3 ms·
Agreed. Where may I read about how to set up an LLM similar to that of Claude, which has the minimum length of Claude's context window, and what are the hardwar
by johnisgood 2y ago
Agreed. Where may I read about how to set up an LLM similar to that of Claude, which has the minimum length of Claude's context window, and what are the hardware requirements? I found Claude incredibly useful.
- j45 2y agoLooking 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.