HopeJW
Please try using the newest version that I uploaded 3 days ago.
Please run the script [make_pb.sh](https://github.com/NVIDIA-AI-IOT/Lidar_AI_Solution/blob/master/CUDA-BEVFusion/src/onnx/make_pb.sh) first.
This is just because the DDS (data dependence shape) feature is not supported in TensorRT(
Could you provide me with a minimal reproduction code?
BEVFusion has been supported by fusion with Camera, Lidar, and Radar.
Please ignore those warnings, thanks.
Hi, could you provide me with the model and data so that I can reproduce this issue?
This is to match the kernel's requirements for the layout of the weights. The best performance will be obtained with the weights in this format.