Bogdan Kulynych

Hi, ShadowModelBundle and AttackModelBundle take a scikit-like object, with `fit`, `predict`, and `predict_proba` methods. You can use [skorch](https://skorch.readthedocs.io/en/latest/) to wrap your torch model in such an API, or mia's own...

If you can add an example, that would be great!

Folks, thanks for your interest. It seems that a bunch of dependencies (specifically, I suspect spacy is one of the culprits) have broken some of the code in this repo....

Hi! Since the search process itself is deterministic, should be because the difference in our trained models.

Ah, I misunderstood. You don't need a pre-trained model to run `python paraphrase.py`. The example on the picture shows an adversarial example that evades the model. Running `python paraphrase.py` does...