AI自己调训练集群，快2.75倍
训练大模型的瓶颈常常不在模型本身，而在「怎么把一堆 GPU 拧成一台机器」——启动参数、算子、通信排布彼此咬合，改一处就 OOM、编译冲突或数值悄悄漂移。这篇让 AI 自己去调这套基础设施：人只给一句开放指令，它先锁死可改范围和「不许作弊」的验收契约(负载、loss、计时不变)，再摸清集群噪声基线，然后同时改启动旋钮和代码补丁，把候选方案拿回基线复验，失败的代价也记账。关键一步是不从单一现状贪心往上爬，而是从多样经验种子出发，对所有试验(包括爆内存的)拟合全局耗时与峰值内存模型，用模型给每个配置的内存开销定价，让吃内存的优化必须搭配省内存的改动。四个单机/多机、稠密与 MoE 负载上最高提速 2.75 倍，平均 MFU 从 17.4% 提到 30.0%，经验还能跨 GPU 迁移。它不是你明天能上手的东西——但如果你在为大模型训练账单发愁，这是「让 AI 优化 AI 生产线」这条路上第一个跑通闭环的证据。
📄 原文摘要(英文)
While Recursive Self-Improvement (RSI) accelerates isolated algorithm and kernel synthesis, distributed LLM training depends on tightly coupled launch settings, operator kernels, and communication schedules, where single edits often hit out-of-memory (OOM) boundaries, compiler conflicts, measurement jitter, or silent numerical drift. We present InfraRSI, the agentic harness that turns open-ended human commands into verified training optimizations via a five-stage workflow: init locks editable scope and a fixed task contract over workload, loss, and timing invariants to prevent reward hacking, doctor profiles baseline phases and calibrates cluster noise, run co-optimizes launch knobs and code patches before re-verifying top candidates against the baseline and logging findings to a cross-hardware experience bank, and report and apply audit source paths, track failed-trial costs, and commit promoted changes. Rather than stepping greedily from a single incumbent, InfraRSI initializes from diverse experience-seeded configurations and fits global execution-time and censored peak-memory models across all trials, including OOMs. Using these models, it prices per-setting memory costs, revises miscalibrated LLM priors from live runs, and pairs memory-hungry optimizations with memory savers and verified code patches. Across four single- and multi-node dense and Mixture-of-Experts (MoE) LLM workloads, InfraRSI achieves a speedup of up to 2.75times and lifts mean Model FLOPs Utilization (MFU) from 17.4% to 30.0%, outperforming existing tuners and coding agents while transferring experience across GPUs. As it improves training infrastructure efficiency, InfraRSI forms a two-level Recursive Self-Improvement loop---revising priors online within runs and distilling experience across clusters---to push foundation model scaling.