ModuleOp module_op = op->getParentOfType<ModuleOp>();

    // Ignore ops without quantization method.
    // Consider adding checks for individual methods.
    if (!op->getAttr(kQuantizationMethodAttr)) return failure();

    // Ignore unquantized ops.
    if (!IsQuantizedXlaCallModuleOp(op)) return failure();

    // For weight-only quantization, op should be hybrid quantized.

      OperationState new_state(candidate_op->getLoc(),
                               candidate_op->getName().getStringRef(), inputs,
                               output_types, candidate_op->getAttrs());
      for (int i = 0; i < candidate_op->getNumRegions(); ++i) {
        new_state.addRegion();
      }
      Operation* quantized_op = rewriter.create(new_state);
      if (candidate_op->getNumRegions() != 0) {

        encoding.getBounds()[i] == ShapedType::kDynamic) {
      return false;
    }
  }
  return true;
}

bool HasSymbolRefAttr(Operation* op) {
  for (const auto& attr : op->getAttrs()) {
    Attribute attr_value = attr.getValue();
    if (mlir::isa<SymbolRefAttr>(attr_value)) {
      return true;
    } else if (auto array_attr = mlir::dyn_cast<ArrayAttr>(attr_value)) {

 private:
  LogicalResult matchAndRewrite(TF::PartitionedCallOp call_op,
                                PatternRewriter& rewriter) const override {
    StringRef function_name =
        mlir::cast<FlatSymbolRefAttr>(call_op.getFAttr()).getValue();
    if (!function_name.starts_with("composite_") ||
        !call_op->hasAttr(kQuantTraitAttrName)) {
      return failure();
    }

    absl::Status check_status;

      }
    }
    OpBuilder builder(call);
    auto new_call = builder.create<CallOp>(
        call.getLoc(), info.decomposed_callee.getFunctionType().getResults(),
        new_operands, call->getAttrs());
    new_call->setAttr(
        "f", SymbolRefAttr::get(
                 builder.getContext(),
                 const_cast<func::FuncOp&>(info.decomposed_callee).getName()));
    for (const auto& entry : info.ret_forward_input) {

  new_call_op->setAttr(kEntryFuncAttrName,
                       rewriter.getStringAttr(new_ref_func_name.getValue()));
  new_call_op->setAttrs(call_op->getAttrs());
  new_call_op->removeAttr(rewriter.getStringAttr(kQuantTraitAttrName));

  FlatSymbolRefAttr new_func_name_attr =
      FlatSymbolRefAttr::get(rewriter.getContext(), new_ref_func_name);
  new_call_op->setAttr(kEntryFuncAttrName, new_func_name_attr);

      new_result_types.push_back(while_op.getResultTypes()[i]);
    }
  }

  auto new_while_op = OpBuilder(while_op).create<WhileOp>(
      while_op.getLoc(), new_result_types, new_operands, while_op->getAttrs());
  new_while_op.getCond().takeBody(while_op.getCond());
  new_while_op.getBody().takeBody(while_op.getBody());

  for (auto i = 0; i < n; ++i) {
    if (!while_op.getResult(i).use_empty()) {

    return failure();
  }

  // Collect all the undefined attributes used in the inputs.
  llvm::SmallVector<StringAttr, 4> undefined_attrs;
  for (auto attr : input_used_attrs) {
    if (!func->getAttr(attr.getValue())) {
      undefined_attrs.push_back(attr);
    }
  }

  // Verify the argument order: tensors, tensor list, attributes; and also
  // verify there is at most one tensor list argument.

        new_operands.push_back(build_list_op.getOut());
      } else {
        auto attr_name = compose_func.getArgAttrOfType<StringAttr>(
            arg.index(), kAttrArgumentNameAttr);
        auto attribute = op->getAttr(attr_name.getValue());
        if (!attribute) {
          attribute =
              compose_func.getArgAttr(arg.index(), kAttrArgumentDefaultAttr);
        }

                                PatternRewriter& rewriter) const override {
    auto input_op = op.getInput().getDefiningOp();
    if (auto q = llvm::dyn_cast_or_null<QuantizeOp>(input_op)) {
      if (!q->getAttr(mlir::quant::kVolatileOpAttrName)) return failure();

      if (remove_volatile_ops_type == kPreserveInputsAndOutputs) {
        // Don't remove leading and trailing QDQ for PTQ workflow, so the io