Feature(Forge): Complete Bullet Dataset for Training

[justinthelaw]

Generate a data set that will be used to train a model for generation of Bullet Forge training data. Generally speaking, this model acts as a bullet interpreter that will take-in a set of the bullets that have been scraped form the internet, and spit back out generalized and context-rich sentences. This "bullet interpreter" will act as a standalone model capable of translating any bullet into easily comprehendible achievements that look as if they have been written by an Airman or Guardian.

There are two tasks involved with this issue:

• Generate a first run of this bullet interpreter model, v0.1.0, using a corpus of 500 select bullets that have been run through ChatGPT
• Generate a further 1000 select bullets and interpreted completions, totaling 1500 now, with human-cleaning involved to produce a v1.0.0 of the bullet interpreter
• Use v1.0.0 of the bullet interpreter to generate all 33,600+ bullets and bullet interpretations to form the final training set for the bullet generation models