This repository is used for storing models and results from the CSIRO summer internship project
All the models here are fine-tuned using the ECHR dataset
Fine-tuned the pretrained GPT2 model
Configuration: undefended model trained on 30% of the original dataset
Performances:
extraction 2500:
extraction 15000:
inference:
Configuration: undefended model trained on 100% of the original dataset
Performances:
extraction 2500:
extraction 15000:
inference:
Configuration: dp-8 model trained on 100% of the original dataset
Performances:
extraction 2500:
extraction 15000:
inference:
Fine-tuned the pretrained Bert model