Show HN: Llama2.c on a Commodore C64

#1

Post by AMICABoard » Sat, Feb 22, 2025, 10:07 PM UTC

Warning: Inference takes ages.

Pics & Media: https://x.com/VulcanIgnis/status/1893420241310335329

Show HN: Llama2.c on a Commodore C64
github.com

Re: Show HN: Llama2.c on a Commodore C64

#2

Post by desertkun » Sat, Feb 22, 2025, 10:13 PM UTC

I once tried to slap TLS/SSL stack onto ZX Spectrum – so HTTPS connection on it worked, but it took solid two minutes to just load a certificate – sometimes you gotta do the "I made a screaming robot - why? - hmmmmmm...."

Re: Show HN: Llama2.c on a Commodore C64

#3

Post by AMICABoard » Sat, Feb 22, 2025, 10:27 PM UTC

desertkun wrote: post #2

I once tried to slap TLS/SSL stack onto ZX Spectrum – so HTTPS connection on it worked, but it took solid two minutes to just load a certificate – sometimes you gotta do the "I made a screaming robot - why? - hmmmmmm...."

It had to be done :) If not us, who else would?

Re: Show HN: Llama2.c on a Commodore C64

#4

Post by Cheyana » Sat, Feb 22, 2025, 11:37 PM UTC

“No? Well, now you can!”

Lol, I’m stealin’ that.

Re: Show HN: Llama2.c on a Commodore C64

#5

Post by strangecasts » Sat, Feb 22, 2025, 11:47 PM UTC

desertkun wrote: post #2

I once tried to slap TLS/SSL stack onto ZX Spectrum – so HTTPS connection on it worked, but it took solid two minutes to just load a certificate – sometimes you gotta do the "I made a screaming robot - why? - hmmmmmm...."

That rules, have you made a writeup of it anywhere?

Re: Show HN: Llama2.c on a Commodore C64

#6

Post by AMICABoard » Sun, Feb 23, 2025, 9:45 AM UTC

Cheyana wrote: post #4

“No? Well, now you can!” Lol, I’m stealin’ that.

You can, it's a meme, I stole it from Karpathy :)

Re: Show HN: Llama2.c on a Commodore C64

#7

Post by desertkun » Sun, Feb 23, 2025, 3:27 PM UTC

strangecasts wrote: post #5

desertkun wrote: post #2

I once tried to slap TLS/SSL stack onto ZX Spectrum – so HTTPS connection on it worked, but it took solid two minutes to just load a certificate – sometimes you gotta do the "I made a screaming robot - why? - hmmmmmm...."

That rules, have you made a writeup of it anywhere?

[dead]

Re: Show HN: Llama2.c on a Commodore C64

#8

Post by HenryBemis » Sun, Feb 23, 2025, 7:43 PM UTC

"How this works" - Slowly!

I was chatting with ChatGPT 4o the other day and I was asking it on "what steps/software/tools/etc. I need to set up my own LLM offline on a PC".

I suggested a few PCs I got, they were all super slow, but for the purpose I want it (numerical analysis) I can make it work. But even my stupid old PCs are times-infinity faster than this C64!!

Re: Show HN: Llama2.c on a Commodore C64

#9

Post by AMICABoard » Tue, Feb 25, 2025, 9:07 PM UTC

HenryBemis wrote: post #8

"How this works" - Slowly! I was chatting with ChatGPT 4o the other day and I was asking it on "what steps/software/tools/etc. I need to set up my own LLM offline on a PC". I suggested a few PCs I got, they were all super slow, but for the purpose I want it (numerical analysis) I can make it work. But even my stupid old PCs are times-infinity faster than this C64!!

lol :)