ray.rllib.core.learner.learner.Learner.register_metrics#

Learner.register_metrics(module_id: str, metrics_dict: Dict[str, Any]) → None[source]#

Registers several key/value metric pairs for loss- and gradient stats.

Parameters:

  • module_id – The module_id to register the metrics under. This may be ALL_MODULES.

  • metrics_dict – A dict mapping names of metrics to be registered (below the given module_id) to the actual values of these metrics. Values might also be tensor vars (e.g. from within a traced tf2 function). These will be automatically converted to numpy values.