Deep learning framework installation (tensorflow & pytorch & paddlepaddle)

One 、 Preface

The steps of installing the deep learning framework explained in this article are based on Anaconda Of , So you want to follow the steps in this article to install the deep learning framework , You need to install Anaconda（ Be careful when changing sources ）.

Experimental environment ：Windows10,NVIDIA GeForce GTX 1050 Ti.

Two 、PaddlePaddle Deep learning framework installation

2.1 brief introduction

PaddlePaddle（ Flying oar ） It is a domestic deep learning framework developed by Baidu , use PaddlePaddle The advantage of this method is that it can be used AI Studio Platform GPU Model training with calculating power , Not only save time, but also free ,PaddlePaddle There are also offers like PaddleSeg Wait for some kits , It is very friendly for beginners to start the project of deep learning model . Of course , Other frameworks can also use some online computing platforms for model training .

2.2 PaddlePaddle CPU Version installed

Step one ： stay Anaconda Create a virtual environment in .

stay cmd in （cmd The opening method of can baidu ） Input “conda create -n paddle_cpu python=3.8”, Creating a virtual environment , Creating a virtual environment can be analogous to creating a virtual environment named “paddle_cpu” Folder .

Input “y”, And then go back .

Step two ： Input “activate paddle_cpu”. Enter the created environment .

Step three ： Input ：

“conda install paddlepaddle==2.2.2 --channel https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/Paddle/”

Conduct paddlepaddle（CPU edition ） Installation of deep learning framework .

Step four : Test for successful installation . Input in sequence “python”, enter ,"import paddle", enter ,"paddle.utils.run_check()", enter .

If the above figure appears , The installation is successful .

2.3 PaddlePaddle GPU Version installed

Step one ： stay Anaconda Create a virtual environment in . stay cmd Input in “conda create -n paddle_gpu python=3.8”, Creating a virtual environment .

Step two ： Input “activate paddle_gpu”. Enter the created environment .

Step three ： Input ：

conda install paddlepaddle-gpu==2.2.2 cudatoolkit=10.1 --channel https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/Paddle/, install paddlepaddle（GPU edition ）,CUDA10.1 And the matching cuDNN.

Input “y”.

As can be seen from the above figure ,CUDA and cuDNN Will be installed together , So there is no need to configure it manually .

appear “done”, Description installation complete .

Step four ： Test for successful installation . Input in sequence “python”, enter ,import paddle, enter ,paddle.utils.run_check(), enter .

If the above figure appears , The installation is successful （ As can be seen from the above figure device 0 as well as cuDNN Version of ）.

3、 ... and 、PyTorch Deep learning framework installation

3.1 brief introduction

PyTorch It's an open source Python Machine learning library , be based on Torch, For natural language processing applications .

3.2 PyTorch CPU Version installed

Step one ： stay Anaconda Create a virtual environment in . stay cmd Input in “conda create -n pytorch_cpu python=3.8”, Creating a virtual environment .

Input “y”.

Step two ： Input “activate pytorch_cpu”. Enter the created environment .

Step three ： Input ：

conda install pytorch torchvision torchaudio cpuonly -c pytorch, install PyTorch CPU Version deep learning framework .

Input “y”.

Waiting for the installation .

appear “done”, Description installation complete .

Step four ： Test for successful installation . Input in sequence python, enter ,import torch, enter ,x = torch.rand(5, 3), enter ,print(x), enter .

If the above figure appears , The installation is successful .

3.3 PyTorch GPU Version installed

Step one ： stay Anaconda Create a virtual environment in . stay cmd Input in “conda create -n pytorch_gpu python=3.8”, Creating a virtual environment .

Step two ： Input “activate pytorch_gpu”. Enter the created environment .

Step three ： Input “conda install pytorch torchvision torchaudio cudatoolkit=10.2 -c pytorch”.

Input “y”.

appear “done”, Description installation complete .

Step four ： Test for successful installation . Input in sequence python, enter ,import torch, enter ,x = torch.rand(5, 3), enter ,print(x), enter , Input torch.cuda.is_available(), enter . among torch.cuda.is_available() It's used to test CUDA Of （ Used to distinguish from CPU Version testing ）.

If the above figure appears , The installation is successful .

Four 、Tensorflow Deep learning framework installation

4.1 brief introduction

TensorFlow yes Google The second generation of open source is used for digital computing (numerical computation) Software library .

4.2 Tensorflow CPU Version installed

Step one ： stay cmd Input in “conda create -n tf2xx_cpu” Creating a virtual environment .

Step two ： Input “activate tf2xx_cpu” Enter the virtual environment , Type the following command to install python and tensorflow（CPU edition ）：

conda install --channel https://mirrors.ustc.edu.cn/anaconda/pkgs/main/ python=3.7 tensorflow==2.1.0 .

Input “y”.

Step three ： Test for successful installation , Input in sequence “python”, enter ,“import tensorflow as tf”, enter ,“import os”, enter ,“os.environ['TF_CPP_MIN_LOG_LEVEL']='2'”, enter ,“print(tf.reduce_sum(tf.random.normal([1000, 1000])))”.

If the same content as above appears , This indicates that the installation was successful .

4.3 Tensorflow GPU Version installed （ Pay attention to your graphics card model , My is 1050TI（ If the graphics card is better than mine , You can also install 2.1.0 edition ）, The graphics card driver is relatively old , So the installation 2.1.0 edition .）

Step one ： stay cmd Input in “conda create -n tf2xx_gpu”, Creating a virtual environment .

Step two ： Input “activate tf2xx_gpu” Enter the virtual environment , And enter the following instructions to install python and Tensorflow-gpu edition ：

conda install --channel https://mirrors.ustc.edu.cn/anaconda/pkgs/main/ python=3.7 tensorflow-gpu==2.1.0.

You can see from the above picture that , The installation includes CUDA and cuDNNN.

Input “y”.

Step three ： Test for successful installation . Input in sequence “python”, enter ,“import tensorflow as tf”, enter ,“print("Num GPUs Available: ", len(tf.config.experimental.list_physical_devices('GPU')))”.

If the same content as above appears , This indicates that the installation was successful .

5、 ... and 、 Conclusion

This paper introduces the installation of three deep learning frameworks , actually , You can choose a deep learning framework to install . In deep learning projects , The framework can only be said to be a tool , Personal programming ability is the core .

6、 ... and 、 The content of the next article

In the next article, I plan to use the deep learning framework to achieve target detection （ Including face recognition ）, The main content is “ Brief introduction ”+“ Detailed code ”, I will try to annotate the code in detail , Try to make beginners understand , If you don't understand, please feel free to contact me . Recently, a confrontation network has been generated (GAN,Generative Adversarial Networks) Comparing the fire , Especially will GAN As the topic of graduation design , So I will try my best to write a few later GAN The content of .

in addition , If you have problems with your undergraduate thesis, you can also contact me （ Free of charge , Because I want to do more good , Gradually become a kind person ）, I reply in the evening , Practice in the company during the day , If the problem is urgent , It can also be direct +q 1031794256 perhaps vx：17860157407.