Installing

Python Version: we require at least Python 3.9.

Environment Installation

We currently do not provide a conda build package of Ocean4DVarNet but only a pypi package.

However, Ocean4DVarNet is based on CUDA/Pytorch. It can be complicated to install this environment from the pypi package.

So the suggested installation is through mamba/conda for the CUDA/Pytorch environment, and then pip form other dependencies.

For Linux the process to make and use a mamba environment is as follows :

mamba create --name "ocean-code" python=3.12 -y
mamba activate ocean-code

Then the requirements must be installed in the environment :

mamba env update -f environment.yaml

More informations about how to deploy a full environment can be found in the Development environment set up section.

To install the package and his dependencies with pip, you can use the following command:

pip install ocean4dvarnet

or

python -m pip install ocean4dvarnet

You can also get a zip or tgz archive of the package from github repository https://github.com/CIA-Oceanix/ocean4dvarnet/releases

pip install ocean4dvarnet-contrib-x.x.x.tar.gz

And you can use the last development version directly from github repository

pip install git+https://github.com/CIA-Oceanix/ocean4dvarnet.git