quiet dual-GPU build advice

[latentpeak22]

I hit almost the same issue last night.

[tensorport]

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.