+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Harnesses, Agents & Orchestration (https://www.datasculpting.ai/inference/forumdisplay.php?fid=11)
+--- Forum: Agents (https://www.datasculpting.ai/inference/forumdisplay.php?fid=44)
+--- Thread: distillation that keeps useful behavior for Llama (/showthread.php?tid=236)
distillation that keeps useful behavior for Llama - quiettoken - 09-12-2026
Posting this in case it saves someone else a little time.
I'm not trying to build a perfect architecture diagram. I just want a sane starting point around SGLang and Threadripper box.
I'm especially interested in mistakes worth avoiding.
I care more about stability than squeezing the last possible tokens/sec out of it.
Re: distillation that keeps useful behavior for Llama - dockember - 09-12-2026
My first guess would be this: If you're on vLLM, I'd confirm the version first and then reduce variables until the behavior is obvious.
Re: distillation that keeps useful behavior for Llama - quiettoken - 09-12-2026
Ran a quick check on my box. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.
Re: distillation that keeps useful behavior for Llama - cobaltlab - 09-13-2026
Interesting, I hadn't tried that. I don't think your conclusion is far off.
Re: distillation that keeps useful behavior for Llama - dockember - 09-13-2026
That matches what I was seeing. I don't think your conclusion is far off.
Re: distillation that keeps useful behavior for Llama - thetagrid15 - 09-14-2026
Thanks, that helps. That matches what I saw on 7900 XTX. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.