Testing nobrainer docker files

[hvgazula]

#273

Requested an ec2 instance ssh'ed into it and ran the following lines

git clone [email protected]:neuronets/nobrainer.git
cd nobrainer
docker build -t neuronets/nobrainer:master-cpu -f docker/cpu.Dockerfile .
docker build -t neuronets/nobrainer:master-gpu -f docker/gpu.Dockerfile .

Both failed.

[hvgazula]

For more info see

and https://github.com/neuronets/build-docker/actions/runs/7917978936/job/21615337005

[hvgazula]

@satra thoughts? I will move on to kwyk retraining.

[satra]

is nobrainer checked out there? the docker builds assume there is a checkout of nobrainer as that's how dockerhub works. have you tried just building locally on your m2 (adding --platform linux/amd64)

[hvgazula]

I tried it on ec2 after cloning and cd'ing and also on github runner via an action. Sure, I can try the above commands again in my mac.

[satra]

currently this repo is set up to autobuild on docker hub whenever changes or releases are created. and both did build fine 4 months again when the repo was last updated. it's possible something has changed in those 4 months. but since the base docker image is the same, i would suggest checking that you should be able to test the build locally to identify the issue. it's possible that some other command is no longer available.

regarding adding datalad to it, you can check how we do it in dandi. we simply do this (https://github.com/dandi/dandi-hub/blob/dandi/docker/Dockerfile#L52C1-L55C46) into a conda environment, instead of trying to get it from neurodebian.

[hvgazula]

is nobrainer checked out there? the docker builds assume there is a checkout of nobrainer as that's how dockerhub works. have you tried just building locally on your m2 (adding --platform linux/amd64)

[satra]

since this is really a datalad issue, let's just remove that for now and bring it back after we update the trained-models.

[hvgazula]

(for later reference) datalad/datalad-osf#161 for debugging

1. drive link is not working
2. datalad get is inconsistent with some files.