kjappelbaum / oximachinerunner

An easy API for using oximachine.
MIT License
7 stars 5 forks source link

Update packaging #87

Open mpougin opened 1 year ago

mpougin commented 1 year ago

Similar changes to those made in https://github.com/kjappelbaum/oximachine_featurizer/pull/48, and also still a draft since it relies on the changes made in that PR.

mpougin commented 1 year ago

Building on this branch I can install the aiida-lsmo package for aiida-core v2.X, see:

https://github.com/mpougin/aiida-lsmo/tree/fix/update-aiida2

I've also run some tests locally so far, and everything seems to work fine!

mpougin commented 11 months ago

@kjappelbaum made a few more changes to the devops tools. I'm not very familiar with the scikit-learn models, so can't be of much help there. Would it be a lot of work to serialise them into JSON? I suppose you'd also have to add a new Zenodo archive?

My main concern here was making sure I can install & run the aiida-lsmo work chains. It seems that just using the 'mof' model doesn't require access to anything that broke with the new sckit-learn version, so installing and running from this branch works fine for now.

Of course, I'd be keen on making an AiiDA v2.X-compatible release (see https://github.com/lsmo-epfl/aiida-lsmo/pull/110) shortly, and for this I'd need a release of oximachine-runner. I understand you wouldn't want to do a broken release, however...

Let me know if there is any other way I can still help get a new release out the door.

mpougin commented 11 months ago

Tried to downgrade the xgboost and scikit-learn dependencies and look for a solution (see this branch), but I'm afraid that'd restrict our Python versions back to <= 3.8. 😅