def main(args): # Get dataset test_files = glob.glob(os.path.join(args.test_path, '*.pickle')) datasource = DataSource(None, test_files, shape=tuple(args.eye_shape), data_format=args.data_format, heatmap_scale=args.heatmap_scale) # Get model learning_schedule = [{ 'loss_terms_to_optimize': { 'heatmaps_mse': ['hourglass'], 'radius_mse': ['radius'], }, 'learning_rate': 1e-3, }] model = CNN(datasource.tensors, datasource.x_shape, learning_schedule) # Get evaluator evaluator = Trainer(model, model_checkpoint=args.model_checkpoint) # Evaluate avg_losses = evaluator.run_eval(datasource) print('Avarage Losses', avg_losses)