nat.experimental.test_time_compute.models.search_config

Classes

SingleShotMultiPlanConfig	Base configuration class for Test Time Compute (TTC) strategy.
MultiLLMPlanConfig	Configuration for a 'multi LLM plan generation' strategy.
MultiQueryRetrievalSearchConfig	Configuration for the MultiQueryRetrievalSearch strategy.

Module Contents

class SingleShotMultiPlanConfig(/, **data: Any)

Bases: nat.data_models.ttc_strategy.TTCStrategyBaseConfig

Base configuration class for Test Time Compute (TTC) strategy. This class is used to define the structure of TTC strategy configurations.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

num_plans: int = None

max_temperature: float = None

min_temperature: float = None

planning_llm: nat.data_models.component_ref.LLMRef | Any | None = None

planning_template: str = None

validate_strategies(values: dict[str, Any]) → dict[str, Any]

Ensure that the required LLMs are provided based on the selected strategies.

class MultiLLMPlanConfig(/, **data: Any)

Bases: nat.data_models.ttc_strategy.TTCStrategyBaseConfig

Configuration for a ‘multi LLM plan generation’ strategy.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

llms: list[nat.data_models.component_ref.LLMRef] = None

plans_per_llm: int = None

max_temperature: float = None

min_temperature: float = None

planning_template: str = None

validate_multi_llm_strategies(values: dict) → dict

class MultiQueryRetrievalSearchConfig(/, **data: Any)

Bases: nat.data_models.ttc_strategy.TTCStrategyBaseConfig

Configuration for the MultiQueryRetrievalSearch strategy. This strategy generates multiple new ‘TTCItem’s per original item, each containing a differently phrased or re-focused version of the original task.

Create a new model by parsing and validating input data from keyword arguments.

Raises [ValidationError][pydantic_core.ValidationError] if the input data cannot be validated to form a valid model.

self is explicitly positional-only to allow self as a field name.

llms: list[nat.data_models.component_ref.LLMRef] = None

query_generation_template: str = None

validate_llms(values)