Private repo for Fair Fairytale Challenge.
Note: The testing python version is 3.8.13.
conda create --name fairy38 python==3.8.13
conda activate fairy38
pip install -r requirements.txt
chmod u+x install_model.sh
./install_model.sh
To run the pipeline on a single story text file (.txt), at the root dir of the project:
export PYTHONPATH=$(pwd)
python utils/run_bias_pipeline_single.py --story-infile /path/to/stories.txt --pipeline-dir /path/to/output/
You can add the following arguments as needed.
| Description | ||
|---|---|---|
| --story-id | -si | Story id used for saving files. If not provided, the story id will be the file name of the story infile. |
| --pipelines | -p | List of pipelines to run. Default is full pipeline. Options: full, booknlp, characters_sentences, attributes, srl, actions_args, supersense, temporal_events, temporal_events_characters |
| --pronoun-only-entities | -poe | Flag to keep character entities only referred to with pronouns. |
| --keep-all-verbs | -kav | Flag to keep stative verbs. |
To run the pipeline on multiple stories, store each story as a separate text file (.txt) in a single directory. At the root dir of the project:
export PYTHONPATH=$(pwd)
python utils/run_bias_pipeline_batch.py --story-dir /path/to/stories/ --pipeline-dir /path/to/output/
You can add the following arguments as needed.
| Description | ||
|---|---|---|
| --pipelines | -p | List of pipelines to run. Default is full pipeline. Options: full, booknlp, characters_sentences, attributes, srl, actions_args, supersense, temporal_events, temporal_events_characters |
| --pronoun-only-entities | -poe | Flag to keep character entities only referred to with pronouns. |
| --keep-all-verbs | -kav | Flag to keep stative verbs. |
At the root dir of the project:
export PYTHONPATH=$(pwd)
python utils/run_FairytaleQA_bias_pipeline.py
The CPU version is expected to run very long time, approximately 20 mins on my Macbook, especially slow for TE_inference.