Skip to content

Adding CETT-based thresholding #53

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

Closed
wants to merge 5 commits into from
Closed

Conversation

kaselby
Copy link
Collaborator

@kaselby kaselby commented Jul 10, 2025

Description

Adds basic support for CETT-based thresholding and refactors activation capture.

kaselby and others added 5 commits July 10, 2025 13:19
* Update to log each separate lora size independently and correctly identify which checkpoints to load based on lora size.

Signed-off-by: Kira Selby <[email protected]>

* Store best f1 scores for each layer in central kv store and save best performing model per layer. Add flag to resume training only from best performing lora sizes for each layer.

Signed-off-by: Kira Selby <[email protected]>

* Bugfixes and move save final predictor into layerwise trainer to avoid saving after early exit due to errors.

Signed-off-by: Kira Selby <[email protected]>

* Remove restart_if_missing and add documentation for load_best_only

Signed-off-by: Kira Selby <[email protected]>

---------

Signed-off-by: Kira Selby <[email protected]>
Signed-off-by: Kira Selby <[email protected]>
@kaselby kaselby closed this Jul 26, 2025
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

1 participant