@Joshua O'Keefe:
Since it seems like you have some experience with these at a certain
level (I've never dealt with LLM models, and most of what I know little
about ANNs were all mainly geared toward SLAM and other robotic
navigation/localization needs) - and you mentioned running smaller
models locally...
It's something that I've wanted to do, but I tend to wonder if I would
have the hardware to do anything useful; the best GPU I have available
is in an older model of the Oryx Pro from System76:
https://system76.com/laptops
It's not horribly out of date, imho - it was bought for me by my last
employer, who then gave it to me as part of my severance, in 2023; it
was bought for my day-to-day working circa Dec/Jan of 2021/22.
I don't recall the specs, just that it was completely maxed out, and the
GPU was some NVidia device with 12 Gig of RAM, and the laptop had 64
Gig, I think. It's basically the best system I own.
I guess I'm just wondering how feasible it would be to use what I have
available; I know my currently box (that I'm typing this on) probably
wouldn't be up to par (I only have 8 gig and an 750ti card in it...my
next system, if I ever get around to finishing building it, will have a
970 in it (I think the main memory was going to be 32 gig - but I can't
really recall; I had spec'd the thing out to be "just enough" to run
Cyberpunk 2077 - which also didn't happen). I also have a couple of
spare 750ti and another 970 plus a 960 floating around.
I guess I'd just love to experiment around with this stuff locally; I am
pretty sure I'm covered on the "inference" side of things (which seems
to require far lower amounts of compute), but training might be too
constrained. But I certainly don't have the money (now, or likely in the
future) to throw at a local system with the compute needed for LLM training.
But maybe I'm wrong, which I why I'm asking your opinion...
Andrew L. Ayers
Glendale, Arizona
phoenixgarage.org
github.com/andrew-ayers
------------------------------
Message: 3
Date: Sun, 7 Dec 2025 10:44:23 -0800
From: Joshua O'Keefe <[email protected]>
To: [email protected]
Subject: Re: [M100] Model 100 - Vibe Coding - Text Sweeper
Message-ID: <[email protected]>
Content-Type: text/plain; charset=utf-8
