prompt length getting silly: worth it?

[turboarch95]

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 small local stacks.

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

I care more about stability than squeezing the last possible tokens/sec out of it.

[mlxgarden]

worth checking the driver first.

[thetaarch43]

What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at MoE serving. I keep promising myself this is the last rebuild.

[mlxgarden]

What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at MoE serving. I keep p

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.

[thetafield]

What exact setup are you running? Is this local-only or are you routing anything externally?

[promptspark58]

What exact setup are you running? Is this local-only or are you routing anything externally?

[paperport64]

I'd approach it in a boring order. If you're on llama.cpp, 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.

[cozywire]

What worked for me was pretty simple. If you're on TensorRT-LLM, I'd confirm the version first and then reduce variables until the behavior is obvious.

[quantyard]

Appreciate the detail. 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.

[promptspark58]

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