Installation

Python environment

Use the existing shine environment:

conda activate shine
python -m pip install -e .

The codebase should still import under uv for tests and packaging checks:

uv sync
uv run python -m compileall euclid_dsps scripts
uv run euclid-dsps --help

FSPS and python-FSPS

The FS2 config requires generated FSPS SSP, gas, AGN component, and compressed runtime grids. Install FSPS and python-FSPS only in the environment used for asset generation:

cd "$HOME/src"
export SPS_HOME="$HOME/src/fsps"
git clone https://github.com/cconroy20/fsps.git "$SPS_HOME"

cd /home/maxime/src/DSPS
export SPS_HOME="$HOME/src/fsps"
python -m pip install fsps

Check that python-FSPS sees the expected libraries:

python -c "import fsps; sp=fsps.StellarPopulation(sfh=0); print(len(sp.wavelengths)); print(sp.isoc_library, sp.spec_library)"

Expected local output for the current assets is 11149 wavelength samples with mist and c3k_a. See Data And Assets for the exact generation commands.

Quality checks

uv run python -m compileall euclid_dsps scripts
uv run pytest tests
uv run --with sphinx --with sphinx-rtd-theme python -m sphinx \
  -W --keep-going -b html docs/source docs/build/html

GPU runtime

The public fit configs are GPU configs. For production, use a CUDA-enabled JAX installation and verify the backend before launching fits:

uv sync --extra gpu
uv run --extra gpu python -c "import jax; print(jax.devices())"

In the shine conda environment the equivalent check is:

conda activate shine
export JAX_PLATFORMS=cuda
export XLA_PYTHON_CLIENT_PREALLOCATE=false
export TF_GPU_ALLOCATOR=cuda_malloc_async
python -c "import jax; print(jax.default_backend()); print(jax.devices())"

If nvidia-smi works but JAX still reports only CpuDevice, check that no stale CPU-only jaxlib install is shadowing the active environment.