Installation tutorial on CUDA+CUDNN in Windows Environment
Anaconda + pychar package is recommended.
This article describes how to install the pytorch and tensorflow frameworks
First of all, you should know which version of your graphics card driver is (take 1660S as an example)
1. Open the NVIDIA control panel, which can be opened by right clicking on the desktop or hiding the icon bar in the lower ...
Posted by poltort on Sun, 05 Dec 2021 01:46:12 -0800
yolov5 decoding is accelerated using GPU
The principle of YOLOv5 will not be described too much here. Start directly from the output head, and then design such as encoding and decoding:
1. The original output of yolov5 series is three head heads. The picture above is the picture with the input resolution of 608 * 608. If the input is changed to the pictu ...
Posted by ajanoult on Tue, 23 Nov 2021 05:10:20 -0800
Ubuntu 18.04 + Anaconda + CUDA + cudnn + Python environment configuration (updated on 2021.10, available for pro testing)
This set of environment is really too complex. There are too many pits. After more than half a year, it finally succeeded today.
Graphics card driver
Direct installation system recommends graphics card driver, with the lowest error probability.
sudo ubuntu-drivers autoinstall
View installation status
nvidia-smi
The graphics card version ...
Posted by Coronet on Fri, 29 Oct 2021 23:07:37 -0700
Pale yellow alchemy furnace (Part 3): Ubuntu 18.04 deep learning Server NVIDIA and CUDA related environment installation
total tips:
1. There is no need to install the graphics card driver first. CUDA toolkit comes with a driver. Install first, but report various errors
2 cudaToolKit must select runFile instead of deb, otherwise an error will be reported and the configuration cannot be selected during installation
3 install gcc. Ubuntu 18.04 installs versi ...
Posted by taldos on Fri, 01 Oct 2021 10:14:32 -0700