在 Amazon SageMaker HyperPod 上微调 GPT-OSS-20B 实现多语言推理
AWS 团队在 Amazon SageMaker HyperPod 上用 SFT 与 GRPO 微调 GPT-OSS-20B,解决其只用英文推理的问题:基座模型在 50 例多语言推理测试中仅 22% 用对目标语言(英语 100%、法语 0%、德语 10%、西班牙语 0%、意大利语 0%)。结论是 SFT 承担大部分工作,GRPO 在模型已有一定概率产出目标行为时才能带来可测量的增益。训练使用 LoRA,数据为 HuggingFaceH4/Multilingual-Thinking(约 949 条训练、50 条验证)。

doesn’t have. GRPO optimizes a behavior the model can already sometimes produce. GRPO works by sampling a group of responses per prompt and pushing the policy toward the above-average ones. If the desired behavior has zero probability in the base distribution, there is very little for GRPO to reinforce. That single idea explains almost every result in this post. The training stack We ran on Amazon SageMaker HyperPod with Amazon Elastic Kubernetes Service (Amazon EKS) orchestration. For the init
匹配字段正文


