router in front of vLLM: anyone testing this?

[batchsize]

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

Just sharing a setup that ended up working reasonably well: Qwen, llama.cpp, Threadripper box, with a lot of attention on router policy.

Maybe useful as a starting point for newer folks.

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.

[benchplease]

Maybe, but I had the opposite result. I've seen people blame batch size when the real issue was much more mundane.

[southpatch]

Maybe, but I had the opposite result. I've seen people blame batch size when the real issue was much more mundane.

I think the bottleneck is somewhere else. The reported numbers can look compelling, but in ordinary usage I didn't see the same benefit. Could be a workload difference, but I wouldn't generalize too fast from one setup.

[batchsize]

worth checking the driver first.

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

[northhall88]

I'd approach it in a boring order. If you're on vLLM, I'd confirm the version first and then reduce variables until the behavior is obvious.

[mangofield33]

What worked for me was pretty simple. If you're on Ollama, I'd confirm the version first and then reduce variables until the behavior is obvious. Once you know the plain path works, then add the faster path back in piece by piece.

There are a few different conversations hiding inside this thread, so being explicit about the goal helps.

[phasestream53]

Maybe, but I had the opposite result. I've seen people blame INT4 when the real issue was much more mundane. Could be a workload difference, but I wouldn't generalize too fast from one setup.

[batchcraft]

Maybe, but I had the opposite result. I've seen people blame INT4 when the real issue was much more mundane. Could be a workload difference, but I wou

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. Worth testing the simpler explanation before adopting the more interesting one.

[cozysmith]

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

[batchsize]

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

Did a small comparison this afternoon. I couldn't reproduce the slowdown until I pushed context much higher. The main thing that moved was latency, not quality.