エラー記録

C++

#tensorflow
ImportError: libcudnn.so.5: cannot open shared object file: No such file or directory
tensorflowのバージョンとcudnnのバージョンが合わない。
対応一覧：https://blog.csdn.net/qq_27825451/article/details/89082978
make benchmark / tensorflow/core/public/session.h: No such file
コンパイルの時でた。the path of tensorflow changed.To solve this problem, add a line below to your Makefile.
CFLAGS += -I/path/to/site-packages/tensorflow/include
#CUDA

  File "/home/Workspace/3d-object-detection_0302/pcdet/ops/iou3d_nms/iou3d_nms_utils.py", line 98, in nms_gpu
    num_out = iou3d_nms_cuda.nms_gpu(boxes, keep, thresh)
RuntimeError: Unrecognized tensor type ID: AutogradCUDA

environment: PyTorch 1.7.0
solution: c++ extension was compiled with PyTorch 1.6. Recompiling fixes the error.

パッケージ

fatal error: GL/glew.h
sudo apt-get install libglew-dev

Could not get lock /var/lib/dpkg/lock-frontend. It is held by process 8722 (unattended-upgr)
ubuntuが無人アップグレード（unattended-upgrade）している
https://ostechnix.com/how-to-disable-unattended-upgrades-on-ubuntu/

画像

TypeError: Cannot handle this data type: (1, 1, 3), <f4
PILのImage.imsave使う時に出た。PILのrgb画像保存は0－255のuintにだけ対応できる。

他

no test dir found testing here:
command: pytest test
テストファイルのフォーマットはtest_xx.pyじゃないと検測されない
py.testならxx_testも大丈夫

CMake

Could not find a package configuration file provided by "" with any of the following names

find_package(CeptionLib REQUIRED common stereo_detection_3d)

to

include_directories(
    include/
    ${CMAKE_SOURCE_DIR}/common/include/
    ${CMAKE_SOURCE_DIR}/modules/stereo_detection_3d/include/
    ${PCL_INCLUDE_DIRS}
)

add_definitions("-Wall -march=native -std=c++17 -g")
add_executable(${PROJECT_NAME} main.cpp)
target_link_libraries (${PROJECT_NAME} ${PCL_LIBRARIES}
                                       ${OpenCV_LIBS}
                                       ${TORCH_LIBRARIES}
                                       ${CMAKE_PROJECT_NAME}::common
                                       ${CMAKE_PROJECT_NAME}::stereo_detection_3d)

Docker

unknown file: Failure
C++ exception with description "Found no NVIDIA driver on your system. Please check that you have an NVIDIA GPU and installed a driver from http://www.nvidia.com/Download/index.aspx

nvidia-smi 結果：　CUDA Version: N/A
DOCKER IMAGEさいいｎ解決

mmdetection3d

##RTX30 series doesn't support CUDA10.2 while pytorch1.6 doesn't support CUDA11
Upgrade to pytorch1.7

version of mmdet

cannot import name 'build' from 'mmdet.models.builder' 

numpy

ValueError: numpy.ndarray size changed, may indicate binary incompatibility. Expected 88 from C header, got 80 from PyObject

Solution 1. installed 1.21

new problem when use create_data.py:

Traceback (most recent call last):
  File "scripts/create_data.py", line 128, in <module>
    max_sweeps=args.max_sweeps,
  File "scripts/create_data.py", line 26, in waymo_data_prep
    root_path, info_prefix, save_path=out_dir, max_sweeps=max_sweeps
  File "/home/tzhong/workspace/autoware_mmdetection3d/scripts/data_converter/kitti_converter.py", line 103, in create_waymo_info_file
    remove_outside=False,
  File "/home/tzhong/workspace/autoware_mmdetection3d/scripts/data_converter/kitti_converter.py", line 59, in _calculate_num_points_in_gt
    indices = box_np_ops.points_in_rbbox(points_v[:, :3], gt_boxes_lidar)
  File "/home/tzhong/workspace/autoware_mmdetection3d/mmdetection3d/mmdet3d/core/bbox/box_np_ops.py", line 434, in points_in_rbbox
    indices = points_in_convex_polygon_3d_jit(points[:, :3], surfaces)
  File "/home/tzhong/workspace/autoware_mmdetection3d/mmdetection3d/mmdet3d/core/bbox/box_np_ops.py", line 765, in points_in_convex_polygon_3d_jit
    normal_vec, d, num_surfaces)
TypeError: expected dtype object, got 'numpy.dtype[bool_]'

https://mmdetection3d.readthedocs.io/en/latest/faq.html
If you face the error shown below, and your environment contains numba == 0.48.0 with numpy >= 1.20.0:

TypeError: expected dtype object, got 'numpy.dtype[bool_]'

please downgrade numpy to < 1.20.0 or install numba == 0.48 from source, because in numpy == 1.20.0, np.dtype produces subclass due to API change. Please refer to here for more details.

Solution 2

downgrage pycocotools from 2.0.2 to 2.0.1

ERROR: pip's dependency resolver does not currently take into account all the packages that are installed. This behaviour is the source of the following dependency conflicts.
torch 1.7.0 requires dataclasses, which is not installed.
torch 1.7.0 requires future, which is not installed.

anaconda

no module called ' ' found

reboot

solving environment failed

conda config --set channel_priority false

torch/_C.cpython-37m-x86_64-linux-gnu.so: undefined symbol

>>> import torch
python: symbol lookup error: /home/tzhong/anaconda3/envs/mmdet3d/lib/python3.7/site-packages/torch/_C.cpython-37m-x86_64-linux-gnu.so: undefined symbol: _Z10initModulev

uninstall pytorch etc.
conda install pytorch==1.8.1 torchvision==0.9.1 cudatoolkit=11.3 -c pytorch -c conda-forge

>>> torch.cuda.is_available()
False
>>> torch.cuda.current_device()
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/home/tzhong/anaconda3/envs/mmdet3d/lib/python3.7/site-packages/torch/cuda/__init__.py", line 388, in current_device
    _lazy_init()
  File "/home/tzhong/anaconda3/envs/mmdet3d/lib/python3.7/site-packages/torch/cuda/__init__.py", line 164, in _lazy_init
    raise AssertionError("Torch not compiled with CUDA enabled")
AssertionError: Torch not compiled with CUDA enable

github

remote: Write access to repository not granted.

https://stackoverflow.com/questions/70538793/remote-write-access-to-repository-not-granted-fatal-unable-to-access
when you create your access token
In expiration: it should say No expiration.
In selecte scopes you mark the repo radio button.
As shown in the image below

gsutils

invalid_grant: Bad Request when using gsutils cp

gcloud auth revoke and gcloud auth login

ubuntu

disk