solvi.core.deciders.adapt¶
Adapting one question: label-bias correction, few-shot shift and scale, temperature.
Part of solvi.core.deciders, which re-exports every name; import from there.
Adapting one question: label-bias correction, few-shot shift and scale, temperature; the question spec. (Part of solvi.core.deciders, which re-exports every name.)
Adaptation
dataclass
¶
Adaptation(bias: list | None = None, n_unlabelled: int = 0, scale: float | None = None, shift: list | None = None, temperature: float = 1.0, other_threshold: float | None = None, n_labelled: int = 0, examples: list = list())
What solvi learned for one question (task, options, kind): the label-bias correction, the few-shot shift / scale, the temperature and the "other" threshold. Part of the fingerprint of every decision that uses it.
params ¶
Everything that changes the output (the examples only through the fitted parameters).
lora_key ¶
The slot a LoRA adapter (solvi.experimental.lora) is kept under: the question — the task, the scored options with
their descriptions (in any order: option_order="average" asks rotations of the same question) and the kind — and,
for a _Spec a decision part asks, the part's name as a fourth element. Two parts asking the same question in one
process have two slots: one part's adapter never answers the other's question. A _Spec.key (or a _Spec with no
owner) gives the question alone; key[:3] is always the question.