This is a simple tutorial to teach you setup your own develop environment.
You can use the file "create_env.sh" to create a virtual environment that contains TensorFlow 1.12.0 with cuda 10
The command to create a new virtual environment is:
./create_env.sh <env name>
Before you start any program, please activate your virtual environment first, otherwise you wouldn't execute it successfully.
The command to activate virtual environment is:
conda activate <env name>
After you activate your virtual environment, you can see "()" at the left most side of your prompt, such as:
(test_env) su@2080-server ~ $
// The test_env would be your virtual environment name
If you forget your virtual environment name, or you want to check all the environment you have, you can use following command to display virtual environments:
conda env list
"vir env" stand for virtual environemnt
Because we need to install packages in your target vir env, there are two approaches to install packages:
- pip
python -m pip install <package>
Don't use pip install <package>
- conda
`conda install
There are also two method to show your packages installed in your virtual environment:
- pip
This method will display the packages you installed by the pip commands
pip freeze
- conda
This method will display the packages you installed by the conda commands
conda list
If you wish to use jupyter notebook to develop your program, you can use following command to start up your jupyter-notebook.
jupyter-notebook -h <host> -p <port>
The you specify is your ip address, otherwise the jupyter will automatically running on the localhost(127.0.0.1) that will not be able to access from other host. The you specify need to be in the range of 1024-49151, if you chose the sample port as others, it would report some error. Just change the port you choose, and try again.