Issues

• Fix stale T2 aggregation documentation in paper_metrics_core.py

  documentation
  Task

  AshleyLab/OpenMHC#115 · Filter by author NarayanSchuetz opened on Jul 25, 2026

• CHANNEL_UNITS in hf_config.py mislabels stored data scale (HR is Hz, not bpm)

  bugdocumentation
  Bug

  AshleyLab/OpenMHC#102 · Filter by author NarayanSchuetz opened on Jul 9, 2026

• Performance footguns: per-call tracemalloc, repeated dataset/parquet reloads, path-keyed caches

  performance
  AshleyLab/OpenMHC#99 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Umbrella: dead code + drifted duplication inventory

  enhancement
  Task

  AshleyLab/OpenMHC#98 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Umbrella: fail-loud violations — swallowed exceptions and silent defaults across the eval + publishing pipeline

  bugmedium priority
  AshleyLab/OpenMHC#96 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Packaging hygiene beyond #17: generic top-level packages, undeclared pyyaml, missing pandas<3 cap

  bugmedium priority
  AshleyLab/OpenMHC#95 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• No CI + zero tests on the stages that define the benchmark (masking generators, Track 3 metric math, split disjointness)

  enhancementhigh priority
  AshleyLab/OpenMHC#94 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• upload_leaderboard_substrate: silent overwrite of existing entries; forecasting bypasses all validation; wrong-file fallback

  bug
  Bug

  AshleyLab/OpenMHC#93 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• openmhc: harden archive extraction — tarfile.extractall without filter="data", no checksum validation of downloads

  bug
  AshleyLab/OpenMHC#92 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• openmhc: evaluate_imputation(version="full", max_samples=N) crashes with an obscure IndexError

  bugmedium priority
  AshleyLab/OpenMHC#91 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• openmhc: _ensure_labels_env silently pins the process to the first dataset root

  bugmedium priority
  AshleyLab/OpenMHC#90 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 3: ZeroToNaNTransform heuristics are inert when applied to full multi-month trajectories

  bugmedium priority
  AshleyLab/OpenMHC#89 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 3: resume-skip keyed only on file existence — silently serves stale or partial predictions

  bugmedium priority
  AshleyLab/OpenMHC#88 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 3: quantile forecasts are never repaired or counted by the seasonal-naive fallback

  bugmedium priority
  AshleyLab/OpenMHC#87 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 3: AutoARIMA baseline is not seasonal (sp never set); raw NaN→0.0 fill; silent last-value fallback

  bug
  Bug

  AshleyLab/OpenMHC#86 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 3: missing split file silently falls back to a RANDOM user split

  bug
  Bug

  AshleyLab/OpenMHC#85 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 2: split loader dropped train/val/test overlap validation (leaky split files accepted silently)

  bug
  Task

  AshleyLab/OpenMHC#84 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 2: Hydra inline-arch construction paths are dead on arrival (LSM2 + PyPOTS)

  bugmedium priority
  AshleyLab/OpenMHC#83 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 2: mhc-impute-eval silently ignores sensitivity=on — no subgroup rows, methods drop out of live fair_skill

  bug
  Bug

  AshleyLab/OpenMHC#82 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 2: MaskCache.load has no sample-alignment validation against the precompute run

  bugmedium priority
  AshleyLab/OpenMHC#81 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 2: compute_fair_skill_scores has no guard on its per-cell-mean input contract; absent baseline yields silently empty output

  bug
  Bug

  AshleyLab/OpenMHC#80 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Track 2: pair writer's float16 cast can mint inf AFTER the finite filter (uncounted user dropout from paired skill)

  bug
  Bug

  AshleyLab/OpenMHC#79 · Filter by author NarayanSchuetz opened on Jul 5, 2026

• Forecasting substrate reducers should honor metric/channel filters

  AshleyLab/OpenMHC#37 · Filter by author NarayanSchuetz opened on Jun 22, 2026

• test_imputation_bca.py imports nonexistent aggregate_imputation_paper_metrics.py (collection fails)

  AshleyLab/OpenMHC#34 · Filter by author NarayanSchuetz opened on Jun 22, 2026

• Stale doc: paper_results README says forecasting (Track 3) Phase 3 bootstrap is 'not yet implemented' — it is

  AshleyLab/OpenMHC#29 · Filter by author NarayanSchuetz opened on Jun 21, 2026