trinity.common.workflows.math_rm_workflow module

We include the math workflow with rm-gallery reward in this file.

class trinity.common.workflows.math_rm_workflow.MathRMWorkflow(*, task: Task, model: ModelWrapper, auxiliary_models: List[OpenAI] | None = None)[source]

Bases: SimpleWorkflow

A workflow for math tasks as introduced in DeepSeek-R1.

__init__(*, task: Task, model: ModelWrapper, auxiliary_models: List[OpenAI] | None = None)[source]
run() → List[Experience][source]

Run workflow and return a list of experiences.