Hi, I’m from Tier4 percption team.
Same as https://github.com/autowarefoundation/modelzoo/pull/3 's conversation, but we tried implementing the tvm version of lidar obstacle detection. https://github.com/tier4/lidar_instance_segmentation_tvm
We would appreciate it if you could give us a feedback.
The model is trained with pytorch using nuscenes data and then converted to tvm.
I’m also interested in a unified story for training models