on every commit when a .py file is changed (but only on the changed py files)
on pull requests (only on changed py files)
singularity container build
runs when a commit changes the .def or periodically once a month (2 at night the first of the month)
scan if configs parse and models can be created (currently only for classification models)
runs on pull requests
Discussion / Future Extensions
Also, the singularity file is cached for faster execution (will be deleted if not used for a week, but can be pulled again then).
However, we might want to discuss whether this would be enough and we drop the package storage and just built it again if needed. Rough time estimates:
load cached container: 30-45s
pull container: 60-90s
build container: 8-10 mins
This should be enough for now. In the future I plan some extensions however:
extend code check using pyright to ensure typing
extend scan for rl configs
To be done before merge:
[x] discuss cache / storage / build stuff
[x] discuss whether any of the future extensions should be implemented already now
[x] make cache available across branches (just realized now that this seems not to be the case)
Added features
Adds three CI pipelines:
Discussion / Future Extensions
Also, the singularity file is cached for faster execution (will be deleted if not used for a week, but can be pulled again then). However, we might want to discuss whether this would be enough and we drop the package storage and just built it again if needed. Rough time estimates:
This should be enough for now. In the future I plan some extensions however:
To be done before merge: