NVIDIA NeMo RL supports FP8 low-precision reinforcement learning post-training, accelerating agent iteration

robot
Abstract generation in progress
ME News 消息,4 月 23 日(UTC+8),NVIDIA AI 近日宣布,其开源库 NVIDIA NeMo RL 新增了一项能力,支持使用低精度 FP8 格式进行强化学习(RL)后训练,以加速相关计算工作负载。根据其发布的信息,在 Qwen3-8B-Base 模型上,使用 FP8 格式可将 RL 工作负载的速度提升 1.48 倍。这一加速旨在为智能体的工具使用和多步骤任务实现更快的迭代周期。(来源:InFoQ)
View Original
This page may contain third-party content, which is provided for information purposes only (not representations/warranties) and should not be considered as an endorsement of its views by Gate, nor as financial or professional advice. See Disclaimer for details.
  • Reward
  • Comment
  • Repost
  • Share
Comment
Add a comment
Add a comment
No comments