Memories of a piece:
programmer black: the interviewer Hello, I'm proficient in java, python, C ++, C , php, mysql, spark, IDEA and other commonly used software installation and uninstallation !
In order to be able to blow water in front of the interviewer in the future, at least have something to say, I have to summarize the installation of these commonly used machine learning libraries, hehe
Update pip:
pip install --upgrade pip
python -m pip install --upgrade pip
Install keras
pip install keras
Install sklearn:
pip install -U scikit-learn
Install tensorflow:
Official website: tensorflow official website 1. CPU
version:
pip install tensorflow-cpu
pip install tensorflow (这样也是默认安装cpu版本)
pip install tensorboard
2.GPU version:
pip install tensorflow-gpu
Install the preview version (non-stable version):
pip install tf-nightly-cpu (不稳定版本)
pip install tf-nightly-cpu (不稳定版本)
3. DDL problem when importing tensorflow:
vc_redist.x64.exe address: download address
4. Tensorflow GPU version pre-cuda and cudnn requirements : GPU support requirements
5. Visual experience of tensorflow deep algorithm process (visual operation to realize deep learning algorithm) : click here
Install pytorch:
Official website: pytorch official website 1. CPU
version:
pip install torch==1.5.1+cpu torchvision==0.6.1+cpu
pip install torch==1.5.1+cpu torchvision==0.6.1+cpu -f https://download.pytorch.org/whl/torch_stable.html
2. GPU version:
add Tsinghua source of putorch:
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/free/
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main/
conda config --set show_channel_urls yes
conda config --add channels https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud/pytorch/
installation:
pip install torch==1.5.1+cu101 torchvision==0.6.1+cu101
官方源安装如下:
pip install torch==1.5.1+cu101 torchvision==0.6.1+cu101 -f https://download.pytorch.org/whl/torch_stable.html
或者直接:(此方式无需配置清华pytorch源)
pip install torch===1.3.0 torchvision===0.4.1 -i https://pypi.tuna.tsinghua.edu.cn/simple
Install Baidu paddlepaddle
Official website: Come in and take a closer look. To be
honest, I personally feel that this framework is really no worse than the above two.
1.CPU version
pip install paddlepaddle -i https://mirror.baidu.com/pypi/simple
conda install paddlepaddle
2. GPU version:
pip install paddlepaddle-gpu==1.8.2.post107 -i https://mirror.baidu.com/pypi/simple
conda install paddlepaddle-gpu cudatoolkit=10.0
The cuda and cudnn installation methods that need to be installed for the GPU version (under naconda):
0. Reference table: Click to enter
1. In anaconda environment, if you use the conda command to install the gpu version of TF or pytorch and other learning frameworks, then cuda and cudnn will be installed automatically: (This method is not slow on the computer and the network speed is good It is recommended to use, otherwise the process of conda searching for packages is very slow, very slow, very slow!)
conda install tensorflow-gpu
2. Download cuda and cudnn to install according to the computing power requirements of the framework and the graphics card version of your computer, and configure the environment variables.
OpenCV installation
pip install opencv-contrib-python
NLTK installation
pip install NLTK
installation of virtualenv
pip install virtualenv
Fabric installation
pip install fabric
Installation of matplotlib
pip install matplotlib
Installation of scikit-image
pip install scikit-image
Installation of pandas and numpy:
pip install pandas
pip install numpy
other:
Domestic source collection-py source configuration
conda common errors and solutions
It is my honour to sum up the bad places, and I hope the elder brother passing by can give me some pointers. Thank you very much.