rlightning.utils.logger.backends

class rlightning.utils.logger.backends.MetricsBackend[source]

Bases: ABC

Abstract base class for a metrics backend.

abstractmethod write(metrics: Dict[str, Any]) → None[source]

Write metrics to the backend.

Parameters:

metrics – A dictionary of metrics to log.

class rlightning.utils.logger.backends.SwanLabBackend(cfg: LogConfig)[source]

Bases: MetricsBackend

Metrics backend for SwanLab.

close()[source]

Finish the SwanLab run.

write(metrics: Dict[str, Any], step: int = None)[source]

Write metrics to SwanLab.

Parameters:

  • metrics – A dictionary of metrics to log.

  • step – The step number to log the metrics at.

class rlightning.utils.logger.backends.TensorBoardBackend(cfg: LogConfig)[source]

Bases: MetricsBackend

Metrics backend for TensorBoard.

close()[source]

Close the TensorBoard writer.

write(metrics: Dict[str, Any], step: int)[source]

Write metrics to TensorBoard.

Only scalar values (int, float) are logged.

Parameters:

  • metrics – A dictionary of metrics to log.

  • step – The global step used when logging scalar metrics.

class rlightning.utils.logger.backends.WandBBackend(cfg: LogConfig)[source]

Bases: MetricsBackend

Metrics backend for Weights & Biases.

close()[source]

Finish the WandB run.

write(metrics: Dict[str, Any], step: int = None)[source]

Write metrics to WandB.

Parameters:

  • metrics – A dictionary of metrics to log.

  • step – The step number to log the metrics at.