when I set the batchsize 1, I found the the shape of feature from the PillarFeatureNet becomes [64] instead of [N, 64]

[qfwysw]

If you do not know the root cause of the problem / bug, and wish someone to help you, please post according to this template:

Instructions To Reproduce the Issue:

1. what changes you made (git diff) or what code you wrote

   samples_per_gpu=1,
   workers_per_gpu=1,

2. what exact command you run: python tools/train.py examples/point_pillars/configs/kitti_point_pillars_mghead_syncbn.py
3. what you observed (including the full logs):

   2021-09-14 16:50:33,540 - INFO - workflow: [('train', 5), ('val', 1)], max: 100 epochs
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([8550, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([10580, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([10413, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([9167, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([12000, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([12000, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([12000, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([10283, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([12000, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([12000, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([11906, 64])
   The shape of feature from the PillarFeatureNet
   input_features.shape:  torch.Size([64])
   Traceback (most recent call last):
   File "/home/wensuinan/program/centerpoint/det_modified/tools/train.py", line 132, in <module>
   main()
   File "/home/wensuinan/program/centerpoint/det_modified/tools/train.py", line 127, in main
   logger=logger,
   File "/home/wensuinan/program/centerpoint/det_modified/det3d/torchie/apis/train.py", line 325, in train_detector
   trainer.run(data_loaders, cfg.workflow, cfg.total_epochs, local_rank=cfg.local_rank)
   File "/home/wensuinan/program/centerpoint/det_modified/det3d/torchie/trainer/trainer.py", line 537, in run
   epoch_runner(data_loaders[i], self.epoch, **kwargs)
   File "/home/wensuinan/program/centerpoint/det_modified/det3d/torchie/trainer/trainer.py", line 404, in train
   self.model, data_batch, train_mode=True, **kwargs
   File "/home/wensuinan/program/centerpoint/det_modified/det3d/torchie/trainer/trainer.py", line 363, in batch_processor_inline
   losses = model(example, return_loss=True)
   File "/home/wensuinan/anaconda3/envs/centerpoint/lib/python3.6/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
   result = self.forward(*input, **kwargs)
   File "/home/wensuinan/program/centerpoint/det_modified/det3d/models/detectors/two_stage.py", line 169, in forward
   out = self.single_det.forward_two_stage(example, return_loss, **kwargs)
   File "/home/wensuinan/program/centerpoint/det_modified/det3d/models/detectors/point_pillars.py", line 74, in forward_two_stage
   x = self.extract_feat(data)     # 3, 384, 248, 216
   File "/home/wensuinan/program/centerpoint/det_modified/det3d/models/detectors/point_pillars.py", line 28, in extract_feat
   input_features, data["coors"], data["batch_size"], data["input_shape"]
   File "/home/wensuinan/anaconda3/envs/centerpoint/lib/python3.6/site-packages/torch/nn/modules/module.py", line 727, in _call_impl
   result = self.forward(*input, **kwargs)
   File "/home/wensuinan/program/centerpoint/det_modified/det3d/models/readers/pillar_encoder.py", line 198, in forward
   voxels = voxel_features[batch_mask, :]
   IndexError: too many indices for tensor of dimension 1

4. please also simplify the steps as much as possible so they do not require additional resources to run, such as a private dataset.

Expected behavior:

If there are no obvious error in "what you observed" provided above, please tell us the expected behavior.

If you expect the model to converge / work better, note that we do not give suggestions on how to train a new model. Only in one of the two conditions we will help with it: (1) You're unable to reproduce the results in model zoo. (2) It indicates a bug in Det3D.