fix metric dimension bug in gaussian_process.py

        model.fit(train_dataset, **hyper_parameters)

        model.save()

        evaluator = Evaluator(model, valid_dataset, output_transformers)

        multitask_scores = evaluator.compute_model_performance([metric])

        score = multitask_scores[metric.name]

        multitask_scores = evaluator.compute_model_performance(metric)

        score = multitask_scores[metric[0].name]

      with open(log_file, 'a') as f:

        # Record performances