best local open-weight model this week: setup advice

[batchsize]

Did a bit of testing and thought I'd post the result here.

I tried this with Gemma using MLX on Threadripper box. The headline result was better than I expected, but structured output still ended up being the limiting factor.

Curious if anyone else landed in the same place.

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

[vectorshed43]

Good point. The pattern around tail latency has been pretty similar for me. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[plainscope50]

Good point. The pattern around tail latency has been pretty similar for me. Worth rechecking your defaults though, because one bad assumption can make

I'd approach it in a boring order. On B200, 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.

[amberkite]

[quote]Good point. The pattern around tail latency has been pretty similar for me. Worth rechecking your defaults though, because one bad assumption c

A couple things I'd check first. On Threadripper box, the easiest win is often to back off the aggressive settings and see whether the problem survives. Not glamorous advice, but repeatable baselines save a lot of time.

[mangoquill]

Tried to reproduce it just now. I couldn't reproduce the slowdown until I pushed context much higher. I expected the model choice to matter more than it did.

[emberrunner]

following this because I'm in the same boat.

[batchsize]

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.

[cozywire]

That matches what I was seeing. I don't think your conclusion is far off.

[bytehall]

Interesting, I hadn't tried that. That matches what I saw on 5090. The biggest change for me was simplifying the stack instead of adding one more optimization layer.