NVIDIA 管理库NVIDIA Management Library
NVIDIA 管理库（NVML）用于监控和管理 NVIDIA GPU 的状态。例如，它可以获取 GPU 的功耗、温度和已分配内存，以及设备的功率限制和功率限制状态等信息。关于这些指标的详细解释（包括如何解读功耗和温度读数），可参考 Modal 文档中的此页面。
NVML 的功能通常可通过 nvidia-smi 命令行工具访问，也能通过程序包装器供代码调用，例如 Python 中的 pynvml 和 Rust 中的 nvml_wrapper。
英文原文(Modal GPU Glossary)
The NVIDIA Management Library (NVML) is used for monitoring and managing the state of NVIDIA GPUs. It exposes, for example, the power draw and temperature of the GPU, the allocated memory, and the device's power limit and power limiting state. For details on these metrics, including how to interpret power and thermal readings, see this page on the Modal docs.
The functions of NVML are frequently accessed via the nvidia-smi command line utility, but are also accessible to programs via wrappers, like pynvml in Python and nvml_wrapper in Rust.
相关词条
本词条改编自 Modal GPU Glossary(CC BY 4.0)· 中文翻译 miter6/gpu-glossary-zh,MAE 整理排版。