Aligning pretrained language models with instruction data generated by themselves. - View it on GitHub