Almost there with the Amazon EC2 environment setup with “git clone”, see below.
Also, is there a way to increase memory?
Thank you for suggestions,
Could it be that you don’t have port 443 open for outbound connections?
Memory is specific to instance type - for more RAM, switch to a different instance type. You can read about various instance types here.
Yeah, you need to put this rule to “Inbound rules”, outbound rules should allow traffic to “0.0.0.0/0”
This was the trick to make it work. Thank you!
Better still, use the -L flag to ssh to create a tunnel - that’s more secure. See the workshop video for details - the bit that @yinterian taught.
0.0.0.0/0 - meant to be for security group outbound rules only.
I just setup an EC2 environment following the instructions at https://course.fast.ai/start_aws.html and I ran into a couple of issues that I wanted to call out. I wanted to just go ahead and create a pull request to update the docs, but I’m using my company’s account and because of the way we have VPCs setup, there are some steps that are different and I can’t get screenshots that would be relevant to a user starting from scratch.
- The relevant version of the AMI now is 23.0. I was able to get a screenshot of this (see below)
- I immediately ran out of storage space going through lesson1. By default, the instance gets 75GB of storage. After installing pytorch and everything, that was almost totally used up. Once I started to untar an image data set, I ran into out of storage errors. I started over and added 150GB of storage as part of the setup, and then things worked.
I’d be curious to know if maybe I did something wrong in the setup to use up so much storage. It did seem odd that just installing those few things would leave me with an instance that was using 75gb. If it really is necessary to add more storage, then I’d suggest updating the docs (sorry that I can’t put in a PR for this).
