I wanted to use VNN-LIB benchmark models to compare against gradient-based algorithms (such as FGSM). However, this looks to be quite challenging as the onnx to pytorch (and tensorflow) converters I’ve used are mainly for inference and are not able to store gradient information.
Do you know of any good modules that can convert onnx to a popular deep learning framework (tf or pytorch) that enable gradient descent well?
Or alternatively, can we get access to these benchmarks in other formats more suitable for this task?
Viewing 1 post (of 1 total)
You must be logged in to reply to this topic.