[Feature] RTMDet Instance model export onnx support

[luohao123]

Motivation

RTMDet get a reasonable small models and fast speed on instance segmentation.

Related resources

No response

Additional context

KeyError: 'RTMDet is not in the model registry. Please check whether the value of RTMDet is correct or it was registered as expected. More details can be found at https://mmengine.readthedocs.io/en/latest/advance\

Currently export just failed.

(RTMDet detection model works though)

[irexyc]

#1662

It seems already supported.

What mmdeploy version are you using? Could you try the latest mmdeploy?

[luohao123]

@irexyc am using mmdeploy latest (refreshly clone yesterday), the problems comes from mmengine which just mim installed.

[luohao123]

@irexyc can u send me a RTMDet instance onnx model if you able to export? kindly share me your export scripts (comands)?

[irexyc]

@luohao123

Hi, this is my command:

python tools/deploy.py  \
configs/mmdet/instance-seg/instance-seg_rtmdet-ins_tensorrt_static-640x640.py \
../mmdetection/configs/rtmdet/rtmdet-ins_s_8xb32-300e_coco.py \
https://download.openmmlab.com/mmdetection/v3.0/rtmdet/rtmdet-ins_s_8xb32-300e_coco/rtmdet-ins_s_8xb32-300e_coco_20221121_212604-fdc5d7ec.pth \
../mmdetection/demo/demo.jpg \
--work-dir /tmp/rtmdet \
--device cuda \
--dump-info

And this is the visualize result of tensorrt

[luohao123]

@irexyc I still get KeyError: 'RTMDet is not in the model registry. Please check whether the value of RTMDet is correct or it was registered as expected. More details can be found at [https://mmengine.readthedocs.io/en/latest/advance\](https://mmengine.readthedocs.io/en/latest/advance%5C)

I have mmengine latest installed via mim

[irexyc]

Please provide the output of python tools/check_env.py