student models for local use: worth it?

[cobaltlab]

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

One thing I keep noticing is that people underestimate INT4 until they actually try to run this stuff for more than one afternoon.

Maybe obvious to some of you, but it clicked for me this week.

[quietstack]

My first guess would be this: On 4090, the easiest win is often to back off the aggressive settings and see whether the problem survives. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[relaysupply90]

Appreciate the detail. 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.

[rocmmixer]

Interesting, I hadn't tried that. That matches what I saw on dual 3090. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[llmtrack26]

Did a small comparison this afternoon. I couldn't reproduce the slowdown until I pushed context much higher. The surprise was that memory pressure showed up before throughput collapsed.

[vramridge]

Thanks, that helps. That matches what I saw on M3 Max.

[plainworks]

Tried to reproduce it just now. I couldn't reproduce the slowdown until I pushed context much higher.

[mangopath4]

A couple things I'd check first. If you're on llama.cpp, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[quiettoken]

I'd approach it in a boring order. If you're on TensorRT-LLM, I'd confirm the version first and then reduce variables until the behavior is obvious. Not glamorous advice, but repeatable baselines save a lot of time.

[cobaltlab]

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