+- 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 with vLLM (/showthread.php?tid=995)
Pages:
1
2
distillation that keeps useful behavior with vLLM - quiettoken - 10-10-2026
Curious what people here are seeing.
I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling tokens/sec.
Would appreciate practical advice rather than theory.
This is mostly for local work rather than public serving.
Re: distillation that keeps useful behavior with vLLM - plainember52 - 10-11-2026
What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at tokens/sec. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.
Re: distillation that keeps useful behavior with vLLM - vramridge - 10-12-2026
that fixed it for me too.
Re: distillation that keeps useful behavior with vLLM - thetagrid15 - 10-12-2026
Need a little more detail here. Is this local-only or are you routing anything externally?
Re: distillation that keeps useful behavior with vLLM - turboyard94 - 10-12-2026
That matches what I was seeing. I ended up in about the same place using SGLang. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.
Re: distillation that keeps useful behavior with vLLM - quietstack - 10-12-2026
Not sure I agree there. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model. Worth testing the simpler explanation before adopting the more interesting one.
Re: distillation that keeps useful behavior with vLLM - amberpatch25 - 10-12-2026
I tested this tonight. My result was mixed: first run looked bad, second run looked normal. I expected the model choice to matter more than it did.
Re: distillation that keeps useful behavior with vLLM - turboyard94 - 10-12-2026
Good point. That matches what I saw on 7900 XTX. The biggest change for me was simplifying the stack instead of adding one more optimization layer.
Re: distillation that keeps useful behavior with vLLM - quietstack - 10-14-2026
What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at tail latency. Once you know the plain path works, then add the faster path back in piece by piece.
Re: distillation that keeps useful behavior with vLLM - batchscope72 - 10-14-2026
I think the bottleneck is somewhere else. I've seen people blame tokens/sec when the real issue was much more mundane.