convlab2

* Seperate test and train domains

* Add progress bars in ontology embedder

* Update custom_util.py

* Fix custom_util things I broke

* Github master

* Save dialogue ids in prediction file

* Fix bug in ontology enxtraction

* Return dialogue ids in predictions file and fix bugs

* Add setsumbt starting config loader

* Add script to extract golden labels from dataset to match model predictions

* Add more setsumbt configs

* Add option to use local files only in transformers package

* Update starting configurations for setsumbt

* Github master

* Update README.md

* Update README.md

* Update convlab/dialog_agent/agent.py

* Revert custom_util.py

* Update custom_util.py

* Commit unverified chnages :(:(:(:(

* Fix SetSUMBT bug resulting from new torch feature

* Setsumbt bug fixes

* Policy config refactor

* Policy config refactor

* small bug fix in memory with new config path

Co-authored-by: Carel van Niekerk <carel.niekerk@hhu.de>
Co-authored-by: Michael Heck <michael.heck@hhu.de>
Co-authored-by: Christian Geishauser <christian.geishauser@hhu.de>