It crashed when training the vad.
Closed this issue · 3 comments
as indicated by the title, and the log are listed bellow,
KeyError: Caught KeyError in DataLoader worker process 4.
Original Traceback (most recent call last):
File "/usr/local/lib/python3.8/dist-packages/torch/utils/data/_utils/worker.py", line 308, in _worker_loop
data = fetcher.fetch(index)
File "/usr/local/lib/python3.8/dist-packages/torch/utils/data/_utils/fetch.py", line 58, in fetch
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/usr/local/lib/python3.8/dist-packages/torch/utils/data/_utils/fetch.py", line 58, in
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/nvme_17t/project/e2e/mmcv/datasets/B2D_vad_dataset.py", line 361, in getitem
data = self.prepare_train_data(idx)
File "/nvme_17t/project/e2e/mmcv/datasets/B2D_vad_dataset.py", line 92, in prepare_train_data
gt_labels,gt_bboxes = self.get_map_info(index)
File "/nvme_17t/project/e2e/mmcv/datasets/B2D_vad_dataset.py", line 253, in get_map_info
map_info = self.map_infos[town_name]
KeyError: 'Junction'
I really do not know what happened...
It seems that you are using an old version of our data. We changed the name of some routes in the new version. You may solve the problem by replacing all "T_Junction" in route name with "TJunction".
data = [self.dataset[idx] for idx in possibly_batched_index]
File "/nvme_17t/project/e2e/mmcv/datasets/B2D_vad_dataset.py", line 361, in getitem
data = self.prepare_train_data(idx)
File "/nvme_17t/project/e2e/mmcv/datasets/B2D_vad_dataset.py", line 91, in prepare_train_data
example = self.pipeline(input_dict)
File "/nvme_17t/project/e2e/mmcv/datasets/pipelines/compose.py", line 40, in call
data = t(data)
File "/nvme_17t/project/e2e/mmcv/datasets/pipelines/formating.py", line 683, in call
results = super(VADFormatBundle3D, self).call(results)
File "/nvme_17t/project/e2e/mmcv/datasets/pipelines/formating.py", line 605, in call
results['gt_labels'] = np.array([
File "/nvme_17t/project/e2e/mmcv/datasets/pipelines/formating.py", line 606, in
self.class_names.index(n) for n in results['gt_names']
ValueError: 'walker.pedestrian.0042' is not in list
another crash....
@MandyMo Thanks for your question! We have updated configs, for VAD we updated NameMapping in https://github.com/Thinklab-SJTU/Bench2DriveZoo/blob/uniad/vad/adzoo/vad/configs/VAD/VAD_base_e2e_b2d.py. Please git pull latest branch.