result_hash, tensorflow::Fingerprint64(concrete_function_serialized));
  }
  // TODO(b/241294832): Complete canonicalization of `object_graph_def.nodes`.
  return result_hash;
}

// Creates a FingerprintDef proto from a SavedModel and the checkpoint meta file
// (.index) in `export_dir`.
absl::StatusOr<FingerprintDef> CreateFingerprintDefPb(
    absl::string_view export_dir, std::string pb_file) {

  nodes_[inst] = node;
  return absl::OkStatus();
}

bool IsEntryFunctionArg(BlockArgument arg) {
  return arg.getParentRegion()->getParentOfType<FuncOp>().getName() == "main";
}

// Creates argument nodes from Block argument. If a name is supplied, that
// name will be used instead of generating a unique name.
Status Exporter::AddArgumentNode(BlockArgument arg, unsigned index,

  nodes_[inst] = node;
  return absl::OkStatus();
}

bool IsEntryFunctionArg(BlockArgument arg) {
  return arg.getParentRegion()->getParentOfType<FuncOp>().getName() == "main";
}

// Creates argument nodes from Block argument. If a name is supplied, that
// name will be used instead of generating a unique name.
Status Exporter::AddArgumentNode(BlockArgument arg, unsigned index,

            /*quant_dim=*/-1, num_bits, narrow_range, is_signed);
        builder.setInsertionPoint(block, insertion_point);
        auto q_op = builder.create<quantfork::QuantizeCastOp>(
            loc, params.getValue(), arg);
        auto dq_op = builder.create<quantfork::DequantizeCastOp>(
            loc, input_type, q_op.getResult());
        arg.replaceAllUsesWith(dq_op.getResult());
        q_op.setOperand(arg);
      }

        // Keep track of IDs of resources that are allocated by ops with
        // `UniqueResourceAllocation` trait, this can be utilized for while-loop
        // parallelization (every iteration creates a new unique resource).
        if (op->hasTrait<OpTrait::TF::UniqueResourceAllocation>()) {
          unique_resource_allocation_ids_.insert(resource_handle.id);
        }
      }

  EXPECT_TRUE(xla::LiteralTestUtil::Equal(
      *literal1, xla::LiteralUtil::CreateR2<int32_t>({{1, 2, 3}})));

  std::shared_ptr<xla::Literal> literal2 = *exec_args[1]->ToLiteralSync();
  EXPECT_TRUE(xla::LiteralTestUtil::Equal(
      *literal2, xla::LiteralUtil::CreateR2<int32_t>({{4, 5, 6}})));
}

}  // namespace
}  // namespace tensorflow

#endif  // PLATFORM_GOOGLE

      if (elem_type && elem_type.hasStaticShape()) return elem_type;
    }
  }
  return std::nullopt;
}

// Creates a ReadVariableOp on a local variable.
Value ReadLocalVariable(Value local_var, OpBuilder builder, Location loc) {
  return builder
      .create<TF::ReadVariableOp>(
          loc,
          ArrayRef<Type>{mlir::cast<TF::ResourceType>(

    const llvm::StringMap<Attribute>& derived_attrs) const {
  // Create the new op
  Location loc = call_op.getLoc();
  rewriter.setInsertionPointAfter(call_op);
  std::string tf_op_name = GetTFOpName(call_op.getCallee());
  OperationState new_state(loc, tf_op_name, inputs, output_types, attr_list);
  Operation* new_op = rewriter.create(new_state);
  if (materialize_derived_attrs_) {

  auto launch = builder->create<tf_device::LaunchOp>(
      before_op->getLoc(), builder->getStringAttr(host_device),
      launch_result_types);
  launch.getBody().push_back(launch_block);

  builder->setInsertionPointToEnd(&launch.GetBody());
  builder->create<tf_device::ReturnOp>(before_op->getLoc(), launch_results);

  return launch;
}

// Creates a new_launch after `before_op` with `ops_to_move` from

  // supported.
  const auto& device_list =
      devices.find(tensorflow::GetDeviceAliasForLogicalCore(0))->getSecond();

  // Create the state variable for each device.
  for (llvm::StringRef device : device_list) {
    state_vars.push_back(builder->create<TF::VarHandleOp>(
        loc,
        llvm::ArrayRef<Type>{RankedTensorType::get(
            {}, TF::ResourceType::get(llvm::ArrayRef<TensorType>{key_type},