3 ms·
Looking into running models locally, maybe a 405B parameter model sounds like the place to start. Once understood you could practice with a private hosted llm
by j45 2y ago
Looking into running models locally, maybe a 405B parameter model sounds like the place to start.
Once understood you could practice with a private hosted llm (run your own model) to tweak and get it dialled in per hour, and then make the leap.
- wkat4242 2y agoAnd now you can get the 405b quality in a 70b according to meta. Costs really come down massively with that. I wonder if it's really as good as they say though.