def validate_zeroshot()

in main.py [0:0]

• 45 lines of code
• 6 McCabe index (conditional complexity)

def validate_zeroshot(val_loader, model, tokenizer, args):
    batch_time = AverageMeter('Time', ':6.3f')
    top1 = AverageMeter('Acc@1', ':6.2f')
    top5 = AverageMeter('Acc@5', ':6.2f')
    progress = ProgressMeter(
        len(val_loader),
        [batch_time, top1, top5],
        prefix='Test: ')

    # switch to evaluate mode
    model.eval()

    print('=> encoding captions')
    cwd = os.path.dirname(os.path.realpath(__file__))
    with open(os.path.join(cwd, 'templates.json')) as f:
        templates = json.load(f)['imagenet']

    with open(os.path.join(cwd, 'labels.json')) as f:
        labels = json.load(f)['imagenet']

    with torch.no_grad():
        text_features = []
        for l in labels:
            texts = [t.format(l) for t in templates]
            texts = tokenizer(texts).cuda(args.gpu, non_blocking=True)
            class_embeddings = utils.get_model(model).encode_text(texts)
            class_embeddings = class_embeddings / class_embeddings.norm(dim=-1, keepdim=True)
            class_embeddings = class_embeddings.mean(dim=0)
            class_embeddings = class_embeddings / class_embeddings.norm(dim=-1, keepdim=True)
            text_features.append(class_embeddings)
        text_features = torch.stack(text_features, dim=0)

        end = time.time()
        for i, (images, target) in enumerate(val_loader):
            images = images.cuda(args.gpu, non_blocking=True)
            target = target.cuda(args.gpu, non_blocking=True)

            # encode images
            image_features = utils.get_model(model).encode_image(images)
            image_features = image_features / image_features.norm(dim=-1, keepdim=True)

            # cosine similarity as logits
            logits_per_image = image_features @ text_features.t()

            # measure accuracy and record loss
            acc1, acc5 = accuracy(logits_per_image, target, topk=(1, 5))
            acc1, acc5 = utils.scaled_all_reduce([acc1, acc5])
            top1.update(acc1.item(), images.size(0))
            top5.update(acc5.item(), images.size(0))

            # measure elapsed time
            batch_time.update(time.time() - end)
            end = time.time()

            if i % args.print_freq == 0:
                progress.display(i)

    progress.synchronize()
    print('0-shot * Acc@1 {top1.avg:.3f} Acc@5 {top5.avg:.3f}'
          .format(top1=top1, top5=top5))
    return {'acc1': top1.avg, 'acc5': top5.avg}