core.MEDIAR.Trainer¶

Module Contents¶

class core.MEDIAR.Trainer.Trainer(model, dataloaders, optimizer, scheduler=None, criterion=None, num_epochs=100, device='cuda:0', no_valid=False, valid_frequency=1, amp=False, algo_params=None)[source]¶

Bases: core.BaseTrainer.BaseTrainer

Abstract base class for trainer implementations

mse_loss[source]¶

bce_loss[source]¶

mediar_criterion(outputs, labels_onehot_flows)[source]¶: loss function between true labels and prediction outputs