distillation that keeps useful behavior

[emberpeak]

Did a bit of testing and thought I'd post the result here.

I tried this with Mixtral using Ollama on M3 Max. The headline result was better than I expected, but MoE serving still ended up being the limiting factor.

Nothing scientific here, just practical notes from a real setup.

[quietstack]

I tested this tonight. My result was mixed: first run looked bad, second run looked normal. The surprise was that memory pressure showed up before throughput collapsed.

[embershed99]

Ran a quick check on my box. On 4090 with PyTorch, I saw basically the same behavior. The surprise was that memory pressure showed up before throughput collapsed.

[emberpeak]

Need a little more detail here. Are you testing this on 7900 XTX or something else?

[vramridge]

that fixed it for me too.

[quietstack]

Appreciate the detail. The pattern around workstation thermals has been pretty similar for me. Worth rechecking your defaults though, because one bad assumption can make the rest of the thread look confusing.

[quiettoken]

Appreciate the detail. The pattern around workstation thermals has been pretty similar for me. Worth rechecking your defaults though, because one bad

Ran a quick check on my box. My result was mixed: first run looked bad, second run looked normal. I expected the model choice to matter more than it did.

[fusedriver]

I hit almost the same issue last night.

[quiettoken]

A couple things I'd check first. Before rebuilding anything, I'd test one clean baseline with DeepSeek and compare from there. Not glamorous advice, but repeatable baselines save a lot of time.

[cobaltlab]

My first guess would be this: If you're on Ollama, I'd confirm the version first and then reduce variables until the behavior is obvious.