fit_ensemble#
- hybrid_aif.train.fit_ensemble(
- system,
- params,
- learn_state,
- batch,
- *,
- row=None,
- member_weights=None,
- lr=0.0001,
- lr_by_slot=None,
- clip=100.0,
- adam_eps=1e-05,
One gradient step of ensemble learning:
(params, learn_state, losses).Pure, jit-able, resumable; steps only its row’s
scoring:<row>address and threads every otherLearnStateentry through.rownames which ofROWSto step;Noneresolves the oneSystemParamscarries and raises if it carries both.member_weightsgoes toensemble_loss().clipis a global-norm clip over all members (Nonedisables it);adam_epsbuilds the optimizer.lr_by_slotis the rate grammar, a static0.0skipping the update at trace time.Raises
ValueErrorwhenlearn_stateisNoneor lacks this row’s optimizer entry (never re-seeded), and whenrowcannot be resolved.