lmsys 用 AMD MI355X 训练 Qwen3-30B-A3B，SWE-bench 从 7.3% 提至 27.5%
LMSYS 在 AMD 显卡上跑通智能体 RL 训练，Qwen3-30B 的 SWE-bench 成绩翻了近 4 倍，工程细节都写在报告里了。
LMSYS 与 AMD 团队用 Miles 框架在 AMD MI355X 上对 Qwen3-30B-A3B 做全异步智能体强化学习训练，技术栈包括 SGLang、Megatron-LM 和 Harbor。训练后模型在 SWE-bench Verified 的 pass@1 从 7.3% 提升到 27.5%（100 任务分层子集、每任务 4 次尝试）。RL 步骤耗时从 850 秒降至 284 秒，提速约 3 倍。rollout 吞吐量从约 860 提升到约 3480 条轨迹/小时，同时运行 384 个智能体沙箱。
🚀 Training SWE Agents with Miles on AMD MI355X
The @AIatAMD and Miles teams trained Qwen3-30B-A3B with fully asynchronous agentic RL using SGLang, Megatron-LM, and Harbor. Keeping long trajectories and hundreds of CPU sandboxes from stalling training took some engineering.
Highlights: - SWE-bench Verified pass@1: 7.3% → 27.5% after SFT + RL (100-task stratified subset, 4 attempts per task) - 3× faster RL steps: 850s → 284s - ~4× rollout throughput (~860 → ~3,480 trajectories/h), with 384 agent sandboxes running at once.
Full report: https://t.co/BbZBN7nhQW