pytc.solver.isdf_xtc_ccsd.profile_isdf_factor_direct_terms_t2

pytc.solver.isdf_xtc_ccsd.profile_isdf_factor_direct_terms_t2(t2, p, grad_p, u1, u3, d, x, *, occupied_pair_batch_size=8, rank_panel_size=128)[source]

Measure every branch and return the exact factor-direct terms.

Profiling is explicitly opt-in. Each profile distinguishes the schedule-panel estimate from XLA’s executable memory analysis. Neither is a process-wide allocator peak.

Parameters:

  • t2 (jax.Array)

  • p (jax.Array)

  • grad_p (jax.Array)

  • u1 (jax.Array)

  • u3 (jax.Array)

  • d (jax.Array)

  • x (jax.Array)

  • occupied_pair_batch_size (int)

  • rank_panel_size (int)

Return type:

tuple[Mapping[str, jax.Array], Mapping[str, FactorDirectProfile]]