OpenRobotLab / EmbodiedScan

[CVPR 2024] EmbodiedScan: A Holistic Multi-Modal 3D Perception Suite Towards Embodied AI
https://tai-wang.github.io/embodiedscan/
Apache License 2.0
395 stars 26 forks source link

[Docs] The training command for MVDET #47

Closed RongkunYang closed 2 months ago

RongkunYang commented 2 months ago

Branch

main branch https://mmdetection3d.readthedocs.io/en/latest/

📚 The doc issue

In the readme.md, this doc provide the command for training: python tools/train.py configs/detection/mv-det3d_8xb4_embodiedscan-3d-284class-9dof.py --work-dir=work_dirs/mv-3ddet --launcher="pytorch"

But this command run get error, because the launcher="pytorch" is set for distributed training. The error is below:

""" anaconda3/envs/embodiedscan/lib/python3.8/site-packages/mmengine/dist/utils.py", line 102, in _init_dist_pytorch rank = int(os.environ['RANK']) File "/anaconda3/envs/embodiedscan/lib/python3.8/os.py", line 675, in getitem raise KeyError(key) from None KeyError: 'RANK' """

It seems that the distributed training parameters are not set during training, could you please share the detail running command about this, thank you very much.

Suggest a potential alternative/fix

No response

mxh1999 commented 2 months ago

The command in README.md has some faults, please remove the --launcher parameter if you are using single GPU for training.

We will update README.md soon, thanks for the report.