Notebook

Easily Install DeePMD-kit Through Offline Mode

2024-11-14

! Deprecated since version 3.0.0, Only used for versions below 3.0.0.

First, we need to find the released version on the DeePMD-kit GitHub page:

https://github.com/deepmodeling/deepmd-kit/tags

Then download the corresponding version of the sh file.

Download the package

Some versions need to download serval files and merge them before they can be installed.

For example, 2.2.9-GPU-CUDA118 need to download these two files:

deepmd-kit-2.2.9-cuda118-Linux-x86_64.sh.0

deepmd-kit-2.2.9-cuda118-Linux-x86_64.sh.1

Merge:

cat deepmd-kit-2.2.9-cuda118-Linux-x86_64.sh.0 deepmd-kit-2.2.9-cuda118-Linux-x86_64.sh.1 > deepmd-kit-2.2.9-cuda118-Linux-x86_64.sh

Set execution permissions

chmod +x deepmd-kit-2.2.9-cuda118-Linux-x86_64.sh

Install

./deepmd-kit-2.2.9-cuda118-Linux-x86_64.sh

Then enter the installation path needed. Sometimes using an absolute path may cause errors, so it is recommended to use the relative path during installation, do not use an existing folder as the installation path, as it will cause an error. A new folder will be automatically created during the installation process.

Then it will automatically install the related programs of DeePMD-kit.

If the end of the installation shows:You currently have a PYTHONPATH environment variable set

Don't worry, when activate this conda environment, it will temporarily override the system's default Python environment.

If the problem occurs, you can check the current PYTHONPATH: echo $PYTHONPATH

Temporarily unset the default PYTHONPATH: unset PYTHONPATH

When start to use, simply execute:

conda activate your_deepmd_installation_path

Alternatively, If used on HPC, it can be packaged as a module and called through module load