// if a special Identity node in the following pattern is clustered in.
    // That is, an Identity node in the loop cond computation is used to drive
    // const nodes consumed by the loop body.  If this Identity node goes into
    // the same cluster with nodes from the loop body, extra dependency is
    // created between the loop cond and body computations and it hinders the

  // It's here so xla::Parameters that are created form block.getArguments will
  // have the proper shapes.
  TF_RETURN_IF_ERROR(RefineShapes(arg_shapes, module_op));

  mlir::func::FuncOp main = module_op.lookupSymbol<mlir::func::FuncOp>("main");
  mlir::Block& block = main.getRegion().front();
  xla::XlaBuilder builder("main");

  // Create xla_params.
  std::vector<xla::XlaOp> xla_params;

    // be categorized as cusom ops despite each of them may require different
    // behaviors. In that case, these ops can be marked in the custom map and
    // treated separately in this pass.

    auto custom_op = llvm::dyn_cast_or_null<CustomOp>(op);
    if (!custom_op) return false;

    // Custom op which is marked in the custom op map is quantizable.

  if (failed(applyPatternsAndFoldGreedily(module_op, std::move(patterns)))) {
    module_op.emitError() << "quant-cast-bf16-ops-to-f32 failed.";
    signalPassFailure();
  }
}

}  // namespace

// Creates an instance of the Cast BF16 ops to F32 pass.
std::unique_ptr<OperationPass<ModuleOp>> CreateCastBf16OpsToF32Pass() {
  return std::make_unique<CastBf16OpsToF32Pass>();
}

      "_xla", "_oc", "cluster", name_attrs, "cluster_rewritten", "host_graph",
      host_compute_core, &fld, &shape_inference_graphs,
      &has_outside_compilation));

  // Check host graph is not created.
  EXPECT_EQ(fld.Find("host_graph"), nullptr);
}

TEST_F(ExtractOutsideCompilationForFunctionTest, OutsideCompilationInIf) {
  // Build the XLA computation func.
  // "const0" (bool)
  // "const1" (int32)

                               "attribute in the quantized composite function.";
      signalPassFailure();
    }
  }
}

}  // namespace

// Creates an instance of the pass for inserting quantized functions.
std::unique_ptr<OperationPass<ModuleOp>> CreateInsertQuantizedFunctionsPass(
    QuantMethod quantization_method, OpSet target_opset) {

    return;
  }

  auto new_func_type =
      builder.getFunctionType(new_input_types, new_output_types);
  func.setType(new_func_type);
}
}  // namespace

// Creates an instance of the TensorFlow Lite modify io nodes pass.
std::unique_ptr<OperationPass<func::FuncOp>> CreateModifyIONodesPass(
    Type input_type, Type output_type) {

    return WalkResult::advance();
  });
  if (result.wasInterrupted()) {
    return failure();
  }

  return builder.getArrayAttr(function_list);
}

// Creates a pruned module containing the XlaCallModule's entry function and
// other functions transitively called by the entry function.
FailureOr<OwningOpRef<ModuleOp>> PruneStablehloModule(

        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) {

  } else {
    builder.setInsertionPointToStart(&block);
  }
  TypeAttr type_attr = TypeAttr::get(new_type);
  auto quantize = builder.create<TFL::QuantizeOp>(value.getLoc(), new_type,
                                                  value, type_attr);
  auto dequantize = builder.create<TFL::DequantizeOp>(
      value.getLoc(), expressed_type, quantize.getOutput());
  value.replaceAllUsesWith(dequantize);