| This is my first attempt at finetuning the GPT AI. | |
| Mia is based on GPT-Neo-125M and was trained on the original training data of the AI Dungeon model_v5 model. | |
| It is a much smaller model than the original model_v5 and will run a lot quicker. | |
| Training seemed to have helped with it not repeating commands back to you, but 125M is not large enough for proper gameplay. | |
| The model was entirely trained on Google Colab (125M is the largest you can train) and ran for almost 3 epoch's (The last 5 mintues I ran out of diskspace but managed to export the model). | |
| - Henk717 |