The standard interface for working with Mathy models and agents.
Mathy( self, model_path: str = None, model: mathy.agents.policy_value_model.PolicyValueModel = None, config: mathy.agents.base_config.BaseConfig = None, silent: bool = False, )
Simplify an input problem using the PolySimplify environment.
Mathy.simplify( self, model: str = 'mathy_alpha_sm', problem: str, max_steps: int = 20, thinking_steps: int = 3, ) -> mathy.agents.episode_memory.EpisodeMemory
- model (str): The input model to use for picking simplifying actions
- problem (str): The ascii math problem text, e.g.
-(4 + 2x) * 8 / 7y^(3 - 2)
- max_steps (int): The maximum number of episode steps to allow the agent to take while solving the problem. Taking more than this is considered a failure.
- thinking_steps (int): The number of timesteps to look at the problem before attempting to solve it. These steps do not count toward the
(EpisodeMemory): The stored episode memory containing the intermediate steps to get to the solution for the input problem.