This model is a fine-tuned version of the 1.3B-parameter GPT-Neo model developed by EleutherAI. As the default GPT-Neo model did not receive any social media data during its pre-training, we fine-tuned it with tweets collected from Twitter from October to November 2021 related to climate change hashtags. The model received data in the format <username> - <tweet> We used an 80/20 train/test split, and to differentiate distinct tweets, we added a start-of-tweet and an end-of-tweet token to the training dataset.

To guide you in using this model, please consult the gpt_neo_1.3B_twitter.ipynb Jupyter Notebook file from this repository.