mrm8488/gpt2-imdb-neg

A small GPT2 (lvwerra/gpt2-imDB-neg) language model fine-tuned to produce negative movie reviews based the IMDB dataset . The model is trained with rewards from a BERT sentiment classifier via PPO . The full experiment setup (for positive samples) in trl repo in the trl repostoia repo for the full experiment . It was trained for 100 optimisation steps with a batch size of 256 which corresponds to 25600 training samples . A few examples of the model response to a query before and after optimisation: "Query. (before) response (after) response) ... ...Query. ...response. (after,