consumer GPU serving economics with SGLang

[stackarch]

Posting this in case it saves someone else a little time.

I'm helping a small team set up a local workflow. We don't need a giant stack, but we do care about speculative decoding.

Would rather learn from your scars than collect my own all at once.

[localpatch78]

That matches what I was seeing. 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.

[orbitbench6]

I'd approach it in a boring order. Before rebuilding anything, I'd test one clean baseline with FLUX and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.

[stackarch]

I hit almost the same issue last night.

[mintquill]

I'd approach it in a boring order. Before rebuilding anything, I'd test one clean baseline with Phi and compare from there. Once you know the plain path works, then add the faster path back in piece by piece.

[rocmmixer]

Good point. I don't think your conclusion is far off. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[tensorfield88]

Maybe, but I had the opposite result. I wouldn't optimize around that first. On my side the bigger swing came from the runtime and not the model.