nat.experimental.test_time_compute.selection.llm_judge_selection#

Attributes#

Classes#

LLMJudgeSelection

A selection strategy that uses a configured Judge LLM to select the best response.

Functions#

register_llm_judge_selection(config, builder)

Module Contents#

logger#

class LLMJudgeSelection( config: nat.experimental.test_time_compute.models.selection_config.LLMJudgeSelectionConfig, )#

Bases: nat.experimental.test_time_compute.models.strategy_base.StrategyBase

A selection strategy that uses a configured Judge LLM to select the best response.

config#

judge_llm_bound = None#

async build_components(builder: nat.builder.builder.Builder) → None#: Builds the Judge LLM configured in the strategy.

supported_pipeline_types() → list[nat.experimental.test_time_compute.models.stage_enums.PipelineTypeEnum]#: Return the stage types supported by this selector.

stage_type() → nat.experimental.test_time_compute.models.stage_enums.StageTypeEnum#: Return the stage type of this strategy.

async ainvoke(items: list[nat.experimental.test_time_compute.models.ttc_item.TTCItem], original_prompt: str | None = None, agent_context: str | None = None, \*\*kwargs) → list[nat.experimental.test_time_compute.models.ttc_item.TTCItem]#: Select the best item using the configured Judge LLM.

async register_llm_judge_selection( config: nat.experimental.test_time_compute.models.selection_config.LLMJudgeSelectionConfig, builder: nat.builder.builder.Builder, )#