icenet 0.2.8

Library for operational IceNet forecasting

icenet

This is the core python library for the IceNet sea-ice forecasting system.

This README will be worked on more, but there's plenty of information around in the icenet-ai organisations repositories, which demonstrate usage of this library.

Table of contents

• Overview
• Installation
• Implementation
• Pipeline
• Contributing to IceNet
• Credits
• License

Installation

We're still working on clear dependency management using pip, Tensorflow is best through pip but obviously you need NVIDIA dependencies for GPU based training. If you're having trouble with system dependencies some advice about environment setup is given by the examples under the pipeline repository.

Please note that icenet has an optional dependency on eccodes which requires a system library and a python wrapper. The system library can be installed via conda.

By default, Tensorflow will be built for CPU only when icenet is installed via pip. So, optionally, Tensorflow can be installed with CUDA support (recommended).

conda install -c conda-forge eccodes

pip install icenet

# To install newer versions of tensorflow (tensorflow>=2.14) with CUDA deps directly via pip:
pip install tensorflow[and-cuda]<2.16

Please consult the tensorflow docs for up-to-date info.

Development installation

Please refer to the contribution guidelines for more information.

Implementation

When installed, the library will provide a series of CLI commands. Please use the --help switch for more initial information, or the documentation.

Documentation

The docs/ directory has a Makefile that builds sphinx docs easily enough, once the requirements in that directory are installed.

Usage Pipeline / Examples

Please refer to the icenet-pipeline repository or the icenet-notebook repository for examples of how to use this library.

Contributing to IceNet

Please refer to the contribution guidelines for more information.

Credits

License

This is licensed using the MIT License

---