Talaria: Interactively Optimizing Machine Learning Models for Efficient Inferenc

Hi, I’m Jochen, one of the authors.<p>We recently did a Show HN (<a href="https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=41463916">https:&#x2F;&#x2F;news.ycombinator.com&#x2F;item?id=41463916</a>) which did not get much traction, so I’m posting this again here:<p>We just released Mycelium, the library that powers Talaria’s graph viewer. You can check it out and play around with it here: <a href="https:&#x2F;&#x2F;apple.github.io&#x2F;ml-mycelium" rel="nofollow">https:&#x2F;&#x2F;apple.github.io&#x2F;ml-mycelium</a><p>I’m happy to answer any questions about Talaria or Mycelium!

Are inference metrics like latency and power measured live from device? To which devices can Talaria be applied?

How does this compare to TVM?

Could you give us a tl;dr on this project? and how could I use something like this work for on-device applications, think &quot;smart home&quot; style applications?