MetricCollector

class cl_gym.utils.callbacks.MetricCollector(num_tasks: int, epochs_per_task: Optional[int] = 1, collect_on_init: bool = False, collect_metrics_for_future_tasks: bool = False, eval_interval: str = 'epoch', eval_type: str = 'classification', tuner_callback: Optional[Callable[[float, bool], None]] = None)[source]

Bases: cl_gym.utils.callbacks.base.ContinualCallback

Collects metrics during the learning. This callback can support various metrics such as average accuracy/error, and average forgetting.

log_metrics(trainer, task_learned: int, task_evaluated: int, metrics: dict, global_step: int, relative_step: int)[source]
on_after_fit(trainer)[source]
on_after_training_epoch(trainer)[source]
on_after_training_task(trainer)[source]
on_before_fit(trainer)[source]
on_before_training_task(trainer)[source]
plot_metrics(logger: Optional[cl_gym.utils.loggers.Logger] = None)[source]
save_metrics()[source]