Pretty sure llama.cpp can already do that | Hacker News

Hacker News new | past | comments | ask | show | jobs | submit

login

		happyPersonR 13 days ago \| parent \| context \| favorite \| on: TScale – Distributed training on consumer GPUs Pretty sure llama.cpp can already do that

TYMorningCoffee 13 days ago [–]

I forgot to clarify dealing with the network bottleneck

moralestapia 12 days ago | [–]

Just my two cents from experience, any sufficiently advanced LLM training or inference pipeline eventually figures out that the real bottleneck is the network!

Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact