Torchscript C++ Inference Error

🐛 Describe the bug

I saved a Detectron2 Torchscript model using the following code:

fields = {
        "proposal_boxes": Boxes,
        "objectness_logits": Tensor,
        "pred_boxes": Boxes,
        "scores": Tensor,
        "pred_classes": Tensor,
        "pred_masks": Tensor,
        "pred_keypoints": torch.Tensor,
        "pred_keypoint_heatmaps": torch.Tensor,
    }
    torch_script_module = scripting_with_instances(torch_model, fields)
    extra_files = {}
    extra_files["module_info.json"] = json.dumps({"input_names": ["data"], "output_names": ["predicted"]})
    torch.jit.save(torch_script_module, os.path.join(outdir, "model.pt"), _extra_files=extra_files)

However, when I tried to load it in C++, I'm getting the following error:

terminate called after throwing an instance of 'torch::jit::ErrorReport'
  what():
Unknown type name 'NoneType':
Serialized   File "code/__torch__/detectron2/modeling/backbone/fpn.py", line 4
  __parameters__ = []
  __buffers__ = []
  _is_full_backward_hook : NoneType
                           ~~~~~~~~ <--- HERE
  in_features : Tuple[str, str, str, str]
  _out_feature_strides : Dict[str, int]

I've tried to remove the _is_full_backward_hook field in torch_script_module, but seems like it will still be initialized as "None" when I load the model (in both Python and C++). What's the best way to resolve this issue?

Versions

Collecting environment information...
PyTorch version: 1.9.0a0+gitdfbd030
Is debug build: False
CUDA used to build PyTorch: 11.0
ROCM used to build PyTorch: N/A

OS: Ubuntu 16.04.7 LTS (x86_64)
GCC version: (Ubuntu 5.4.0-6ubuntu1~16.04.12) 5.4.0 20160609
Clang version: Could not collect
CMake version: version 3.5.1
Libc version: glibc-2.2.5

Python version: 3.7.5 (default, Aug 26 2021, 16:53:13) [GCC 5.4.0 20160609] (64-bit runtime)
Python platform: Linux-4.15.0-1065-aws-x86_64-with-debian-stretch-sid
Is CUDA available: True
CUDA runtime version: 11.0.221
GPU models and configuration:
GPU 0: Tesla V100-SXM2-32GB
GPU 1: Tesla V100-SXM2-32GB
GPU 2: Tesla V100-SXM2-32GB
GPU 3: Tesla V100-SXM2-32GB
GPU 4: Tesla V100-SXM2-32GB
GPU 5: Tesla V100-SXM2-32GB
GPU 6: Tesla V100-SXM2-32GB
GPU 7: Tesla V100-SXM2-32GB

Nvidia driver version: 450.142.00
cuDNN version: Probably one of the following:
/usr/lib/x86_64-linux-gnu/libcudnn.so.8.0.5
/usr/lib/x86_64-linux-gnu/libcudnn_adv_infer.so.8.0.5
/usr/lib/x86_64-linux-gnu/libcudnn_adv_train.so.8.0.5
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_infer.so.8.0.5
/usr/lib/x86_64-linux-gnu/libcudnn_cnn_train.so.8.0.5
/usr/lib/x86_64-linux-gnu/libcudnn_ops_infer.so.8.0.5
/usr/lib/x86_64-linux-gnu/libcudnn_ops_train.so.8.0.5
HIP runtime version: N/A
MIOpen runtime version: N/A

Versions of relevant libraries:
[pip3] botorch==0.4.0
[pip3] gpytorch==1.5.1
[pip3] mypy-extensions==0.4.3
[pip3] mypy-protobuf==2.4
[pip3] numpy==1.19.4
[pip3] pytorch-lamb==1.0.0
[pip3] torch==1.9.0a0+gitdfbd030
[pip3] torch-tb-profiler==0.3.1
[pip3] torchfile==0.1.0
[pip3] torchmetrics==0.3.1
[pip3] torchscript==0.2.10
[pip3] torchvision==0.8.0a0+2f40a48
[conda] Could not collect

Start with the reported torch.jit.save flow and the C++ model-loading path, then inspect the serialized code/torch/detectron2/modeling/backbone/fpn.py entry where _is_full_backward_hook is declared as NoneType. Done means the saved model loads in C++ without this error and the inference path works.