Add self hosted aws gpu testing

[mikemhenry]

Description

Resolves: #560

I'm sure some secrets will be missing, which we can make org secrets instead of repo secrets so we can use this in other repos

Todos

• Implement feature / fix bug
• Add tests
• Update documentation as needed
• Update changelogNotable points that this PR has either accomplished or will accomplish.

Status

• Ready to go

[mikemhenry]

@mjw99 I thought you would want to take a look at this one, please as any questions you have about how this works! I'd like to start using this as a template to adding GPU testing to other repos, so I'd like to add comments explaining what is going on in the file.

[mikemhenry]

Also, in order to really test this and see what fails we will need to merge it, but I'd like to wait to do that until we get this file documented!

[mjw99]

Hi Mike,

Generally looks good.

I would pin the version of Ubuntu at 20.04 since 22.04 has just arrived; you probably do not want this to latently change whilst you're initially testing this. Also actions/checkout is now at version 3. Also, one super minor thing in devtools/conda-envs/test_env.yaml , it seems to have cython quoted twice.

[mjw99]

Hi Mike,

Generally looks good.

I would pin the version of Ubuntu at 20.04 since 22.04 has just arrived; you probably do not want this to latently change whilst you're initially testing this. Also actions/checkout is now at version 3. Also, one super minor thing in devtools/conda-envs/test_env.yaml , it seems to have cython quoted twice

[mikemhenry]

Good idea! Github is kinda slow when rolling these updates out, but I rather not worry about it actions/runner-images#5490

[mikemhenry]

Also I know this makes the diff messy, but I sorted the packages in each section so it is easier to catch dupes

[mjw99]

LGTM!

[mikemhenry]

Sweet! I expect something to fail post merge, I might have to add the PAT to this repo, but I think it's better to see what fails and fix it iteratively

[mjw99]

Hmmm on merge; I'm seeing:
Error: Input required and not supplied: aws-region

Are the relevant AWS secrets set OK on the master?

[mikemhenry]

Not sure what you mean, it looks like I will need to add secrets.AWS_ACCESS_KEY_ID secrets.AWS_SECRET_ACCESS_KEY secrets.AWS_REGION and secrets.GH_PERSONAL_ACCESS_TOKEN

They are set correctly in the repo that is currently using this yaml file https://github.com/choderalab/perses/
I thought that maybe I had set them on the org level, but per-repo level is nice if we ever need/want to use different aws accounts for security reasons.

[mikemhenry]

I think I know what you mean, no these are not set OK on master so it will fail until I add them, which I will do right now.