graphorge.gnn_base_model.train.training.save_loss_history

save_loss_history(model, n_max_epochs, loss_nature, loss_type, training_loss_history, lr_scheduler_type=None, lr_history_epochs=None, validation_loss_history=None)

Save training process loss history record.

Loss history record file is stored in model_directory under the name loss_history_record.pkl.

Overwrites existing loss history record file.

Parameters:

• model (torch.nn.Module) – Model.

• n_max_epochs (int) – Maximum number of epochs of training process.

• loss_nature (str) – Loss nature.

• loss_type (str) – Loss function type.

• training_loss_history (list[float]) – Training process training loss history (per epoch).

• lr_scheduler_type ({'steplr', 'explr', 'linlr'}, default=None) – Type of learning rate scheduler.

• lr_history_epochs (list[float], default=None) – Training process learning rate history (per epoch).

• validation_loss_history (list[float], default=None) – Training process validation loss history (e.g., early stopping criterion).