Xiaoyang Wu
Xiaoyang Wu
@SpeedyGonzales949 Thx again.
Hi, when building the CUDA code in the local environment, we need to ensure that the CUDA path is successfully exported. Could you check the output of: nvcc -V ?
Hey friend. The OA-CNNs is a bit old now. I think we also don't have weight on our machine. Please consider using the latest PTv3, which is still actively enhanced...
Maybe you can adjust the PTv3 parameters to fit the scale you want to deploy. Deployment PTv3 to robotic environment is also something we care, and we chat a lot...
We are working on PTv4, which will be fully PyTorch-based. We will make the code/weight public once they are ready, even before the paper.
> Could you please provide a pre-trained model on ScanNet with Ocformer? Then I can still use the Pointcept repo. Yes, I can take a look at this.
Hey, sorry for the late response. I don't think this is a big problem. We also have some discussion here: https://github.com/Pointcept/Pointcept/issues/429
Hi, I also haven't validated the full workflow for the Waymo test set and submission. I will do this after finishing the release work of PTv3.
Hi, check the updates in v1.5.2
Hey, it was just because I only make the TTA testing process support bs1/gpu, it needs some additional effort to make it support bs > 1 / GPU.