LLM alignment: SFT with KL-divergence and DPO - View it on GitHub