search

PJLab-ADG / neuralsim

neuralsim: 3D surface reconstruction and simulation based on 3D neural rendering.
MIT License
582 stars 31 forks source link

RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu) #51

Open amoghskanda opened 5 months ago

amoghskanda commented 5 months ago

Getting this error when I try to train streetsurf on seg-10061. Config used : withmask_nolidar.240219.yaml. preload_on_gpu is false(by default). Cache is getting loaded onto cpu. Error on Line411(_ret.update({k: v[frame_ind, h, w].to(device)) for k,v in _ret_image_raw.items()}) in dataloader/baseloader.py.

File "/app/neuralsim/dataio/data_loader/image_loader.py", line 258, in getitem ret = self.sample(scene_id, cam_id, cam_fi) File "/app/neuralsim/dataio/data_loader/image_loader.py", line 213, in sample ground_truth = self.scene_loader.get_image_and_gts(scene_id, cam_id, cam_fi, hw[..., 0], hw[..., 1], device=self.device) File "/app/neuralsim/dataio/data_loader/base_loader.py", line 614, in get_image_and_gts gt = self.get_image_and_metas(scene_id, cam_id, frame_ind, h, w, device=device) File "/app/neuralsim/dataio/data_loader/base_loader.py", line 411, in get_image_and_metas _ret.update({k: v[frame_ind, h, w].to(device) for k,v in _ret_image_raw.items()}) File "/app/neuralsim/dataio/data_loader/base_loader.py", line 411, in _ret.update({k: v[frame_ind, h, w].to(device) for k,v in _ret_image_raw.items()}) RuntimeError: indices should be either on cpu or on the same device as the indexed tensor (cpu)

I only changed the loss from MonoDepthLoss to MonoSDFLoss, nothing else. Did anybody face this issue? Thanks