This repo contains files to fuzz a GPT neo model from Hugging Face.
The fuzzing is very primitive and just generates random string inputs for the model. This is no serious approach in "cracking" the model or finding bugs in the stack. I was only curious what such a model would do with some random input.
Install python 3 and pip. Run install.sh
.
Run one of the run_xxx.py
files. They will generate a CSV file with the in- and outputs.