Training/Evaluation codes mostly come from this Turotials
The original model comes from microsoft/DialoGPT-medium and is finetuned on a discord-conversation-dataset
The dataset is not pre-processed as it's just a large chunks of conversation dialogue in a text file. Regex is mainly used to extract the useful data from the file.
- Limit the frequency of sending requests to discord
- IDK
pip install torch
pip install transformers
pip install discord