-
Notifications
You must be signed in to change notification settings - Fork 104
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
CU-8694fwyje pre load config(s) #447
Conversation
Note: still need to add translation layers (along with tests) that allow loading and converting old format configs. |
…stant key in original
Best I can tell, before (i.e when actions was successful, e.g after 49d25ba) the config(s) loaded off disk weren't really being applied due to having no remapping. As such, some specific tests weren't being run as expected since they used default config instead. This may also have something to do with the other config-related PR #449 since that might mean that some things aren't saved and/or loaded correctly on this branch. In any case, ran out of time. |
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I've checked this for MetaCAT, looks good to me!
RelCAT (old version) should be good too :) |
Decided to go with #473 instead |
EDIT:
This PR has an alternative approach in #473 .
Separate config entries that only (truly) take effect if set before a model is loaded or initialised.
Some of these config entries are used elsewhere in the codebase as well. But most of the time it looks (at least to me) like their main use is at init/load time.
EDIT: This will also add translation layers to the configs for legacy mappings when loading the various configs (main, MetaCAT, TNER, RelCAT).
Also removed some of the unused config entries in RelCAT config.
This affects
NOTE:
I'm open to discussions on which config entries should stay where they were as well as which ones shouldn't be removed from RelCAT config.
PS:
This might create issues downstream (i.e in tutorials and WWC). So if/when this PR will be approved, I'll look into creating PRs for those to go along with this one.
PPS:
The main config's
linking.subsample_after
option is also unused. However, there is commented code for it to be used incontext_based_linker.Linker._train
so I left it in for now.