sakura.utils.logger.Logger.log_loss

Logger.log_loss(trainer_output: dict, tick, loss_name_prefix='', selected_loss_group=['loss', 'regularization'])

Record hierarchical loss structure to TensorBoard

Parameters:

• trainer_output – Nested loss structure containing: - pheno_loss: Dictionary of phenotype-specific losses - signature_loss: Dictionary of signature-specific losses - main_latent_loss: Core latent space loss components

• tick (int) – Training iteration counter for x-axis scaling

• loss_name_prefix (str, optional) – Namespace prefix for loss grouping, defaults to ‘’

• selected_loss_group (list, optional) – Loss types to log from [‘loss’, ‘regularization’], defaults to both

Returns:

None