NVIDIA Management LibraryNVML
A library for monitoring and managing the state of NVIDIA GPUs, such as power draw, temperature and memory use.
The NVIDIA Management Library (NVML) monitors and manages the state of NVIDIA GPUs. It exposes readings such as:
- power draw and temperature
- allocated memory
- the device's power limit and power-limiting state
For what these metrics mean, including how to interpret power and thermal readings, see this page in the Modal docs.
Most people reach NVML through the nvidia-smi command line tool. Programs can also use it through wrappers like pynvml for Python and nvml_wrapper for Rust.
Related#
- nvidia-smi — the command line front end to NVML
- libnvml.so — the shared library that implements it