I see the overview says that ARES supports TensorFlow & PyTorch models with the same interface. but there is no direct function or interface for a trained Pytorch model to use. Does that mean user need to retrain their model with the decorator @pytorch_classifier_with_logits so that they could use the function in ARES to processing robustness testing?
尊敬的开发人员你们好，我想知道ARES是否无法直接使用我训练好的Pytorch模型进行他的鲁棒性测试？我看到我们的简介里说ARES以相同的接口支持tf和torch的模型。但是我仔细阅读我们的文档和代码后，并没有发现有类似最新版cleverhans的那种直接面向torch模型的接口。如果是我有所遗漏，请您指出我该如何操作。