Training step refactor - Githubissues

raehik commented 11 months ago

this builds off of some shared work in the data step refactoring #85

[ ] replace code obsoleted by previous data step refactoring
- subdomain (bounding box) loading
[ ] improve CLI
- [x] load forcings via either MLflow or filepath #84
- [ ] fix learning rates configuring #90
[ ] replace CLI
- lower priority
[ ] decouple internals
- maybe not in scope. decoupling functions is one thing, but training is accomplished by building up classes and integrating with existing Torch classes in unobvious ways. the Pytorch model itself is one of the larger examples of this (see the DetectOutputSizeMixin class, used as a superclass in FullyCNN)

mondus commented 11 months ago

This should be going into dev

raehik commented 11 months ago

needs rebase (onto dev branch, which should mostly be #85 )

raehik commented 10 months ago

Most likely to be merged with #97 due to overlapping changes (the later steps are fairly interlinked with previous ones).

raehik commented 9 months ago

Re-done in #97. Closing.

m2lines / gz21_ocean_momentum