structured output failures

[paperrunner]

I figured I'd ask here before changing everything.

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

Would appreciate practical advice rather than theory.

[cudacrate]

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

[mellowsupply]

I hit almost the same issue last night.

[promptpoint]

following this because I'm in the same boat.

[fusedtinker]

Thanks, that helps. I don't think your conclusion is far off. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[macrostack69]

A couple things I'd check first. Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at long context. Once you know the plain path works, then add the faster path back in piece by piece.

[dockkite69]

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. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[orbitgarden]

Can you fill in a bit more context? What model, what runtime, and roughly what context size?

[deepseeked]

I don't think that's the whole story. I get why that seems right, but on 5090 the bottleneck felt elsewhere. Turns out 'just one more test' was a lie.