Zml-smi: universal monitoring tool for GPUs, TPUs and NPUs

 (zml.ai)

61 points | by steeve 5 days ago

5 comments

  • imcritic 23 minutes ago
    Is it capable of exposing metrics in Prometheus format?
    [-]
    • steeve 11 minutes ago
      consider it done
  • rdyro 8 hours ago
    Looks cool!

    nvtop can actually support TPUs too via https://github.com/rdyro/libtpuinfo/ https://github.com/Syllo/nvtop/blob/76890233d759199f50ad3bdb...

  • serialx 3 hours ago
    Look into all-smi https://github.com/lablup/all-smi It supports all GPUs thinkable including Apple Silicon and many AI accelerator cards.
  • mrflop 5 days ago
    Renaming fopen64 to intercept library calls feels like a brittle hack masquerading as "sandboxing." Why not just upstream this hardware support to nvtop instead of fragmenting the ecosystem?
    [-]
    • steeve 5 days ago
      sadly, sandboxing is something that can't be upstreamed. this way, sandboxing is kept in zml instead of patching mesa.

      as for nvtop, great program, but we missed a few features (such as sandboxing)

      [-]
      • pstuart 6 hours ago
        It looks cool and I was excited to get monitoring for the NPU on my Ryzen AI 395+, unfortunately it does not show. NPU support in linux really seems to be an afterthought.
        [-]
        • steeve 6 hours ago
          Weird, because we tried it. It doesn’t show anything?

          We use the amdsmi to get metrics. I’ll investigate.

    • marwanet 6 hours ago
      If this logic were pushed into nvtop, wouldn't the codebase become unmaintainable? Each vendor's interception method is going to be different.
      [-]
  • 152334H 4 hours ago
    "NPU" seems to refer to trainium only?