# PyPhi configuration (2.0 layered format) # ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ # See the documentation for ``pyphi.config`` for the full list of options # and their defaults. Each top-level key corresponds to one of the three # config layers: formalism (what PyPhi computes), infrastructure (how it # runs — parallelism, caching, logging), and numerics (precision). # # The ``formalism`` layer is split into two sub-namespaces: ``iit`` for # IIT-formalism knobs and ``actual_causation`` for AC-framework knobs. # # Only override fields you actually want to change; omitted fields use the # default values defined on the layer dataclasses. --- formalism: # The default IIT formalism is IIT 4.0 (2026): version "IIT_4_0_2026" with # system_phi_measure "INTRINSIC_INFORMATION", which applies the Eq. 23 # intrinsic-information cap φ_s = min{φ_c, φ_e, ii(s)}. Mechanism and # specification measures are GENERALIZED_INTRINSIC_DIFFERENCE. # # Leaving ``iit`` empty inherits these defaults from the layer dataclass # (the single source of truth). For another formalism prefer a preset via # ``pyphi.config.override(**pyphi.presets.iit4_2023)`` / ``.iit3`` rather than # setting an active override here — an active ``version:`` in this file would # shadow the dataclass default during in-repo runs. iit: {} actual_causation: {} infrastructure: {} numerics: {}