AWS setup procedure pitfalls


(Tom Halasz) #1

After spending more than a week trying to set up my Windows 7 machine on AWS, I want to offer some advice for prepping yourself for the course:

  1. I don’t recommend installing Cygwin to use as a Linux terminal as Jeremy does in the (rather dated) setup video, particularly if you already use Git Bash. The installation alone created conflicts that rendered Git Bash unusable; a nightmare within my nightmare. I backed out of the problem by restoring my Win 7 system to an earlier point.

  2. The Github repository that someone kindly posted, https://github.com/reshamas/fastai_deeplearn_part1/blob/master/tools/aws_ami_gpu_setup.md, was the most helpful resource I found, although it is a bit sketchy for those unfamiliar with working in Linux shell. For instance, one step calls for editing the jupyter_configuration.py file after generating it. I don’t know how to do this (not everyone taking the course majored in CS, did they?). Wasted a lot of time trying to figure out how, and I’m not sure if I successfully did it. Things seem to be working, in any case, but I’m still just at the beginning.

  3. Difficulties aside, it’s probably worth the pain to familiarize oneself with using a service like AWS, as it’s likely to be something you use in the future. I’m easily frustrated, but I learned a lot by calling the billing department and asking many questions even though they don’t really have much technical knowledge. Incidentally, paying for “premium” support might not be a bad way to go because once you’re up and running you can cancel the contract and only be charged for it on a prorated basis ($29/month). I probably would have learned more and faster if I’d done that.