Hacker News new | past | comments | ask | show | jobs | submit
xienze | on: Qwen 3.8 27B
Well there's a lot of knobs to turn if you want to improve performance. You can always point an LLM at the model card, give it your info, and have it write up the command.
Auracle | parent
Sure, but shouldn’t the programs to run the LLMs go “the user has this much vram and the model is this size, so I’ll start with sensible defaults based on that”?

You could override, obviously.

zargon | root | parent
Yes, llama.cpp does that.