bnnr 0.2.6

BNNR — Train → Explain → Improve → Prove

BNNR (Bulletproof Neural Network Recipe)

BNNR is a PyTorch toolkit that makes vision models production-ready through a closed loop: train a model, explain its decisions with XAI, improve it with intelligent augmentation, and prove the result with structured reports and a live dashboard.

Supported tasks (v0.2.6):

• Single-label classification (task: classification)
• Multi-label classification (task: multilabel)
• Object detection (task: detection) — COCO-mini / YOLO pipelines; see examples/detection/

Quickstart

1) Install

pip install bnnr

With optional extras:

pip install "bnnr[dashboard]"       # live dashboard (React + FastAPI)
pip install "bnnr[gpu]"             # GPU-accelerated augmentations (Kornia)
pip install "bnnr[albumentations]"  # Albumentations integration

For development from source:

git clone https://github.com/bnnr-team/bnnr.git
cd bnnr
# Dashboard UI must exist before editable install (hatch force-include).
(cd dashboard_web && npm ci && npm run build)
pip install -e ".[dev,dashboard]"

Examples, notebooks, and markdown docs

The PyPI wheel ships the bnnr package only. Runnable scripts (examples/), notebooks, and the documentation tree (docs/) live in the GitHub repository. After cloning, follow docs/examples.md (for example PYTHONPATH=src python3 examples/... from the repo root).

2) Create a minimal config

cat > /tmp/bnnr_quickstart.yaml <<'YAML'
m_epochs: 1
max_iterations: 1
metrics: [accuracy, f1_macro, loss]
selection_metric: accuracy
selection_mode: max
checkpoint_dir: checkpoints_quickstart
report_dir: reports_quickstart
xai_enabled: false
device: auto
seed: 42
candidate_pruning_enabled: false
YAML

3) Run a short training with live dashboard

python3 -m bnnr train \
  --config /tmp/bnnr_quickstart.yaml \
  --dataset cifar10 \
  --max-train-samples 128 \
  --max-val-samples 64 \
  --preset light \
  --with-dashboard \
  --dashboard-port 8080

Expected terminal markers:

• BNNR PIPELINE SUMMARY
• BASELINE TRAINING
• TRAINING COMPLETE
• Report JSON : reports_quickstart/run_.../report.json
• Dashboard : http://127.0.0.1:8080/

After training, keep the process running and open dashboard:

• desktop: http://127.0.0.1:8080/
• mobile: scan QR code printed in terminal (same Wi-Fi)

Stop dashboard with Ctrl+C when done.

4) Inspect the report

RUN_DIR=$(ls -1dt reports_quickstart/run_* | head -n 1)
python3 -m bnnr report "$RUN_DIR/report.json" --format summary

If you prefer one-shot (no live dashboard), run the same command with --without-dashboard.

Main CLI commands

python3 -m bnnr --help
python3 -m bnnr train --help
python3 -m bnnr report --help
python3 -m bnnr list-datasets
python3 -m bnnr list-augmentations -v
python3 -m bnnr list-presets
python3 -m bnnr dashboard serve --run-dir reports --port 8080
python3 -m bnnr dashboard export --run-dir reports/run_YYYYMMDD_HHMMSS --out exported_dashboard

Python API entry points

For custom models and dataloaders:

• BNNRConfig
• BNNRTrainer
• SimpleTorchAdapter
• quick_run()

See docs/api_reference.md and docs/golden_path.md.

Repository docs

• docs/README.md (documentation index and source of truth)
• docs/getting_started.md
• docs/dashboard.md
• docs/configuration.md
• docs/cli.md
• docs/api_reference.md
• docs/golden_path.md
• docs/detection.md
• docs/augmentations.md
• docs/examples.md
• docs/notebooks.md
• docs/artifacts.md
• docs/troubleshooting.md

Requirements

From pyproject.toml:

• Python >=3.10
• Core dependencies include torch, torchvision, numpy, typer, pydantic, pyyaml
• Dashboard extra adds fastapi, uvicorn, websockets, qrcode

License

MIT — free and open source, forever. BNNR is an academic and community-driven project built for learning, research, and fun. No paid tiers, no commercial offerings — just open-source tools for the CV community.