mayank31398 / VRAG

This repository contains code for the paper Variational Learning for Unsupervised Knowledge Grounded Dialogs.

Flags used to run the code:

1. --params_file Path to params file
2. --eval_only If passed, models only generate responses (no training is done)
3. --checkpoint Checkpoint name to load from
4. --knowledge_file Path to knowledge.jsonl
5. --labels_file File to evaluate (only to be used when --eval_only is used)
6. --output_file Path to dump generated outputs (only to be used when --eval_only is used)
7. --model_path Path to save/load the model from
8. --prior_path Path to load the prior from (only use if model_path is not specified)
9. --posterior_path Path to load the posterior from (only use if model_path is not specified)
10. --decoder_path Path to load the decoder from (only use if model_path is not specified)
11. --build_index Builds index and exits (needs to be run before training)
12. --n_gpus Number of GPUs to use (defaults to 1) (May not work)
13. --dialog To be passed if dataset is a dialog dataset
14. --save_every Save every nth step (Never tested this argument so don't use. By default saves on each epoch)
15. --multitask Pass to use a classifier loss with decoder to classify CANNOTANSWER instances
16. --weight Weight for CANNOTANSWER instances (only use if --weigh_cannot_answer is passed)
17. --weigh_cannot_answer Weigh CANNOTANSWER instances
18. --skip_cannot_answer Skips CANNOTANSWER instances
19. --fix_DPR Fix both prior and posterior during training
20. --fix_prior Fix only prior during training
21. --fix_posterior Fix only posterior during training
22. --fix_decoder Fix only decoder during training

run prepare_data.sh to uncompress the data. Look at train.sh and val.sh for how to train and evaluate the models respectively.