Issues · CAIIVS/raitap · GitHub

Archived Development.md
#55 · stanlrt opened on Apr 12, 2026

Labels Milestones

Tracking/mlflow absent from generated config schema (swallowed circular import)

#357

· stanlrt opened

on Jul 8, 2026

Support non-HuggingFace text models / decouple the tokenizer seam

#348

· stanlrt opened

on Jul 5, 2026

·

🟡 P2: Medium

Implement adversarial robustness for object detection (task family currently opts out)

#343

· stanlrt opened

on Jun 25, 2026

·

🟡 P2: Medium

CLI bootstrap fast-path: skip hardware probe + uv sync + re-exec on provisioned envs (design options)

#329

· stanlrt opened

on Jun 10, 2026

·

Eager matplotlib.pyplot imports defeat lazy-import design: every startup loads matplotlib

#328

· stanlrt opened

on Jun 10, 2026

·

🟡 P2: Medium

Per-chunk gc.collect() + cuda.empty_cache() in batch loops (runs on CPU-only too)

#327

· stanlrt opened

on Jun 10, 2026

·

Robustness assessors repeat the clean forward pass and run verdict forwards unbatched

#326

· stanlrt opened

on Jun 10, 2026

·

🟡 P2: Medium

SHAP explainer rebuilt from scratch for every mini-batch (background forwards repeated per chunk)

#325

· stanlrt opened

on Jun 10, 2026

·

🟡 P2: Medium

Replace algorithm-name string branches in default invokers with declarative registry-spec fields

#324

· stanlrt opened

on Jun 10, 2026

·

🟡 P2: Medium

Per-adapter budget-kwarg vocabulary: robustness semantics hardcodes torchattacks/foolbox spellings

#323

· stanlrt opened

on Jun 10, 2026

·

🟡 P2: Medium

Reproducibility sweep iterates hardcoded [transparency, robustness] instead of generic phase_results

#322

· stanlrt opened

on Jun 10, 2026

·

🟡 P2: Medium

forward_batch_size: phantom run.* config path, consumption-site validation and default

#321

· stanlrt opened

on Jun 10, 2026

·