torch>=2.4
numpy

[cuda]
triton>=3.0
cupy-cuda13x

[dev]
pytest
pytest-cov
ruff>=0.14

[mpi]
mpi4py
cffi

[perceval]
perceval-quandela
