Greetings brilliant minds!
Today’s the day! My boss finally sent me one of 2 Sparks he bought for our team, we are ‘beta testing them’ before turning them over for work. I have read the entire users guide manual several times and will read it again during my unboxing/booting up.
My question is compared to all of you I an an “infant” when it comes to AI. The terminology, the ways of running it everything. Mind blowing. I’m trying to learn at a rapid pace. I’m lucky enough to have a Macbook Pro Ultra Max with 64GB of unified memory so I am able to run some models on it and play with it. I’m eager to see how “bigger model’s” run!
That all said, got any tips/tricks for a total AI newbie. Best tooling to use to start off with to get my feet wet, recommended models (interested in models good at python programming as well automation of IT), Any common landmines to avoid? ect…?
I obviously know my way around a computer and a Linux environment. But I know these can be fragile in terms of blowing them up with the wrong combo of code ect..
So yes I’ve read the manual, plan to read it again, and for someone with my seriously low knowledge of AI, what’s a good place for me to begin playing?
On my Mac I use either LM Studio or Inference to load different models and just play with them. Mostly python related. I have a dumb question but I’ll ask anyways. Be gentle!
I know that the ALX models from inference or LM Studio are tailed for Apple silicon, and that GUUF is another that runs I guess anywhere? Is there one specific for the DGXs HW? Is that what vLLM is?
Dont worry I have to give the Spark back to the company so its not like they bought this expensive device for me exclusively. We bought two and we plan to cluster them, run a few medium sized models and use other software to have it spawn bots that can query other bots for data/information to get an end result. Who knows how well this will go. But we will have the horsepower.
Another random and completely off topic question, how well do these Sparks do for something like Folding@home? Can you get all that horsepower loaded into Folding@home and have it crunch data at an insane speed compared to say a laptop or even a 1U server? Just curious I gotta try it out :)
So any beginner tips/tricks/software/models/ ect you recommend or should avoid let me know please. I want to maximize my time with the Spark testing out things rather than having to factory restore it because I tried to do something advanced and messed it all up :)
Cheers.