backport_manager.cpp

#include <ATen/core/ivalue.h>
#include <c10/util/Exception.h>
#include <caffe2/serialize/file_adapter.h>
#include <caffe2/serialize/inline_container.h>
#include <torch/csrc/jit/mobile/compatibility/backport_manager.h>
#include <torch/csrc/jit/mobile/compatibility/model_compatibility.h>
#include <torch/csrc/jit/mobile/import.h>
#include <torch/csrc/jit/mobile/module.h>
#include <torch/csrc/jit/serialization/export.h>
#include <torch/csrc/jit/serialization/import.h>
#include <torch/csrc/jit/serialization/pickler.h>
#include <cstddef>
#include <sstream>

namespace torch {
namespace jit {

using caffe2::serialize::FileAdapter;
using caffe2::serialize::IStreamAdapter;
using caffe2::serialize::PyTorchStreamReader;
using caffe2::serialize::PyTorchStreamWriter;
using caffe2::serialize::ReadAdapterInterface;

// Current support bytecode version
namespace {
constexpr int64_t kBytecodeVersionV4 = 0x4L;
constexpr int64_t kBytecodeVersionV5 = 0x5L;
constexpr int64_t kBytecodeVersionV6 = 0x6L;
constexpr int64_t kBytecodeVersionV7 = 0x7L;
constexpr int64_t kBytecodeVersionV8 = 0x8L;
} // namespace

/********************** Utility Functions **********************/

// Utility function that can be reused by backport_vn_to_vn-1(). If any utility
// function can be reused by other backport function, move it here.
namespace {
// Copy files from source to destination except the files and dirs
void selective_copy(
    PyTorchStreamReader& reader,
    PyTorchStreamWriter& writer,
    const std::unordered_set<std::string>& excluded_files,
    const std::unordered_set<std::string>& excluded_dirs) {
  auto records = reader.getAllRecords();
  for (const auto& record : records) {
    // Don't copy archive in excluded_files, usually archive `version` and
    // `bytecode`. Archive `version` will be written when PyTorchStreamWriter is
    // going to finalize and run writeEndOfFile()

    // records is the list of all files names in the zip file, and each record
    // is one file with path to parent folder, the example records is:
    // data.pkl
    // code/__torch__/___torch_mangle_5.py
    // code/__torch__/___torch_mangle_5.py.debug_pkl
    // constants/140245072983168.storage
    // constants.pkl
    // bytecode.pkl
    // version
    bool skip = excluded_files.count(record) > 0;

    // Skip dirs, find the last '/' and compare it with record
    for (const auto& excluded_dir : excluded_dirs) {
      std::size_t found = record.find_last_of("/\\");
      auto path = record.substr(0, found);
      if (excluded_dir == path) {
        skip = true;
        break;
      }
    }
    if (!skip) {
      auto data_ptr = reader.getRecord(record);
      auto data = std::get<0>(data_ptr).get();
      auto size = std::get<1>(data_ptr);
      writer.writeRecord(record, data, size);
    }
  }
}

// Copy all content from reader to stringstream
void get_model_stream(PyTorchStreamReader& reader, std::stringstream& out) {
  auto writer_func = [&](const void* buf, size_t nbytes) -> size_t {
    out.write(static_cast<const char*>(buf), nbytes);
    return !out ? 0 : nbytes;
  };
  PyTorchStreamWriter writer(writer_func);

  selective_copy(
      reader,
      writer,
      std::unordered_set<std::string>(),
      std::unordered_set<std::string>());
}

// The write_archive_current function is used for bytecode from version v5 to
// v7 (the latest bytecode version). pre-v5 we serialized things differently.
// This write archive function may change in export_module.cpp, however we don't
// have a way to keep the old export function in the codebase. To be able to
// export the model in old format, we keep a record of the export function here.
void write_archive_current(
    PyTorchStreamWriter& writer,
    const IValue& value,
    const std::string& archive_name,
    const std::string& archive_dir,
    const std::string& tensor_dir,
    bool use_storage_context,
    SerializationStorageContext& storage_context) {
  std::vector<char> data;
  // Vector to capture the run-time class types during pickling the IValues
  std::vector<c10::ClassTypePtr> memoizedClassTypes;
  std::vector<std::string> tensor_names;
  Pickler data_pickle(
      [&](const char* buf, size_t size) {
        data.insert(data.end(), buf, buf + size);
      },
      nullptr,
      nullptr,
      &memoizedClassTypes,
      [&](const at::Tensor& tensor) {
        // returns a string to use in picker.cpp as storage obj key
        if (use_storage_context) {
          std::string string_id =
              std::to_string(reinterpret_cast<std::intptr_t>(
                  tensor.storage().unsafeGetStorageImpl()));
          tensor_names.push_back(string_id + ".storage");
          storage_context.getOrAddStorage(tensor.storage());
        } else {
          tensor_names.push_back(std::to_string(tensor_names.size()));
        }
        return tensor_names.back();
      });
  data_pickle.protocol();
  data_pickle.pushIValue(value);
  data_pickle.stop();
  // write out tensor data
  size_t i = 0;
  std::string prefix = archive_name + "/";

  TORCH_INTERNAL_ASSERT(tensor_names.size() == data_pickle.tensorData().size());
  const std::unordered_set<std::string>& pre_serialized_files =
      writer.getAllWrittenRecords();

  for (const auto& td : data_pickle.tensorData()) {
    WriteableTensorData writable_td = getWriteableTensorData(td);
    std::string fname = tensor_dir + tensor_names[i++];
    if (use_storage_context &&
        pre_serialized_files.find(fname) != pre_serialized_files.end()) {
      // storage has been serialzed already, skip
      continue;
    }
    writer.writeRecord(fname, writable_td.data(), writable_td.sizeInBytes());
  }

  std::string fname = archive_dir + archive_name + ".pkl";
  writer.writeRecord(fname, data.data(), data.size());
}

/*
inputs: 1) bytecode tuple from bytecode.pkl 2) the output bytecode version,
return: A boolean to indicate whether bytecode tuple is updated successfully
*/
bool update_bytecode_version(
    std::vector<at::IValue>& bytecode_values,
    const int64_t to_version) {
  if (!bytecode_values.empty() && bytecode_values[0].isInt()) {
    bytecode_values[0] = c10::IValue(to_version);
    return true;
  }
  return false;
}

/*
inputs: 1) input model stringstream 2) the output bytecode version,
return: model stringstream with updated bytecode version in bytecode.pkl

Example bytecode.pkl:
(${bytecode_version},
  ('__torch__.m.forward',
    (('instructions',
      (('STOREN', 1, 2),
       ('DROPR', 1, 0),
       ('MOVE', 2, 0),
       ('OP', 0, 0),
       ('RET', 0, 0))),
     ('operators', (('aten::Int', 'Tensor'),)),
     ('constants', ()),
     ('types', ()),
     ('register_size', 2))))
*/
std::stringstream update_bytecode_version(
    std::stringstream& input_model,
    const int64_t to_version) {
  PyTorchStreamReader reader_bytecode(&input_model);
  auto constants_values =
      std::move(*readArchive(kArchiveNameConstants, reader_bytecode).toTuple())
          .elements();

  std::vector<IValue> bytecode_values = get_bytecode_ivalues(reader_bytecode);
  std::unordered_set<std::string> excluded_files{
      "constants.pkl", "bytecode.pkl"};

  std::unordered_set<std::string> excluded_dirs{
      "constants",
      "bytecode",
  };

  std::stringstream ouput_model_stream;
  auto writer_func = [&](const void* buf, size_t nbytes) -> size_t {
    ouput_model_stream.write(static_cast<const char*>(buf), nbytes);
    return !ouput_model_stream ? 0 : nbytes;
  };

  PyTorchStreamWriter writer_bytecode(writer_func);

  selective_copy(
      reader_bytecode, writer_bytecode, excluded_files, excluded_dirs);

  update_bytecode_version(bytecode_values, to_version);
  auto bytecode_tuple = c10::ivalue::Tuple::create(std::move(bytecode_values));
  SerializationStorageContext storage_context;
  write_archive_current(
      writer_bytecode,
      c10::ivalue::Tuple::create(std::move(constants_values)),
      /*archive_name=*/"constants",
      /*archive_dir=*/"",
      /*tensor_dir=*/"constants/",
      /*use_storage_context=*/true,
      storage_context);
  write_archive_current(
      writer_bytecode,
      bytecode_tuple,
      /*archive_name=*/"bytecode",
      /*archive_dir=*/"",
      /*tensor_dir=*/"constants/",
      /*use_storage_context=*/true,
      storage_context);

  return ouput_model_stream;
}
} // namespace

/******************** backport_v{i}_to_v{i-1} Functions **********************/

/*
 To add next backport function, for example, backport_vn_to_vn-1, create an
 anonymous namespace with a backport_vn_to_vn-1 function + other necessary
 customized function. If a function can be reused by other backport functions,
 move it to the utility function group. It will be easier to split out
 backport_manager.cpp to smaller files when it grows too long.

 How to add backport_v{i}_to_v{i-1} ?
 There are two options:
 1) [Format change only, recommended] Constrcut a reader with the
 input_model_stream, modify the file, and use PyTorchWriter to write it to
 output_model_stream. See backport_v5_to_v4.

 2) [Both format and content change] ]Use torch.jit.load() to load the stream,
 and save it to output_model_stream.

 The first option is preferred, because it will be purely format change, and
 the model doesn't need to go through inline again and model content will
 remain the same.

 A note for manipulate stringstream, it's recommend to declare a new
 stringstream, tmp_stream, and swap it with the argument output_model_stream
 once it's ready, output_model_stream.swap(tmp_stream). Do not use
 output_model_stream.clear(). It only clears out error state flag
 (https://www.cplusplus.com/reference/ios/ios/clear/), while the content is the
 same. It's cleaner to just declare a new one and swap.

*/

namespace {

/*
The following functions needed for backport model from v5 to v4.
Backport function bytecode v5 that deduplicate constanst table.
Previously, in v4, constant table will be exported twice, in both archive
bytecode and archive constants, and majority (almost all) are duplicates.
Currently, in v5, JIT and mobile will share archive constants, and all
constant tensors will be exported in this archive. The bump was needed
because the v5 bytecode export the tensor storage path in the schema, since
the runtime code is now able to query which archive this tensor is stored at
and query the correct archive.
For example, Previously, in v4, we deserialize tensor as without archive
path, and mobile will always read tensor from bytecode archive:
(torch._utils._rebuild_tensor_v2(pers.obj(('storage', torch.DoubleStorage,
'0', 'cpu', 8),),
   0,
   (2, 4),
   (4, 1),
   False,
   collections.OrderedDict()),
 1)),
 So, if the program defines: torch.add(x, h, out=x)
Currently, in v5, we deserialize the bytecode with the archive path, and
mobile can read tensor from the given path:
(torch._utils._rebuild_tensor_v2(pers.obj(('storage', torch.DoubleStorage,
'constants/0', 'cpu', 8),),
   0,
   (2, 4),
   (4, 1),
   False,
   collections.OrderedDict()),
 1)),
Thus, the backport is necessary such that the runtime can read tensor from
the correct archive.
*/
std::stringstream backport_v5_to_v4(std::stringstream& input_model_stream) {
  // 1) read from archive `bytecode` archive
  PyTorchStreamReader reader(&input_model_stream);
  std::vector<IValue> bytecode_values = get_bytecode_ivalues(reader);
  auto constants_values =
      std::move(*readArchive(kArchiveNameConstants, reader).toTuple())
          .elements();

  // 2) Copy everything to new output, except some specific files and dirs
  // (usually version, bytecode.pkl and bytecode folder are skipped)
  std::unordered_set<std::string> excluded_files{
      "constants.pkl", "bytecode.pkl"};

  std::unordered_set<std::string> excluded_dirs{
      "constants",
      "bytecode",
  };

  std::stringstream ouput_model_stream;
  auto writer_func = [&](const void* buf, size_t nbytes) -> size_t {
    ouput_model_stream.write(static_cast<const char*>(buf), nbytes);
    return !ouput_model_stream ? 0 : nbytes;
  };

  PyTorchStreamWriter writer(writer_func);

  selective_copy(reader, writer, excluded_files, excluded_dirs);

  // 3) write `bytecode` archive
  // Update the bytecode version in bytecode.pkl
  update_bytecode_version(bytecode_values, kBytecodeVersionV4);
  // Construct the list of ivalues to a big tuple
  auto bytecode_tuple = c10::ivalue::Tuple::create(std::move(bytecode_values));

  // The export function to generate bytecode.pkl for version 4. After bytecode
  // version bump, the old export function doesn't exist anymore, so keep a copy
  // here for backport pupose.
  auto writeArchiveV4 = [](PyTorchStreamWriter& writer,
                           const std::string& archive_name,
                           const c10::IValue& value) {
    std::vector<char> data;

    // Vector to capture the run-time class types during pickling the IValues
    std::vector<c10::ClassTypePtr> memoizedClassTypes;
    Pickler data_pickle(
        [&](const char* buf, size_t size) {
          data.insert(data.end(), buf, buf + size);
        },
        nullptr,
        nullptr,
        &memoizedClassTypes);
    data_pickle.protocol();
    data_pickle.pushIValue(value);
    data_pickle.stop();
    size_t i = 0;
    std::string prefix = archive_name + "/";

    for (const auto& td : data_pickle.tensorData()) {
      WriteableTensorData writable_td = getWriteableTensorData(td);
      std::string fname = prefix + c10::to_string(i++);
      writer.writeRecord(fname, writable_td.data(), writable_td.sizeInBytes());
    }
    std::string fname = archive_name + ".pkl";
    writer.writeRecord(fname, data.data(), data.size());
  };

  // write `bytecode` archive
  writeArchiveV4(writer, kArchiveNameBytecode, bytecode_tuple);
  // write `constants` archive
  auto constants_tuple =
      c10::ivalue::Tuple::create(std::move(constants_values));
  writeArchiveV4(writer, kArchiveNameConstants, constants_tuple);
  return ouput_model_stream;
}

/*
Backport function bytecode v6 that introduced support for operators with default
arguments in mobile. Previously, in v5, there is no number of specified
arguments for operators in bytecode operator table. In v6, operators are aware
of the number of specified arguments being present in the schema.

The bump was needed because the v6 bytecode specifies number of specified
arguments for operators in the schema, since the runtime code is now able to
query the number of specified arguments and supports default arguments.

For example, aten::foo's schema in v5 is
foo(Tensor a, Tensor b) -> Tensor
and in v6, it's
foo(Tensor a, Tensor b, int groups=1) -> Tensor

Accordingly, the operator table in v5 is:
('operators', (('aten::foo', ''),))
and in v6, it's
('operators', (('aten::foo', '', 2),))

Thus, the backport is necessary such that the bytecode operator table contains
number of specified arguments.
*/
std::stringstream backport_v6_to_v5(std::stringstream& input_model_stream) {
  std::shared_ptr<IStreamAdapter> rai =
      std::make_shared<IStreamAdapter>(&input_model_stream);
  auto reader = std::make_shared<PyTorchStreamReader>(rai);

  // If there are debug info files in the original model file, it should also
  // show up in the backported model
  bool hasBytecodeDebug = reader->hasRecord("mobile_debug_handles.pkl");

  // extra_files are kept
  auto records = reader->getAllRecords();
  ExtraFilesMap extra_files;
  for (const auto& record : records) {
    std::size_t found = record.find_last_of("/\\");
    auto path = record.substr(0, found);
    if ("extra" == path) {
      extra_files.emplace(record.substr(found + 1), "");
    }
  }
  // Loading the TS module is required for this backport, because bytecode needs
  // to be re-emitted (refer to the comments below)
  Module torch_script = torch::jit::load(rai, c10::nullopt, extra_files);

  // The RAII guard to change the flag, emitBytecodeDefaultInputs, to true, so
  // that TS stores the default argument values in the constant table, and emits
  // the instructions (LOADC, for example), to push the values to the stack. It
  // restores the behavior of V5 and before. For V6, the default arg values are
  // resolved at runtime init stage for better operator compatibility.
  std::stringstream intermediate_model_stream;
  {
    BytecodeEmitModeGuard argNumGuard(
        true /*emit_default_input_instructions*/,
        false /*enable_defaults_args_with_out_args*/,
        false /*enable_emit_promoted_ops*/);
    torch_script._save_for_mobile(
        intermediate_model_stream, extra_files, hasBytecodeDebug);
  }

  // Update the bytecode version (from 6 to 5)
  std::stringstream output_model_stream =
      update_bytecode_version(intermediate_model_stream, kBytecodeVersionV5);
  return output_model_stream;
}

/*
Backport function bytecode v7 that introduced support for operators with out
arguments. Previously, in v6, operators with out arguments forced the
serialization of all arguments in the schema, even when optional arguments
were not provided (as they had default values). Currently, in v7, operators
are aware of out arguments being present in the schema (always appended),
allowing the serialization of only required arguments (as default values will
be provided by the runtime).

The bump was needed because the v7 bytecode specifies less arguments for ops
with out arguments in the schema, since the runtime code is now able to query
whether an argument is of type "out" and insert the necessary default values in
the right order in the interpreter stack (i.e. before the out arguments).

For example schema is: torch.add(x, h, alpha=1.0, out=x) So, if the program
defines: torch.add(x, h, out=x) Previously, in v6, we serialized the bytecode to
contain all 4 arguments. Currently, in v7, we serialize the bytecode with only 3
arguments, since alpha is optional and has a default value that the runtime will
push in the stack. Thus, the backport is necessary such that the bytecode
contains all the arguments as before.
*/
std::stringstream backport_v7_to_v6(std::stringstream& input_model_stream) {
  std::shared_ptr<IStreamAdapter> rai =
      std::make_shared<IStreamAdapter>(&input_model_stream);
  auto reader = std::make_shared<PyTorchStreamReader>(rai);
  auto constants_values =
      std::move(*readArchive(kArchiveNameConstants, *reader.get()).toTuple())
          .elements();

  // If there are debug info files in the original model file, it should also
  // show up in the backported model
  bool hasBytecodeDebug = reader->hasRecord("mobile_debug_handles.pkl");

  // extra_files are kept
  auto records = reader->getAllRecords();
  ExtraFilesMap extra_files;
  for (const auto& record : records) {
    std::size_t found = record.find_last_of("/\\");
    auto path = record.substr(0, found);
    if ("extra" == path) {
      extra_files.emplace(record.substr(found + 1), "");
    }
  }
  // Loading the TS module is required for this backport, because bytecode needs
  // to be re-emitted (refer to the comments below)
  Module torch_script = torch::jit::load(rai, c10::nullopt, extra_files);

  // The RAII guard to change the flag, emit_default_input_instructions, to
  // false to keep the same behavior in bytecode version 6. Change the flag,
  // enable_defaults_args_with_out_args, to deserialized the number of specified
  // operators which allowing both out arguments and default arguments to
  // #all_args, instead of (#all_args - #default_args)
  std::stringstream intermediate_model_stream;
  {
    BytecodeEmitModeGuard argNumGuard(
        false /*emit_default_input_instructions*/,
        false /*enable_defaults_args_with_out_args*/,
        false /*enable_emit_promoted_ops*/);
    torch_script._save_for_mobile(
        intermediate_model_stream, extra_files, hasBytecodeDebug);
  }

  // Update the bytecode version (from 7 to 6)
  std::stringstream output_model_stream =
      update_bytecode_version(intermediate_model_stream, kBytecodeVersionV6);
  return output_model_stream;
}

std::stringstream backport_v8_to_v7(std::stringstream& input_model_stream) {
  std::shared_ptr<IStreamAdapter> rai =
      std::make_shared<IStreamAdapter>(&input_model_stream);
  auto reader = std::make_shared<PyTorchStreamReader>(rai);
  // extra_files are kept
  auto records = reader->getAllRecords();
  bool hasBytecodeDebug = reader->hasRecord("mobile_debug_handles.pkl");
  ExtraFilesMap extra_files;
  for (const auto& record : records) {
    std::size_t found = record.find_last_of("/\\");
    auto path = record.substr(0, found);
    if ("extra" == path) {
      extra_files.emplace(record.substr(found + 1), "");
    }
  }
  Module torch_script = torch::jit::load(rai, c10::nullopt, extra_files);
  std::stringstream intermediate_model_stream;
  {
    BytecodeEmitModeGuard argNumGuard(
        false /*emit_default_input_instructions*/,
        true /*enable_defaults_args_with_out_args*/,
        false /*enable_emit_promoted_ops*/);
    torch_script._save_for_mobile(
        intermediate_model_stream, extra_files, hasBytecodeDebug);
  }

  // Update the bytecode version (from 8 to 7)
  std::stringstream output_model_stream =
      update_bytecode_version(intermediate_model_stream, kBytecodeVersionV7);

  return output_model_stream;
}

} // namespace

/********************** BackportManager **********************/

// A generic contract for backport logic to the previous bytecode version.
// Args:
// * PyTorchStreamReader has access to the input model from N bytecode version.
// * PyTorchStreamWriter has access to the output model backported to the
// previous N-1 bytecode version. Returns true if successful, false otherwise.
using BytecodeBackportFunction =
    std::function<std::stringstream(std::stringstream&)>;

BackportManager::BackportManager() {
  registerBytecodeBackportFunction(kBytecodeVersionV5, backport_v5_to_v4);
  registerBytecodeBackportFunction(kBytecodeVersionV6, backport_v6_to_v5);
  registerBytecodeBackportFunction(kBytecodeVersionV7, backport_v7_to_v6);
  registerBytecodeBackportFunction(kBytecodeVersionV8, backport_v8_to_v7);
}

std::unordered_map<
    int64_t,
    std::function<std::stringstream(std::stringstream&)>>&
BackportManager::bytecodeBackportFunctions() const {
  static std::unordered_map<
      int64_t,
      std::function<std::stringstream(std::stringstream&)>>
      backport_functions;
  return backport_functions;
}

bool BackportManager::hasBytecodeBackportFunction(
    const int64_t from_version) const {
  return bytecodeBackportFunctions().count(from_version);
}

void BackportManager::registerBytecodeBackportFunction(
    const int64_t from_version,
    const BytecodeBackportFunction& backport_function) {
  TORCH_CHECK(
      !hasBytecodeBackportFunction(from_version),
      "Backporting from version ",
      from_version,
      " is already registered.");
  bytecodeBackportFunctions()[from_version] = backport_function;
}

// The main function to run backport from version n to version i.
// All models (file or buffer) will be converted stream first, and
// istream_adapter has access to it. During the backport process,
// the intermediate result will be stored with stream.
bool BackportManager::backport(
    std::shared_ptr<IStreamAdapter> istream_adapter,
    PyTorchStreamWriter& final_writer,
    int64_t from_version,
    int64_t to_version) const {
  PyTorchStreamReader start_reader(istream_adapter);

  if (from_version <= to_version) {
    TORCH_WARN(
        "backport donesn't support backporting model to new version. It's trying to backport from version ",
        from_version,
        " to version ",
        to_version);
    return false;
  }
  int64_t bytecode_version = from_version;
  bool backport_success = true;

  // 1) Given an istream_adapter (an adapter with access to the input model, the
  // model can be from istream, file and etc), copy all model content to
  // stringstream
  std::stringstream oss;
  get_model_stream(start_reader, oss);
  std::stringstream input_model_stream(oss.str());
  std::stringstream output_model_stream;

  // 2) backport model, backport_v{i}_to_v{i-1} function's argurment is
  // (input_model_stream and output_model_stream)
  while (bytecode_version > to_version) {
    // Swap input and output if it's not the first time and output_model_stream
    // has value.
    if (!output_model_stream.str().empty()) {
      input_model_stream.swap(output_model_stream);
      // reset output_model_stream
      output_model_stream.str("");
    }

    if (!hasBytecodeBackportFunction(bytecode_version)) {
      return false;
    }

    auto input_model_stream_version =
        _get_model_bytecode_version(input_model_stream);

    if (static_cast<int64_t>(input_model_stream_version) != bytecode_version) {
      TORCH_WARN(
          "The bytecode version of input model stream is supposed to be ",
          bytecode_version,
          ", but it gets ",
          input_model_stream_version);
      return false;
    }

    // Keep backporting till request version
    std::stringstream backport_model_stream =
        bytecodeBackportFunctions()[bytecode_version--](input_model_stream);

    output_model_stream.swap(backport_model_stream);
    auto output_model_stream_version =
        _get_model_bytecode_version(output_model_stream);

    if (static_cast<int64_t>(output_model_stream_version) != bytecode_version) {
      TORCH_WARN(
          "The bytecode version of output model stream is supposed to be ",
          bytecode_version,
          ", but it gets ",
          output_model_stream_version);
      return false;
    }
  }

  // 3) Write the final output_model_stream to final_writer, final_writer has
  // access to the final model destination (file, ostream and etc)
  if (output_model_stream.str().empty()) {
    TORCH_WARN("No output model from backport.");
    return false;
  }
  PyTorchStreamReader last_model_reader(&output_model_stream);
  selective_copy(
      last_model_reader,
      final_writer,
      std::unordered_set<std::string>(),
      std::unordered_set<std::string>());

  return backport_success;
}

} // namespace jit
} // namespace torch