numpy==1.23.4
pandas==1.5.3
pyro_ppl==1.8.4
torch==1.12.1
tqdm==4.64.1
matplotlib==3.3.4
sacred==0.8.4
The sacred library is used in the training script to save the trained model in a run-specific folder, along with the config file and some metrics.
The training can be run on different datasets by running the following commands
To train on the highschool dataset, run
python train.py with config/highschool.yaml
To train on the toy dataset, run
python train.py with config/toy.yaml
In notebooks, we provide examples of how to visualize the 2d embeddings obtained through TGNE on the toy dataset and the highschool dataset.
For any questions, please contact Raphaël Romero at raphael.romero@ugent.be.