test model error

[MingFengHill]

Thanks to the author for sharing excellent work. I have the following error when running the test program, using cuda 11.7, python 3.9, torch 2.0.1:

inst_iou:  0.9999999999891441 type_iou:  1.0
ID:2604 | inst_iou: 0.9999999999891441 type_iou: 1.0 inst_recall: 1.0
inst_iou:  0.9217833057396648 type_iou:  1.0
ID:2605 | inst_iou: 0.9217833057396648 type_iou: 1.0 inst_recall: 0.6666666666666666
inst_iou:  0.9999999999835414 type_iou:  1.0
ID:2606 | inst_iou: 0.9999999999835414 type_iou: 1.0 inst_recall: 1.0
inst_iou:  0.926522588942739 type_iou:  1.0
ID:2607 | inst_iou: 0.926522588942739 type_iou: 1.0 inst_recall: 1.0
inst_iou:  0.2727425572723341 type_iou:  0.5
ID:2608 | inst_iou: 0.2727425572723341 type_iou: 0.5 inst_recall: 1.0
Traceback (most recent call last):
  File "/home/wang/recon/SED-Net/generate_predictions_aug.py", line 372, in <module>
    embedding = hpnet_process(embedding.transpose(1, 2), points, normals, id=val_b_id,
  File "/home/wang/recon/SED-Net/src/smooth_normal_matrix.py", line 198, in hpnet_process
    v = torch.lobpcg(affinity_matrix_normal, k=edge_topk, niter=10)[1]
  File "/home/wang/anaconda3/envs/sednet/lib/python3.9/site-packages/torch/_lobpcg.py", line 567, in lobpcg
    return _lobpcg(
  File "/home/wang/anaconda3/envs/sednet/lib/python3.9/site-packages/torch/_lobpcg.py", line 673, in _lobpcg
    worker.run()
  File "/home/wang/anaconda3/envs/sednet/lib/python3.9/site-packages/torch/_lobpcg.py", line 837, in run
    self.update()
  File "/home/wang/anaconda3/envs/sednet/lib/python3.9/site-packages/torch/_lobpcg.py", line 767, in update
    self._update_ortho()
  File "/home/wang/anaconda3/envs/sednet/lib/python3.9/site-packages/torch/_lobpcg.py", line 924, in _update_ortho
    E_, Z = _utils.symeig(_utils.qform(self.A, S_), largest)
  File "/home/wang/anaconda3/envs/sednet/lib/python3.9/site-packages/torch/_linalg_utils.py", line 86, in symeig
    E, Z = torch.linalg.eigh(A, UPLO="U")
torch._C._LinAlgError: linalg.eigh: The algorithm failed to converge because the input matrix is ill-conditioned or has too many repeated eigenvalues (error code: 8).

Does anyone have the same problem?