Florian Tramer

I agree with Nicholas that this discrepancy in notation for R+FGSM is very confusing. Here's my take: - In our original paper on R+FGSM, \epsilon refers to the *total* perturbation...

This code is just the standard way of calculating a model's loss over a given set of tokens. Say you have a sentence of tokens (t1, ..., tn). Then if...

Where is this code from? The code we use is a standard way of calculating perplexity in huggingface.

This really depends on the total number of parameters in the network, and whether you assume access to the exact class probabilities or just the predicted class labels.. In some...

Can you try commenting out this import here? https://github.com/ftramer/slalom/blob/1badff914657398733b4136325066cf7c2f66168/python/slalom/models.py#L7

Hi Robert, Thanks this worked, with a few caveats: 1) This doesn't work with the latest Tensorflow (2.1.0). With tensorflow=1.15.0 and dm-sonnet=1.36, the errors disappear 2) There seems to be...