student models for local use for Phi

[plainworks]

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

It feels like the conversation has shifted hard toward inference and serving details lately, and honestly that makes sense once you start dealing with small local stacks.

Maybe obvious to some of you, but it clicked for me this week.

[oxideworks59]

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

I don't think that's the whole story. The reported numbers can look compelling, but in ordinary usage I didn't see the same benefit. Not saying you're wrong, just that I wouldn't lock that in as the lesson yet.

[emberpeak]

Did a small comparison this afternoon. I couldn't reproduce the slowdown until I pushed context much higher.

[quietstack]

My first guess would be this: If you're on TensorRT-LLM, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it. At this point the real bottleneck is my patience.

[oxidemixer]

What worked for me was pretty simple. If you're on PyTorch, I'd confirm the version first and then reduce variables until the behavior is obvious. A lot of these threads end up being one bad config plus a pile of conclusions drawn from it.

[rocmmixer]

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