-
Notifications
You must be signed in to change notification settings - Fork 52
Developer calls
Leif Denby edited this page Jan 13, 2025
·
8 revisions
Present: JO , SA, HS, SKC, SW and LCD
Agenda
- Organising, where to host roadmaps:
- In future we will use github milestones to track upcoming milestones rather than separate issues/wiki with roadmap details
- Deprecate neural-lam wiki with roadmap, point to milestones on repo
- Remove mllam-data-prep and weather-model-graphs issues that served to host roadmaps before
- In future we will use github milestones to track upcoming milestones rather than separate issues/wiki with roadmap details
- Next releases,
- Neural-lam:
- v0.3.0: waiting for tests to pass, then @leifdenby will release
- v0.4.0: already decided what will be part of this release
- v0.5.0:
- writing zarr-based output for neural-lam added to roadmap https://github.com/mllam/neural-lam/issues/89
- v0.6.0:
- Not planned for now
- mllam-data-prep:
- existing planned releases (v0.6.0 and v0.7.0) will remain unchanged
- weather-model-graphs:
- v0.3.0 will be next release
- Mathias at DMI will attempt to fix bug https://github.com/mllam/weather-model-graphs/issues/33 and will create patch release v0.2.1 if successful before work for planned v0.3.0 is complete
- Neural-lam:
- Patch updates (bugfixes only):
- Issues/PRs don’t need to go on roadmap (Github milestones)
- Patch releases coordinate on slack
- On testing
- Currently the tests take a long time to execute, JO thinks this is because the test dataset is too big
- Plan is to reduce size of DANRA by cropping to say a domain around Copenhagen. Once this new data is ready then using these smaller datasets can go in on the next release planned.
- Could in future look into caching tests, SKC has experimented with that before
- Could also group tests in future by only running the longest-running tests
- Currently the tests take a long time to execute, JO thinks this is because the test dataset is too big
Link: https://liu-se.zoom.us/j/62128717764?pwd=QnYyRVFOeEtQZHJGSEpHUlJBcTlNZz09
All roadmap changes decided on and discussions can be found in https://docs.google.com/spreadsheets/d/1Q94qpVXeQhywaTuMDrNunIkWJJ_oA4oJ0XXmlEBG_Hc/edit?usp=sharing . Listing only some major points here:
- We will aim for more frequent releases of new versions going forward. To facilitate this we will make sure to have these dev meetings every month.
- Dev meetings will happen every 4th week, at the same time as today. This means that the next one will be on December 16, 13:00-14:00 CET (add to your calendar as recurring, every 4th week :slightly_smiling_face:)
- Meeting link will be https://liu-se.zoom.us/j/62128717764?pwd=QnYyRVFOeEtQZHJGSEpHUlJBcTlNZz09
- I will also post reminders here on slack
- Review policies:
- When planned change is added to roadmap people can write up their name as reviewer. Then that person should approve of change before merging.
- We are not updating the roadmaps with the entries and reviewers from the meeting document above. If there is a roadmap entry where you want to be added as reviewer let us know by replying to this message on slack.
- Apart from reviewers listed in roadmap we only require one approving review before merge.
Next meeting will be 1400 CEST on https://liu-se.zoom.us/j/62128717764?pwd=QnYyRVFOeEtQZHJGSEpHUlJBcTlNZz09
- Brief discussion between atendees. No noteworthy changes decided.
- Progress towards next tagged version (v0.2.0): How close are we? When do we expect to have this complete?
- https://github.com/mllam/neural-lam/pulls?q=milestone%3Av0.2.0
- @leifdenby will complete PR this week and ask for reviews
- Discussion for following release (v0.3.0): Are all changes we want included, or are some missing? How are discussions progressing?
- https://github.com/mllam/neural-lam/pulls?q=milestone%3Av0.3.0
- likely that zarr-reading will be merged in first, generalising graph construction will be based on this
- discussion around graph construction (e.g. work in https://github.com/mllam/neural-lam/issues/4) will continue on weather-model-graphs
- Longer term: Are you working on anything not currently on the roadmap? Is there something in neural-lam that you are missing? Are you working on something, but not sure how to fit it into neural-lam?
- (we didn't have time to discuss this item)
- Is there a need for some more defined roles around the Neural-LAM codebase (e.g. "maintainers", "core developers"). And if so, who should have these roles?
- no strict roles for now, we want to motivate people to review things they have interest and understanding in. Please tag people you think have an interest. @leifdenby is happy to do reviews in general, and ... (feel free to add yourself here)
- What do we require for merging PRs? An ok from one or multiple such "maintainers"?
- we agreed that nothing will be merged without being on the roadmap
- to get things on the roadmap we will use the develop calls to discuss and adapt the roadmap, please come to the developer calls to explain your idea so we can build the thing you need together 😄
- AOB
- next meeting 1400 CEST 5 august 2024