[11 / 1 / 1]
Quoted By: >>1475328
I originally tried to post the questions on LocalLLaMA on reddit, but I have severe communication difficulties and almost never post or respond, so my article was automatically deleted the moment I sent it, so sad.
Just trying my luck and hoping someone here can help me... or just watching and hoping I overload the wires and burn down the entire apartment. This is Fine.
I don't understand English, and I'm so lazy that I only use Google Translate for this article. My incoherent articles will probably be as painful as torture for you to read.
Background:
The top floor of a small apartment in Germany (I heard it seemed to have been renovated in 2005) will turn into an oven when exposed to direct sunlight in the summer.
I have absolutely no experience or knowledge in assembling computers, and such a haphazard upgrade method means that even if I am willing to pay, no regular computer store is willing to help me assemble and upgrade it. (No, I have no friend, and my family doesn't understand this, so there's really no one who can help me upgrade and assemble my computer.)
I don't train LLM, I just want to use Q8 or 8bpw 103B (and above) model and run the unquantized original llama3 70B in 10K or 12K contexts, only for ERP and RP. Yes, pathetic, I know.
For reference, I have attached photos of the transformer box and a sketch of the approximate location of the sockets.
[The article is quite long so I will post it in sections.]
Just trying my luck and hoping someone here can help me... or just watching and hoping I overload the wires and burn down the entire apartment. This is Fine.
I don't understand English, and I'm so lazy that I only use Google Translate for this article. My incoherent articles will probably be as painful as torture for you to read.
Background:
The top floor of a small apartment in Germany (I heard it seemed to have been renovated in 2005) will turn into an oven when exposed to direct sunlight in the summer.
I have absolutely no experience or knowledge in assembling computers, and such a haphazard upgrade method means that even if I am willing to pay, no regular computer store is willing to help me assemble and upgrade it. (No, I have no friend, and my family doesn't understand this, so there's really no one who can help me upgrade and assemble my computer.)
I don't train LLM, I just want to use Q8 or 8bpw 103B (and above) model and run the unquantized original llama3 70B in 10K or 12K contexts, only for ERP and RP. Yes, pathetic, I know.
For reference, I have attached photos of the transformer box and a sketch of the approximate location of the sockets.
[The article is quite long so I will post it in sections.]
