def train()

in src/train.py [0:0]

• 22 lines of code
• 5 McCabe index (conditional complexity)

    def train(self):
        best_eval = float('inf')

        # Begin!
        for epoch in range(self.start_epoch, self.start_epoch + self.args.epochs):
            self.logger.info(f'Starting epoch, Rank {self.args.rank}, Dataset: {self.args.data[self.args.rank]}')
            self.train_epoch(epoch)
            self.evaluate_epoch(epoch)

            self.logger.info(f'Epoch %s Rank {self.args.rank} - Train loss: (%s), Test loss (%s)',
                             epoch, self.train_losses(), self.eval_losses())
            self.lr_manager.step()
            val_loss = self.eval_total.summarize_epoch()

            if val_loss < best_eval:
                self.save_model(f'bestmodel_{self.args.rank}.pth')
                best_eval = val_loss

            if not self.args.per_epoch:
                self.save_model(f'lastmodel_{self.args.rank}.pth')
            else:
                self.save_model(f'lastmodel_{epoch}_rank_{self.args.rank}.pth')

            if self.args.is_master:
                torch.save([self.args,
                            epoch],
                           '%s/args.pth' % self.expPath)

            self.logger.debug('Ended epoch')