sakura.utils.logger.Logger

class sakura.utils.logger.Logger(log_path='./logs/', suppress_tensorboardX=False)

Bases: object

SAKURA Pipeline Logging Controller

Handles multi-dimensional logging including:

  • Real-time loss tracking via TensorBoard

  • Latent space embedding exports

  • Model configuration persistence

Parameters:

  • log_path (str, optional) – Root directory for all logging artifacts, defaults to ‘./logs/’

  • suppress_tensorboardX (bool, optional) – Whether to disable TensorBoard integration, defaults to False

Methods

dump_latent_to_csv

Export latent representations from the output of the controller to structured storage.

log_loss

Record hierarchical loss structure to TensorBoard

log_metric

( to be implemented )

log_parameter

( to be implemented )

save_config

Persist pipeline configuration to JSON.

save_splits

Serialize data splits to pickle format.