Error during inference

[julesmorel]

Thank you for your recent fixes, the inference is now starting. However it still ends with the following error:

Running the container Input directory: /home/datascience/tmp_in_folder Output directory: /home/datascience/tmp_out_folder Script directory: /home/nibio/mutable-outside-world Renamed: /home/datascience/tmp_out_folder/input_data/sample_sub01.laz to /home/datascience/tmp_out_folder/input_data/sample_sub01.laz Processing 1 files... Processing in utm2local: /home/datascience/tmp_out_folder/input_data/sample_sub01.laz Saving min values to: /home/datascience/tmp_out_folder/utm2local/sample_sub01_out_min_values.json Output files are saved in: /home/datascience/tmp_out_folder/utm2local Clearing: /home/datascience/tmp_out_folder/utm2local/treeinsfused/processed_0.2_test Path does not exist: /home/datascience/tmp_out_folder/utm2local/treeinsfused/processed_0.2_test /usr/local/lib/python3.8/dist-packages/wandb/init.py:998: SyntaxWarning: "is not" with a literal. Did you mean "!="? if name is None and resume is not "must": /usr/local/lib/python3.8/dist-packages/requests/init.py:102: RequestsDependencyWarning: urllib3 (1.26.7) or chardet (5.2.0)/charset_normalizer (2.0.6) doesn't match a supported version! warnings.warn("urllib3 ({}) or chardet ({})/charset_normalizer ({}) doesn't match a supported " /usr/local/lib/python3.8/dist-packages/MinkowskiEngine/init.py:36: UserWarning: The environment variable OMP_NUM_THREADS not set. MinkowskiEngine will automatically set OMP_NUM_THREADS=16. If you want to set OMP_NUM_THREADS manually, please export it on the command line before running a python script. e.g. export OMP_NUM_THREADS=12; python your_program.py. It is recommended to set it below 24. warnings.warn( [2024-11-11 01:07:36,313][torch_points3d.trainer][INFO] - DEVICE : cuda [2024-11-11 01:07:36,570][torch_points3d.metrics.model_checkpoint][INFO] - Loading checkpoint from /home/nibio/mutable-outside-world/model_file/PointGroup-PAPER.pt Processing... area_name: sample_sub01_out.ply data: Data(instance_labels=[9383781], pos=[9383781, 3], y=[9383781]) test_data_list: [[Data(instance_labels=[9383781], pos=[9383781, 3], y=[9383781])]] [2024-11-11 01:07:37,725][torch_points3d.datasets.segmentation.treeins][INFO] - pre_collate_transform ... [2024-11-11 01:07:37,726][torch_points3d.datasets.segmentation.treeins][INFO] - Compose([ SaveOriginalPosId, GridSampling3D(grid_size=0.2, quantize_coords=False, mode=last), ]) Done! test_data_list: [[Data(instance_labels=[9383781], pos=[9383781, 3], y=[9383781])]] [2024-11-11 01:07:40,606][torch_points3d.datasets.segmentation.treeins][INFO] - pre_collate_transform ... [2024-11-11 01:07:40,606][torch_points3d.datasets.segmentation.treeins][INFO] - Compose([ SaveOriginalPosId, GridSampling3D(grid_size=0.2, quantize_coords=False, mode=last), ]) [2024-11-11 01:07:46,299][torch_points3d.applications.minkowski][WARNING] - Minkowski API is deprecated in favor of the SparseConv3d API. It should be a simple drop in replacement (no change to the API). [2024-11-11 01:07:46,300][torch_points3d.applications.modelfactory][INFO] - The config will be used to build the model [2024-11-11 01:07:46,403][torch_points3d.applications.minkowski][WARNING] - Minkowski API is deprecated in favor of the SparseConv3d API. It should be a simple drop in replacement (no change to the API). [2024-11-11 01:07:46,403][torch_points3d.applications.modelfactory][INFO] - The config will be used to build the model [2024-11-11 01:07:46,452][torch_points3d.applications.minkowski][WARNING] - Minkowski API is deprecated in favor of the SparseConv3d API. It should be a simple drop in replacement (no change to the API). [2024-11-11 01:07:46,452][torch_points3d.applications.modelfactory][INFO] - The config will be used to build the model Traceback (most recent call last): File "eval.py", line 13, in main trainer = Trainer(cfg) File "/home/nibio/mutable-outside-world/torch_points3d/trainer.py", line 48, in init self._initialize_trainer() File "/home/nibio/mutable-outside-world/torch_points3d/trainer.py", line 91, in _initialize_trainer self._model: BaseModel = self._checkpoint.create_model( File "/home/nibio/mutable-outside-world/torch_points3d/metrics/model_checkpoint.py", line 181, in create_model model = instantiate_model(OmegaConf.create(run_config), dataset) File "/home/nibio/mutable-outside-world/torch_points3d/models/model_factory.py", line 44, in instantiate_model model = model_cls(model_config, "dummy", dataset, modellib) File "/home/nibio/mutable-outside-world/torch_points3d/models/panoptic/PointGroup3heads.py", line 99, in init thing_classes = dataset.thing_classes AttributeError: 'TreeinsFusedDataset' object has no attribute 'thing_classes'

[julesmorel]

I just saw this has been fixed in a pull request

[keillernogueira]

Hi @julesmorel , I am still facing this issue. Have you done anything else (besides pulling the updated code)?

[dluks]

I'm also encountering this issue with the latest commit. It looks like commit 8088462d6960d24ca71f6cde2e7589744a3ae1fb from #9 tries to address this, but the author also later said they're still encountering an issue.

[keillernogueira]

So, should we reopen this issue then?

[dluks]

I think generally an issue shouldn't be closed until a fix has been merged into main, so I'd vote to re-open. I can confirm that the fix proposed in #9 does appear to address this specific issue, but another issue arises when saving the segmentation results which may or may not be linked to issue reported here. Meanwhile, I think conversation could be moved to #9.

[julesmorel]

Indeed, I might have closed the issue a bit too fast.

On my side, the modifications torch_points3d/datasets/panoptic/treeins.py have solved the issue.

[dluks]

Great that you got it working on your end! Were the segmentation results reasonable? And I'm guessing you didn't encounter the same issue as reported at the end of #9 (type casting issue when merging and saving the points and segmentations)?