Train_logger

class conformer_rl.logging.train_logger.TrainLogger(tag: str, dir: str = 'data', use_tensorboard: bool = True, use_cache: bool = True, use_print: bool = True)

Used by agent for logging agent metrics during training.

Parameters

• tag (str) – Unique tag for identifying the logging session.

• dir (str) – Path to root directory for where logging results should be saved.

• use_tensorboard (bool) – Whether or not to save metrics to Tensorboard.

• use_cache (bool) – Whether or not to keep a running cache of the metrics.

• use_print (bool) – Whether or not to print metrics when logged.

add_scalar(key: str, scalar_value: float, global_step: Optional[int] = None, walltime: Optional[float] = None) → None

Logs a single scalar value.

Parameters

• key (str) – The key associated with the logged value.

• scalar_value (float) – The value to be logged.

• global_step (int) – The current agent step when logging the metric.

• walltime (float) – The current time elapsed since the start of agent training.