jaxvacua.flux_utils.dedup_key

Contents

jaxvacua.flux_utils.dedup_key#

dedup_key(moduli, tau, fluxes, n_digits=6)#

Build a hashable dedup key from a single (moduli, tau, fluxes) triple.

Details

The fluxes vector is rounded to integers and packed as raw bytes (it represents an integer flux quantum in practice). The continuous moduli and tau are rounded to n_digits decimal places to absorb numerical noise from solver convergence. The result is hashable, so it can be used as a Python set element or dict key.

Note that a single flux vector can admit multiple distinct critical points (because the moduli equations are non-linear), so the key deliberately includes the moduli + tau, not just the flux.

Distinct from FluxVacuaFinder.deduplicate_vacua(), which is a batched jnp.unique pass: this helper is for streaming / incremental dedup of one solution at a time.

Parameters:

  • moduli (Any) – Complex structure moduli. Real and imaginary parts are each rounded to n_digits decimal places.

  • tau (Any) – Axio-dilaton. Real and imaginary parts are each rounded to n_digits decimal places.

  • fluxes (Any) – Flux vector. Real part is rounded to int32; imaginary part is ignored.

  • n_digits (int) – Decimal precision for rounding continuous components. Defaults to 6.

Returns:

  • tuple – 4-tuple (flux_bytes, moduli_re_bytes, moduli_im_bytes, tau_tuple),

  • hashable for use in a ``set`` / ``dict``.

Return type:

Tuple[bytes, bytes, bytes, Tuple[float, float]]

Example:

seen = set()
for vac in candidate_vacua:
    k = dedup_key(vac.moduli, vac.tau, vac.fluxes)
    if k in seen:
        continue
    seen.add(k)
    ...
Contents