crop_yield_prediction/train_c3d.py [145:180]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
                    start = i * batch_size
                    end = start + batch_size if i != n_batches - 1 else n_samples
                    predictions[start:end] = pred

            if cuda:
                predictions = predictions.cpu()
                predictions = predictions.data.numpy()

            rmse, r2, corr = cal_performance(predictions, y_test)

            if 'epoch' in model_file:
                print('  - {header:12} epoch: {epoch: 5}, rmse: {rmse: 8.3f}, r2: {r2: 8.3f}, corr: {corr: 8.3f}'.
                      format(header=f"({'Test'})", epoch=checkpoint['epoch'], rmse=rmse, r2=r2, corr=corr), file=f, flush=True)
            else:
                print('  - {header:12} best selected based on validation set, '
                      'rmse: {rmse: 8.3f}, r2: {r2: 8.3f}, corr: {corr: 8.3f}'.
                      format(header=f"({'Test'})", rmse=rmse, r2=r2, corr=corr), file=f, flush=True)

    return predictions, rmse, r2, corr


def eval_test_best_only(test_dataloader, y_test, batch_size, model, epoch, log_file):
    cuda = torch.cuda.is_available()

    model.eval()
    if cuda:
        model.cuda()

    n_batches = len(test_dataloader)
    n_samples = len(y_test)

    predictions = torch.zeros(n_samples)

    with torch.no_grad():
        for i, (batch_X, batch_y) in enumerate(test_dataloader):
            batch_X, batch_y = prep_data(batch_X, batch_y, cuda)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -



crop_yield_prediction/train_cross_location.py [227:262]:
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
                    start = i * batch_size
                    end = start + batch_size if i != n_batches - 1 else n_samples
                    predictions[start:end] = pred

            if cuda:
                predictions = predictions.cpu()
                predictions = predictions.data.numpy()

            rmse, r2, corr = cal_performance(predictions, y_test)

            if 'epoch' in model_file:
                print('  - {header:12} epoch: {epoch: 5}, rmse: {rmse: 8.3f}, r2: {r2: 8.3f}, corr: {corr: 8.3f}'.
                      format(header=f"({'Test'})", epoch=checkpoint['epoch'], rmse=rmse, r2=r2, corr=corr), file=f, flush=True)
            else:
                print('  - {header:12} best selected based on validation set, '
                      'rmse: {rmse: 8.3f}, r2: {r2: 8.3f}, corr: {corr: 8.3f}'.
                      format(header=f"({'Test'})", rmse=rmse, r2=r2, corr=corr), file=f, flush=True)

    return predictions, rmse, r2, corr


def eval_test_best_only(test_dataloader, y_test, batch_size, model, epoch, log_file):
    cuda = torch.cuda.is_available()

    model.eval()
    if cuda:
        model.cuda()

    n_batches = len(test_dataloader)
    n_samples = len(y_test)

    predictions = torch.zeros(n_samples)

    with torch.no_grad():
        for i, (batch_X, batch_y) in enumerate(test_dataloader):
            batch_X, batch_y = prep_data(batch_X, batch_y, cuda)
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -