ops->filesystem_ops = static_cast<TF_FilesystemOps*>(
      plugin_memory_allocate(TF_FILESYSTEM_OPS_SIZE));
  ops->filesystem_ops->init = tf_posix_filesystem::Init;
  ops->filesystem_ops->cleanup = tf_posix_filesystem::Cleanup;
  ops->filesystem_ops->new_random_access_file =
      tf_posix_filesystem::NewRandomAccessFile;

int TF_CheckpointReaderHasTensor(TF_CheckpointReader* reader,
                                 const char* name) {
  return reader->HasTensor(name);
}

const char* TF_CheckpointReaderGetVariable(TF_CheckpointReader* reader,
                                           int index) {
  return reader->variable_list[index].c_str();
}

int TF_CheckpointReaderSize(TF_CheckpointReader* reader) {

  TF_RETURN_IF_ERROR(
      env->GetSymbolFromLibrary(dso_handle, "TF_InitGraph", &dso_symbol));

  // Step 2: Call `TF_InitPlugin`
  auto init_fn = reinterpret_cast<TFInitGraphPluginFn>(dso_symbol);
  return InitGraphPlugin(init_fn);
}

absl::Status InitGraphPlugin(TFInitGraphPluginFn init_fn) {
  TP_OptimizerRegistrationParams params{
      TP_OPTIMIZER_REGISTRATION_PARAMS_STRUCT_SIZE};

    params->device_type = "Device1";
  };

  TF_ASSERT_OK(InitGraphPlugin(plugin_init_0));
  TF_ASSERT_OK(InitGraphPlugin(plugin_init_1));
  ASSERT_EQ(PluginGraphOptimizerRegistry::CreateOptimizers(
                std::set<string>{"Device0", "Device1"})
                .size(),
            2);
}

TEST(Grappler, DeviceTypeNotSet) {
  auto plugin_init = [](TP_OptimizerRegistrationParams* const params,

  }

  {
    std::unique_ptr<ParallelTensor> int_tensors =
        parallel_device.ScalarsFromSequence<int>({5, 6}, context.get(),
                                                 status.get());
    ASSERT_EQ(TF_GetCode(status.get()), TF_OK) << TF_Message(status.get());
    ExpectScalarEq<int>(int_tensors->tensor(0), 5);
    ExpectScalarEq<int>(int_tensors->tensor(1), 6);
  }
}

}  // namespace parallel_device

#include "tensorflow/core/protobuf/tensorflow_server.pb.h"

namespace {

using ::tensorflow::string;

void ReplaceTaskInServerDef(tensorflow::ServerDef* server_def, int task_index) {
  tensorflow::JobDef* job_def = server_def->mutable_cluster()->mutable_job(0);
  int port = tensorflow::testing::PickUnusedPortOrDie();
  job_def->mutable_tasks()->at(task_index) =
      tensorflow::strings::StrCat("localhost:", port);
}

                         {12, 15, 18, 12, 15, 18, 12, 15, 18},
                         {6, 6, 6, 15, 15, 15, 24, 24, 24},
                         {6, 15, 24, 6, 15, 24, 6, 15, 24}};

  for (int i{}; i < 4; ++i) {
    bool transpose_a = transpose_a_vals[i];
    bool transpose_b = transpose_b_vals[i];
    Model MatMulModel =
        [transpose_a, transpose_b](
            AbstractContext* ctx,

}
Status SetAttrStringList(AbstractOperation* op_, const char* attr_name,
                         const void* const* values, const size_t* lengths,
                         int num_values, ForwardOperation* forward_op_) {
  std::vector<StringPiece> v(num_values);
  for (int i = 0; i < num_values; ++i) {
    v[i] = StringPiece(static_cast<const char*>(values[i]), lengths[i]);
  }
  forward_op_->attrs.Set(attr_name, v);

// data.
bool IsValidStrideCompactRowMajorData(int64_t* shape_arr, int64_t* stride_arr,
                                      int ndim) {
  bool valid = true;
  int64_t expected_stride = 1;
  for (int i = ndim - 1; i >= 0; --i) {
    // Empty tensors are always compact regardless of strides.
    if (shape_arr[i] == 0) return true;

}

TF_Tensor* BoolTensor(bool v) {
  const int num_bytes = sizeof(bool);
  bool* values = new bool[1];
  values[0] = v;
  return TF_NewTensor(TF_BOOL, nullptr, 0, values, num_bytes, &BoolDeallocator,
                      nullptr);
}

TF_Tensor* Int8Tensor(const int64_t* dims, int num_dims, const char* values) {
  int64_t num_values = 1;
  for (int i = 0; i < num_dims; ++i) {
    num_values *= dims[i];
  }