Fix/multi pathogen training - #89

Merged

jeffreyHoelzel merged 12 commits into

mainfrom

fix/multi-pathogen-training

Jun 23, 2026

Commits on May 22, 2026

docs: added markdown descriptions of bugs/areas to refactor to improve multi-dataset training/validation
jeffreyHoelzel
committed
d5e1b28
fix: implemented check to verify actionable embeddings exist before back prop and optimizer step
jeffreyHoelzel
committed
b432ead
fix: implemented DDP global valid-residue loss fix so training backprops a globally-normalized residue loss
jeffreyHoelzel
committed
3a00253
fix: refactored validation to not use overlapping windowed residues, reducing chances of duplicated residues being counted as one
jeffreyHoelzel
committed
ca4b910
refactor: implemented train-only automatic positive class weighting using train loader instead of labels, removed stale CLI args
jeffreyHoelzel
committed
a7ceb9e

Commits on May 26, 2026

refactor/feat: added threshold-policy selection, added majority vs mean-prob aggregation to prediction/evaluation CLIs
jeffreyHoelzel
committed
1b7e173

Commits on May 27, 2026

Commits on Jun 8, 2026

refactor: implemented CLI flags for appending sequence length to end of each residue embedding; none, raw (i.e., 1476), or inverse (i.e., 1 / 1476)
jeffreyHoelzel
committed
b494bbb

Commits on Jun 22, 2026

Commits on Jun 23, 2026

feat: added subset environment variable to training for testing
jeffreyHoelzel
committed
393cc18