docker build and run problem

[yrt1993]

problems.txt

In the attached file are problems I met when I build docker and run the command in the CTRL_instructions.md. There is still several problems not solved:

1. when I runpython ./tools/ctrl/generate_track_input.py ./tools/ctrl/data_configs/fsd_base_vehicle.yaml --process 8

I want to know the mmcv-full version of yours.

2. when I run python ./tools/ctrl/generate_train_gt_bin.py

when I run python ./tools/ctrl/extract_poses.py

I want to know how to get the ./data directory

[Abyssaledge]

My environment is:

cumm-cu113                    0.3.4
mmcv-full                     1.3.9              
mmdet                         2.14.0
mmdet3d                       0.15.0            
mmsegmentation                0.14.1
spconv-cu113                  2.2.3
torch                         1.8.0+cu111
torchvision                   0.9.0+cu111
torch-scatter                 2.0.9

For the data, it is generated by the official MMDet3D script: ./tools/create_data.py. For this part, you could follow the instructions of MMDet3Dv0.15.

For idx2timestamp.pkl, please follow https://github.com/tusen-ai/SST/blob/main/docs/overall_instructions.md#fast-waymo-evaluation

[Abyssaledge]

To be more clear, waymo_format/training directory contains raw waymo data (xxx.tfrecord).

[yrt1993]

To be more clear, waymo_format/training directory contains raw waymo data (xxx.tfrecord).

That's what I want to know. and it seems pytorch >=1.8.1is needed to use torch.profiler

https://discuss.pytorch.org/t/problem-to-load-the-torch-profiler-module/127956

[yrt1993]

I got another problem when I runpython tools/create_data.py waymo --out-dir ./data/waymo

it seems tensorflow version does not match with my cuda version. This is what I used, and I want to know the tensorflow-gpu and waymo-open-dataset-tf version of yours

[Abyssaledge]

My tensorflow and waymo version is:

tensorflow                    2.4.0
waymo-open-dataset-tf-2-4-0   1.4.1

Besides, I usually follow the following shell script to setup my environment:

pip config set global.index-url https://pypi.tuna.tsinghua.edu.cn/simple
pip install tensorflow==2.4.0
pip3 install waymo-open-dataset-tf-2-4-0 --user
pip install torch-1.8.0+cu111-cp38-cp38-linux_x86_64.whl
pip install torchvision-0.9.0+cu111-cp38-cp38-linux_x86_64.whl
cd ./mmcv-1.3.9
TORCH_CUDA_ARCH_LIST="8.6" MMCV_WITH_OPS=1 pip install -e .
cd ..
pip install mmdet==2.14.0
pip install mmsegmentation==0.14.1
cd ./SST
pip install -e .

Hope it helps.

[yrt1993]

and your cuda version?

[Abyssaledge]

11.2

[20210726]

what's your python version?

[Abyssaledge]

3.8.x @20210726

[atlantins]

can you show the folder ./data/waymo, is there will be two folder kitti format and waymo format?

[atlantins]

can you show the pip list folder? there is always some conflicts in my configuration such as when i cd SST/ pip install -e.,
tensorflow 2.4.0 requires typing-extensions~=3.7.4, but you have typing-extensions 4.7.1 which is incompatible.

[Abyssaledge]

Tensorflow is always annoying. Just ignore the warning from TensorFlow, I guess it is not a big deal.

[atlantins]

Can you tell me the space occupied after converting to kittiformat? I have 2.8T, but during the conversion process, libpng error: Write Error is displayed

[Abyssaledge]

You could delete the raw .tar files to save the space. 2.8T should be enough. @atlantins

[atlantins]

When I generate my waymoopen, I often generate a directory larger than 2.8T when generating training. This is the directory generated by my training folder, and I have deleted some images. Our algorithm requires at least which folders. I don't know if you have encountered this problem, and if i have my own waymo-open dataset ,how to generate "The detection results should be in official WOD submission format(.bin)" , through mmdetection to inference traning scenes to generate bin?how do you get it?Can you share your running command with me, Thanks!

[Abyssaledge]

Please open a new issue. Otherwise, I may forget to answer your question when I am free. Thanks! @atlantins

[atlantins]

thanks a lot ! please see here https://github.com/tusen-ai/SST/issues/141