few-shot when you actually need it

[thetafield]

Spent some time on this today and wanted to compare notes.

Two paths both seem viable, but they break down differently once long context becomes important.

Trying to avoid rebuilding the whole thing twice.

[quiettoken]

Tried to reproduce it just now. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.

[mlxgarden]

What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at HBM constraints.

[quantyard]

Thanks, that helps. I ended up in about the same place using SGLang. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[mintstation99]

Thanks, that helps. I ended up in about the same place using SGLang. Worth rechecking your defaults though, because one bad assumption can make the re

I think the bottleneck is somewhere else. I've seen people blame embeddings when the real issue was much more mundane. Could be a workload difference, but I wouldn't generalize too fast from one setup.

[thetaarch43]

What worked for me was pretty simple. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at consumer GPUs.

[deltastation]

I'd start with the simpler stack.

[mlxgarden]

Interesting, I hadn't tried that. I don't think your conclusion is far off.

[northgarden37]

Ran a quick check on my box. My result was mixed: first run looked bad, second run looked normal. I expected the model choice to matter more than it did.

[thetafield]

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