Troubles running CLUE about carla

[JohanvandenHeuvel]

Hi,

Unfortunately some methods only work for either Tensorflow or PyTorch. It's a bit unintuitive, we should probably throw a warning when this happens. For now you can find which recourse model is supported by which framework here

The following should fix it

# load artificial neural network from catalog
model = MLModelCatalog(dataset, "ann", backend="pytorch")

Also note that right now your factuals contain samples with negative and positive predicted labels. This may lead to unintended behaviour. If you want only the factuals that have a negative predicted label you can use the following.

# get factuals
factuals = predict_negative_instances(model, dataset.df).iloc[:1000]

[drobiu]

Thank you! That fixed it. :)

Also thanks for the negative predicted label tip.