【PointSet】の 填坑指南

---

问题1：No CUDA runtime is found, using CUDA_HOME=’/usr/local/cuda-9.0’

原因分析：

可能是CUDA 版本与pytorch不匹配，可以先通过以下指令确定CUDA的实际版本，如果不一致时需要对文件进行修改。

nvcc -V
cat /usr/local/cuda/version.txt
nvidia-smi

我的cuda版本一开始查看时也不一致：

解决方案：

修改根目录下的 .bashrc 文件，在文件末尾添加以下两行，使版本一致

export PATH="/usr/local/cuda-10.1/bin:$PATH"
export LD_LIBRARY_PATH="/usr/local/cuda-10.1/lib64:$LD_LIBRARY_PATH"

然后再source一下即可：

source ~/.bashrc

最后 .bashrc 文件末尾是介个样子的。

然后根据 CUDA 版本安装好对应的 pytorch 版本即可。

参考链接：

• 安装多版本cuda时，nvcc和cuda版本不一致问题
• 安装CUDA时，nvcc --version和cat /usr/local/cuda/version.txt版本不一致
• FCOS出现No CUDA runtime is found, using CUDA_HOME=’/usr/local/cuda-10.0’

---

问题2：ImportError: deform_conv_cuda…undefined symbol: cudaSetupArgument

ImportError: /data3/zhanghuiyong/PointSetAnchor/mmdet/ops/dcn/deform_conv_cuda.cpython-36m-x86_64-linux-gnu. so: undefined symbol: cudaSetupArgument

解决方案：

解决方法同问题一，都是由于CUDA版本不同造成的。修改.bashrc文件即可。

---

问题3：CondaHTTPError: HTTP 000 CONNECTION FAILED for url …

CondaHTTPError: HTTP 000 CONNECTION FAILED for url https://repo.continuum.io/pkgs/main/linux-64/repodata.json.bz2
Elapsed: -
An HTTP error occurred when trying to retrieve this URL.
HTTP errors are often intermittent, and a simple retry will get you on your way.
ConnectTimeout(MaxRetryError(“HTTPSConnectionPool(host=‘repo.continuum.io’, port=443): Max retries exceeded with url: /pkgs/main/linux-64/repodata.json.bz2 (Caused by ConnectTimeoutError(<urllib3.connection.VerifiedHTTPSConnection object at 0x7fa22dc99b70>, ‘Connection to repo.continuum.io timed out. (connect timeout=9.15)’))”,),)

解决方案：

将根目录下 .condarc 中的默认源删除。

参考链接：

• 解决Anaconda出现CondaHTTPError: HTTP 000 CONNECTION FAILED for url问题
• 有效解决CondaHTTPError: HTTP 000 CONNECTION FAILED for url问题

---

问题4：Connection to pypi.org timed out. (connect timeout=15)’): /simple/cython/

WARNING: Retrying (Retry(total=4, connect=None, read=None, redirect=None, status=None)) after connection broken by ‘ConnectTimeoutError(<pip._vendor.urllib3.connection.HTTPSConnection object at 0x7f745ad06a90>, ‘Connection to pypi.org timed out. (connect timeout=15)’)’: /simple/cython/

解决方案：

更换 python 镜像源：

临时：
pip命令 + -i + 镜像地址。
例如 pip install numpy -i https://mirrors.aliyun.com/pypi/simple/
永久：
在更新到最新版本pip后，pip config set global.index-url+国内镜像地址，
例如 pip config set global.index-url https://mirrors.aliyun.com/pypi/simple/

参考链接：

• 在python中安装包出现Retrying (Retry(total=4, connect=None, read=None, redirect=None, status=None))
• Python下载库超时Connection to files.pythonhosted.org timed out问题解决（换源）

---

问题5：ERROR: Command errored out with exit status 1: python setup.py egg_info Check the logs for full command output.

ERROR: Command errored out with exit status 1:
command: /usr/bin/python -c ‘import sys, setuptools, tokenize; sys.argv[0] = ‘"’"’/tmp/pip-install-c2wuc3/gpustat/setup.py’"’"’; file=’"’"’/tmp/pip-install-c2wuc3/gpustat/setup.py’"’"’;f=getattr(tokenize, ‘"’"‘open’"’"’, open)(file);code=f.read().replace(’"’"’\r\n’"’"’, ‘"’"’\n’"’"’);f.close();exec(compile(code, file, ‘"’"‘exec’"’"’))’ egg_info --egg-base /tmp/pip-pip-egg-info-eC0Jl9

解决方案：

在终端执行以下命令：

pip install --upgrade pip
pip install --upgrade setuptools
pip install django-import-export

参考链接：

• ERROR: Command errored out with exit status 1: python setup.py egg_info Check the logs for full command output.

---

问题6：It seems that scikit-image has not been built correctly.Your install of scikit-image appears to be broken

ImportError: dlopen: cannot load any more object with static TLS.It seems that scikit-image has not been built correctly. Your install of scikit-image appears to be broken. Try re-installing the package following the instructions at: https://scikit-image.org/docs/stable/install.html

解决方案：

降低 scikit-image 版本到 0.14.2

pip uninstall scikit-image
pip install scikit-image==0.14.2

降低 imagecorruptions 版本到 0.1.0

pip uninstall imagecorruptions
pip install imagecorruptions==0.1.0

可以根据自己的实际情况降低版本，对我而言之前的版本一直报错，只有降到这个版本时才没错。

---

问题7：It seems that scikit-learn has not been built correctly.

解决方案：

降低 scikit-learn 版本到 0.20.4

pip uninstall scikit-learn
pip install scikit-learn==0.20.4

参考链接：

• python：It seems that scikit-learn has not been built correctly

---

问题8：ImportError: cannot import name ‘get_dist_info’

解决方案：

修改 mmdet/datasets/loader/sampler.py 文件
在文件开头附近，将 from mmcv.runner.utils import get_dist_info 修改为 from mmcv.runner import get_dist_info

参考链接：

• ImportError: cannot import name ‘get_dist_info’ #1862

---

问题9：ModuleNotFoundError: No module named ‘mmcv.cnn.weight_init’

ModuleNotFoundError: No module named ‘mmcv.cnn.weight_init’

解决方案：

对于我的问题而言，是因为没有把mmcv文件夹放在 PointSetAnchor 的根目录下，所以才会找不到 weight_init 文件。

---

问题10：cv2.error: OpenCV(4.1.1) /io/opencv/modules/highgui/src/window.cpp:627: error

cv2.error: OpenCV(4.1.1) /io/opencv/modules/highgui/src/window.cpp:627: error: (-2:Unspecified error) The function is not implemented. Rebuild the library with Windows, GTK+ 2.x or Cocoa support. If you are on Ubuntu or Debian, install libgtk2.0-dev and pkg-config, then re-run cmake or configure script in function ‘cvShowImage’

解决方案：

安装 opncv-contrib 包

pip install --upgrade pip
pip install opencv-contrib-python

参考链接：

• The function is not implemented. Rebuild the library with Windows, GTK+ 2.x or Carbon support.

---