if (input_names.size() != main_func.getNumArguments() ||
      output_names.size() != main_func.getNumResults()) {
    module_op.emitError()
        << "Number of inputs and outputs in the tf.entry_function attribute "
           "mismatched. [Input] Expected: "
        << input_names.size() << ", got: " << main_func.getNumArguments()
        << ". [Output] Expected: " << output_names.size()

  if (!input_shapes_attr) return success();

  if (input_shapes_attr.size() != func_op.getNumArguments()) {
    return func_op->emitError(
               "Number of arguments and 'tf._input_shapes' "
               "attribute size do not match. ")
           << "Num args: " << func_op.getNumArguments()
           << ", tf._input_shapes size: " << input_shapes_attr.size();
  }

  return success();
}

  if (!return_op)
    return function.emitError() << "expects function '" << function.getName()
                                << "' to have a MLIR ReturnOp";

  llvm::SmallVector<ResourceInfo, 4> resources(function.getNumArguments());
  auto argument_types =
      llvm::to_vector<4>(function.getFunctionType().getInputs());
  bool has_resources = false;
  auto add_resource_argument = [&](BlockArgument arg,

  fused_lstm_func_.dump();

  // verify transpose
  EXPECT_EQ(
      fused_lstm_func_->getAttrOfType<StringAttr>(kTFImplements).getValue(),
      convert.GetCompositeOpName());
  EXPECT_EQ(fused_lstm_func_.getNumArguments(), 5);
  EXPECT_EQ(fused_lstm_func_.getFunctionType().getNumResults(), 1);

  auto transpose_op = fused_lstm_func_.getBody().front().begin();
  transpose_op++;

  // passing the platform index.
  FunctionType function_type = func_op.getFunctionType();
  ArrayRef<Type> new_input_types =
      function_type.getInputs().take_back(func_op.getNumArguments() - 1);
  func_op.setFunctionType(
      FunctionType::get(ctx, new_input_types, function_type.getResults()));
  func_op.getBody().eraseArgument(0);
}

}  // namespace

             << init_type;
    }

    init_types.insert(init_type);
  }

  return success();
}

static bool HasAnyTfSavedModelArgAttr(func::FuncOp func) {
  for (int i = 0, e = func.getNumArguments(); i < e; i++) {
    if (func.getArgAttr(i, kTfSavedModelIndexPathAttr) ||
        func.getArgAttr(i, "tf_saved_model.bound_input")) {
      return true;
    }
  }

// scatter like ops.
template <>
LogicalResult MatchBinaryReduceFunction<void>(mlir::Region& function) {
  Block& body = function.front();
  if (body.getNumArguments() != 2) return failure();

  mhlo::ReturnOp return_op = dyn_cast<mhlo::ReturnOp>(body.back());
  if (!return_op) return failure();
  if (return_op.getNumOperands() != 1) return failure();

        func_.getLoc(), output_type, lookup, value);

    builder.create<mlir::func::ReturnOp>(func_.getLoc(), op.getResult());
  }

  LogicalResult VerifySignature() {
    if (func_.getNumArguments() != 2) {
      return func_.emitWarning()
             << "Invalid number of arguments in the embedding "
                "matmul composite function";
    }
    if (func_.getFunctionType().getNumResults() != 1) {

  // Verify the quantization specification is expected for quantizing the
  // current function.
  bool IsLegalQuantSpecs(func::FuncOp func) {
    if (func.getName() == quant_specs_.target_func) {
      return func.getNumArguments() == quant_specs_.input_ranges.size();
    }
    return true;
  }

  // Get the min and max values from the quantization specification for the

  // current function.
  bool IsLegalQuantSpecs(func::FuncOp func) {
    if (func.getName() == quant_specs_.target_func) {
      return (quant_specs_.disable_set_input_nodes_quantization_params ||
              func.getNumArguments() == quant_specs_.input_ranges.size());
    }
    return true;
  }

  // Get the min and max values from the quantization specification for the