def train()

in eval_linear.py [0:0]

• 46 lines of code
• 4 McCabe index (conditional complexity)

def train(model, reglog, optimizer, loader, epoch):
    """
    Train the models on the dataset.
    """
    # running statistics
    batch_time = AverageMeter()
    data_time = AverageMeter()

    # training statistics
    top1 = AverageMeter()
    top5 = AverageMeter()
    losses = AverageMeter()
    end = time.perf_counter()

    model.eval()
    reglog.train()
    criterion = nn.CrossEntropyLoss().cuda()

    for iter_epoch, (inp, target) in enumerate(loader):
        # measure data loading time
        data_time.update(time.perf_counter() - end)

        # move to gpu
        inp = inp.cuda(non_blocking=True)
        target = target.cuda(non_blocking=True)

        # forward
        with torch.no_grad():
            output = model(inp)
        output = reglog(output)

        # compute cross entropy loss
        loss = criterion(output, target)

        # compute the gradients
        optimizer.zero_grad()
        loss.backward()

        # step
        optimizer.step()

        # update stats
        acc1, acc5 = accuracy(output, target, topk=(1, 5))
        losses.update(loss.item(), inp.size(0))
        top1.update(acc1[0], inp.size(0))
        top5.update(acc5[0], inp.size(0))

        batch_time.update(time.perf_counter() - end)
        end = time.perf_counter()

        # verbose
        if args.rank == 0 and iter_epoch % 50 == 0:
            logger.info(
                "Epoch[{0}] - Iter: [{1}/{2}]\t"
                "Time {batch_time.val:.3f} ({batch_time.avg:.3f})\t"
                "Data {data_time.val:.3f} ({data_time.avg:.3f})\t"
                "Loss {loss.val:.4f} ({loss.avg:.4f})\t"
                "Prec {top1.val:.3f} ({top1.avg:.3f})\t"
                "LR {lr}".format(
                    epoch,
                    iter_epoch,
                    len(loader),
                    batch_time=batch_time,
                    data_time=data_time,
                    loss=losses,
                    top1=top1,
                    lr=optimizer.param_groups[0]["lr"],
                )
            )

    return epoch, losses.avg, top1.avg.item(), top5.avg.item()