The code was made to run on the Compute Canada (Gentoo Linux) environment. Some modifications may be required to run in other environments
On Compute Canada run:
source ./activate.sh
You can run each file like:
python3 A_train_question_generator_control.py single_model_with_soft_prompt_t5_lm 5e-5 quail large 1
Some programs can also be run as vscode notebooks.
The interface for each program is not fully documented. If you are having problems, please create a GitHub issue.
Most of the code in this repo is licensed until the Apache 2.0 license. Files in some folders have a difference license, please refer to the licenses in these folders for more details.
Cite as:
@article{von_der_Ohe_Fyshe_SoftSkillQG,
author = {von der Ohe, Spencer McIntosh and Fyshe, Alona},
journal = {Proceedings of the Canadian Conference on Artificial Intelligence},
year = {2024},
month = may,
url = {https://caiac.pubpub.org/pub/n3kbetb8},
publisher = {Canadian Artificial Intelligence Association (CAIAC)},
title = {SoftSkillQG: Robust Skill-Targeted Reading Comprehension Question Generation Using Soft-prompts},
}