by

Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Andrew Nelson
    @andyfaff
    Hi all, I made a repo for a possible test suite for the community. The idea is that different programs can test for correctness using it (which I believe was one of the outcomes of the meeting). I have no problems with such a repo being hosted somewhere more suitable.
    https://github.com/andyfaff/orso
    Brian Benjamin Maranville
    @bmaranville
    Thanks @andyfaff for getting things rolling... I think a set of tests is essential for any work. Would you mind if I pull all your files into https://github.com/reflectivity/analysis ?
    Andrew Nelson
    @andyfaff
    Sure. It was more of an effort to get things off the ground. So I'm happy for any reorgs on the layout and on the tests themselves.
    Brian Benjamin Maranville
    @bmaranville
    @andyfaff The files have been added under the validation/ folder in the reflectivity/analysis repo on github (I just invited you to have "maintainer" access on that repo)
    Andrew Nelson
    @andyfaff
    Nice. If you have any ideas of how to organise the files in a better fashion please go for it. What do you think about the way the files with the layer parameters are written?
    Andrew Nelson
    @andyfaff
    Perhaps we could use CI to run all the test examples we make against refl1d/refnx/{any package who wants to amend the CI scripts to test against}.
    Andrew McCluskey
    @arm61
    Hey all, was trying to work on a logo for ORSO. I have made a repo to keep ideas for the logo (if anyone else fancies making one) and eventually all of the logo assets. All comments/design suggestion on mine are welcome. https://github.com/reflectivity/logo/tree/master/arm
    Andrew Nelson
    @andyfaff
    @arm61, everyone, I was wondering if we could get squatters rights on a docker organisation name. It seems orso and reflectivity are already taken.
    Wojciech Potrzebowski
    @wpotrzebowski
    @arm61 I like the logo!
    Andrew Nelson
    @andyfaff
    I've been having a bash on the analysis repo on github. The idea behind this is to check that various calculators are producing the same reflectivity for the same structure.
    New changes include:
    • change/rebuild the dockerfile to provide a common container for everyone running the same tests
    • refactor the test collection and validation scripts to reduce duplication
    • check (black) that the validation scripts are linted
    • create a github actions to run the validation. This could become a cron job that can run every night.
    The last point installs refnx/refl1d and builds+installs BornAgain, then runs the validation scripts for refnx/refl1d. The validation scripts are not yet activated for BornAgain because it uses a different roughness approach (I'm investigating this at the moment).
    I'd then like for resolution smearing to be added to the validation.
    Andrew Nelson
    @andyfaff
    Happy for more analysis packages to be examined (GenX?), but they need to be Linux'able and testable from a script.
    Also I'm testing BornAgain and refl1d from tip of their respective HEAD, which is less than ideal. refl1d because then one doesn't need to install wxpython (it takes ages for it to build), and BornAgain because one needs the develop branch to use different RoughnessModels (i.e. NevotCroce) to be able to compare.
    Andrew Nelson
    @andyfaff
    Now running:
    • BornAgain validation
    • on a cron job at midnight (as well as pull requests)
      could do with some more validation tests though.
    Brian Benjamin Maranville
    @bmaranville
    @andyfaff : are you getting the same answer from all the engines you're testing?
    Andrew Nelson
    @andyfaff
    It's early days but yes. Had trouble installing refl1d because wxpython took forever to build. Now I'm pip installing from git and happy days. Bornagain was different, no problems installing, but they've only just got nevot Croce roughnesses. Now all 3 are working and validating. Could do with some more tests.
    Next step resolution. Turns out that exact resolution kernel important for relaxed tof measurements.
    (why doesn't gitter send my phone messages that there are updated conversations)
    Would be good to get exact resolution into data file. I.e q, p(q) info, where p(q) is probability. Could still do this with flat text file.
    Andrew Nelson
    @andyfaff
    This is the badge for the validation. Didn't put it into README.md because it implies things (like it's ok to validate on the development hashes rather than a release tag).
    Validation workflow
    jfkcooper
    @jfkcooper
    We have a had a few enquiries (in the NR group at ISIS) about a library of models for various analysis programs, to give users a place to start from. I thought the gitlab here would be the most appropriate place. Any thoughts?
    acaruana2009
    @acaruana2009

    Thanks Jos. The basic idea would be to start with basic models and some form of description text with it to explain what it does. For the more complex models maybe break it down in to code snippits with the total model included in the same folder.

    So for example for Refl1d this would allow for users to play around with different functions independent of the total model, say for simulations.

    Also I would be happy to include the Refl1d data loaders I have for the ISIS beamlines.

    jfkcooper
    @jfkcooper
    I have made some folders, I think put files in them if you like, and add more folders for analysis programs I haven't added
    Andrew McCluskey
    @arm61
    I like this idea! Potentially for refnx just linking to the refnx_models repo will be sufficient
    acaruana2009
    @acaruana2009
    Thanks! I will start adding some of the simpler models I have.
    jfkcooper
    @jfkcooper
    @arm61 I have no idea how to do that, but sounds good, would this somehow enable others to add simultaneously to both, or just always upload to the refnx one?
    Andrew McCluskey
    @arm61
    I was thinking more just having a normal html link to the refnx-models repo
    jfkcooper
    @jfkcooper
    that makes sense
    jfkcooper
    @jfkcooper
    and is now done
    Andrew Nelson
    @andyfaff
    @bmaranville the week of 5-11 October is not possible for me, I'm on holidays then. I don't reckon I'll be able to make any of the 0100 to 0500 timeslots either.
    Brian Benjamin Maranville
    @bmaranville
    @andyfaff what timeslots would be possible for you? I think other working groups have split into two meetings, to handle the widely disaparate time zones.
    Andrew Nelson
    @andyfaff
    I marked down some possible spots on the doodle poll. The week starting 12th is also difficult as I'm running a mail-in expt by myself, but we'll have a go.