Feature/evaluation pipeline#55

Merged

milanagm merged 54 commits intomainfrom

feature/evaluation-pipeline

Apr 30, 2026

Commits on Apr 11, 2026

Add evaluation framework scaffold with activity task
milanagm
committed

Commits on Apr 12, 2026

removed stubs to make review easier
milanagm
committed

Commits on Apr 15, 2026

removed activity related components - will push on new branch
milanagm
committed
Revert README changes from evaluation PR
milanagm
committed
Remove notebooks figures, defered to activity related eval that will be pushen on another branch
milanagm
committed
Write predictions as CSV while keeping metrics as JSON
milanagm
committed
Removed masking bootstrap metric failures / removed Exception Catch
milanagm
committed
Clean up README files and docstring wording
milanagm
committed
Add ordinal metrics and align BMI label handling - expanded metrics
milanagm
committed
removed seperat weekly metabolic dataset script to expand the weekly integration into existing mhc metabolic qa dataset
milanagm
committed
removed local env name
milanagm
committed
updated requirements (added opentslm dependancy)
milanagm
committed

Commits on Apr 16, 2026

Commits on Apr 19, 2026

Commits on Apr 20, 2026

Store values as float32 ndarrays instead of nested Python lists
milanagm
committed

Commits on Apr 21, 2026

Commits on Apr 23, 2026

Commits on Apr 24, 2026

Commits on Apr 28, 2026

Allow peft's LoraConfig in torch.load(weights_only=True) checkpoints.
milanagm
committed

Commits on Apr 29, 2026

removed smoke test shell scripts for merge w PR and the text only eval benchmark
milanagm
committed

Commits on Apr 30, 2026

Commit explicit Answer: as model intent — return unknown when candidate is invalid
milanagm
committed