neuralqx.hilbert.utils package

check_one_state_jax(state, constraints, q_min, q_max, q_step)

Check all gauge-fixing constraints for a single configuration.

Given a state state of shape (N,) and a preprocessed list of constraint specifications, this function verifies that every constraint is satisfied. Each constraint is represented as a pair (inc_data, out_data), where both entries are lists of (idx, sgn) pairs. The value state[idx] is multiplied by its sign sgn and then combined using a modular sum.

For a single constraint, define the signed incoming and outgoing aggregates

\[S_{\mathrm{in}} = \bigoplus_{(i, s) \in \mathrm{inc}} (s\,\sigma_i), \qquad S_{\mathrm{out}} = \bigoplus_{(j, s) \in \mathrm{out}} (s\,\sigma_j),\]

where \(\sigma_i = \texttt{state}[i]\) and \(\oplus\) denotes the modular addition implemented by neuralqx.utils.jit.mod_sum_jax() on the discrete domain [q_min, q_max] with step q_step. The constraint is satisfied iff

\[S_{\mathrm{in}} = S_{\mathrm{out}}.\]

All constraints must pass for the state to be accepted.

Parameters:

  • state – One configuration array of shape (N,).

  • constraints – Preprocessed constraints as produced by the corresponding preprocessing helper. Each item is (inc_data, out_data) with inc_data and out_data lists of (idx, sgn) pairs.

  • q_min – Minimum allowed local quantum number for the modular domain.

  • q_max – Maximum allowed local quantum number for the modular domain.

  • q_step – Step size between allowed local quantum numbers.

Returns:

A boolean JAX scalar indicating whether all constraints are satisfied.

make_hashable(obj)

Recursively convert an object into a hashable representation.

This helper is intended for building stable cache keys from nested container structures commonly encountered in constraint specifications and preprocessed data.

Conversion rules:

  • list/tuple are converted into tuples of recursively converted items;

  • dict is converted into a sorted tuple of (key, value) pairs, with values converted recursively;

  • array-like objects exposing .tolist() (for example JAX or NumPy arrays) are converted via their Python list representation and then processed recursively;

  • all other objects are returned unchanged.

Parameters:

obj (Any) – Any Python object (possibly nested).

Return type:

Any

Returns:

A hashable equivalent suitable for use as a dictionary key.

Subpackages