| .. | ||
| models | ||
| results | ||
| utils | ||
| .gitignore | ||
| environment.yml | ||
| README.md | ||
| run_test.sh | ||
| run_train.sh | ||
| tbd_dataloader.py | ||
| test.py | ||
| train.py | ||
TBD
Data
The original code can be found here. The dataset is not directly available but must be requested using the link to the Google form provided in the README.
Installing Dependencies
Run conda env create -f environment.yml.
Train
source run_train.sh.
Test
source run_test.sh. Make sure to use the same random seed used for training, otherwise the splits will be different and you will likely have a data leakage.
Visualisations
The plots are made using utils/fb_scores_err.py (false belief analysis) and utils/similarity.py (PCA of latent representations).