caffe2 export onnx model, Gemm operator dismissing

If you do not know the root cause of the problem, please post according to this template:

## Instructions To Reproduce the Issue:

When I export onnx model use caffe2, the onnx model do not contain Gemm operator; The model I have exported successfully a month ago. But I dont know what let the model wrong. The code has not rewritten.

1. Full runnable code or full changes you made:
```
python export_model.py --sample-image 1344x1344.jpg --config-file ../../configs/COCO-InstanceSegmentation/mask_rcnn_R_50_FPN_3x.yaml --export-method caffe2_tracing --format onnx --output output/ MODEL.WEIGHTS model_final_f10217.pkl MODEL.DEVICE cuda

```
**correct onnx model:**
**wrong onnx model**

## Expected behavior:

If there are no obvious crash in "full logs" provided above,
please tell us the expected behavior.

If you expect a model to converge / work better, we do not help with such issues, unless
a model fails to reproduce the results in detectron2 model zoo, or proves existence of bugs.

## Environment:
latest detectron2 version;
sys.platform linux
Python 3.8.8 (default, Feb 24 2021, 21:46:12) [GCC 7.3.0]
numpy 1.22.0
detectron2 0.6 @/dfs/data/LCM_TRT/detectron2/detectron2
Compiler GCC 9.3
CUDA compiler CUDA 11.3
detectron2 arch flags 5.2, 6.0, 6.1, 7.0, 7.5, 8.0, 8.6
DETECTRON2_ENV_MODULE
PyTorch 1.9.0a0+2ecb2c7 @/opt/conda/lib/python3.8/site-packages/torch
PyTorch debug build False
GPU available Yes
GPU 0 Tesla V100-PCIE-32GB (arch=7.0)
Driver version 515.48.07
CUDA_HOME /usr/local/cuda
TORCH_CUDA_ARCH_LIST 5.2 6.0 6.1 7.0 7.5 8.0 8.6+PTX
Pillow 8.2.0
torchvision 0.9.0a0 @/opt/conda/lib/python3.8/site-packages/torchvision
torchvision arch flags 5.2, 6.0, 6.1, 7.0, 7.5, 8.0, 8.6
fvcore 0.1.5.post20220512
iopath 0.1.9
cv2 3.4.11
---------------------- -------------------------------------------------------------
PyTorch built with:
- GCC 9.3
- C++ Version: 201402
- Intel(R) Math Kernel Library Version 2019.0.4 Product Build 20190411 for Intel(R) 64 architecture applications
- Intel(R) MKL-DNN v1.8.0 (Git Hash N/A)
- OpenMP 201511 (a.k.a. OpenMP 4.5)
- NNPACK is enabled
- CPU capability usage: AVX2
- CUDA Runtime 11.3
- NVCC architecture flags: -gencode;arch=compute_52,code=sm_52;-gencode;arch=compute_60,code=sm_60;-gencode;arch=compute_61,code=sm_61;-gencode;arch=compute_70,code=sm_70;-gencode;arch=compute_75,code=sm_75;-gencode;arch=compute_80,code=sm_80;-gencode;arch=compute_86,code=sm_86;-gencode;arch=compute_86,code=compute_86
- CuDNN 8.2
- Magma 2.5.2
- Build settings: BLAS_INFO=mkl, BUILD_TYPE=Release, CUDA_VERSION=11.3, CUDNN_VERSION=8.2.0, CXX_COMPILER=/usr/bin/c++, CXX_FLAGS= -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp -DNDEBUG -DUSE_FBGEMM -DUSE_QNNPACK -DUSE_PYTORCH_QNNPACK -DUSE_XNNPACK -O2 -fPIC -Wno-narrowing -Wall -Wextra -Werror=return-type -Wno-missing-field-initializers -Wno-type-limits -Wno-array-bounds -Wno-unknown-pragmas -Wno-sign-compare -Wno-unused-parameter -Wno-unused-variable -Wno-unused-function -Wno-unused-result -Wno-unused-local-typedefs -Wno-strict-overflow -Wno-strict-aliasing -Wno-error=deprecated-declarations -Wno-stringop-overflow -Wno-psabi -Wno-error=pedantic -Wno-error=redundant-decls -Wno-error=old-style-cast -fdiagnostics-color=always -faligned-new -Wno-unused-but-set-variable -Wno-maybe-uninitialized -fno-math-errno -fno-trapping-math -Werror=format -Werror=cast-function-type -Wno-stringop-overflow, FORCE_FALLBACK_CUDA_MPI=1, LAPACK_INFO=mkl, PERF_WITH_AVX=1, PERF_WITH_AVX2=1, PERF_WITH_AVX512=1, TORCH_VERSION=1.9.0, USE_CUDA=ON, USE_CUDNN=ON, USE_EXCEPTION_PTR=1, USE_GFLAGS=OFF, USE_GLOG=OFF, USE_MKL=ON, USE_MKLDNN=ON, USE_MPI=ON, USE_NCCL=ON, USE_NNPACK=ON, USE_OPENMP=ON,