Best Hardware/Environment Setups for the Average Joe?

We were having this discussion on another post, and I figured it’d be better to make this its own thread. So I have a 2-aprt question. What is the best solution for the average joe who wants to:

  1. Just learn ML
    or
  2. Compete in kaggle competitions

Obviously for #1 (learning ML), you just need a setup that works reasonably well; no need to pay $$$ for a state of the art 1000-core supercomputer. I started learning ML on a $250 laptop without an nvidia GPU, so basic CNNs took like 10 hours for me. Obviously I could not continue with that setup, so I looked around for a better option.

For #2 (being competitive), I would think you’d need some high-quality hardware or setups, so for those experts out there, what sorts of hardware do you use?

I’ll give the first answer on my own thread, for question #1, since I am reasonably happy with a solution I found that for me is 2x as fast as the standard AWS P2 solution, and costs me less, and that is: I use a paperspace p5000 hourly instance

1 Like

Take a look at this forum thread :
http://forums.fast.ai/t/making-your-own-server/?source_topic_id=3962

1 Like

+1 for paperspace. pain free economical cloud GPU.