Show HN: Minimal LLM Post-Training Experiments on an 8GB GPU (SFT, DPO, GRPO)

(github.com) by popopanda | Aug 1, 2026 | 0 comments on HN