做LLM微调优化的团队终于可以省下GPU时间了——把ZO微调当推理跑，vLLM直接提速8倍，建议做低成本微调的人点开看看实现细节。
研究发现，大语言模型的零阶（ZO）微调本质上是推理密集型负载，而非传统训练。现有实现将ZO算法运行在训练循环中，导致工作负载与运行时的不匹配。研究者通过将ZO微调的重复评分阶段部署在推理运行时（如vLLM）上，在OPT-13B模型上实现了8.13倍加速，且精度几乎无损。该方法在多个模型规模下获得2.34-7.72倍加速，并支持MeZO风格的高秩分解实验。这项工作为将轻量级适配作为推理类负载调度提供了实用路径。
LLM Zeroth-Order Fine-Tuning is an Inference Workload
Zeroth-order (ZO) fine-tuning is attractive for large language models because it replaces backpropagation with forward objective evaluations. Existing implementations nevertheless execute ZO algorithms inside conventional training loops, even though their dominant work is repeated scoring under nearby parameter states. This creates a workload-runtime mismatch: the algorithm asks for structured inference-style scoring, while the system exposes a sequence of fragmented training-loop steps. We show that LLM ZO fine-tuning is an inference-dominated workload and execute its repeated scoring phase through a serving runtime. On OPT-13B SST-2, the resulting vLLM execution path completes the 20k-step LoZO run in 0.51 estimated training hours versus 4.15 hours for the official LoZO baseline under the matched LoRA-only setting, an 8.13x speedup, while reaching 0.922 final evaluation accuracy and 0.931 final full-validation accuracy. In core-step scaling experiments across OPT-1.3B to OPT-13B, the same runtime reorganization gives 2.34x--7.72x speedups. A MeZO-style high-rank factorized experiment shows that the same runtime paradigm can track a MeZO-like loss trajectory while running up to 2.55x faster. More broadly, representing ZO updates as dynamic adapter states suggests a practical path toward inference-time training, where lightweight adaptation can be scheduled as an inference-like workload rather than as a separate training job.