fhein

Assuming they already own a PC, if someone buys two 3090 for it they'll probably also have to upgrade their PSU so that might be worth including in the budget. But it's definitely a relatively low cost way to get more VRAM, there are people who run 3 or 4 RTX3090 too.

For LLMs it entirely depends on what size models you want to use and how fast you want it to run. Since there's diminishing returns to increasing model sizes, i.e. a 14B model isn't twice as good as a 7B model, the best bang for the buck will be achieved with the smallest model you think has acceptable quality. And if you think generation speeds of around 1 token/second are acceptable, you'll probably get more value for money using partial offloading.

If your answer is "I don't know what models I want to run" then a second-hand RTX3090 is probably your best bet. If you want to run larger models, building a rig with multiple (used) RTX3090 is probably still the cheapest way to do it.

Is max tokens different from context size?

Might be worth keeping in mind that the generated tokens go into the context, so if you set it to 1k with 4k context you only get 3k left for character card and chat history. I think i usually have it set to 400 tokens or something, and use TGW's continue button in case a long response gets cut off

llama.cpp uses the gpu if you compile it with gpu support and you tell it to use the gpu..

Never used koboldcpp, so I don't know why it would it would give you shorter responses if both the model and the prompt are the same (also assuming you've generated multiple times, and it's always the same). If you don't want to use discord to visit the official koboldcpp server, you might get more answers from a more llm-focused community such as [email protected]

A static website and Immich

So it's supposed to be 15 hours/month included with your premium subscription? Since I'm not familiar with how Spotify audio books work, I thought you meant that you had a free account and was allowed to listen 15 hours to books that would be included/unlimited with a premium subscription. Contact support if it ate through your monthly credits faster that it should. If you're a paying customer supports are usually quite helpful.

I've been using Intel NUCs, even though they have a lot of issues and start failing after about 3 years of heavy use. Previously used Kodi on Arch, but with the latest NUC I decided to go with Xubuntu and for some reason video playback doesn't work in Kodi now. So instead I just use VLC media player for TV/movies and a web browser for everything else. Got a Logitech K400 Plus wireless keyboard which makes it easy to control the computer from the couch.

Less than a year after that mail Swedish laws were rewritten to make copying music and movies illegal.

Because USA told us to.

There are tons of options for running LLMs locally nowadays, though none come close to GPT4 or Claude 2 etc. One place to start is /c/[email protected]

We mostly use discord since it's difficult to convince people to sign up for new services.. Have to do a workaround to stream desktop audio on Linux, since their client still only supports that for Windows, but other than that it usually works.

Tried https://twoseven.xyz/ a few times during a period when Discord streaming was lagging a lot. It supports desktop streaming with a browser plugin, and sync watching on various streaming services. As far as I can remember it worked ok but had a few issues, though that was a while ago so those might've been fixed.

Also tried to get https://sfu.mirotalk.com/ working but for some reason video wouldn't show up..

From what I can find, Plex downloads subs from opensubtitles.org and they already exist there. I think the problem is that it treats "Star Wars" and "Star Wars Despecialized Edition" as the same movie

I put all the subs in a zip file, in case anyone finds that easier than hunting them down individually on opensubtitles: https://www.swisstransfer.com/d/2ab10863-e9f9-442b-9d2c-44f0711f8280

Max validity was 30 days, so if someone has the possibility host them more permanently others might appreciate it in the future.

The only info I have about the actual video files is that Star Wars is supposedly Despecilized Edition v2.5, while ESB and RotJ only come with a text file crediting Harmy. Perhaps the latter two are also v2.5 but I have no note of it.