routing between local and hosted

[fastbench]

This keeps coming up in conversation so I figured I'd start a thread.

The more I work with local setups, the more it seems like structured output decides everything downstream.

Curious whether that lines up with what other people are seeing.

[cudapoint]

Did a small comparison this afternoon. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.

[pixelspark]

Tried to reproduce it just now. On Threadripper box with Ollama, I saw basically the same behavior. The surprise was that memory pressure showed up before throughput collapsed.

[stackarch]

Interesting, I hadn't tried that. 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.

[prompttinker44]

That matches what I was seeing. The pattern around reranking has been pretty similar for me. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[vectorshed43]

Good point. I ended up in about the same place using llama.cpp. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[kernelrunner]

A couple questions before I guess too hard. Is this local-only or are you routing anything externally?

[turboscope]

Good point. I ended up in about the same place using llama.cpp. The biggest change for me was simplifying the stack instead of adding one more optimiz

following this because I'm in the same boat.

[visionforge]

I'd approach it in a boring order. On 7900 XTX, the easiest win is often to back off the aggressive settings and see whether the problem survives. Once you know the plain path works, then add the faster path back in piece by piece.

[nanobrook]

What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at prefill/decode split. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.