recipes for a good local stack with MLX

[mintstation99]

Did a bit of testing and thought I'd post the result here.

After a few runs on A6000, I ended up with decent performance, but the setup only really made sense once I adjusted around batch size.

Nothing scientific here, just practical notes from a real setup.

This is mostly for local work rather than public serving.

The part that surprised me was how much the boring details mattered more than the model choice.

[vectorfield]

Did a bit of testing and thought I'd post the result here.

I don't think that's the whole story. I've seen people blame prefill/decode split when the real issue was much more mundane. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.

[rocmriley]

A couple things I'd check first. On mini PC, the easiest win is often to back off the aggressive settings and see whether the problem survives. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[kernelgrid]

A couple questions before I guess too hard. Do you see the same thing with a smaller batch or shorter prompt?

[rocmriley]

That matches what I was seeing. I don't think your conclusion is far off. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[cachebench]

I think the bottleneck is somewhere else. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model. Could be a workload difference, but I wouldn't generalize too fast from one setup. I keep promising myself this is the last rebuild.

[rocmhouse90]

That matches what I was seeing. I ended up in about the same place using Ollama. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[kernelpeak]

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with Mistral and compare from there.

[rocmriley]

Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly. The surprise was that memory pressure showed up before throughput collapsed.