MultiEvaluator

class maze.train.trainers.common.evaluators.multi_evaluator.MultiEvaluator(evaluators: List[maze.train.trainers.common.evaluators.evaluator.Evaluator])

Evaluates the given policy using multiple different evaluators (ran in sequence).

Useful when evaluating a policy in different scenarios. E.g., during behavioral cloning, we might want to evaluate the policy first on a validation dataset and then through an evaluation rollout.

Parameters

evaluators – Evaluators to run.

evaluate(policy: maze.core.agent.torch_policy.TorchPolicy) → None

(overrides Evaluator)

Evaluate given policy using the given evaluators.

param policy

Policy to evaluate