Getting Started#
ml4xcube
is a comprehensive Python-based toolkit designed for researchers and developers in the field of machine learning with an emphasis on xarray
data cubes. This toolkit is engineered to provide specialized and robust support for data cube management and analysis, operating with the state-of-the-art machine learning libraries (1) scikit-learn
, (2) PyTorch
and (3) TensorFlow
.
Installation#
Get started with ml4xcube
effortlessly by installing it directly through pip:
pip install ml4xcube
or conda:
conda install -c conda-forge ml4xcube
Features#
- Data preprocessing and postprocessing functions
- Filling masked data and gap filling features
- Dataset creation and train-/ test splitting techniques
- Trainer classes for
sklearn
,TensorFlow
andPyTorch
- Distributed training framework compatible with
PyTorch
- chunk utilities for working with data cubes
Requirements#
Package | Versions |
---|---|
dask | ≥2023.2.0 |
numpy | ≥1.24 |
pandas | ≥2.2 |
scikit-learn | >1.3.1 |
xarray | >2023.8.0 |
zarr | >2.11 |
rechunker | ≥0.5.1 |
Make sure you have Python version 3.8 or higher.
If you're planning to use ml4xcube
with TensorFlow or PyTorch, set up these frameworks properly in your conda environment.