GPRM's picture
Update README.md
4fde0fb verified
A straightforward implementation of the **Backward Model Myx**, inspired by the paper "Self-Alignment with Instruction Backtranslation."
This model was fine-tuned on **LLaMA-2-hf** using (output, instruction) pairs {(yi, xi)} from the OpenAssistant-Guanaco training dataset.
The fine-tuning process was conducted using **LoRA**, and the uploaded model is provided in its **merged** form.