what are you all running this week for Phi

[ggufwire22]

This keeps coming up in conversation so I figured I'd start a thread.

The more I work with local setups, the more it seems like prefix caching decides everything downstream.

Curious whether that lines up with what other people are seeing.

For context, I'm not chasing leaderboard numbers. I mostly want the system to feel consistent day to day.

[quiettoken]

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

[mangofield33]

Not sure I agree there. I get why that seems right, but on 5090 the bottleneck felt elsewhere. Could be a workload difference, but I wouldn't generalize too fast from one setup.

[plaingarden18]

Ran a quick check on my box. The issue went away once I changed one setting and restarted the stack cleanly.

[thetagrid15]

Appreciate the detail. I ended up in about the same place using vLLM. The biggest change for me was simplifying the stack instead of adding one more optimization layer.

[kernelmike]

Appreciate the detail. That matches what I saw on B200. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[marbletrack]

Interesting, I hadn't tried that. The pattern around small local stacks has been pretty similar for me. I got better results once I stopped chasing one metric and just looked at whether the workflow felt usable.

[tensorkite]

My first guess would be this: Start by verifying the driver/runtime combo, then test with a smaller prompt, then look at NVFP4.

[rocmplane]

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

[plaingarden18]

Tried to reproduce it just now. On 4090 with vLLM, I saw basically the same behavior.