Pytoch yolov5 runs bug solution from 0:

Directory hyperlink ：

• Q: You need to install the corresponding version of pytorch 、 torchvision 、opencv Wait for a series of libraries , Where to install it ？（ How to change the source perfectly ）
• Q: import cv2 appear libGL.so.1: cannot open shared object file: No such file or directory
• Q:AttributeError: ‘Upsample’ object has no attribute ‘recompute_scale_factor’

technological process ： docker After creating the image , Installation in mirror anaconda after , Create good yolov5 Environment

Q: You need to install the corresponding version of pytorch 、 torchvision 、opencv Wait for a series of libraries , Where to install it ？（ How to change the source perfectly ）

Details： however Anacodna The default source is foreign , It will cause us to install things very slowly ,pip (python The source of package management tools is also foreign , We need to change the source 、 Or when it's installed , Temporary source change )

resolvent ：

1. anaconda When adding sources , Be sure to write http Do not write https

#  Execute the command on the terminal command line 
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/menpo/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --add channels http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --set show_channel_urls yes

After execution , You'll find out home There is one more... In the catalog .condarc The file of

• vim ~/.condarc
  The contents are as follows ：

channels:
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/menpo/
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/bioconda/
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/msys2/
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/conda-forge/
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
  - http://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/

thus , Source change successful , According to the official website pytorch Install the command to install .

• Bear in mind , Remove the end of -c Ptyorch ( -c, Refer to -channel from Pytorch Official website installation , Particularly slow , There is a high probability of failure )
• Pytorch Official website ：
  • https://pytorch.org/
  • The old version ：https://pytorch.org/get-started/previous-versions/

2. pip Source change

# tsinghua ：
https://pypi.tuna.tsinghua.edu.cn/simple
# Alibaba cloud ：
http://mirrors.aliyun.com/pypi/simple/
# douban ：
http://pypi.douban.com/simple/

• yolov5 Under the installation command （ Directly pull and install yolov5 All libraries under the corresponding version ）：

pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple –-trusted-host pypi.douban.com

Q：import cv2 appear libGL.so.1: cannot open shared object file: No such file or directory

resolvent ：

pip install opencv-python-headless

Reference from ：https://blog.csdn.net/qq_50195602/article/details/124188467

This brother explained more thoroughly , Even talked about how to dockerfile Add installation instructions , To avoid this problem .

Q:AttributeError: 'Upsample' object has no attribute 'recompute_scale_factor'

describe ： My reasoning yolov5 The error occurred when demonstrating the example image .

reason ： torch and torchvision Version mismatch caused by ;

resolvent ：

1. Install the matching version
2. change upsample.py The code in , Remove the parameter of the function

Reference resources ： https://blog.csdn.net/qq_40280673/article/details/125095353