NVML is not supported on Jetson platforms, e.g. Jetson Nano. Supporting Jetson platforms can be very useful for people who do ML/MLSys research on embedded platforms. Related discussion entry: https://github.com/ml-energy/zeus/discussions/102
We can make this a new zeus.device.
Things to figure out:
[ ] Get access to an actual Jetson device
[ ] Try out the sysfs interface for power measurement (ref).
[ ] Is this for GPU only? Or is this CPU + GPU? Or the whole device?
[ ] What is the granularity of information exposed? How often does it update? Are there any overhead related to reading this?
[ ] What about the energy consumption of DRAM, which is shared by the CPU and GPU?
[ ] Are there any control knobs on the platform, like power limit?
NVML is not supported on Jetson platforms, e.g. Jetson Nano. Supporting Jetson platforms can be very useful for people who do ML/MLSys research on embedded platforms. Related discussion entry: https://github.com/ml-energy/zeus/discussions/102
We can make this a new
zeus.device
.Things to figure out: