benchmarks that map to real use

[marbletrack]

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

If you were starting from scratch today for this use case, what would you choose first and what would you ignore?

Simple and boring is a valid answer.

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

[tensorfield88]

That matches what I was seeing. I ended up in about the same place using vLLM.

[macropoint]

subscribed, curious where this lands.

[audionode8]

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

[circuitwire58]

worth checking the driver first.

[deepseeked]

Good point. I ended up in about the same place using PyTorch.

[cacheline]

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.

[localcraft17]

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

[plainember52]

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

[batchscope72]

Not sure I agree there. The reported numbers can look compelling, but in ordinary usage I didn't see the same benefit.