Penny-1.7B

Penny-1.7B is an AI model capable of translating modern text into the style of the Irish Penny Journal. This model was trained using GRPO (Generalized Reinforcement Learning with Policy Optimization) in less than a day on a single A6000 GPU, without any Supervised Fine-Tuning (SFT).

The training involved creating a synthetic dataset of sentences from the Irish Penny Journal and their modern translations. A small encoder (MiniLM2) was trained to classify these texts, and its output was used to guide the training of the larger Penny-1.7B language model.