undefined | Better HN

0 pointsottah4d ago0 comments

That's actually pretty cool, but I'd hate to freeze a models weights into silicon without having an incredibly specific and broad usecase.

0 comments

patapong3d ago

Depends on cost IMO - if I could buy a Kimi K2.5 chip for a couple of hundred dollars today I would probably do it.

whatever13d ago

I mean if it was small enough to fit in an iPhone why not? Every year you would fabricate the new chip with the best model. They do it already with the camera pipeline chips.

superxpro123d ago

Sounds like just the sort of thing FGPA's were made for.

The $$$ would probably make my eyes bleed tho.

chrsw3d ago

Current FPGAs would have terrible performance. We need some new architecture combining ASIC LLM perf and sparse reconfiguration support maybe.

0x4573d ago

Wouldn't it be the opposite of freezing weights?

j / k navigate · click thread line to collapse