Jetson still a thing

[homelabhenry]

Good point. I don't think your conclusion is far off. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[promptspark58]

I tested this tonight. On M3 Max with PyTorch, I saw basically the same behavior. The main thing that moved was latency, not quality.

[vectorcrate98]

A couple questions before I guess too hard. Do you see the same thing with a smaller batch or shorter prompt? The logs were somehow calmer than I was.

[mintquill]

Appreciate the detail. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.