Most of the code templates for training and evaluation of GPT-2 model were derived from the existing base repository (private)
Business logic of training, data preprocessing and evaluation was implemented manually
- Dataset Exploration
- MultiWOZ 2.2 Loader
- Finetuning GPT-2 on MultiWOZ
- MultiWOZ 2.2 DB + State
- Two-stage decoding
The code here requires Python 3 and pip.
For a basic installation, clone the repository and run:
cd <your-repo>; pip install [--user] -r requirements.txt
However, you probably also want to use the repo as a set of libraries (e.g. import packages from it). This will be used in some of the test scripts for the individual assignments. Unless you always run these scripts from your repository's base directory, you can do a full in-place install of your cloned repository:
cd <your-repo>; pip install [--user] -e .
Use --user to install into your user directories (recommended unless you're using
a virtualenv or conda).