simonko

a little bit skeptic about 1m parameters though

Plausible on server cpus, but maybe lower for normal cpus idk

Yeah on server cpus actually makes sense

A laptop i5 8th gen has around 6mb cache, you can run like a 200k llm to generate tiny stories data, so it's possible on normal laptops and pcs too

a little bit skeptic about 1m parameters though

Plausible on server cpus, but maybe lower for normal cpus idk

make a like 1m param model for vision and put it into the cpus cache, here is another idea