Tip: Install the corresponding CUDA version according to your graphics card type . Turing (Turing) and Ampere (Ampere) are two advanced GPU architectures of Nvidia Turing GPU
( 12nm manufacturing process): RTX20 series and GTX16 series
Ampere GPU (8nm manufacturing process): RTX30 series, including GeForce RTX3090, RTX3080, RTX3070, etc.
Article directory
foreword
With the continuous development of artificial intelligence, the technology of machine learning is becoming more and more important. Many people have started to learn machine learning. This article introduces the detailed installation tutorial of paddlepaddle (flying paddle) under the domestic Baidu under the windows system.
1. Enter Flying Paddle official website
Click the link to see the following screenshot:
After opening the connection, execute the following command to install in the installation information bar and copy it to the Dos window to download the paddlepaddle-gpu package
2. Install CUDA
1. CUDA download address
Compare your graphics card type according to the prompt at the top. CUDA10.2 version is recommended for GPUs with non-ampere architecture , and CUDA11
or later for GPUs
with ampere architecture Select the recommended installation option, and your system environment will be detected during the CUDA installation process. If you have previously installed other versions of CUDA and need to uninstall it in advance, you can uninstall it yourself on Baidu, otherwise it cannot be installed. In addition, if your computer does not have a VS environment, there will be a warning about the lack of visual studio in the middle of the process. After checking this warning, you can ignore it and continue the next installation. The screen will flicker during installation, which is a normal phenomenon. Restart the computer after installation
2. CUDA environment variable
The above are the environment variables that need to be configured after the default installation path of CDUA
2. Install cuDNN
1.cuDNN download address
Unfortunately, to download cuDNN, you must register a free NVIDIA developer account. After visiting the link below, click the Join now button to register, and just verify it in the email you received halfway (Note: The delay in receiving emails may be a bit high, Wait patiently!). When you have an account and log in, you need to fill in some reasons for downloading, just fill in as you like! When you log in, if you need email verification and the page keeps turning in circles, wait for the email verification to complete before you can enter the download page
This time, take the cuDNN download corresponding to CUDA10.2 version as an example. The cuDNN version corresponding to different versions of CUDA has prompt instructions in the download link.
2. Copy the cuDNN file to the CUDA corresponding file
After the installation is complete, you can use python to enter the python interpreter, enter import paddle, and then enter paddle.utils.run_check()
If PaddlePaddle is installed successfully! appears, it means you have successfully installed
Summarize
The Windows system Paddle GPU version is installed! ! !
For the installation of the TensorFlow framework, please refer to my other blog post:
https://blog.csdn.net/weixin_47834823/article/details/107893851