Train without Lidar

[Mariusmarten]

When leaving out the --with-lidar flag during training, I get the following error:

Traceback (most recent call last): File "train.py", line 1853, in <module> main() File "train.py", line 1236, in main train_metrics = train_one_epoch( File "train.py", line 1374, in train_one_epoch output = model(input) File "/usr/local/lib/python3.8/dist-packages/torch/nn/modules/module.py", line 1501, in _call_impl return forward_call(*args, **kwargs) File "/app/InterFuser/interfuser/timm/models/interfuser.py", line 987, in forward lidar = x["lidar"] KeyError: 'lidar'

Is there a specific flag for training without lidar? How can I reproduce the ablation results in the paper where training was done without Lidar?

[deepcs233]

Hi!

We haven't completed the related functions about this flag. You may delete https://github.com/opendilab/InterFuser/blob/e4f0314482124bb06a475c3f6fb4bfe3a2701c4d/interfuser/timm/models/interfuser.py#L987 and https://github.com/opendilab/InterFuser/blob/e4f0314482124bb06a475c3f6fb4bfe3a2701c4d/interfuser/timm/models/interfuser.py#L1008

[Mariusmarten]

Thank you for the quick response! Commenting out lidar throughout the models/interfuser.py file solved the problem.