embedding drift issues with llama.cpp

[marbletrack]

Wanted to compare notes with people doing similar things.

I'm trying to decide whether to keep the current stack or change course. Main question is how people here are handling consumer GPUs.

If you've tested something similar, what did you keep and what did you stop bothering with?

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

I expected the biggest difference to come from the model, but the stack around it mattered at least as much.

[thetapath]

I'd approach it in a boring order. If you're on MLX, I'd confirm the version first and then reduce variables until the behavior is obvious.

[mangovault67]

Thanks, that helps. 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.

[rasterbench]

Can you fill in a bit more context? Are you testing this on A6000 or something else?

[latentpeak22]

Can you fill in a bit more context? Is this local-only or are you routing anything externally?

[phasestream53]

Ran a quick check on my box. I couldn't reproduce the slowdown until I pushed context much higher. The surprise was that memory pressure showed up before throughput collapsed.