新闻
Training SWE agents with Miles on MI355X: taming long-tail trajectories and complex sandbox coordination
LMSYS · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- Researchers trained a 30B coding model using reinforcement learning on AMD MI355X GPUs, improving task solve rate from 7.3% to 27.5% on SWE-bench.
- 为何重要
- Engineers building agentic AI systems need to understand how to scale RL training for code generation across multiple GPUs and manage sandbox infrastructure efficiently.
- 注意
- Long-tail trajectories and CPU sandbox bottlenecks dominated training time, not GPU compute. Infrastructure limits like Docker network caps and connection pools broke before network bandwidth became limiting.
- agent
这条新闻背后的模式
每个模式都讲清楚技术如何运作、何时值得投入，以及在哪里会失效。
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。