do small edge boxes make sense with PyTorch

[southspark]

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

It feels like the conversation has shifted hard toward inference and serving details lately, and honestly that makes sense once you start dealing with token cost.

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

[thetaember38]

A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at consumer GPUs.

[emberpeak]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at continuous batching.

[deepseeked]

A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with Gemma and compare from there. At this point the real bottleneck is my patience.

[amberpatch25]

Ran a quick check on my box. The issue went away once I changed one setting and restarted the stack cleanly. I expected the model choice to matter more than it did.

[emberpeak]

I think the bottleneck is somewhere else. I get why that seems right, but on B200 the bottleneck felt elsewhere.

[llmarch51]

Thanks, that helps. I ended up in about the same place using LM Studio.

[homelabhenry]

Can you fill in a bit more context? What model, what runtime, and roughly what context size?

[mangopeak72]

I tested this tonight. My result was mixed: first run looked bad, second run looked normal.

[routeline23]

Thanks, that helps. I ended up in about the same place using LM Studio.

My first guess would be this: On dual 3090, 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.