Open Wayne3521 opened 6 months ago
I meet the same question
I haven't solved it yet
Generally, this issue is likely caused by your graphics card's CUDA and Torch versions being too old. In my case, using a 4090 with CUDA 11.6 and torch 1.12 resolved the error.
I haven't solved it yet
I solved this issue and didn't change my envirment setting.
det = torch.det(V.cpu() @ Ut.cpu())
eye[:, -1, -1] = torch.sign(det.cuda())
According to the Readme: Ubuntu 20.04, GCC 9.3.0, Python 3.8, PyTorch 1.7.1, CUDA 11.1 and cuDNN 8.1.0.torch==1.7.1+cu110, the following problems occur:
Traceback (most recent call last): File "trainval.py", line 62, in
main()
File "trainval.py", line 58, in main
trainer.run()
File "/home/wl/code/GeoTransformer/geotransformer/engine/epoch_based_trainer.py", line 180, in run
self.train_epoch()
File "/home/wl/code/GeoTransformer/geotransformer/engine/epoch_based_trainer.py", line 95, in train_epoch
output_dict, result_dict = self.train_step(self.epoch, self.inner_iteration, data_dict)
File "trainval.py", line 41, in train_step
output_dict = self.model(data_dict)
File "/home/wl/anaconda3/envs/geotransformer/lib/python3.8/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, kwargs)
File "/home/wl/code/GeoTransformer/experiments/geotransformer.3dmatch.stage4.gse.k3.max.oacl.stage2.sinkhorn/model.py", line 198, in forward
ref_corr_points, src_corr_points, corr_scores, estimated_transform = self.fine_matching(
File "/home/wl/anaconda3/envs/geotransformer/lib/python3.8/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(*input, *kwargs)
File "/home/wl/code/GeoTransformer/geotransformer/modules/geotransformer/local_global_registration.py", line 231, in forward
ref_corr_points, src_corr_points, corr_scores, estimated_transform = self.local_to_global_registration(
File "/home/wl/code/GeoTransformer/geotransformer/modules/geotransformer/local_global_registration.py", line 171, in local_to_global_registration
batch_transforms = self.procrustes(batch_src_corr_points, batch_ref_corr_points, batch_corr_scores)
File "/home/wl/anaconda3/envs/geotransformer/lib/python3.8/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
result = self.forward(input, kwargs)
File "/home/wl/code/GeoTransformer/geotransformer/modules/registration/procrustes.py", line 84, in forward
return weighted_procrustes(
File "/home/wl/code/GeoTransformer/geotransformer/modules/registration/procrustes.py", line 56, in weighted_procrustes
eye[:, -1, -1] = torch.sign(torch.det(V @ Ut))
RuntimeError: CUDA error: operation not supported when calling
cusparseCreate(handle)