this post was submitted on 28 Jan 2025
28 points (96.7% liked)

askchapo

22938 readers
174 users here now

Ask Hexbear is the place to ask and answer ~~thought-provoking~~ questions.

Rules:

  1. Posts must ask a question.

  2. If the question asked is serious, answer seriously.

  3. Questions where you want to learn more about socialism are allowed, but questions in bad faith are not.

  4. Try !feedback@hexbear.net if you're having questions about regarding moderation, site policy, the site itself, development, volunteering or the mod team.

founded 4 years ago
MODERATORS
 

Is there anything I need to know ? Where can I get it clean ? Is it type and go ? Its free right and open source ?

top 10 comments
sorted by: hot top controversial new old
[–] gay_king_prince_charles@hexbear.net 12 points 2 months ago (1 children)

https://www.deepseek.com/ is the "intended" experience and has the highest quality, but it isn't local. If you want to run it locally, install ollama (https://ollama.com/) and run ollama pull deepseek-r1:<size>. I doubt you will be able to run 617b, but 14b preforms brilliantly.

[–] 9to5@hexbear.net 5 points 2 months ago* (last edited 2 months ago) (1 children)

What would I gain from it being local ?

[–] FumpyAer@hexbear.net 12 points 2 months ago (1 children)

Not having data collected from you and unlimited tokens.

[–] 9to5@hexbear.net 5 points 2 months ago (1 children)

What are tokens ? Sorry I dont know much about this stuff beyond the surface level.

[–] fox@hexbear.net 7 points 2 months ago (1 children)

When you message an LLM that input is parsed into tokens, and a response is generated that's built out of tokens. You get a certain number of tokens with online models per day.

[–] 9to5@hexbear.net 6 points 2 months ago (2 children)

ok so in short the amount of questions you can ask is limited ?

[–] peeonyou@hexbear.net 5 points 2 months ago* (last edited 2 months ago)

tokens are what the model breaks input and output into, and how it maps what words should come next in any given sentence.. so basically when you use one of these models online they limit the number of tokens you can input/output in a given time because it is costly to run and people will abuse it otherwise. If you're running it on your own machine then you don't have to worry about it because you can use your machine to the max if you want.

[–] Vingst@hexbear.net 4 points 2 months ago

tokens are words, more or less

[–] FumpyAer@hexbear.net 10 points 2 months ago* (last edited 2 months ago)

You need a GPU with like 12 or 16 GB of VRAM minimum to run it locally. Or else it will run very, very, very slowly.

If you just want to play with it you need to give them your email to make an account or log in with Google.

[–] ClathrateG@hexbear.net 2 points 2 months ago

If you just want to see what it is and play around you can get it from the app/play store