Awesome video, been trying to do something similar but with malware analysis and this just inspired me to continue with that pursuit
@afterglow528516 күн бұрын
Nice series!. This is certainly a rabbit hole.
@enmanuelmorales966717 күн бұрын
Hi John! just been in a meeting were you talked and everything you are doing is very very promising, Ive got a question that I wanted to ask you if possible
@petermorvay184219 күн бұрын
Hi John, thanks for this video. Looking forward where this is going. Does it make sense to give to LLMs some personality/system prompt? Like you are a network expert, focus on the details of the protocols, names and specific of the particular packet, inluding, "..." ?
@johncapobianco252718 күн бұрын
@@petermorvay1842 yes I prime the prompt with some additional context as you suggest
@thegtlab16 күн бұрын
why is my comment being deleted?
@johncapobianco252716 күн бұрын
?
@thegtlab16 күн бұрын
@@johncapobianco2527 hey! not sure what is happening but I made a comment about this not being the real deepseekv3 model. The real one is 650B, the one you are pulling is just llama 3.2 under the hood. Just search "nezahatkorkmaz/deepseek-v3" in google and you will see a X thread about it. To run the real one locally you will need TONS of VRAM or RAM.
@thegtlab16 күн бұрын
@@johncapobianco2527 youtube is auto-deleting my comment and I don't know what word is triggering it. Basically that is not the real DS3 model. It's running a different model under the hood. The real one is 650B.
@johncapobianco252716 күн бұрын
@ very cool call out - yes this is not the DS3 that runs in their cloud
@thegtlab16 күн бұрын
@ np...under the hood, it's actually llama3.2; I'm not sure why the user published it, perhaps to confuse people