AttributeError: 'NoneType' object has no attribute 'cluster_poses'

[sonnyhe]

I encountered the same problem when using 5060TI.

1. I used Docker image ,
   https://hub.docker.com/r/yanglicb/foundationpose-cuda128

docker pull yanglicb/foundationpose-cuda128

2. changed the code as follows,
   [(https://github.com/Add RTX 50 series GPU compatibility via PyTorch 2.7 and CUDA 12.8 #369)]

3. And then we encountered this problem,
   If this is not desired, please set os.environ['TORCH_CUDA_ARCH_LIST']. warnings.warn( [reset_object()] self.diameter:0.19646325799497472, vox_size:0.009823162899748735 [reset_object()] self.pts:torch.Size([607, 3]) [reset_object()] reset done [make_rotation_grid()] cam_in_obs:(42, 4, 4) [make_rotation_grid()] rot_grid:(252, 4, 4) Segmentation fault (core dumped)

4. Annotated this line of code, it's normal now
   FoundationPose/estimater.py, line 120
   # rot_grid = mycpp.cluster_poses(30, 99999, rot_grid, self.symmetry_tfs.data.cpu().numpy())