimizeIntGraph::runOnOperation() {
  RewritePatternSet patterns(&getContext());
  populateWithGenerated(patterns);
  auto func = getOperation();

    // the textual format (on the commandline for example).
    return "quant-optimize";
  }
  StringRef getDescription() const final {
    // This is a brief description of the pass.
    return "Applies optimization after quantization";
  }

  void runOnOperation() override;
};

#include "tensorflow/compiler/mlir/quantization/tensorflow/passes/optimize.inc"

void OptimizePass::runOnOperation() {

// RUN: tf-tfrt-opt -optimize-tf-for-tfrt -split-input-file -verify-diagnostics %s | FileCheck %s

// CHECK-LABEL: @fold_device_index
func.func @fold_device_index() -> tensor<i32> {
  // CHECK-NOT: tf.DeviceIndex
  // CHECK: tf.Const
  // CHECK-SAME: value = dense<1> : tensor<i32>
  %0 = "tf.DeviceIndex"() {device = "/device:CPU:0", device_names = ["GPU", "CPU"]} : () -> tensor<i32>
  func.return %0 : tensor<i32>
}

// -----

// RUN: tf-quant-opt %s -quant-optimize -allow-unregistered-dialect | FileCheck %s

func.func @remove_redundant_cast(%arg0: tensor<1x100x100x1xf32>) -> (tensor<1x96x96x1xf32>) {
  %cst = "tf.Const"() {value = dense<-128> : tensor<i32>} : () -> tensor<i32>
  %cst_0 = "tf.Const"() {value = dense<0.0235294122> : tensor<f32>} : () -> tensor<f32>
  %cst_1 = "tf.Const"() {value = dense<0.00708661414> : tensor<1xf32>} : () -> tensor<1xf32>

#include "tensorflow/compiler/mlir/tensorflow/utils/dynamic_shape_utils.h"

namespace mlir {
namespace TFL {

//===----------------------------------------------------------------------===//
// The actual Optimize Pass.
namespace {
#define GEN_PASS_DEF_OPTIMIZEPASS
#include "tensorflow/compiler/mlir/lite/transforms/passes.h.inc"

constexpr char kRelu[] = "RELU";
constexpr char kRelu6[] = "RELU6";

// RUN: odml-to-stablehlo-opt %s -split-input-file -mhlo-optimize | FileCheck %s

// CHECK-LABEL: testDotToDotGeneralVectorVector
func.func @testDotToDotGeneralVectorVector(%arg0: tensor<3072xf32>, %arg1: tensor<3072xf32>) -> tensor<f32> {
  %0 = "mhlo.dot"(%arg0, %arg1) : (tensor<3072xf32>, tensor<3072xf32>) -> tensor<f32>
  func.return %0 : tensor<f32>

// CHECK:      %[[RES:.*]] = "mhlo.dot_general"(%arg0, %arg1) <{

// RUN: tf-opt %s -tfl-prepare-quantize -canonicalize -tfl-quantize -canonicalize -tfl-optimize -canonicalize | FileCheck %s

// CHECK-LABEL: fuseMulIntoPerTensorConv2dWithQDQs
func.func @fuseMulIntoPerTensorConv2dWithQDQs(%arg0: tensor<256x32x32x3xf32>) -> tensor<256x8x7x3xf32> {
  %cst = arith.constant dense<1.5> : tensor<3xf32>
  %cst_0 = arith.constant dense<[1.0, 2.0, 3.0]> : tensor<3xf32>
  %w = arith.constant dense<2.0> : tensor<3x3x3x3xf32>

// RUN: stablehlo-quant-opt -optimize-int-graph -split-input-file %s -verify-diagnostics | FileCheck %s

// CHECK-LABEL: func @convolution_add_add
func.func @convolution_add_add(
    %lhs: tensor<?x3x2x1xi8>, %rhs: tensor<2x1x1x1xi8>,
    %zp_offset: tensor<?x2x2x1xi32>, %bias: tensor<1xi32>
  ) -> tensor<?x2x2x1xi32> {
  // CHECK-DAG: %[[conv:.*]] = mhlo.convolution
  // CHECK-DAG: %[[combined:.*]] = chlo.broadcast_add %[[zp_offset:.*]], %[[bias:.*]]