Our group consisted of Bjarke Larsen, Ethan Osborne, and Marj Cuerdo. To run this project, you need to open it in Google Colab. Here is the original: https://colab.research.google.com/drive/1VLG8e7YSEwypxU-noRNhsv5dW4NfTGce. You can upload your own text file to train it on whatever you want. It says to make sure you are running in Chrome, but I ran it in Firefox and had no issues.
This is an implementation of the GPT2 learning model using Destiny Grimoire Lore. It is trained for 1000 steps and here are images to show how the model does every 200 steps. The constraints that were implemented were the input only being a part of the Destiny Lore as a whole. Interestingly enough, there was reference to an Intel processor 800 steps in, which seems out of constraint and we were not sure what led to that. Our process in creating this good output was to let it run for as long as the original model had set up and luckily for us it led to some quality fake Destiny Lore. A full example can be found at Destiny 2 Text.txt
Step 200
Step 400
Step 600
Step 800
Fully Trained
