for (int i : llvm::seq<int>(0, metadata.result_devices.size())) {
      func_op.setResultAttr(
          i, kTFDeviceAttr,
          StringAttr::get(context, metadata.result_devices[i]));
    }

    func_op->setAttr(kHostAttr, StringAttr::get(context, host));
    func_op.setPublic();
    Block *block = func_op.addEntryBlock();

    // Clones and moves the operations into the function's body. And the cloned

                                     ModuleOp& module) {
  const auto ex_attr_value = module->getAttrOfType<IntegerAttr>(attr_name);
  if (ex_attr_value == nullptr) {
    module->setAttr(attr_name, attr_value);
    return success();
  }
  if (ex_attr_value == attr_value) {
    return success();
  }
  return op->emitOpError() << "module already contains an attribute "

  // LINT.ThenChange(../../../../core/common_runtime/inline_function_utils.cc)

  for (Block &block : inlinedBlocks) {
    block.walk([&](Operation *op) {
      if (op->getDialect() == call->getDialect())
        op->setAttr(device_id, node_device(op));
    });
  }
}

struct TFInlinerInterface : public DialectInlinerInterface {
  using DialectInlinerInterface::DialectInlinerInterface;

  // Check that there are no duplicated exported_names.
  DenseMap<StringRef, Operation *> exported_name_to_op;
  for (auto &op : module) {
    auto attr = op.getAttr(exported_names_ident);
    if (!attr) continue;
    // If this verifier is called before we verify the
    // 'tf_saved_model.exported_names' attribute, then it might be invalid.

class ConvertEmbeddedLookupFunc {
 public:
  explicit ConvertEmbeddedLookupFunc(func::FuncOp func) : func_(func) {}

  void RewriteFunc() {
    func_->setAttr(kTFImplements,
                   StringAttr::get(func_.getContext(), "embedding_lookup"));
    Value lookup = func_.getArgument(1);
    Value value = func_.getArgument(0);

    if (!preceding_qcast) return failure();

    auto new_qcast = rewriter.create<quantfork::QuantizeCastOp>(
        q_op.getLoc(), q_op.getType(), preceding_qcast.getArg());
    new_qcast->setAttr(kVolatileOpAttrName, rewriter.getUnitAttr());
    q_op->replaceAllUsesWith(new_qcast);
    return success();
  }
};

bool PrepareQuantizePass::ContainsQuantizeOps(func::FuncOp func) {

            isinstance(field_info, params.Param)
            and getattr(field_info, "in_", None) is None
        ):
            field_info.in_ = params.ParamTypes.query
        use_annotation_from_field_info = get_annotation_from_field_info(
            use_annotation,
            field_info,
            param_name,
        )
        if not field_info.alias and getattr(field_info, "convert_underscores", None):

        // tensorflow/lite/kernels/fully_connected.cc. The kernels will handle
        // the asymmetric_quantize_inputs attribute in the builtin option.
        dynamic_range_quantized_user->setAttr(
            kAsymmetricQuantizeInputsAttr,
            BoolAttr::get(rewriter.getContext(), true));
        changed = true;
      }
    }
    return changed;
  }

  xla::OpSharding sharding =
      manual_sharding ? ::xla::sharding_builder::Manual()
                      : ::xla::sharding_builder::AssignDevice(kShardingTpuCore);
  op->setAttr(kShardingAttr,
              StringAttr::get(op->getContext(), sharding.SerializeAsString()));
}

// Assigns frontend attributes holding information about data type and

    for (NamedAttribute attr : op->getAttrs()) {
      auto symref = mlir::dyn_cast<FlatSymbolRefAttr>(attr.getValue());
      if (!symref) continue;
      if (symref.getValue() != func.getName()) continue;
      op->setAttr(attr.getName(),
                  FlatSymbolRefAttr::get(op->getContext(), cloned.getName()));
      break;
    }
  }

  // Traverse results backward so that indices to be deleted stay unchanged.