NewsShow HN: Minimal LLM Post-Training Experiments on an 8GB GPU (SFT, DPO, GRPO)August 1, 20261 min readSource: Hacker NewsHN score 6，comments 0