Restructure as DMPKBench with ADMET-MPO and pk-interpretation subsets
#3
by zkoch1 - opened
Rename the packaging to DMPKBench and replace the single default/train split with two configs:
- ADMET-MPO (650 MCQ items from DMPKBench public ADMET-MPO.json)
- pk-interpretation (existing 99 open-answer Result Interpretation items)
Each subset has train/val/test at ~70/15/15 (seed=42).
Repo rename EdisonScientific/dmpkbench-result-interp → EdisonScientific/DMPKBench follows this PR.
zkoch1 changed pull request status to merged