q yao

Oh, err... sorry, master branch requires mmdet 2.18.0+ , I forgot to update the Readme.md. Would you mind updating mmdetection and trying again?

You need to enable the flag to convert a model with instance segmentation. please read [this](https://github.com/grimoire/mmdetection-to-tensorrt/blob/master/docs/getting_started.md#instance-segmentation-supportexperimentation) for more detail. And as your input is `x = torch.ones([1,3,320,320])`, I think the...

Hi, Could you provide the script,model file and test image data?

Hi I have test the image you provided. Seems convertor works. here is my test script: ``` python demo/inference.py \ test.jpg \ retinanet_r50_fpn_1x_coco.py \ retinanet_r50_fpn_1x_coco_20200130-c2398f9e.pth \ retina.pth ``` And result:...

[execute_async_v2](https://docs.nvidia.com/deeplearning/tensorrt/developer-guide/index.html#perform_inference_python) is the inference entry of tensorrt. The error is happening inside the model. The project has been changed a lot since my last reply, please reinstall [torch2trt_dynamic](https://github.com/grimoire/torch2trt_dynamic), [amirstan_plugin](https://github.com/grimoire/amirstan_plugin), [mmdetection-to-tensorrt](https://github.com/grimoire/mmdetection-to-tensorrt)...

Please try the latest torch2trt_tynamic and amirstan_plugin.

Errr, use command `git log` to see if the top commit is the same as the one on github. Or just `git pull origin master` and wait.

Hi Can you reproduce this error on yolox_s? MMDetection did not provide a checkpoint of yolox_l.

Sorry for the late reply. It seems that you have not set the `opt_shape_param` when converting the model. Since the input shape of Yolox is different from other models in...

Hi Could you provide the model config?