Search results

Threshold minimum precision is not configurable

featureNew feature or requestNew feature or request

Status: Open.

#88 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

--use-pos-weight in Optuna appears misleading

bugSomething isn't workingSomething isn't working

Status: Open.

#87 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Optuna best checkpoint copy path mismatch

bugSomething isn't workingSomething isn't working

Status: Open.

#86 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Raw protein sequence length is appended as an unnormalized feature

maybe

refactor

Status: Open.

#85 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Model has no local sequence modeling beyond ESM embeddings

featureNew feature or requestNew feature or request

refactor

Status: Open.

#84 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Grouped splits prevent family leakage but are not label-stratified

refactor

Status: Open.

#83 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Threshold policy is hard-coded and can be overly conservative

refactor

Status: Open.

#82 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Positive class weight is stale or fold-inappropriate

refactor

Status: Open.

#81 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Training uses overlapping windows while evaluation uses full proteins

bugSomething isn't workingSomething isn't working

Status: Open.

#80 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

DDP and batch loss weighting are likely biased by local valid-residue counts

bugSomething isn't workingSomething isn't working

Status: Open.

#79 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Sparse or zero-valid windows still participate in training

bugSomething isn't workingSomething isn't working

Status: Open.

#78 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026

Label/objective mismatch for residue-level prediction

maybe

refactor

Status: Open.

#77 In LadnerLab/PepSeqPred;

· jeffreyHoelzel opened on May 22, 2026