Sandeep Menon

@mattgiamou Any update on this issue?

@hopef When I export-scn from a "non ptq model" and try to load it using `load_engine_from_onnx` in https://github.com/NVIDIA-AI-IOT/Lidar_AI_Solution/blob/87fb0cc6fcf38d0cf998bf0cdcbd039e6732d928/CUDA-BEVFusion/src/bevfusion/lidar-scn.cpp#L38C1-L39C1 I get the error ```bash [libprotobuf FATAL /usr/include/google/protobuf/repeated_field.h:1506] CHECK failed: (index) <...

> Hi sandeepnmenon, > > I can't see the bias of the SparseConvolution layer in your onnx. This may be the root cause. Thank you. I was using the lidar...

Hi @hopef When I export after loading the state dict of the model and run the exptool this is being caused. But if I run it through the quantisation module...

Can we run the `export-camera.py` etc scripts to export the onnx modules from a non quantized model? Would those onnx models work in the inference pipeline of bevfusion.cpp just slower?

@cdefg You can use the original repo's model to run the export scripts and each script extracts out the components like model/fuser.onnx, the camera backbone etc. The `ptq.py` is not...

I see the confusion. I dont think `bevfusion/tools/export.py` works because there are some operations that do not directly export to onnx and TRT (Hence this repo). I meant you can...

Glad you were able to deploy! I was able to run the swint backbone model but without quantization ( not running `ptq.py` and just running the exports) As for the...

Seems like the `qat/ptq.py` is written for the resnet50 backbone. So I think even for custom dataset, as long as the model architecture doesn't change, the script should work as...

+1 I am also interested in the map segmentation head with the `BEVGridTransform` and `BEVSegmentationHead`