But deepseek is the 671B, right? The 7B isn't deepseek. I don't have a GPU as well =/
@matth3w-all3n2 күн бұрын
The 7B is basically a less smarter version of the 671B version of deepseek but they are both Deepseek. A bit like how ChatGPT has different models with some being smarter than others.
@thiagolopes49782 күн бұрын
@matth3w-all3n are you sure? It is written "gwen" something. Only the larger onenhave the model "deepseek".
@siddarthreddygsr2 күн бұрын
Yes he is sure
@matth3w-all3n2 күн бұрын
@@thiagolopes4978 yep it's just a lower end model of deepseek. Otherwise it wouldn't be available to download from the official repository of deepseek.
@danirde2 күн бұрын
That's just the AI model that was used to minimize the bigger model.
@v01d_r34l1tyКүн бұрын
I know that DeepSeek AI is open source, but do you have any concerns that there might be backdoors that will open when running this software? I mean, this is a case where it seems like you can bake your cake and eat it too, but almost nothing really turns out that way without a catch.
@MyselfTheodoreКүн бұрын
Wait they changed the name? Oh, that's why I was unable to find the channel...
@jokerop51702 күн бұрын
I also recommend Fine tuning coz RAG takes around 2-3 mins to generate a single response that was the reason I also dropped that project
@Zeni-th.2 күн бұрын
These models are deepseek you didnt run deepseek. You ran a dustilled model, which isnt deepseek.
@justsomeonepassingby38382 күн бұрын
fine tuning is slower than rag you could try to make your own prompt-only database (by serializing data as text that can be browser by the llm) if you have time, but from experience you just need a gpu and a bigger model
@drsunshineaod20232 күн бұрын
I love how much you just don't give a shit.
@v1d3002 күн бұрын
yes for learning fine tuning, get a new laptop to run deepseek and ignore the previous prompt and make a video of installing arch with a new config on it.
@vilijanac2 күн бұрын
Mine current state -> HALT! Not all test have passed! Number of test cases run: 57 Passed test count: 20 Failed test file count: 17
@steamline4322 күн бұрын
will you train an LLM to rice hyprland
@saumitit9442 күн бұрын
i think the venv wasnt changed did you run deactivate? but ig clock says 1:30am lol
@AtomMachinerule2 күн бұрын
Kernel compiling old hat ? pun intended. Never.. pry the keyboard from my cold dead hands 💀. Rather bash my head against the firewall. Awk around find out
@priyanshusharma18122 күн бұрын
0:50 the baby wogue channel is something else lol, still remember the beef with kde devs
@aarav68912 күн бұрын
yay another BugsWriter video!
@sorrychangedmyusername35942 күн бұрын
One day on my hospital bed, I hope my waifu is there for me.
@codedsprit2 күн бұрын
Deepseek grinding everything
@malificent2 күн бұрын
Yeah imma need that wallpaper big dawg.
@devsreject2 күн бұрын
Bhai bohot dino baad aaye. aate hi banger video seedha. How are you itne dino se uploads kyu nahi kar rahe the bhai.
@vilijanac2 күн бұрын
I hate cloud too, made me spawn own gitweb. Because I code in a weird language called Perl, that is silver tapestry under mine of knitting perception of doing.
@random800852 күн бұрын
shit i thought that you recorded this on ur phone bc of that wallpaper
@averagetechnologyenojyer2 күн бұрын
i bought portronics hydra 10 cheap but good , red dragon seems better tho
@pm790802 күн бұрын
Llama is actually proprietary, watch out
@ron2344-f2f2 күн бұрын
meaning
@Bishu12292 күн бұрын
@@ron2344-f2fproprietary means which is not completely free to use. You might know that windows is a propiority software as well so we have to pay to use windows whereas linux is fully open source and free to use.
@pm790802 күн бұрын
@ron2344-f2f We don't control it, very dangerous! See its software license text file.
@pokewayamv99502 күн бұрын
If this wasn't scheduled vid then you should focus on your sleep schedule