Algorithm.evaluate(parallel_train_future: ThreadPoolExecutor | None = None) dict | NestedDict[source]#

Evaluates current policy under evaluation_config settings.


parallel_train_future – In case, we are training and avaluating in parallel, this arg carries the currently running ThreadPoolExecutor object that runs the training iteration. Use parallel_train_future.done() to check, whether the parallel training job has completed and parallel_train_future.result() to get its return values.


A ResultDict only containing the evaluation results from the current iteration.