recipes for a good local stack

[mlxgarden]

Thought I'd post a small recipe because a lot of threads ask similar things.

Maybe useful as a starting point: TensorRT-LLM on 7900 XTX, keeping structured output in mind from the start.

Feel free to tear it apart if you see something dumb.

[audionode8]

My first guess would be this: On Threadripper box, the easiest win is often to back off the aggressive settings and see whether the problem survives.

[rocmriley]

What worked for me was pretty simple. On GB200, the easiest win is often to back off the aggressive settings and see whether the problem survives. Not glamorous advice, but repeatable baselines save a lot of time.

[audionode8]

Interesting, I hadn't tried that. I ended up in about the same place using SGLang. Turns out 'just one more test' was a lie.

[mintstation99]

Can you fill in a bit more context? Are you testing this on 5090 or something else?

If you post your exact settings, people here can probably narrow it down pretty fast.

[rocmcrate26]

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.

[deltagrid]

Thanks, that helps. That matches what I saw on M3 Max. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[llmpeak]

My first guess would be this: Before rebuilding anything, I'd test one clean baseline with Kimi and compare from there.

[rocmriley]

My first guess would be this: Before rebuilding anything, I'd test one clean baseline with Kimi and compare from there.

I'd start with the simpler stack.

[thetaplane]

Ran a quick check on my box. The issue went away once I changed one setting and restarted the stack cleanly. I expected the model choice to matter more than it did.