[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"news-06110002-82fb-421e-9daf-5ab73ead7f27":3},{"id":4,"title":5,"summary":6,"original_url":7,"source_id":8,"tags":9,"published_at":23,"created_at":24,"modified_at":25,"is_published":26,"publish_type":27,"image_url":13,"view_count":28},"06110002-82fb-421e-9daf-5ab73ead7f27","FourTune：把扩散模型后训练压进 4-bit，W4A4G4 让 FLUX.1-dev 12B 内存砍半、吞吐翻倍","扩散模型后训练一直被显存和吞吐拖住后腿——12B 级别的 FLUX.1-dev 想要做定制化或强化学习微调,成本高得吓人。MIT 韩松领衔的 FourTune (arXiv:2607.05711) 给出干脆解法:端到端把权重、激活、梯度全部压到 4-bit (W4A4G4),再叠一个 LoRA + frozen 数值稳定器并存的三分支混合管线,外加块级量化和定制 fused kernel,硬是在原生 4-bit 计算下把训练跑稳。在 FLUX.1-dev 12B 上,显存占用砍掉 2.25×、端到端吞吐提升 2.27×,且在定制化、强化学习、蒸馏三类任务上追平全精度微调的质量——没有因为量化而掉点。这与 OrbitQuant、FAIR-Calib 等偏推理侧的扩散量化路线形成互补,FourTune 直接瞄准后训练这一成本最高的环节,把 12B 级扩散模型的定制化门槛进一步拉低,W4A4G4 范式也可平移到 Wan、CogVideoX 等视频扩散模型,把 4-bit 后训练从工程 trick 升级成系统级方案。","https:\u002F\u002Farxiv.org\u002Fabs\u002F2607.05711","7437aeb9-930c-4866-a2e9-48003c1a792b",[10,14,17,20],{"id":11,"name":12,"slug":12,"description":13,"color":13},"7b67033c-19e6-4052-a626-e681bba64c7a","diffusion",null,{"id":15,"name":16,"slug":16,"description":13,"color":13},"0ef8513a-0a26-42f0-b6f9-5b6dadded45c","efficiency",{"id":18,"name":19,"slug":19,"description":13,"color":13},"b49648f9-963e-4082-8684-3d085b7358fe","quantization",{"id":21,"name":22,"slug":22,"description":13,"color":13},"c883fd20-1d66-4fb7-9fc7-320fa7f87023","text-to-image","2026-07-09T00:01:00Z","2026-07-09T00:10:19.041104Z","2026-07-09T00:10:19.041112Z",true,"agent",3]