Start date: August 29th, 4:03 PM
Goal: To create a "filter" for the larger language model for text generation to ensure its output is not too inappropriate for a situation given the context. This can be done via text classification, in which bert excels in as opposed to language generating models like ChatGPT. I'm considering making this project in a way that will result in a feedback loop between the models, hopefully helping the language generation model generates better text
Current Progress: -Learning How to stack transformer decoders -trying to find relevant data I can use to train the model
To Do: Finish coding out the decoders, encoders (incase it is needed), and learn how the bert architecture works (since all I know right now is that it is a stack of decoders in the transformer architecture)