# Dependencies for the OpenH-RF eSAF sub-dataset package.
# Install with:  python -m pip install -r requirements.txt
#
# reconstruct.py follows the OpenH-RF example templates and runs the processing
# chain defined in pipeline.yaml through the zea.Pipeline — zea (>= 0.1.0, the
# version the files were written with via zea.File.create) and one Keras backend
# (jax below; tensorflow/torch also work, set KERAS_BACKEND accordingly) are
# required. Run with:  KERAS_BACKEND=jax python reconstruct.py
numpy>=1.24
scipy>=1.10
matplotlib>=3.6
h5py>=3.8
pyyaml>=6.0
zea>=0.1.0
jax

# Notes:
# - verify_submission.py's structural checks (deliverables, HDF5 schema, manifest,
#   pipeline.yaml) need only numpy/h5py/pyyaml; its reconstruction smoke test runs
#   reconstruct.py and therefore needs zea + a backend like everything else.
# - extras/esaf_demo.py (optional eSAF demonstration) needs only numpy/scipy/
#   matplotlib/h5py.