知识卡片

普通读书笔记卡 · 6759.a · 生成模型:GPT-6

资源估算区分权重与运行状态

理解等级:中级

权重内存依参数量、数据类型和量化表示计算,运行还需激活、缓存、临时空间及框架开销。书中34GB是特定系统观察,不是所有7B模型的固定要求;下载大小、常驻内存与峰值内存也不能互换。

这张卡片看不懂?

已有反馈 · 0 条

    还没有人提交反馈。

    在整书笔记中的位置

    原书出处 · 1 处

    • · 查看原书依据

      Tip Our system required a total of 34 GB of memory to load Llama 2 into RAM. When the model weights are decompressed, Llama 2 requires at least 28 GB of memory, and your operating system and running applications may require several more additional gigabytes of memory. Our Linux system required 6 GB to run several applications, including Python. Monitor your RAM usage when loading

    参考来源

    - 位置:《Natural Language Processing in Action, Second Edition》第10章《真实世界中的大语言模型》(来源项目:a19c7933-4eda-4176-aebd-30fe7e1ad292;源文件:extracted/Natural Language Processing in Action, Second Edition/_epub-src/OEBPS/Text/chapter-10.html) - 结论依据:摘录中关于资源估算区分权重与运行状态的定义、比较或条件支撑卡片;原书表述不严谨的范围已在覆盖记录逐项限定,不把案例结果外推为普遍结论。 - 原始内容:Tip Our system required a total of 34 GB of memory to load Llama 2 into RAM. When the model weights are decompressed, Llama 2 requires at least 28 GB of memory, and your operating system and running applications may require several more additional gigabytes of memory. Our Linux system required 6 GB to run several applications, including Python. Monitor your RAM usage when loading