desktop apps around local AI with LM Studio

[rocmcrate26]

Thought I'd post a small recipe because a lot of threads ask similar things.

Just sharing a setup that ended up working reasonably well: Phi, vLLM, dual 3090, with a lot of attention on tail latency.

Maybe useful as a starting point for newer folks.

My use case is pretty ordinary: local chat, some document work, and occasional batch jobs.

[rocmriley]

Thought I'd post a small recipe because a lot of threads ask similar things.

What worked for me was pretty simple. Before rebuilding anything, I'd test one clean baseline with FLUX and compare from there. Not glamorous advice, but repeatable baselines save a lot of time.

[audionode8]

Thought I'd post a small recipe because a lot of threads ask similar things.

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

[mintstation99]

A couple questions before I guess too hard. What model, what runtime, and roughly what context size?

[batchshed]

A couple things I'd check first. On 4090, the easiest win is often to back off the aggressive settings and see whether the problem survives.

[amberpatch25]

Tried to reproduce it just now. The issue went away once I changed one setting and restarted the stack cleanly. The main thing that moved was latency, not quality.

[mangoquill]

Need a little more detail here. Do you see the same thing with a smaller batch or shorter prompt?

[mlxgarden]

A couple things I'd check first. On dual 3090, the easiest win is often to back off the aggressive settings and see whether the problem survives.

[rocmhouse90]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at KV cache pressure. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[cachebench]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at KV cache pressure. A lot of t

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.