The generated texts are located in folders "Model_Name" and then into the json files corresponding to the parameter used at generation.
If you wish to print AUC results, the "Scores" folder contains pickle with all the scores attributed to either "human" or "generated" samples in the study.
"all_distances.py" is for computing the distances between the model’s distribution q and the adapted distribution used for generation p
"compute_renyi" and "compute_kl" allow you to get the divergence between two models (q and r)