+- Inference (https://www.datasculpting.ai/inference)
+-- Forum: Harnesses, Agents & Orchestration (https://www.datasculpting.ai/inference/forumdisplay.php?fid=11)
+--- Forum: Routers (https://www.datasculpting.ai/inference/forumdisplay.php?fid=45)
+--- Thread: dataset format recommendations (/showthread.php?tid=487)
Pages:
1
2
Re: dataset format recommendations - forgemixer - 10-05-2026
Thanks, that helps. 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.
Re: dataset format recommendations - phasenode - 10-05-2026
I'd approach it in a boring order. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at workstation thermals.
Re: dataset format recommendations - routecraft - 10-05-2026
Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher. The main thing that moved was latency, not quality.
Re: dataset format recommendations - forgemixer - 10-05-2026
Quote:Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher. The main thing that moved was latency, not quality.
I tested this tonight. On Threadripper box with MLX, I saw basically the same behavior. I expected the model choice to matter more than it did.
Re: dataset format recommendations - routecraft - 10-05-2026
Quote:[quote]Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher. The main thing that moved was latency, not q
Interesting, I hadn't tried that. I ended up in about the same place using SGLang. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.
Re: dataset format recommendations - plainworks - 10-05-2026
Quote:[quote][quote]Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher. The main thing that moved was latency
I don't think that's the whole story. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model.