nemo_microservices.types.customization.library_judge_math_environment_param#

Module Contents#

Classes#

API#

class nemo_microservices.types.customization.library_judge_math_environment_param.LibraryJudgeMathEnvironmentParam#

Bases: typing_extensions.TypedDict

judge_temperature: float#

None

Sampling temperature for judge responses.

Higher values (e.g., 1.0) increase randomness, lower values (e.g., 0.1) make output more deterministic. Temperature of 0 is equivalent to greedy sampling.

judge_top_p: float#

None

Nucleus sampling parameter (top-p) for judge responses.

Only tokens with cumulative probability >= top_p are considered. 1.0 means no filtering; lower values (e.g., 0.9) increase quality by filtering unlikely tokens.

name: typing_extensions.Literal[library_judge_math]#

None

Name of the task-specific environment that the dataset schema is designed to be used with. This field is automatically added to Dataset Rows based on the Environment selection.

should_use_judge: bool#

None

Whether to use a judge for the responses.